0018-8646/2001/$5.00  (C)  2001 IBM

Influence of trapped and interfacial charges in organic multilayer 
light-emitting devices

by W. Riess, H. Riel, T. Beierlein, W. Brutting, P. Muller, and P. F. Seidler

Trapped and interfacial charges have significant impact on the performance of 
organic light-emitting devices (OLEDs). We have studied devices consisting of 20 
nm copper phthalocyanine (CuPc) as the buffer and hole-injection layer, 50 nm 
N,N'-di(naphthalene-1-yl)-N,N'-diphenyl-benzidine (NPB) as the hole transport 
layer, and 65 nm tris(8-hydroxyquinolinato)aluminum (Alq[sub]3[/sub]) as the 
electron transport and emitting layer sandwiched between a high-work-function 
metal and a semitransparent Ca electrode. Current-voltage measurements show that 
the device characteristics in the negative bias direction and at low positive 
bias below the built-in voltage are influenced by trapped charges within the 
organic layers. This is manifested by a strong dependence of the current in this 
range on the direction and speed of the voltage sweep. Low-frequency 
capacitance-voltage and static charge measurements reveal a voltage-independent 
capacitance in the negative bias direction and a significant increase between 0 
and 2 V in the given device configuration, indicating the presence of negative 
interfacial charges at the NPB/Alq[sub]3[/sub] interface. Transient experiments 
show that the delay time of electroluminescence at low voltages in these 
multilayer devices is controlled by the buildup of internal space charges, which 
facilitates electron injection, rather than by charge-carrier transport through 
the organic layers. To summarize, our results clearly demonstrate that the 
tailoring of internal barriers in multilayer devices leads to a significant 
improvement in device performance. 

1. Introduction 

Early work on molecular crystals clearly demonstrated that the relevant 
mechanism of electroluminescence (EL) in organic solids is of the injection type 
[1]. Injection luminescence requires several steps, including the injection, 
transport, capture, and radiative recombination of positive and negative charge 
carriers inside an organic layer with an energy gap suitable for yielding 
visible light output. A very successful approach to separately optimizing these 
individual steps is the concept of organic multilayer light-emitting devices 
(LEDs) using hetero-interfaces between different organic materials [2]. The 
simplest organic LED (OLED) of this kind incorporates a hetero-interface between 
a hole-conducting material (usually a triphenyl-amine derivative) and an 
electron-conducting aluminum chelate complex (Alq[sub]3[/sub]), where light 
emission is generated in the Alq[sub]3[/sub] layer close to the organic-organic 
interface [3]. 

The optimization of these devices requires on the one hand an improvement in the 
fluorescence yield of the emitting material, which can be achieved by doping, 
and on the other hand a well-balanced injection of positive and negative charge 
carriers. Moreover, since these devices are current-driven, good 
charge-carrier-transport capabilities are considered a prerequisite for a high 
luminous efficiency. For example, low charge-carrier mobilities ([muon] << 1 
cm[sup]2[/sup]/Vs) can lead to space-charge-limited currents, especially if one 
contact is able to inject more carriers than would be present in thermal 
equilibrium [4], and thus limit the device performance [5-8]. Furthermore, in 
the presence of an organic-organic interface in a hetero- or multilayer device, 
the energy level offset at the interface can be an additional source of the 
buildup of interfacial space charges. Even in the absence of these energy 
barriers, the abrupt differences of charge-carrier mobilities in the respective 
layers of typical multilayer devices (see Section 4) lead in effect to the 
presence of mobility barriers at hetero-interfaces, which in turn can be the 
source of interfacial space-charge formation. Obviously, knowledge of the charge 
and field distributions inside a device as well as their variation with the 
applied voltage is crucial for the physical understanding of device operation. 
Also, from the viewpoint of technical applications, the avoidance of space 
charge is important for device optimization with respect to efficiency, temporal 
response, and long-term stability. 

In the following sections we describe our experimental evidence for the presence 
of space charges in metal anode/CuPc/NPB/Alq[sub]3[/sub]/Ca devices by 
current-voltage, capacitance-voltage, and transient electroluminescence 
measurements. 

2. Device preparation and experimental methods 

Our devices are built on glass substrates (Schott AF 45) precoated with a 
high-work-function anode such as ITO, Pt, Ir, Ni, or Pd (~~75 nm thick). The 
organic multilayer structure consists of CuPc as the buffer layer, NPB as the 
hole transport layer, and Alq[sub]3[/sub] as the electron-transporting and 
emitting layer (Figure 1). The thicknesses of the different organic layers are 
given in the body of the paper. When a metal anode was used, EL was observed 
through a 20-nm- thick Ca cathode. The active area of each of our devices was 2 
mm X 3 mm. Prior to use, all organic materials were purified by vacuum 
sublimation. Depositions were carried out in a high-vacuum system (Leybold) by 
thermal evaporation from resistively heated tantalum and tungsten boats. The 
base pressure in the chamber ranged between 4 X 10[sup]-7[/sup] and 1 X 
10[sup]-6[/sup] mbar. Typical deposition rates for the organic compounds and the 
metal were 1 [Ampere]/s. The evaporation chamber is attached directly to an 
argon glove-box system, which allows devices to be fabricated, characterized, 
and encapsulated under inert conditions. Current-voltage (I-V) and 
luminance-voltage (EL-V) characteristics were measured with a Hewlett-Packard 
parameter analyzer (HP 4145B) and a sensitive Si photodiode (Hamamatsu S2281). 
The luminance calibration of the photodiode was obtained with a Photo Research 
PR704 spectroradiometer. Transient electroluminescence (TEL) was measured in a 
specially designed setup that allows one to detect simultaneously time-dependent 
EL, current, and voltage across the device. The OLEDs were characterized in a 
modified HP 16058A test fixture with a Hamamatsu photomultiplier 5783-01 (time 
resolution ~~ 0.65 ns) located directly on top of the emitting area to detect EL 
intensity. An HP 8116A dc pulse/function generator (50 MHz, rise time ~~ 7 ns, 
decay time ~~ 10 ns) was used to apply rectangular voltage pulses to the device. 
The pulse length was varied between 8 and 300 ms, with a duty cycle of 10%. The 
function generator permits the OLED to be driven with various positive and 
negative offset voltages before the rectangular voltage pulses are applied. The 
photomultiplier was connected to the 50-[Omega] input resistance of a digital 
oscilloscope (Tektronix 2440) to record the EL signal. A second digital 
oscilloscope (Tektronix 2440) allowed one to monitor the voltage pulse and the 
time-dependent current flow through the device simultaneously with the EL 
signal. Impedance (Z) and capacitance-voltage (C-V) measurements were carried 
out with a frequency response analyzer (Solartron Instruments SI1260). Typically 
the oscillator level was set to 50 mV, and the measurement was averaged over 100 
cycles of the respective frequencies. Static discharge measurements were 
performed with a Keithley electrometer (EM 617) and the device affixed in an HP 
8116A test fixture. Charging of the OLED was performed for a distinct time of 
one minute by the built-in voltage source of the electrometer. Directly after 
charging, the device was separated from the source and connected to the Coulomb 
input of the electrometer to measure the total stored charge of the OLED. 
Switching was done manually, and because of the excellent electrical isolation 
of the setup and the high impedance of the device, discharging effects became 
negligible at low voltages. All measurements were carried out with encapsulated 
devices at room temperature. 

3. Experimental results 

Current-voltage characteristics 

Figure 2 shows the dependence of I-V and EL-V characteristics on sweep direction 
and sweep speed for a metal anode/CuPc/NPB/Alq[sub]3[/sub]/Ca device. In Figure 
2(a) the voltage has been incremented in steps of 50 mV from -3 to 7 V and back 
again. Current and brightness were detected with the parameter analyzer set to a 
"medium" integration time[foot1] and the delay time between individual data 
points to 0. In this mode, the time required for acquisition of a single data 
point depends on the magnitude of the detected current and ranges from 10 to 100 
ms for the given device. Figure 2(b) was measured with a "long" integration time 
and an additional delay of 10 s between individual data points. In both cases 
the current increases strongly in the forward bias direction above a threshold 
voltage of 2 V and is virtually unaffected by sweep direction and speed. The 
onset of EL determined at 0.1 mcd/m[sup]2[/sup] is 2.1 V. This onset is not 
affected by external parameters. We note that the onset voltage for detectable 
EL is significantly lower than the optical gap of Alq[sub]3[/sub] (2.7 eV) and 
rather corresponds to the built-in voltage (i.e., difference in work function) 
of the two electrodes. Remarkable differences in the current flow are observed 
below 2 V, where no detectable EL and thus no double carrier injection takes 
place. The measurement with integration time "medium" shows a strong hysteresis 
between up and down sweeps [Figure 2(a)]. In particular, the voltage at which 
the current passes through zero is not at zero bias but at about -1.4 V for 
increasing voltage and around +1.9 V for the other sweep direction. Furthermore, 
under these measurement conditions a step-like structure at about 0.5 V is 
observed in both sweep directions, which we found to be characteristic for a 
multilayer structure containing CuPc. We do not observe this feature if the CuPc 
layer is omitted. The delay time of 10 s between subsequent data points used in 
the slow measurement is sufficient to achieve zero crossing of the current at 0 
V. In addition, the structure at about 0.5 V disappears. The described 
hysteresis of the I-V characteristics is a clear indication of the presence of 
space charges in the device. During these measurements we observe a higher 
current for decreasing voltage between 2 and 0.5 V and below -1 V. In these 
regimes the current is also relatively noisy. This is most likely caused by 
leakage currents, which can occur after the device has been operated for longer 
times at higher positive bias. However, these leakage currents are reversible 
and can be suppressed by applying negative voltage for a certain time. 

Bias-dependent capacitance 

The measurement of bias- and frequency-dependent device capacitance is a 
well-established technique for the investigation of conductivity, doping 
concentration, and trap states/density in organic semiconductors. For some 
polymeric semiconductors, this method has been successfully used to prove the 
existence of p-type doping leading to Schottky junctions with low-work-function 
metals [10-12]. Furthermore, in combination with temperature-dependent 
measurements, the energy of shallow acceptor states and deep trap states has 
been determined [13]. The materials used in this study, especially NPB and 
Alq[sub]3[/sub], are expected to have very few intrinsic carriers and extrinsic 
dopants. In addition, they have been purified before usage. Only CuPc is known 
to be easily doped by atmospheric oxygen, leading to p-type conducting behavior 
[14, 15]. Thus, in the absence of free or trapped charges inside the device, we 
expect to measure a bias- and frequency-independent device capacitance 
corresponding to a series of three dielectric media with dielectric constants 
[epsilon][sub]r,i[/sub], thickness d[sub]i[/sub], and area A. Given the active 
device area and thickness of the individual layers together with the dielectric 
constants of the materials in the range of 3.5 to 4, one can estimate the device 
capacitance to be about 1.5 to 2 nF. 

Figure 3 shows the capacitance as a function of frequency of the same device as 
in Figure 2 for different applied bias values and as a function of bias for a 
fixed frequency of 10 Hz. The frequency-dependent measurements in Figure 3(a) 
show that at zero and negative bias the capacitance is essentially 
frequency-independent up to some 10[sup]4[/sup] Hz, with a value in the 
neighborhood of 2 nF, as estimated above. Above 10[sup]5[/sup] Hz the 
capacitance drops rapidly to a value much lower than the one corresponding to 
the dielectric constants of the materials. This can be ascribed to parasitic 
effects due to lead/contact resistances and capacitances, which are not 
discussed further here. The frequency dependence of the capacitance is almost 
identical for all applied bias values above ~~200 Hz. Below 200 Hz and for 
positive bias (1 and 2 V), an increase in the capacitance is observed. At low 
frequencies (f < 10 Hz), the capacitance seems to saturate at a value 
independent of the bias. To study the bias dependence in more detail, the 
applied bias was varied at a fixed frequency of 10 Hz. This frequency is in most 
cases low enough to monitor the saturation value of the capacitance at low 
frequencies and still gives a good signal-to-noise ratio. The curves shown in 
Figure 3(b) were taken from -5 to 3 V in steps of 25 mV, with an additional 
delay of 10 s between the individual data points. No difference was observed for 
the opposite sweep direction. Whereas there is only a very weak bias dependence 
in the negative direction, the capacitance increases significantly for positive 
bias at about 0.2 V and reaches a maximum at 2.1 V, which coincides with the 
onset of EL. Above this voltage the capacitance decreases sharply to a value of 
about 2 nF before it finally drops further as the bias exceeds 2.5 V. It is 
important to note that in the regime of double carrier injection accompanied by 
recombination, the capacitance is not well defined; thus, no measurements for 
voltages above 3 V have been performed. 

To determine the low-frequency saturation value of the device capacitance, we 
have performed static discharge measurements. Figure 4 shows the measured charge 
Q for different charging voltages V and the calculated capacitance using the 
relation 

    [Delta]Q
C = -------- .                                                              (1)
    [Delta]V

Although the scattering of data is considerable, it is clear that the 
capacitance in the positive bias direction increases and reaches a maximum at 2 
V with a value about twice as high as in the negative direction. At 2 V, the 
maximum value of 5 nF is slightly higher than that obtained in 
frequency-dependent measurements, indicating that in the latter the capacitance 
has not reached its saturation value at the frequency of 10 Hz used here. In the 
negative bias direction, the capacitance has approximately the same constant 
value of 2 nF as in the ac measurement. The slight decrease for increasing 
negative bias can be explained as a consequence of leakage currents leading to a 
partial discharge during the time of switching between the voltage source and 
the Coulomb meter. Consequently, the measuring error increases with higher 
negative voltage. 

Transient electroluminescence (TEL) 

In TEL measurements one is interested in the time-resolved EL response to a 
pulse excitation, usually a rectangular voltage pulse. Figure 5 shows typical 
current and EL signals upon excitation with a rectangular voltage pulse (4 V). 
The EL signal is characterized by a finite delay time between the application of 
the voltage pulse and the first appearance of EL, an extrapolated rise time to 
reach a steady-state value, and finally the plateau value itself. The 
time-dependent behavior of the current is characterized by a fast initial 
decrease from a high starting point (the cutoff in Figure 5 is caused by the 
limited dynamic range of the oscilloscope) to the steady-state value, which 
corresponds to the current in the I-V characteristics at the given voltage. The 
initial decrease of the current also contains the charging current of the 
device. However, this process is expected to be much faster than the displayed 
decay, because with the given device capacitance of 2 nF and a series resistor 
of 100 [Omega] (to measure the current), the resulting charging current should 
have decayed after roughly 5 RC time constants, yielding about 1 [muon]s for 
this process. 

The observation of delayed EL has been used to extract charge-carrier mobilities 
and their field dependence in polymeric and organic LEDs [16-20]. Additional 
information about the kinetics of charge-carrier recombination and trapping can 
be obtained from the rise to the steady state and the decay of the EL signal 
after the voltage pulse has been switched off. However, especially in multilayer 
structures, the interpretation of these transient experiments is not as 
straightforward as in a single-layer device because the injection of both types 
of carriers may not occur simultaneously [21]. 

Figure 6 shows normalized TEL traces for voltage pulses with different 
amplitudes and durations on a logarithmic time scale. The high sensitivity of 
the setup allows us to detect TEL with a reasonably good signal-to-noise ratio 
at a pulse amplitude as low as 2.5 V, which is only 0.4 V higher than the onset 
voltage detected in dc measurements. At this voltage the delay time is almost 10 
ms. It rapidly becomes shorter with increasing pulse amplitude, and at a voltage 
of 5 V the delay time is only 3 [muon]s, i.e., about three orders of magnitude 
lower. At 14.2 V (the highest voltage accessible with the pulse generator used), 
the delay time is less than 500 ns. Figure 7 shows the dependence on the applied 
voltage of the determined delay times as defined in Figure 5. Remarkably, the 
delay times vary over more than three orders of magnitude in a narrow voltage 
range between 2.5 and 5 V. 

Figure 8 shows the transient response of the EL signal, and the inset shows the 
device current at a 5-V pulse amplitude on a logarithmic time scale extending 
from 1 [muon]s to 300 ms. After reaching a first plateau value in the time range 
from about 40 to 400 [muon]s, the EL signal and the current increase further--it 
takes more than 10 ms before they finally achieve a stationary value. The small 
bump in the EL signal at about 10 ms has no corresponding feature in the 
current. Detailed investigations have shown that this is an artifact of the 
photomultiplier system. 

4. Discussion 

Because electron-hole recombination in a multilayer device takes place at the 
interface between the hole-transporting and emitting layers, the 
NPB/Alq[sub]3[/sub] interface is of crucial importance for device operation. To 
enhance device reliability and long-term stability, however, we inserted an 
additional CuPc hole-injecting/buffer layer. For the sake of simplicity we treat 
these devices as two-layer systems composed of a hole and an electron-conducting 
compartment having only one organic hetero-interface. Where necessary, we 
mention the peculiarities of the CuPc/NPB interface. 

Steady-state device characteristics of organic light-emitting devices 

From the I-V and EL-V characteristics shown in Figure 2, it is immediately 
obvious that the operating voltage range of the OLEDs investigated can be 
divided into two regimes; in the forward bias regime V > 2 V, light emission is 
observed, and current flow is orders of magnitude higher than in the second 
regime, the region V < 2 V, where no light is emitted. The voltage in the 
forward direction, where the onset of significant current flow is detected, is 
commonly considered to be the built-in potential V[sub]BI[/sub], which is to a 
first approximation given by the difference between the work functions of the 
anode and cathode metals. In a system of one or more organic materials without 
free carriers sandwiched between two metal electrodes having different work 
functions, the external voltage must overcome the resulting built-in voltage 
before a net drift current can flow through the device. Therefore, the relevant 
quantity is not the applied voltage V but the corrected effective value V - 
V[sub]BI[/sub] [22]. For our experiment, we estimated a built-in potential 
V[sub]BI[/sub] of about 2 V by measuring the maximum open-circuit voltage of 
illuminated samples. 

The observed dependence of the current on sweep speed and direction indicates 
that extremely slow processes are involved. Numerical simulations on NPB 
single-layer devices by Nguyen et al. show that the hysteresis observed in the 
I-V characteristics can qualitatively be explained by the presence of deep 
traps, which require large time constants to reach thermal equilibrium after a 
variation of the voltage across the device.[foot2]

A crucial point for a more quantitative understanding of the device 
characteristics is therefore the knowledge of the electric field distribution 
inside the individual layers. This is of overall importance because the 
hole-conducting (NPB) and electron-conducting materials (Alq[sub]3[/sub]) of 
this device show quite different charge-carrier transport properties. Whereas 
NPB and related triphenyl-amines are purely hole-conducting materials with 
relatively high hole mobilities and very weak dependence of the mobility on the 
electric field F [[muon][sub]h,NPB[/sub] = 6.1 X 10[sup]-4[/sup] 
cm[sup]2[/sup]/Vs o exp(1.5 X 10[sup]-3[/sup] [square root](cm/V) o [square 
root]F)], Alq[sub]3[/sub] shows predominantly electron transport with lower 
carrier mobilities and a strong electric field dependence 
[[muon][sub]e,Alq[sub]3[/sub][/sub] = 7.1 X 10[sup]-10[/sup] cm[sup]2[/sup]/Vs o 
exp(1.2 X 10[sup]-2[/sup] [square root](cm/V) o [square root]F)] [23, 24]. On the 
other hand, recent time-of-flight and TEL measurements have shown that hole 
transport in Alq[sub]3[/sub] cannot be neglected, although the low-field hole 
mobility is considerably lower than the electron mobility 
[[muon][sub]h,Alq[sub]3[/sub][/sub] = 6 X 10[sup]-11[/sup] cm[sup]2[/sup]/Vs o 
exp(9 X 10[sup]-3[/sup] [square root](cm/V) o [square root]F)] [25, 26]. Owing to 
the strong field dependence of the charge-carrier mobility of holes and 
electrons in the Alq[sub]3[/sub] layer, the time scale to reach a stationary 
charge distribution depends strongly on the electric field distribution in this 
layer. A further point concerns the injection of carriers from the electrodes 
into the organic and from one organic layer into the adjacent one, where the 
electric field in the individual layers and at the interfaces also plays an 
important role. 

To a first approximation, one can assume from the energy-level diagram of the 
OLED device shown in Figure 1 that the voltage drops homogeneously over the 
entire device. Since NPB and Alq[sub]3[/sub] have similar dielectric constants, 
the electric field in the hole- and electron-conducting layers is given by 
F[sup](a)[/sup] = (V - V[sub]BI[/sub])/d[sub]tot[/sub], where d[sub]tot[/sub] is 
the total thickness of both organic layers. The capacitance in this case would 
be C[sup](a)[/sup] = 
[epsilon][sub]r[/sub][epsilon][sub]0[/sub]A/d[sub]tot[/sub], where A is the 
device area and [epsilon][sub]r[/sub] is the permittivity. However, such a 
situation is not very realistic for the OLED operated under forward bias 
conditions, because from the large difference in hole and electron mobility in 
NPB and Alq[sub]3[/sub], respectively, one has to expect that the 
Alq[sub]3[/sub] layer limits the current, where consequently most of the applied 
voltage will drop. Indeed, recent electroabsorption measurements by Rohlfing et 
al. have shown that the voltage drop at the Alq[sub]3[/sub] layer for positive 
bias is almost ten times larger than that at the NPB layer [27]. Thus, the 
electric field in the NPB layer is to a first approximation negligible compared 
to the field in the Alq[sub]3[/sub] layer; in the extreme example, 
F[sup](b)[/sup][sub]NPB[/sub] = 0, and in the Alq[sub]3[/sub] layer, 
F[sup](b)[/sup][sub]Alq[/sub] = (V - V[sub]BI[/sub])/d[sub]Alq[/sub], which is 
larger than F[sup](a)[/sup]. Such a discontinuity in the electric field must be 
accompanied by an accumulation of charge carriers at the interface (the simplest 
case being in an infinitely thin layer) via 

                             [sigma]
[Delta]F = ------------------------------------------ ,                     (2)
           [epsilon][sub]r[/sub][epsilon][sub]0[/sub]

where [sigma] is the number of charges per unit area sitting at the interface, 
and equal dielectric constants [epsilon][sub]r[/sub] in both materials have been 
assumed. The capacitance is now determined by the Alq[sub]3[/sub] thickness 
alone and is, with C[sup](b)[/sup] = 
[epsilon][sub]r[/sub][epsilon][sub]0[/sub]A/d[sub]Alq[/sub], larger than 
C[sup](a)[/sup]. 

From our capacitance-voltage (C-V) measurements (Figure 3) we can indeed see 
that there is a transition from a situation similar to case (a), with the 
capacitance equal to the geometrical capacitance of the device for V < 0 V, to 
the situation in case (b) for V approaching 2 V, where the capacitance is now 
about twice as large as in the reverse bias direction. This corresponds well to 
the thickness of the Alq[sub]3[/sub] layer of 65 nm compared to about 70 nm for 
the sum of the CuPc and NPB layer thicknesses. By a systematic variation of both 
organic layer thicknesses on ITO/NPB/Alq[sub]3[/sub]/Ca devices, Berleb et al. 
have recently found that the value of the capacitance in the reverse bias 
direction always corresponds to the total thicknesses of all organic layers, 
while the value at 2 V is determined solely by the thickness of the 
Alq[sub]3[/sub] layer [28]. Moreover, the increase in the capacitance at a 
voltage well below the built-in voltage was explained by the presence of 
negative interfacial charges at the NPB/Alq[sub]3[/sub] interface, resulting in 
a discontinuity of the electric field as discussed above. Our C-V measurements 
also prove that under a sufficiently large negative bias, the device actually 
behaves like a dielectric with no mobile charges inside the organic layers. The 
weak voltage dependence of the capacitance observed in this range can be 
attributed to the presence of the CuPc layer (presumably some residual doping) 
and is not seen if this layer is omitted. When the bias voltage is increased 
above 0 V, the capacitance increases, indicating that the NPB layer has reached 
the flat-band condition, and its resistance drops drastically [28]. With 
increasing bias, holes are injected from the anode, gradually reducing immobile 
negative charges at the NPB/Alq[sub]3[/sub] interface until at V[sub]BI[/sub] 
the negative interfacial charge is fully compensated and the device is in the 
flat-band condition. The additional positive charges inside the device are 
directly measured in the discharge experiments shown in Figure 4. The crossover 
from the dielectric to the charged state of the device is observed at a voltage 
V[sub]c[/sub], which is given by the magnitude of the negative interfacial 
charges [sigma] and the thickness of the Alq[sub]3[/sub] layer via 

           V[sub]c[/sub] - V[sub]BI[/sub]     
[Delta]F = ------------------------------ 
           d[sub]Alq[sub]3[/sub][/sub]     

                            [sigma]
         = ------------------------------------------ .                     (3)
           [epsilon][sub]r[/sub][epsilon][sub]0[/sub]

Using V[sub]c[/sub] ~~ 0 V, V[sub]BI[/sub] ~~ 2 V, and [epsilon][sub]r[/sub] ~~ 
3.5, the interfacial charge density can be estimated to be [sigma] ~~ -5.9 X 
10[sup]11[/sup] e/cm[sup]2[/sup], which is in good agreement with the value of 
[sigma] ~~ -6.8 X 10[sup]11[/sup] e/cm[sup]2[/sup] obtained on 
ITO/NPB/Alq[sub]3[/sub]/Ca devices [28]. 

Transient response of organic light-emitting devices 

The scenario described above has important consequences for the transient 
response of OLEDs, because the compensation of the negative interfacial charges 
cannot be expected a priori to occur instantaneously. This is directly seen in 
the frequency dependence of the capacitance, which shows that the enhanced 
capacitance between 0 and 2 V resulting from the redistribution of the electric 
field can only be monitored up to frequencies of ~~ 200 Hz. Furthermore, in the 
forward direction (V > V[sub]BI[/sub]), the electric field inside the 
Alq[sub]3[/sub] layer and at the Alq[sub]3[/sub]/cathode interface is enhanced 
by the presence of the positive space charge at the NPB/Alq[sub]3[/sub] 
interface [29]. From this field enhancement at the cathode side, one must expect 
a strong influence on the electron-injecting properties of the cathode into 
Alq[sub]3[/sub], even if the injection barrier is assumed to be small, as in the 
case of Ca as cathode. 

Transferring the scenario described above to TEL experiments, several processes 
must occur following the application of a positive voltage pulse with V > 
V[sub]BI[/sub] before light emission can take place: injection of holes, 
compensation of negative charges, and buildup of positive space charge at the 
NPB/Alq[sub]3[/sub] interface; injection of electrons at the cathode; and 
transport of electrons to the NPB/Alq[sub]3[/sub] interface, where they can 
ultimately recombine radiatively with holes injected from NPB into 
Alq[sub]3[/sub]. Therefore, it is not as straightforward to obtain 
charge-carrier mobilities from the delayed onset of EL observed in a 
time-resolved EL experiment in these OLEDs, as has been reported for 
single-layer devices. The important question here is whether the observed delay 
times (as defined in Figure 5) originate from electron transport through the 
Alq[sub]3[/sub] layer to the recombination zone, or from the buildup of the 
internal space charge and the concomitant redistribution of the electric field 
inside the device. A direct check whether the buildup of space charges 
contributes to the observed temporal response of the OLED can be obtained by 
superimposing a dc offset bias on the applied voltage pulse. Figure 9 compares 
two EL traces obtained with the same voltage amplitude of 4 V but different 
offset biases: 0 and 2 V. The bias value of 2 V is chosen just below the onset 
of double carrier injection in order to guarantee the compensation of negative 
charges at the NPB/Alq[sub]3[/sub] interface, as has been proved by 
capacitance-voltage measurements. Figure 9 clearly shows that the delay times of 
the EL signal to a steady-state value are at least a factor of 2 shorter if a 
positive bias of 2 V is applied prior to the application of the 4-V voltage 
pulse; this directly proves that the buildup of space charge at the 
NPB/Alq[sub]3[/sub] interface has a significant influence on the temporal 
response of light emission in this OLED structure. Also, the decay of the 
current to the steady state is much faster in the presence of the positive bias, 
which in turn shows that the initial high current and its temporal decay after 
the application of a voltage pulse are not given by the RC time constant of the 
setup, as discussed in Section 3. Instead, this temporal behavior of the current 
directly reflects the buildup of the internal space charge in the device. As the 
existence of positive space charge in the NPB layer reduces the electric field 
in the vicinity of the hole-injecting contact, the initial current at t = 0 in 
the absence of space charges is much higher than the steady-state current for V 
> V[sub]BI[/sub]. By taking the integral of the current over time, one can 
directly estimate the accumulated charge injected into the device. The 
integration from t = 0 to 100 [muon]s yields a charge of 7.3 nC for zero offset 
bias and only 4.5 nC for 2 V positive bias. The difference of almost 3 nC is a 
crude estimate for the space charge built up by the positive bias of 2 V prior 
to the voltage pulse. This is in the same range as the values measured directly 
from the discharge experiments [Figure 4(a)]. 

Bearing in mind that the buildup of space charges has a non-negligible influence 
on TEL, one can nevertheless use the delay times obtained from the EL transients 
displayed in Figure 6 to calculate charge-carrier mobilities in Alq[sub]3[/sub]. 
In our device structure, EL first occurs when the leading charge carriers of 
each polarity meet in the Alq[sub]3[/sub] layer close to the NPB/Alq[sub]3[/sub] 
interface. At a given electrical field F and layer thickness d, the delay time 
t[sub]d[/sub] depends on the mobility [muon] as 

                   d
t[sub]d[/sub] = ------- .                                                   (4)
                [muon]F

If the mobilities of both charge-carrier types in Alq[sub]3[/sub] are very 
different, the higher drift mobility rather than the ambipolar mobility [muon] = 
[muon][sub]h[/sub] + [muon][sub]e[/sub] determines the delay time. From the 
delay time of the EL signal, one can calculate the mobility by taking into 
account the built-in voltage as 

               d                 V - V[sub]BI[/sub]
[muon] = -------------- with F = ------------------ .                       (5)
         t[sub]d[/sub]F                   d

Because the mobility of holes in NPB is approximately three orders of magnitude 
higher than that of electrons in Alq[sub]3[/sub] ([muon][sub]h,NPB[/sub] >> 
[muon][sub]e,Alq[sub]3[/sub][/sub]) and because most of the applied voltage 
drops at the Alq[sub]3[/sub] layer as discussed above, it is reasonable to 
consider only the thickness of the Alq[sub]3[/sub] layer in the calculation of 
an electric field. Owing to the higher mobility of electrons compared to holes 
in Alq[sub]3[/sub] ([muon][sub]e,Alq[sub]3[/sub][/sub] >> 
[muon][sub]h,Alq[sub]3[/sub][/sub]), recombination takes place in the 
Alq[sub]3[/sub] layer close to the NPB/Alq[sub]3[/sub] interface; thus, the 
delay of EL is determined by the transit time of electrons in the 
Alq[sub]3[/sub] layer [23, 25, 26]. 

Using the measured delay times (Figure 7), one can calculate a field-dependent 
electron mobility in Alq[sub]3[/sub], as shown in Figure 10. A plot of the 
logarithm of the mobility vs. the square root of the electric field has been 
chosen because these amorphous materials are known to show a field-dependent 
mobility of the form [muon] [proportional] exp([beta][square root]F) [30], which 
is explained by a disorder formalism [31, 32] or the phenomenological 
Poole-Frenkel model [33]. In the logarithmic plot of the mobility versus [square 
root]F (see Figure 10), two different slopes can be observed, with a crossover 
at about [square root]F = 700[square root](V/cm). In the low-field regime, a 
pronounced field dependence is observed, in which the calculated mobility seems 
to vary over more than two orders of magnitude; this is followed by the second 
regime, in which the determined mobility changes by only a factor of 2. This 
crossover between regions with different field dependences indicates that 
various processes are involved in the temporal EL response of these multilayer 
devices. In calculating the mobility from the experimental data, we have so far 
neglected the influence of the internal interface on the transient response. As 
discussed above, space charges build up with time at the NPB/Alq[sub]3[/sub] 
interface, leading to an enhancement of the electric field in the 
Alq[sub]3[/sub] layer and facilitating electron injection at the cathode. The 
time required for building up the space charge depends on the applied voltage 
pulse (i.e., the amount of current flow). Consequently, if the buildup of the 
internal space charge takes significantly longer than the transit of electrons 
through the Alq[sub]3[/sub] layer, the delay time will be determined by the time 
required to build up the space charge and will then contain no information about 
the electron mobility in Alq[sub]3[/sub]. Therefore, we attribute the strong 
increase of the calculated mobility in the low-field region to a charging effect 
at internal interfaces rather than to a real field-dependent behavior of the 
mobility. 

The control of the TEL response in heterolayer OLEDs by the buildup of an 
internal space charge inside the device has been observed experimentally and 
verified by numerical simulations on polymeric bilayer OLEDs by Nikitenko et al. 
[21]. They have shown that the temporal response (delay time of EL) depends very 
sensitively on the injection barriers of electrons and holes at the cathode and 
anode sides, respectively. For sufficiently large injection barriers, the 
buildup of the internal space charge can eventually fully control the transient 
response of the device. 

Finally, we note that the delay time of the TEL signal does not contain all of 
the information about the time-dependent response of these OLED devices. As seen 
in Figure 8 after a first plateau in the time range of some 10 to 100 [muon]s 
has been reached, there is a further increase of EL and current on a much longer 
time scale up to about 10 ms. At present there is no definite explanation for 
the observed behavior. A possible origin could be dispersive transport of charge 
carriers leading to a spread of transit times over many orders of magnitude in 
time. However, if this is the dominant factor, one would expect to see a 
continuous rise of EL and current rather than the relatively distinct two steps 
observed. Probably the complex interplay among the buildup of the internal space 
charge, the concomitant redistribution of the electric field, and their feedback 
on charge-carrier injection and transport is responsible for this behavior. A 
possible shift or broadening of the recombination zone due to a non-negligible 
hole mobility in Alq[sub]3[/sub] [25, 26] may also contribute. Thus, further 
experiments coupled with numerical simulations are required in order to clarify 
the origin of the long-term response. 

5. Conclusion 

Our experimental results show that the electrical characteristics of organic 
light-emitting devices incorporating the hetero-interface between 
hole-conducting triphenyl-amine derivatives (e.g., NPB) and aluminum 
hydroxy-quinoline (Alq[sub]3[/sub]) can be separated into three regimes. First, 
the built-in potential V[sub]BI[/sub] of approximately 2 V caused by the contact 
potential difference of the metal electrodes separates the voltage range into 
the forward bias regime (V > V[sub]BI[/sub]), where double carrier injection, 
transport, and radiative recombination in the Alq[sub]3[/sub] layer close to the 
NPB interface occur. Second, for high reverse bias (V << V[sub]BI[/sub]), the 
devices behave like a dielectric, with insulating material sandwiched between 
two metal electrodes. In particular, our capacitance-voltage measurements show 
that between these two regimes there is a third regime between a critical value 
V[sub]c[/sub] depending on the Alq[sub]3[/sub] thickness (V[sub]c[/sub] ~~ 0 V 
for a 65-nm-thick Alq[sub]3[/sub] layer) and the built-in voltage. The 
observation of an enhanced device capacitance indicates the presence of negative 
interfacial charges at the NPB/Alq[sub]3[/sub] interface, which are gradually 
compensated by injected holes as the built-in voltage is approached. The 
important consequences of this process are seen in the TEL response, which at 
low forward voltages is completely governed by the buildup of the internal space 
charge rather than by the transport of carriers through the Alq[sub]3[/sub] 
layer. 

Acknowledgments 

The authors thank S. F. Alvarado, S. Barth, and S. Karg for valuable 
discussions, and M. Tschudy for preparing the substrates. W. B. acknowledges 
financial support from IBM Ruschlikon for hosting his visit. 

Footnotes

[foot1] It is difficult to give an accurate definition of a "medium," "long," or 
"fast" time, since the actual time depends not only on the integration time but 
also on the current range. Therefore, the total time depends on the impedance of 
the device structure and the integration time (see HP Manual 4145B).

[foot2] P. H. Nguyen, S. Scheinert, S. Berleb, W. Brutting, and G. Paasch 
(preprint).


References

1. W. Helfrich and W. G. Schneider, "Recombination Radiation in Anthracene 
Crystals," Phys. Rev. Lett. 14, 229 (1965). 

2. C. W. Tang and S. A. Van Slyke, "Organic Electroluminescent Diodes," Appl. 
Phys. Lett. 51, 913-915 (1987). 

3. C. W. Tang, S. A. Van Slyke, and C. H. Chen, "Electroluminescence of Doped 
Organic Thin Films," J. Appl. Phys. 65, 3610-3616 (1989). 

4. M. A. Lampert and P. Mark, Current Injection in Solids, Academic Press, Inc., 
New York, 1970. 

5. C. Giebeler, H. Antoniadis, D. D. C. Bradley, and Y. Shirota, "Influence of 
the Hole Transport Layer on the Performance of Organic Light-Emitting Diodes," 
J. Appl. Phys. 85, 608-615 (1999). 

6. S. Berleb, W. Brutting, M. Schwoerer, R. Wehrmann, and A. Elschner, "Effect 
of Majority Carrier Space Charges on Minority Carrier Injection in Dye Doped 
Polymer Light-Emitting Devices," J. Appl. Phys. 83, 4403-4409 (1998). 

7. P. W. M. Blom, M. J. de Jong, and J. J. M. Vleggaar, "Electron and Hole 
Transport in Poly(p-phenylene Vinylene) Devices," Appl. Phys. Lett. 68, 
3308-3310 (1996). 

8. L. Bozano, S. A. Carter, J. C. Scott, G. G. Malliaras, and P. J. Brock, 
"Temperature- and Field-Dependent Electron and Hole Mobilities in Polymer 
Light-Emitting Diodes," Appl. Phys. Lett. 74, 1132-1134 (1999). 

9. W. Brutting, H. Riel, T. Beierlein, and W. Riess, "Influence of Trapped and 
Interfacial Charges in Organic Multilayer Light-Emitting Devices," J. Appl. 
Phys. (in press). 

10. J. Kanicki, in Handbook of Conducting Polymers, T. A. Skotheim, Ed., Marcel 
Dekker, New York, 1986, Ch. 17. 

11. D. M. Taylor, H. Gomes, A. E. Underhill, S. Edge, and P. I. Clemenson, 
"Effect of Oxygen on the Electrical Characteristics of Field Effect Transistors 
Formed from Electrochemically Deposited Films of Poly(3-methylthiophene)," J. 
Phys. D: Appl. Phys. 24, 2032-2038 (1991). 

12. M. Meier, S. Karg, and W. Riess, "Light-Emitting Diodes Based on 
Poly-p-phenylene-vinylene: II. Impedance Spectroscopy," J. Appl. Phys. 82, 
1961-1966 (1997). 

13. J. Scherbel, P. H. Nguyen, G. Paasch, W. Brutting, and M. Schwoerer, 
"Temperature Dependent Broadband Impedance Spectroscopy on 
Poly-(p-phenylene-vinylene) Light-Emitting Diodes," J. Appl. Phys. 83, 5045-5055 
(1998). 

14. A. K. Hassan and R. D. Gould, "The Effects of Exposure to Oxygen and 
Annealing on the Conductivity of Copper Phthalocyanine Thin Films," J. Phys.: 
Condens. Matter 1, 6679-6684 (1989). 

15. H. Riel, "Ladungstragerinjektion, Transport und Rekombination in organischen 
LEDs: Von Einschiebestrukturen zu effizienten blauen OLEDs," Diplomarbeit, 
Universitat Erlangen, Germany, 1998. 

16. C. Hosokawa, H. Tokailin, H. Higashi, and T. Kusumoto, "Transient Behavior 
of Organic Thin Film Electroluminescence," Appl. Phys. Lett. 60, 1220-1222 
(1992). 

17. H. Vestweber, J. Oberski, A. Greiner, W. Heitz, R. F. Mahrt, and H. Bassler, 
"Electroluminescence from Phenylenevinylene-Based Polymer Blends," Adv. Mater. 
Opt. Electron. 2, 197-204 (1993). 

18. S. Karg, V. Dyakonov, M. Meier, W. Riess, and G. Paasch, "Transient 
Electroluminescence of PPV LEDs," Synth. Met. 67, 165-168 (1994). 

19. H. Nakamura, C. Hosokawa, and T. Kusumoto, "Transient Behavior of Organic EL 
Cells," in Inorganic and Organic Electroluminescence, R. H. Mauch and H.-E. 
Gumlich, Eds., Wissenschaft & Technik Verlag, Berlin, 1996, pp. 95-100. 

20. S. Barth, P. Muller, H. Riel, P. F. Seidler, W. Riess, H. Vestweber, and H. 
Bassler, "Electron Mobility in Alq[sub]3[/sub] Determined via Transient 
Electroluminescence from Single- and Multilayer Organic Light-Emitting Diodes," 
J. Appl. Phys., in press. 

21. V. R. Nikitenko, Y.-H. Tak, and H. Bassler, "Rise Time of 
Electroluminescence from Bilayer Light-Emitting Diodes," J. Appl. Phys. 84, 
2334-2340 (1998). 

22. J. C. Scott, S. Karg, and S. A. Carter, "Bipolar Charge and Current 
Distributions in Organic Light-Emitting Diodes," J. Appl. Phys. 82, 1454-1460 
(1997). 

23. R. G. Kepler, P. M. Beeson, S. J. Jacobs, R. A. Anderson, M. B. Sinclair, V. 
S. Valencia, and P. A. Cahill, "Electron and Hole Mobility in 
Tris(8-hydroxyquinolinolato-N1,O8) Aluminum," Appl. Phys. Lett. 66, 3618-3620 
(1995). 

24. T. Tsutsui, H. Tokuhisa, and M. Era, "Charge Carrier Mobilities in Molecular 
Materials for Electroluminescent Diodes," in Polymer Photonic Devices, Proc. 
SPIE 3281, 230-239 (1998). 

25. A. G. Muckl, S. Berleb, W. Brutting, and M. Schwoerer, "Transient 
Electroluminescence Measurements on Organic Hetero Layer Light Emitting Diodes," 
Synth. Met. 111-112, 91-94 (2000). 

26. S. Naka, H. Okada, H. Onnagawa, J. Kido, and T. Tsutsui, "Time-of-Flight 
Measurement of Hole Mobility in Aluminum(III) Complexes," Jpn. J. Appl. Phys. 
38, L-1252-L-1254 (1999). 

27. F. Rohlfing, T. Yamada, and T. Tsutsui, "Electroabsorption Spectroscopy on 
Tris-(8-hydroxyquinoline) Aluminum-Based Light-Emitting Diodes," J. Appl. Phys. 
86, 4978-4984 (1999). 

28. S. Berleb, W. Brutting, and G. Paasch, "Interfacial Charges and Electric 
Field Distribution in Organic Hetero-Layer Light-Emitting Devices,"Organic 
Electron.  1, 41-47 (2000). 

29. M. Matsamura, A. Ito, and Y. Miyamae, "Accumulation of Positive Charges in 
Organic Light-Emitting Diodes with a Double-Layer Structure," Appl. Phys. Lett. 
75, 1042-1044 (1999). 

30. P. M. Borsenberger and D. S. Weiss, Organic Photoreceptors for Imaging 
Systems, Marcel Dekker, New York, 1993. 

31. H. Bassler, "Charge Transport in Disordered Organic Photoconductors," Phys. 
Stat. Sol. B 175, 15 (1993). 

32. D. H. Dunlap, P. E. Parris, and V. M. Kenkre, "Charge-Dipole Model for the 
Universal Field Dependence of Mobilities in Molecularly Doped Polymers," Phys. 
Rev. Lett. 77, 542-545 (1996). 

33. W. D. Gill, "Drift Mobilities in Amorphous Charge-Transfer Complexes of 
Trinitrofluorenone and Poly-n-vinylcarbazole," J. Appl. Phys. 43, 5033-5040 
(1972). 

Received May 21, 2000; accepted for publication September 25, 2000


Biographical sketches of authors  

Walter Riess   

IBM Research, Zurich Research Laboratory, 8803 Ruschlikon, Switzerland 
(wri@zurich.ibm.com).  Dr. Riess received a degree in physics (Dipl. Phys.) from 
the University of Bayreuth, Germany, in 1987. In 1991, he received a Ph.D. 
degree from the University of Bayreuth for his thesis on nonlinear conduction 
phenomena in organic charge density wave conductors. From 1992 to 1995, he was a 
research assistant in the Physical Institute of the University of Bayreuth, 
where he investigated charge-carrier injection, transport, and recombination 
processes in organic light-emitting diodes. During his stay as a visiting 
scientist at the IBM Zurich Research Laboratory in 1994, he initiated a project 
on organic light-emitting diodes and, in October 1995, he joined the IBM Zurich 
Research Laboratory as a Research Staff Member in the Display Technology group. 
Dr. Riess finished his habilitation in 1996 with a thesis on polymeric 
light-emitting diodes. In October 1998, he became manager of the Display 
Technology group and continues to focus on display applications of 
electroluminescent organic materials. 

Heike Riel   

IBM Research, Zurich Research Laboratory, 8803 Ruschlikon, Switzerland 
(hei@zurich.ibm.com).  Mrs. Riel studied physics at the University of 
Erlangen-Nuremberg, Germany. In 1996, she pursued practical training at the 
Institute of Technical Physics in Erlangen, working on the optical 
characterization of GaAs/AlGaAs heterostructures. In August 1996 she began a 
four-month internship at the Hewlett-Packard Research Laboratory, Palo Alto, 
working on the incorporation of metal/semiconductor/metal photodetector arrays 
with microlens arrays. In April 1997 she joined the Display Technology group at 
the IBM Zurich Research Laboratory to conduct her master's thesis work, focusing 
on the injection, transport, and recombination processes of charge carriers in 
organic light-emitting devices. In August 1998 she received a degree in physics 
(Dipl. Phys. Univ.) from the University of Erlangen-Nuremberg. Since then, she 
has been pursuing work on her Ph.D. thesis on organic electroluminescent devices 
for display applications within the Display Technology group at the IBM Zurich 
Research Laboratory. 

Tilman Beierlein   

IBM Research, Zurich Research Laboratory, 8803 Ruschlikon, Switzerland 
(tib@zurich.ibm.com).  Mr. Beierlein studied electrical engineering at the 
University of Ulm, Germany. After receiving an undergraduate degree, he was with 
the Optoelectronics Department of the University of Ulm, where he pursued work 
on photoassisted wet chemical etching of gallium nitride. In October 1996 he 
began a six-month internship in the Display Technology group of the IBM Zurich 
Research Laboratory. In August 1997 he rejoined the IBM Zurich Research 
Laboratory to perform his master's thesis on the fabrication and 
characterization of thin InGaN films for application in organic light-emitting 
devices. In February 1998 he received a degree in electrical engineering (Dipl. 
Ing.) from the Optoelectronics Department of the University of Ulm. Since 
mid-1998 he has been working on his Ph.D. thesis in the Display Technology group 
of the IBM Zurich Research Laboratory on research and technology of organic 
electroluminescent devices for display applications. 

Wolfgang Brutting   

Experimental Physics II, University of Bayreuth, 95440 Bayreuth, Germany 
(wolfgang.bruetting@uni-bayreuth.de).  Dr. Brutting studied physics at the 
Universities of Erlangen-Nuremberg and Bayreuth, Germany, graduating in 1992. He 
received a Ph.D. degree from the University of Bayreuth in 1995 for his thesis 
on charge transport in quasi-one-dimensional charge density wave systems. Since 
then, he has been a research assistant in experimental physics at the University 
of Bayreuth, where he heads the organic semiconductors group. Dr. Brutting has 
been a visiting scientist at Kyushu University, Japan, and at the IBM Zurich 
Research Laboratory. His current research interests include charge transport in 
organic semiconductor devices, especially light-emitting diodes from conjugated 
polymers and low-molecular-weight materials. 

Peter Muller   

IBM Research, Zurich Research Laboratory, 8803 Ruschlikon, Switzerland 
(pmu@zurich.ibm.com).  Mr. Muller is a Research Staff Member at the IBM Zurich 
Research Laboratory. He is currently involved in the OLED project, where his 
main interests are electrified interfaces. He joined IBM in 1987 after having 
received a degree in computer science from the Brugg-Windisch Polytechnical 
Institute in Switzerland. From 1994 to 1996, he was a guest scientist at the 
Exploratory Research and Technology Organization (ERATO) at Tohoku University, 
Japan. 

Paul F. Seidler   

IBM Research, Zurich Research Laboratory, 8803 Ruschlikon, Switzerland 
(pfs@zurich.ibm.com).  Dr. Seidler is manager of Science and Technology at the 
IBM Zurich Research Laboratory. His department pursues research in the fields of 
micromechanics, molecular-scale engineering, display technology, optical 
communications, advanced materials processing and characterization, and 
computational materials science. Dr. Seidler received a B.S. degree in chemistry 
from the California Institute of Technology in 1980 and a Ph.D. degree in 
chemistry from the University of California at Berkeley in 1985 for his thesis 
in the field of mechanistic organometallic chemistry. After a postdoctoral 
fellowship at Exxon Corporate Research, he joined the IBM Thomas J. Watson 
Research Center, where he worked on a variety of scientific studies related to 
semiconductor processing technology, in particular the kinetics and mechanisms 
of metal chemical vapor deposition. After serving a year as Technical Assistant 
to the Vice President, Systems, Technology and Science for the IBM Research 
Division, he moved to the Zurich Research Laboratory to become head of the 
Display Technology group and IBM's organic light-emitting diode project.