The type 1 devices, which had a bottom contact/bottom gate (BC/BG, see Figure ) configuration, were ready for characterization directly after nanofiber transfer and annealing using the underlying highly doped silicon as the gate electrode. The type 2 devices had a top contact/bottom gate (TC/BG, see Figure ) configuration, and were prepared by depositing gold electrodes in high vacuum (range of 10-6
mbar) on top of the transferred and annealed nanofibers through a nanostencil [28
] with a pattern that gives top electrodes with the same dimensions as those used for the bottom contacts. In both bottom and top contact configurations, the contacts had dimensions of 10 μm × 200 μm, separated by a channel length of around 2 μm. Figure shows an illustration of a TC/BG device with top contacts prepared by deposition through a stencil. The device type 3 was also a staggered configuration in a bottom contact/top gate (BC/TG, see Figure ) geometry.
The three different configurations used: (a) BC/BG, (b) TC/BG, and (c) BC/TG. (d) Drawing of a device with TC/BG configuration prepared by deposition of the top contacts through a nanostencil.
Figure show the nanofibers integrity and also the sharpness of the electrode edges on top of the nanofibers (TC configuration) (Figure ). The stencil used had 2 μm channel length but because of a blurring effect [29
] during electrode deposition, a channel length of only approximately 1.5 μm is observed in the SEM image.
Figure 2 Nanofibers in top contacts configuration. (a) Fluorescence microscope image of nanofibers in the top contacts configuration. (b) White light microscope image of the sharp top contacts on nanofibers. (c) Scanning electron microscope image of the electrodes (more ...)
Figure shows the measured transfer characteristics, i.e., current versus gate voltage for a drain-source voltage of -15 V for p6P nanofibers on a BC/BG device. The inset in Figure is the Mott-Schottky energy scheme at negative gate and drain voltages which, however, do not account for interface traps states that could further reduce the current. The source-drain field allows only holes injected from the source electrode or electrons injected from the drain electrode to pass through the device and the measured characteristics clearly show that the transport is p-type, i.e., holes are injected from the source (see Figure inset).
Figure 3 Measured transistor characteristics for BC/BG nanofibers. (a) Current versus gate voltage for Vds = -15 V. Inset shows schematic Mott-Schottky energy scheme for negative gate and drain voltages. (b) Current versus drain-source voltage for zero gate voltage. (more ...)
Figure shows the current versus drain-source voltage for zero gate voltage for the same device. The inset schematically shows the energy level positions: the work function levels for the gold drain and source electrodes and the LUMO and HOMO levels for p6P. In Figure , current flow is observed only for positive Vds. This must mean that the electrical characteristics are dominated by an injection barrier between the injecting metal electrode and the organic material. This is not unexpected given the energy levels shown in the inset that suggest an injection barrier for holes of around 0.9 eV. As shown in Figure , a positive Vds then leads to downward band bending near the drain electrode and thereby a lowering of the hole injection barrier, while a negative Vds does not cause a similar band bending at the source electrode as would be required for hole injection in the opposite direction since the band bending again occurs at the drain electrode (see Figure ).
A hysteresis effect can also be observed in Figure where the forward sweep is higher than the reverse sweep. This is assumed to be caused by trapping of the charge carriers [26
]. We propose that the observed hysteresis is due to hole trapping close to the interface region between the injecting electrode and the organic material creating a space charge that reduces the band bending and thereby limits further hole injection, causing a lower back sweep current. We will elaborate on this aspect below.
Figure shows current versus drain-source voltage for zero gate voltage for transferred p
6P nanofibers for BC/BG, BC/TG, and TC/BG configurations, while the inset shows the same data plotted with a different current scale. Considering that approximately the same number of nanofibers was present in all the samples, the coplanar (BC/BG) configuration exhibits a lower output current than the staggered geometries due to a high contact resistance associated with the high injection barrier to the organic material [32
]. In the staggered geometries (BC/TG and TC/BG), the charges are injected not only from the edge of the electrode but also from the surface of the contacts in the region where the source-drain electrodes overlap with the gate electrode and consequently charges are injected over a larger area leading to a lower contact resistance than in the coplanar (BC/BG) geometry [26
Current versus drain-source voltage for zero gate voltage for (a) p6P nanofibers transferred from mica to a transistor platform and (b) p6P thin films for BC/BG, BC/TG and TC/BG configurations.
The TC/BG configuration exhibits the highest output current. We propose that this is due to the smaller contact resistance between the nanofibers and the electrodes due to deposition of the electrodes under vacuum, which prevents water residues in the nanofiber-electrode interface in contrast to the bottom contact devices where the nanofiber-electrode interface is created under humid conditions during the transfer. As suggested by Bao and co-workers [33
], moisture residing at the interface between the electrode and the organic material is expected to cause an increased contact resistance. Although our devices are annealed after fabrication, this can presumably not eliminate all water or water-transferred contaminants residing at the interface, since hysteresis is observed even after prolonged annealing. Also, metal penetrating into the organic material during electrode deposition can enable a better electrical contact [34
The symmetric characteristics of the TC/BG device as opposed to the asymmetric behavior of the bottom connected devices can be observed in the inset of Figure . Since no n-type behavior has been observed, this must mean that in the TC/BG devices the source electrode is injecting holes for negative drain-source voltages. The situation depicted in Figure with band bending at the drain electrode is thus not valid for the top contact devices. Here, the main current limiting factor is the bulk nanofiber resistance giving rise to the observed symmetric output curve.
In Figure , essentially no hysteresis is observed for the TC/BG configuration. Since these output characteristics are dominated by the nanofiber bulk as described previously, this suggests that the traps that cause the hysteresis must be spatially located near the injection region that governs the behavior of the BC devices.
Figure shows the output characteristics for a 30 nm thick p
6P film on similar transistor platforms. Around eight times more material was used to form the films compared to the material used to grow the nanofibers. The higher current density for the p
6P nanofibers in comparison with the film must be consequence of the crystallinity of the nanofibers, i.e., p
6P nanofibers have a long range order compared with thin films which is believed to favor a high charge-carrier mobility as a result of the π-conjugated coupling between the packed molecules [36
] (see Figure ). The asymmetric curve observed for the thin film FET also in the TC/BG configuration in Figure must be the result of a high contact resistance compared to the resistance of the film bulk. This implies that the contact resistance in TC devices is significantly lower for the crystalline nanofibers than for the amorphous film. In addition, the significant hysteresis observed for the injection limited thin film devices further support our conclusion of the traps being spatially located at the surface.
In Figure , a drain current saturation is not observed. The channel length used was around 2 μm and the gate dielectric was 0.2 μm thick. It is well-known that if the channel length of a transistor is less than ten times the thickness of the gate dielectric, the space-charge-limited bulk current will be dominated by the lateral field due to the source-drain voltage preventing saturation since the gate voltage will not determine the charge distribution within the channel and consequently the "on" or "off" state of the transistor will not be observed [26
Figure shows the transfer characteristics, i.e., gate voltage sweep at a certain Vds for both p6P thin films and nanofibers. Figure shows that the nanofibers conduct better than the thin films (as mentioned previously the film cross-sectional area is around eight times the nanofiber cross-section) and current saturation is not observed reinforcing the conclusion from Figure .
Current versus gate voltage at Vds = -15 V for p6P (a) nanofibers and thin films in TC/BG configuration and (b) for nanofibers in BC/BG, BC/TG, and TC/BG configurations.
From Figure , the subthreshold swings (S
] were obtained from the transfer characteristics of the p
6P nanofibers on different transistor configurations to elaborate on the switching behavior.
From the data in Figure , the subthreshold swing (S
) for the nanofibers on BC/BG, BC/TG, and TC/BG configurations were found to be 13.7, 9.5, and 7.5 V/decade, respectively. The TC/BG configuration exhibits the lowest subthreshold swing being almost half that of the BC/BG device. For comparison, Klauk et al. [38
] have studied the electrical characteristics for pentacene transistors with 100 nm SiO2
as the gate dielectric and found a subthreshold swing of only 0.7 V/decade. Our results is around a decade above this, however, this is not unexpected since the p
6P mobility is significantly below that found in pentacene [21
] and since our device geometry (here particularly the gate dielectric thickness) was not optimized for efficient switching.