The impact of OECT channel size on performance and yield
The performance of the printed OECTs was evaluated using the ON/OFF current ratio as one prime figure of merit for transistor current modulation. The ON/OFF current ratio is one of the most crucial transistor parameters and is of great importance as it impacts the logic switching characteristics in digital circuits. In one full print run, a total number of 760 OECTs are manufactured on a ~A4-sized substrate area (200 × 300 mm2, Fig. 1). Statistical investigations of the device performance are essential in order to determine the ON/OFF-based yield as a function of OECT design. Moreover, from a digital circuitry standpoint, only 20% of the total area of the printed A4-sized sheet is required to realize complex circuits such as decoders and shift registers20, circuits which typically consist of more than 100 OECTs. In this study we evaluate, in depth, all the OECTs printed onto one single sheet, even though one print batch typically contains an (statistical) ensemble of several sheets. We expect the variation between sheets to be very small since the most crucial fabrication step, the alignment between layers, was tuned for every sheet with respect to the alignment marks of the previously printed layers. Besides the ON/OFF current ratio vs. OECT design, our investigation will also provide information about the quality of the screen printing process across the total substrate area, by determining the variation of the ON/OFF current ratio vs. the OECT location on the particular sheet. The test design includes 120 blocks, where each block contains five OECTs and each OECT has a channel area of 200 × 200 µm2 (referred to as Small OECTs). Additionally, the design contains 32 blocks of five relatively larger sized OECTs, with a channel dimension of 400 × 400 µm2 (referred to as Large OECTs). In total, 600 Small OECTs and 160 Large OECTs are available on the full test design.
For all OECTs studied, two carbon features serve as the source and drain electrodes, i.e., injecting and extracting holes transported through the PEDOT channel, respectively (Fig. 2). A rectangular area of each carbon contact is not covered by any insulator layer and is therefore also in direct contact with the electrolyte layer. The Small OECT structure has a total carbon area of 70,000 µm2 that is exposed to the electrolyte. The large OECTs are divided into two kinds: (i) large OECTs with a 110,000 µm2 carbon area exposed to the electrolyte: referred to as Large Version 1; and (ii) large OECTs with an exposed carbon area of 280,000 µm2: referred to as Large Version 2. The purpose of designing two sets of large OECTs is to study the effect of the carbon area exposed to the electrolyte while keeping the channel dimensions (OECT size) constant. Therefore, the layout includes three sets of OECT structures (Small, Large Version 1, Large Version 2). Top and cross-sectional views of the OECT structures are given in Fig. 1. The complete screen printing process to manufacture the OECTs is given in Fig. 2a. The approximate footprint of a single OECT, omitting any conducting routing lines, measures ~1 × 1 mm2 (marked with a dashed square in the figure). In Fig. 2b, the three OECT structures are depicted; yellow rectangles illustrate the carbon contacts exposed to the electrolyte and red squares mark the active channel areas.
All 760 OECTs were evaluated by measuring the ON/OFF current ratio. Figure 3 shows typical transfer and output characteristics, respectively, for a block of five adjacent Small OECTs. While measuring the transfer characteristics, a fixed VD = −1 V was applied while VG was swept from 0 → 1.5 → 0 V by using a voltage step of 10 mV. In the output measurements, VD was swept from 0 to −1.25 V by using a voltage step of 10 mV, for VG values ranging from 0 to 1.5 V with 250 mV increments. For all measurements, an integration time of 20 ms was used for each data point.
The results in Fig. 3a show that the ON/OFF (ID(VG = 0)/ID(VG = 1.5 V)) current modulation reaches approximately 10,000 for the Small OECTs. The gate current (IG,OFF) equals ID,OFF as the PEDOT phase in the channel is “fully” reduced close to its neutral state (at VG > 1.3 V). For the transfer curve measurements ID,ON and ID,OFF are picked from the forward sweep, the latter value is picked at VG equal to 1.5 V. The channel includes a capacitive component that will result in an initial current drop, which is caused by a potential gradient along the channel upon applying VD. Therefore, to avoid such initial capacitive-like behavior in the ON/OFF current ratio estimations, ID,ON values were extracted at VG equal to 0.1 V. Figure 3b shows the typical output characteristics of the Small OECTs where the linear (low VD) and saturated (elevated VD) regimes are clearly observed for different VG values.
Figure 4 includes the histograms of the ID,ON, ID,OFF, IG,OFF values and the drain current ON/OFF ratio for all the 600 Small OECTs recorded from the forward sweeps of each transfer curve. The empirical mean values of ID,ON, ID,OFF, IG,OFF, and the ON/OFF current ratio is −0.4 mA, −70 nA, 73 nA, and 7800, respectively. Based on these empirical mean values and corresponding standard deviations, Gaussian distributions are given. The OECTs, here studied, are designed and aimed to be used in circuits. Circuits relying on p-type depletion mode OECTs are typically based on the concept of voltage division in a resistor ladder (R1, R2, R3), and it is assumed that a functional circuit requires at least a channel resistance that can be switched between values that are ~20 times lower (ON) and ~20 times higher (OFF) than the value of a specific resistor (R3) in the ladder. Hence, a minimum ON/OFF current ratio of ~400 is thus required19. Therefore, an ON/OFF threshold value of 400 is used to define functional OECTs throughout this report, and this threshold was applied as the criterion to determine the manufacturing yield. All the 600 Small OECTs were operational and exhibited ON/OFF current ratios varying between 2200 (measured minimum value) and 32,000 (maximum value) with an empirical mean value of 7800. This demonstrates consistency of the screen printing process, and also indicates that it is possible to obtain similar switching functionality in all OECTs across the relatively large substrate. Notably, the lowest ON/OFF value obtained from the set of Small OECTs is 2200, which certainly is sufficient to guarantee successful propagation of a signal through cascaded sub-circuits in a screen-printed large-area OECT-based circuit. Thus, we classify all the 600 OECTs as operational, i.e., the manufacturing yield is here 100%.
Next, the performance parameters of the Large Version 1 and Large Version 2 OECTs were examined, especially paying attention to the effect of the size of the carbon areas exposed to the electrolyte. The transfer characteristics for these OECTs (Supplementary Fig. 1) indicate that there is a negative effect on the ID,OFF and the IG,OFF current levels as the area of the carbon contacts exposed to the electrolyte is increased.
In Fig. 5, the performance summary of the Large Version 1 and Large Version 2 OECTs is presented as histograms for the ID,ON, ID,OFF, IG,OFF values and the drain current ON/OFF ratio for transfer curves recorded in the forward sweeps. Among the total of 160 large OECTs, there were only two of the Version 2 OECTs that were not functional, i.e., they were exhibiting an ON/OFF value below 400. (The microscope images of these two OECTs are shown in Supplementary Fig. 2) These two malfunctioning OECTs are included when calculating the ON/OFF-based manufacturing yield for the printed sheet examined (the percentage of functional OECTs), but excluded from the performance statistics presented in Fig. 5. The empirical mean value of ID,ON for all large OECTs, i.e., the 158 operational large OECTs, is −0.4 mA whereas the mean value of ID,OFF is −230 nA and −500 nA for the Large Version 1 and Large Version 2, respectively. Consequently, the empirical mean value of the ON/OFF current ratio for Large Version 1 and Large Version 2 OECTs is 2250 and 720, respectively. Based on these empirical mean values and the corresponding standard deviations, Gaussian distributions were plotted. As seen in Fig. 5, the two sets of large OECTs display two distinctly different, but closely situated, ID,ON distributions with only 0.09 mA difference between their mean values. This is due to that the active channel areas are identical for the two designs. However, a relatively much larger difference is found for the ON/OFF current ratios when comparing Large Version 1 and 2. The difference stems from the fact that the ID,OFF values are not equal. We hypothesize that the different sizes of the carbon source and drain contact areas, exposed to the electrolyte, lead to different levels of parasitic electrochemical reactions. The most probable parasitic electrochemical reactions are associated with water splitting at the drain and/or the source electrodes at an elevated voltage difference between the gate electrode and the carbon pads serving as the drain and source electrodes. Upon applying a potential difference between the source and drain contacts, electrochemical side reaction is inevitable, and this effect should scale with the contact area exposed to the electrolyte (cf. Fig. 2b). So, for IG,OFF, ID,OFF, and the ON/OFF current ratio two distinctive sets of Gaussian distributions are representing the Large Version 1 (green counts) and Large Version 2 (blue counts) OECTs, see Fig. 5. The minimum and maximum values of the ON/OFF current ratio for the 80 operational Large Version 1 OECTs are ~900 and ~6900, respectively, with the empirical mean value of 2250. For the 78 operational Large OECTs Version 2 OECTs, ~400 and ~900 are the minimum and maximum ON/OFF values, respectively, with an empirical mean value of 720. From this comparison it is obvious that the Large Version 1 OECTs are superior as compared to the Large Version 2 OECTs. It is worth mentioning that even the lowest ON/OFF current ratio measured (400, as obtained from the Large Version 2 OECTs) most probably will work in an OECT-based circuit where a minimum ON/OFF current ratio of ~400 is required.
All three OECT structures (Small, Large Version 1, Large Version 2) exhibit more or less the same ID,ON values (−0.4 mA) and related distributions (cf. Figs. 4a and 5a) due to identical channel resistor values, which stems from the fact that all channels are shaped as squares, making the channel resistance equal to the PEDOT:PSS sheet resistance of ~1 kΩ/□. Furthermore, it can be observed that enlarging the OECT channel and the carbon area exposed to the electrolyte by a factor of four (Small vs. Large Version 2 OECTs) results in significantly increased ID,OFF values. The empirical mean value of ID,OFF for the Small OECTs and Large Version 2 OECTs is −70 and −500 nA, respectively. In Table 1, empirical mean, minimum, and maximum values of the ON/OFF current ratios of the three OECT structures are summarized. Following the assumption of having an ON/OFF ratio threshold value of ~400 to define the manufacturing yield, all 600 Small OECTs are positioned well above this limit. Although the large OECTs were designed for the purpose of studying the effect of the carbon areas exposed to the electrolyte, they also satisfy the assumption. The overall manufacturing yield of the 760 OECTs is 99.7% with only two malfunctioning OECTs.
When applying printed OECTs into various applications, such as in digital switches and circuits, the dynamic switching characteristics is of importance. In the dynamic switching tests conducted here, VG was sourced from a function generator in the form of a square wave alternating between 0 and 1.5 V at 500 mHz frequency, with a 50% duty cycle, and at the same time a constant VD of −1 V was applied. ID was then measured using an integration time of 640 µs. Figure 6 displays the dynamic switching characteristics of two blocks of Small OECTs along with their corresponding transfer characteristics. In the dynamic switching tests, the cycle time was set to 2 s. The sampling frequency differs slightly from sample to sample due to limitations of the parameter analyzer (HP/Agilent 4155B) and therefore the time axis of each curve has been normalized with respect to its sampling frequency. The two Small OECT blocks were selected in a manner that the transfer curves of all the OECTs within one block either match very well with one another (cf. Fig. 6a) or differ significantly from each other (cf. Fig. 6b). It is observed that the dynamic switching behavior is consistent regardless if the transfer curves are overlapping or not. This is due to the fact that the ID,ON and ID,OFF levels are almost identical for the transfer curves even when the different transfer curves do not overlap. Based on these results, the fabricated OECTs, here reported, are found to be good candidates for future development of digital circuits of various kinds, mainly due to that maximum ON/OFF current ratio is obtained even with slight variations in transfer characteristics.
The switching time is one of the key parameters in printed circuit applications, as it dictates the overall speed of circuit operation. For OECTs, the switching times typically fall in the 20 to 200 ms range5,21. For the Small OECTs, here presented in Fig. 6, the typical switching times for ON-to-OFF (ID switches from 90 to 10% of its maximum value), and OFF-to-ON (ID switches from 10 to 90% of its maximum value), extracted from the dynamic switch measurements, are found to be 20 and 225 ms, respectively (dynamic switching measurements of the two large versions of OECTs are provided in Supplementary Fig. 1). The Large Version 1 OECT gives higher current modulation than the Large Version 2. For Large Version 1 OECTs, the typical switching times for ON-to-OFF and OFF-to-ON are found to be 240 ms and 1.4 s, respectively. The ON-to-OFF and OFF-to-ON switching times for Large Version 2 OECTs are 320 ms and 1.8 s, respectively. The Small OECTs exhibit the shortest switching times due to their small channel area along with their small carbon areas exposed to the electrolyte. Comparing the two sets of large OECT versions, the switching time is prolonged as the carbon areas exposed to the electrolyte are increased.
Because of the excellent manufacturing yield and performance of the Small OECTs, additional miniaturization steps were undertaken, in order to further push the limits of OECTs manufactured by screen printing. OECTs with channel dimensions of 150 × 100 µm2 and carbon areas exposed to the electrolyte of 45,000 µm2 were designed and printed, referred to as Tiny OECTs. The performance of the Tiny OECTs was evaluated based on drain current ON/OFF ratios. While promising results were achieved for some of the OECTs in this design, precision limitations of the printing method frequently came into play at these dimensions and the results were not entirely consistent.
ON/OFF current ratios higher than 100,000 were observed for some of these OECTs. As an example, the performance of a block of six printed OECTs (Supplementary Fig. 3a) gives ID,ON and ID,OFF values of ~1 mA and ~5 nA, respectively. The relatively higher ID,ON, as compared to the values reported above, is explained by that the channel length and width are 100 and 150 µm, respectively, which results in a lower channel resistance value than for the square-shaped channels. Further, the Tiny OECT design has also a smaller carbon area exposed to the electrolyte, which leads to relatively lower parasitic current contribution, and hence lower values of ID,OFF and IG,OFF. This led to ON/OFF current ratios of more than five orders of magnitude. However, the results in terms of ID,ON and ID,OFF levels were not consistent for the entire sheet and the manufacturing yield in terms of ON/OFF current ratio decreased to about 98% (based on the yield threshold of ON/OFF current ratio > 400). The reason for the decreased yield is probably directly tied to the small dimensions used, which introduces more malfunctioning devices. Printing several consecutive layers on top of each other requires excellent alignment accuracy between all the different layers. When the printed dimensions are truly small with respect to the accuracy limitation of the printing method, even a slight misalignment will have detrimental impact on the performance of the printed OECT device. Unintended contacts between layers will affect the performance, leading to large spread in terms of ON/OFF current ratio (exemplified in Supplementary Fig. 3b) across the sheet, or even inoperative devices, both being equally bad from a circuit design perspective. For instance, having an insulator layer slightly misaligned with respect to the carbon pads (serving as the source and drain electrodes) may create a different OECT architecture, and hence utterly different switching performance22. Having this in mind, it is critical to ensure that the same type of OECT architecture can be printed across the whole substrate area; hence, screen printing of the smallest possible OECT structures is not recommended. Choosing the Small OECT design results in more reliable and reproducible devices and is preferable over the Tiny OECT design, even though the trade-off leads to higher leakage current and lower ON/OFF current ratio.