In recent years, organic thin-film transistors (OTFTs) have attracted a great deal of attention due to their potential applications in low cost sensors,1 memory cards,2 and integrated circuits.3 Great efforts are under way to design OTFTs with high performance, high stability, high reproducibility, and low cost.4 Two of the most crucial device parameters are the charge carrier mobility and the threshold voltage (VTh). Concerning the mobility, the main goals for most applications is its maximization.5 For VTh, the situation is more complex: for example, for integrated circuits it would be desirable to tune VTh over a broad range,6 e.g., for inverter applications. In silicon technology, complementary circuits that consist of p-channel and n-channel transistors are typically used.7 There have been many attempts to adapt this technology to OTFTs and fabricate organic complementary inverters.2, 8, 9 They, however, suffer from poor n-type transistor performance and/or air instability of n-type semiconductor materials. An alternative approach is the adaptation of unipolar depletion-load inverters enabling simplified processing, even if they do not provide the low power consumption and the simple circuit design intrinsic to complementary logic.10, 11 Depletion-load inverters consist of an enhancement-mode driver transistor and a depletion-mode load transistor and can be realized using only p-type OTFTs. So far there have been attempts to achieve this target by using a level shifter12, 13 or a dual gate structure.14 The main objective is to find a reproducible method to realize driver and load transistors with equivalent device characteristics (in particular mobilities), but different VTh values.
Over the years, a number of methods have been developed to tune threshold voltages. For example, oxygen plasma treatment of an organic gate dielectric (parylene) results in a large shift of VTh to more positive values.15 This shift is related to the generation of charged surface states at the dielectric-semiconductor interface but the exact mechanism is only poorly understood. The same holds for UV-ozone treatments16 of the parylene and for the UV treatment of a poly-4-vinylphenol layer,17 where the latter has been used to realize and eventually to optimize depletion-load inverters. VTh can also be tuned by changing the capacitance of the dielectric,18 by inserting a polarizable layer into the dielectric,19, 20, 21 or by using this polarizable layer as an encapsulation.22 With the latter methods, VTh can be controllably and reversibly shifted over a wide range. The drawback of such approaches is, however, that a high “programming” voltage is needed to tune VTh. Another possibility to tune VTh is by insertion of self-assembled monolayers23, 24 or chemically reactive thin layers.25, 26 The latter have been shown to tune VTh by a local channel doping using acid groups that can be combined with de-doping reactions using bases. The disadvantage of the latter methods is that a local patterning of threshold voltages is not straightforward. Such a patterning is, however, a prerequisite for the realization of integrated electronic circuits.
Here, we significantly refine the concept of chemical channel doping by replacing the covalently bonded silane layers bearing sulfonic acid groups used in Refs. 25, 26 with photoacid generator polymers.27 The goal hereby is to use them as an interface-modification layer (see Figure1a), whose properties can be patterned photochemically, because, in contrast to the molecules used previously25, 26, the acid group is formed only upon illumination. This paves the way for a photolithographic patterning of the interfacial doping and, thus, for controlling which transistors in a circuit operate in depletion or in enhancement mode. In other words, it enables an accurate local control of VTh through the illumination dose by a method that is fully compatible with lithographic techniques omnipresent in conventional semiconductor industry.
To show the versatility of our approach, we have synthesized two qualitatively different polymers, namely, poly(endo,exo-bicyclo[2.2.1]hept-5-ene-2,3-(2-nitrobenzyl) dicarboxylate) (PBHND) and poly-[(endo,exo-N-hydroxy bicyclo[2.2.1]hept-5-ene-2,3-dicarboximide perfluoro-1-butanesulfonate)-co-(endo,exo- bicyclo[2.2.1]hept-5-ene-2,3-dicarboxylic acid, dimethyl-ester)] (PHDBD). The chemical structures and the main reaction products upon UV illumination are shown in Figure 1b,c. From PBHND, a carboxylic acid is formed that remains chemically linked to the polymer backbone, while in PHDBD a sulfonic acid is part of the leaving group. Both photochemical reactions can be followed by IR spectroscopy (see Supporting Information).
In the device, the deprotonation of the acidic groups due to the reaction with the organic semiconductor results in the formation of a space-charge region at the interface. This gives rise to a threshold voltage shift, since mobile holes have to compensate the conjugated bases. This has been shown by drift-diffusion based modelling.28, 29 From the data in 29 it can be estimated that the deprotonation of only a few percent of the acid groups at the interface is sufficient to shift VTh by several tens of volts. An alternative way to understand the underlying process suggested for a polythiophene active layer25 is to view it as acid doping, in analogy with the situation in PEDOT/PSS.30 Indeed, a photoinduced protonation of pentacene by acids has already been shown, albeit only at temperatures around 3.5 K with the product being thermally unstable.31 Consequently, the occurrence of stable protonated pentacene molecules appears unlikely for the present devices and the actual fate of the protons needs to be investigated further.
The results in Figure 1d show that varying the illumination time allows tuning the threshold voltage over several volts by increasing the UV illumination time from 0 to 5 seconds in a PBHND-containing OTFT. The shape of the curves is essentially the same for the different illumination times; in particular the slopes (and thus the mobilities) of the OTFTs remain constant. Also the drain current in the output characteristics (Figure 1e) increases with illumination, consistent with an increased channel doping. The hysteresis of the transistors becomes somewhat larger but remains small even after 5 s of illumination. The gate voltage at a drain current of ID = 0.10 mA changes by ΔVG = 2 V. Also the off-current remains in the range of nA and is barely altered by the UV treatment. The devices, for which the data are shown in Figure 1e, were kept in an argon glove box throughout all process steps so no detrimental effects due to contact with O2 or H2O occured.
While the VTh shifts shown in Figure 1d are sufficient for the realization of a depletion load inverter (vide infra), the impact of longer illumination times (see left part of Figure2) was also explored. The time scales for illumination cannot be directly compared, as different lamps with different emission spectra were used for the two different polymers (see Experimental Section). The observed threshold voltages changes reach close to 50 V both for PBHND and PHDBD interfacial layers. However, the threshold values without illumination differ significantly between the two photoacid generator polymers (only the PHBDB devices are actually switched from enhancement to depletion mode operation). Moreover, the VTh shifts for PHDBD interfacial layers are not stable for repeated measurement cycles, as shown in the right plot in Figure 2.32
For the PBHND-based OTFTs, even after more than 200 measurements the threshold voltage did not shift back considerably. The small decrease during the first measurements was mostly due to a decrease of hysteresis in these devices. The hysteresis is at least in part a result of the need to briefly transport the devices through air before the illumination (which, again, happened under inert atmosphere). One also has to keep in mind that the chosen measurement range for the device characteristics (from −60 V to 60 V) induces a significantly larger amount of bias stress than in any inverter-based application. The small offset between the final data point in the left plot and the first data point in the right plot is a consequence of using devices from different samples. The difference in device stability for the illuminated PBHND- and PHDBD-containing devices can be attributed to the different positions of the acids formed in the photochemical process. As mentioned above, in PBHND the acid is linked directly to the polymer backbone, while it is on the leaving group in PHDBD.
An observed disadvantage of extended illuminations is that, in addition to the VTh shift, it results in some deterioration of the device characteristics. These effects, however, play a role only for illumination times far beyond the 3 s needed for the optimum inverter operation. They include an increase of the off current, a larger hysteresis, and a deterioration of the drain current at large negative gate bias. A detailed discussion can be found in the Supporting Information. All measurements (threshold voltage shift and stability) for the PHDBD devices were reproduced for several samples. The PHBND measurements were reproduced for two batches of freshly synthesized polymer and for various samples within one batch.
To verify that the observed effects are indeed a consequence of interfacial acid doping, two test experiments were performed. First, the devices were exposed to a strong base (NH3 as in Refs. 25, 26). As expected, this treatment results in a neutralization of the acid, i.e., a de-doping of the channel, and after 10 min of exposure to ammonia the devices were switched back to a situation similar to that prior to UV exposure. As a second test, devices containing no interfacial layer were also illuminated; here, even after 20 min of UV exposure no changes in the transfer and output characteristics were observed. More details about these experiments are included in the Supporting Information.
This provides us with all the tools necessary to perform the next step in the direction of integrated p-type organic electronic devices, which is the fabrication of a (tuneable) depletion-load inverter. The wiring diagram of such an inverter is plotted as an inset in Figure3. The switch transistor is realized by a non-illuminated PBHND-containing device that has a negative threshold voltage of around VTh = –10 V. The load transistor (also a device containing a PBHND interfacial layer) was illuminated in steps of 1 s in an argon glove box after the inverter wiring was realized. As seen in Figure 3, the inverter characteristic is very poor prior to illumination, the gain of the inverter is negligible, and the achievable output voltage at the low level is only −4 V. This is because the non-illuminated load transistor is normally off and works in enhancement mode. By increasing the illumination time of the load transistor and shifting its threshold towards the positive voltage regime, the inverter characteristic improves significantly. The turn-on voltage of the inverter is shifted to more negative values and the gain of the inverter increases. After illuminating the load transistor for 3 s, an optimum value of VTh,load with respect to the threshold voltage of the switch-transistor is reached, resulting in a steep inverter transition with a maximum gain of about 40 (see bottom graph in Figure 3). Further increasing the illumination time results in a deterioration of the inverter performance. At this point it should be noted that no attempts for optimizing the inverter characteristics other than tuning VTh,load were made. Therefore, further significant improvements can be expected by adapting the width-to-length ratio of the channel between the load and switch and by optimizing the performance of individual transistors with respect to mobility, gate leakage, etc.
In conclusion, we have demonstrated an easy and reproducible way to switch OTFTs from enhancement to depletion mode by a photochemical reaction using photoacid generators as interfacial layers and demonstrated that this allows the fabrication of good quality depletion-load inverters with tunable characteristics. The suggested fabrication method can easily be adapted to monolithically integrated circuits by illuminating individual transistors through shadow- or photomasks when patterned gate electrodes are used (in this way simultaneously reducing leakage and parasitic effects).