Electrical Details

For a combinatorial analysis of optoelectronic devices it is necessary to address each of the devices individually both optically and electrically.

For this purpose the ITO substrate is patterned into 8 rows by chemical etching with zinc powder and HCl solution [Sch01]. After the evaporation of the functional layers, an array of 8×8 electrodes is evaporated on top to form overlapping areas with the ITO of 0.12 cm² as schematically depicted in Figure3.21. The aluminium electrodes are contacted with gold-coated spring contacts on locations where the ITO had been etched away in order to prevent short contacts through the evaporated layers. The ITO is contacted on each side of the patterned rows to prevent a voltage drop along the ITO layer. This is important for the simultaneous testing of OLEDs as the voltage drop would affect theI–Vcharacteristics of an OLED if another OLED in the same row were simultaneously driven at high current.

For the most recent measurements we have further optimised the setup with the help of pointed spring contacts that contact the ITO in the middle of the rows through the organic layers as indicated by the small circles.

Figure 3.21: Electrical setup. The ITO has been patterned on rows so that 8 devices share a common anode which is contacted at each side of the substrate. The aluminium cathodes are individually addressed by spring contacts at locations without an ITO layer.

Each cathode can be switched to either the SMU, a current source or an off-position.

The latter can be set by a dip switch to either short the device or to leave the cathode unconnected.

60 Chapter 3. Combinatorial Techniques To separate the devices electrically, we chose to characterise the devices serially by multi-plexing them to a Source Measure Unit (SMU, Keithley 236), while the rest of the devices is operated under constant conditions. We did so for the following reasons: First, the mul-tiplication of a reliable source measuring system to the number of devices would be very costly. Second, the simultaneous recording of electroluminescence curves is non-trivial, as the influence of scattered light from neighbouring devices can hardly be excluded. If conditions for the other devices are kept constant, scattered light can be corrected for by background subtraction. Third, the measuring of the optoelectronic characteristics does not present a bottleneck in the combinatorial analysis: for a short-term comparative analysis, the devices that are currently not under testing can be completely switched off or shorted. For a long-term study, particularly for OLEDs where a continuous current is supplied, it may be desirable not to operate the device before having taken the first I–V curve. We have therefore implemented the option to only switch a device to continuous operation after the first I–V measurement.

The multiplexing unit was realised by computer controlled reed relays that allow switching between the SMU, an ‘off’-position and a constant-current source. The latter possibility is designed for comparative long-term testing of OLEDs: since the shelf lifetime of OLEDs is much higher than their operational lifetime, the main part of the degradation must be induced by charge flow and by heat. Comparable testing conditions are therefore best achieved by ensuring identical current flow for all devices. For this purpose we have integrated 64 voltage–current converters that are controlled by a common control voltage. Applying an identicalvoltage to the diodes would instead lead to higher currents for thinner devices, which in consequence will degrade faster. In our prototype setup, where 20 TPD/Alq₃ devices with a linear thickness gradient were simultaneously tested (see [Zet00]), we were indeed able to observe this effect and we have adapted the final setup to overcome this problem. For the same reason it is not always desirable to use the same voltage range when taking I–V curves of different devices. We have therefore implemented different I–V measuring protocols: a) linear voltage steps, b) linear current steps, c) non-linear current steps

Additionally, each off-position can be set manually to either short the device or to inter-rupt the connection by a dip switch (S1 in Figure3.21). This does not make a difference for OLEDs but is desirable for solar cells. As mentioned above, charge flow is one of the important degradation factors for OLEDs. In a short-circuited (grounded) device, there will be a short-circuit current (ISC) flowing, whereas in a disconnected (floating) device no current flow occurs. Instead the device will produce the open-circuit voltage (V_OC) between its electrodes under illumination. If there is any current-induced degradation, we would expect different degradation curves for differently connected devices. In a first attempt we set the switches in a chessboard pattern (Figure3.22A). It turned out, how-ever, that the V_OC for floating devices was slightly lower than for grounded devices, even without the devices having been driven over a long period. So the effect must originate from the electrical setup rather than from the devices themselves. A possible explana-tion is cross-talk between the devices: the nearest neighbours of a floating device are all grounded and will lower the measuredVOC, whereas the nearest neighbours of a grounded device are all floating and hence leave the open-circuit voltage of the device under test unperturbed. In a similar way, the whole I–V curve will be influenced by the

contact-3.6. Combinatorial Setup 61

Figure 3.22: Dip switch configurations. In the chessboard pattern (A), a grounded device has floating nearest neighbours and vice versa. In the diagonal pattern (B), both the grounded and the floating device have an identical environment up to their second nearest neighbours.

0.394 V

0.341 V

A) B)

Chessboard Pattern Diagonal Pattern

Figure 3.23: Open-circuit voltage of a solar-cell array taken with the chessboard (A) and the diagonal switch configuration (B). A clear correlation between the configuration pattern and the V_OC can be only be seen for the chessboard pattern.

ing of the neighbouring devices. To reduce such influence, we chose a configuration that provides a neighbourhood as similar as possible for both grounded and floating devices:

a diagonal pattern as shown in Figure3.22B leads to an identical number of floating and grounded devices as nearest and second nearest neighbours. We have documented the influence of the neighbourhood by changing the contacting pattern between two success-ive measurement cycles of a solar-cell array from the chessboard pattern to the diagonal pattern (Figure3.23, the layout of the solar-cell array is given in section 3.9). While we can easily recognise a chessboard pattern in the V_OC in the left-hand image, there is no such clear correlation between the contacting pattern and theV_OCin the right-hand image, which is a first success. To quantify the improvement further, we calculated the mean values of the grounded and floating devices (Table3.4) and found a decrease in the mean difference of V_OC by a factor of 4. This result shows that the cross-talk interpretation is correct and that the influence of neighbouring solar cells can be reasonably minimised by an appropriate choice of the contacting pattern while retaining the possibility of having half the devices grounded and the other half floating.

62 Chapter 3. Combinatorial Techniques

Pattern hV_OCi_grounded hV_OCi_floating hV_OCi_grounded− hV_OCi_floating

Chessboard 0.3783 V 0.3615 V 0.0168 V

Diagonal 0.3724 V 0.3676 V 0.0048 V

Table 3.4: Mean values of V_OC for grounded and floating devices and their difference for different connection patterns of switched-off devices. The difference in the mean values is 4 times less with the diagonal pattern than with the chessboard pattern.