High Frequency Solution‐Processed Organic Field‐Effect Transistors with High‐Resolution Printed Short Channels

Organic electronics is an emerging technology that enables the fabrication of devices with low‐cost and simple solution‐based processes at room temperature. In particular, it is an ideal candidate for the Internet of Things since devices can be easily integrated in everyday objects, potentially creating a distributed network of wireless communicating electronics. Recent efforts allowed to boost operational frequency of organic field‐effect transistors (OFETs), required to achieve efficient wireless communication. However, in the majority of cases, in order to increase the dynamic performances of OFETs, masks based lithographic techniques are used to reduce device critical dimensions, such as channel and overlap lengths. This study reports the successful integration of direct written metal contacts defining a 1.4 µm short channel, printed with ultra‐precise deposition technique (UPD), in fully solution fabricated n‐type OFETs. An average transition frequency as high as 25.5 MHz is achieved at 25 V. This result demonstrates the potential of additive, high‐resolution direct‐writing techniques for the fabrication of organic electronics operating in the high‐frequency regime.


Introduction
[3] The main advantage is the use of low-cost, solution based and high throughput fabrication processes.In particular, digital additive direct-writing approaches offer efficient patterning of different materials on a large variety of substrates without the use of any mask, strongly reducing the cost and time required of pattern change and reducing the waste of precious materials.However, despite of the advantages, current direct-writing techniques such as inkjet and higher-resolution jetting strategies show several limitations, especially where high operational frequencies are required, as for Internet of Things (IoT) applications.This can be understood considering the expression of the so-called transition frequency (f T ), which is largely adopted to determine the maximum operational frequency of an organic field-effect transistor (OFET), a fundamental building block of any organic electronic circuit and an ideal component for rectifiers in the IoT field.f T is defined as the frequency at which the small-signal gate and drain currents become equal, and in saturation it can be expressed as follow [4] : where g m is the device transconductance, C g the total gate capacitance, μ eff the effective charge carrier mobility, [5] V TH the threshold voltage, L ch and L ov the channel and overlap length, respectively.Hence, to make faster printed circuits one should employ semiconductors performing well with short channels and downscale transistor dimensions.The first is made difficult by the contact resistance R C , responsible of performance degradation, in terms of μ eff reduction, when the channel length is shortened. [6]he second point is more a technological limitation: finding a solution based additive direct-written fabrication method characterized by high resolution, able to preserve at the same time a low level of complexity and a high throughput.Inkjet is a widely used direct-writing, printing technique and has the advantage to be very versatile and already demonstrated to be scalable at industrial level. [7,8]However the achievable resolution with stateof-the-art tools is typically ≈ 10 μm.Some strategies were developed to improve the resolution.One way is to locally modify the surface properties of the substrate, such as ink wettability, creating an alternation of hydrophobic and hydrophilic regions by either a physical or chemical treatment. [9,10]Along this strategy, self-aligned printing (SAP) exploits the wettability modification of the first printed electrodes, for example using self-assembled monolayers (SAMs), in order to repel the second ones, leaving a small gap with a length in the range of hundreds of nm. [11,12]owever, such methods do not improve the resolution of the printed lines, but only their mutual alignment.This does not allow to directly tackle the scaling of the parasitic overlap among electrodes (L ov ).[15][16] However, EHD presents some constraints: the ink must be conductive or at least polar and, very importantly, the method relays on a fine control on the electric field distribution on the whole substrate, complicating the design of multi-nozzles systems and the scaling up of the technique. [17]Moreover, the high electric fields used can be deleterious for previously patterned components.[20][21][22] In this case the waste of materials is higher with respect to inkjet, since it is not a fully additive process, and its scaling up requires a higher complexity opto-mechanical system.Recently a novel printing method was introduced for high resolution patterning, called Ultra-Precise Deposition (UPD), which is able to retain the simplicity and versatility of inkjet printing but with an improved spatial resolution, comparable with laser assisted printing.It allows, without the use of external electric fields like in EHD, high resolution printing of either conductive or insulating inks on different types of substrates, which may also present high aspect ratio topographical features.Mateusz Łysień et al. recently introduced this technique to print silver lines with resolution down to 700 nm with a constant and well-defined line separation, also down to 700 nm. [23]The process can be described as a pressure assisted inkjet printing of an highly concentrated paste placed inside a nozzle with a fine diameter (between 0.5 and 10 μm).Thanks to the non-Newtonian properties of the ink at the tip of the nozzle, the paste can be ejected in an extremely confined way with resolution ranging from hundreds of nm up to tens of μm.Typically, the thickness of UPD printed silver electrodes is around hundreds of nanometers, much higher than what obtained with inkjet.This feature is advantageous in several cases as it allows low electrical resistance, but it may represent an issue in the case of OFETs, which are characterized by functional layers in the tens of nanometers range.Since such aspect ratio electrodes are not common in OFET, UPD has not been adopted before to fabricate high performance, downscaled organic transistors.
In this work, we report fully solution fabricated n-type, polymer based OFETs operating in the high-frequency (HF) range thanks to the adoption of UPD printed silver electrodes with mi-cron scale resolution.The frequency performance is enabled by an average electrode line width as low as 3.3 μm, reducing the parasitic overlap length, and a controlled channel length of the transistors, of 1.4 μm on average, increasing the device transconductance, in combination with a suitable interfacial modification to improve charge injection.Our results show that, despite the unusual electrode thickness (maximum thickness of ≈ 330 nm), performing OFETs for HF applications can be fabricated with digital high resolution patterning methods such as UPD.An important aspect to achieve optimized devices with UPD patterned electrodes is the use of a molecular n-type dopant to retain excellent transport properties in the thin semiconducting layer, with ideal and reproducible device characteristics.An average value of transition frequency as high as 25.5 MHz was measured, which is the best achievement obtained up to now in organic devices with direct-written, printed source and drain electrodes.In perspective, by further reduction of the overlap length by scaling of the gate line as well, a transition frequency above 100 MHz appears feasible, opening a path for printing-based, direct-writing schemes for the manufacturing of high-performance OFETs for HF and even ultra-HF applications, relevant for communication purposes in IoT.

Results and Discussion
We printed interdigitated silver (XTPL Ag Nanopaste CL85) contacts by UPD on glass substrates to be used as source and drain electrodes in transistors (Figure 1a).In Figure S1 (Supporting Information), an example of an array of interdigitated contacts is shown.For each transistor, five silver electrodes were printed with an average line width of 3.34 ± 0.19 μm, for a total channel width of 2 mm and an average channel length of ≈ 1.47 ± 0.24 μm.As reported in ref., [23] direct-written silver contacts printed by UPD and characterized by a resolution down to 2 μm, are characterized by an electrical conductivity of ≈ 2.5 × 10 7 S m −1 (40% of bulk A g ).Therefore, the expected electrical resistance of the UPD contacts deposited in this work is ≈ 74 Ω.The electrodes are characterized by a round shape and an average maximum thickness of 330 ± 30 nm.After printing, a short drying of the silver contacts was performed, heating the substrates at 200 °C for 10 min.A more detailed description of the printing method and parameters to achieve such spatial resolution can be found in a previous work. [23]o assess the potential use of the UPD printed silver contacts in OFETs, we fabricated staggered top gate-bottom contact (TGBC) field-effect transistors based on the well-known, reference co-polymer poly[N,N ʹ -bis(2-octyldodecyl)-naphthalene-1,4,5,8-bis(dicarboximide)−2,6-diyl]-alt-5,5 ʹ -(2,2 ʹ -bithiophene) (P(NDI2OD-T2)) as electron transporting semiconductor.A sketch of the device architecture is presented in Figure 1b.P(NDI2OD-T2) was deposited from a 7 g l −1 toluene solution by off-centered spin-coating perpendicularly to channel, in order to achieve uniaxial alignment, which ensures optimal charge transport. [18,20,24]Poly(methyl methacrylate) (PMMA) was used as a solution processable dielectric since it is known to form an optimal interface with the chosen semiconductor. [25,26] gate contact made of poly(2,3-dihydrothieno-1,4-dioxin)poly(styrenesulfonate) (PEDOT:PSS) was ink-jet printed on top of the dielectric, in full overlap with source and drain electrodes.The realized device with downscaled channel length is therefore fully solution fabricated: the process flow is schematized in Figure 1d.
[30] P(NDI2OD-T2) was selected as semiconductor since it is a well-studied material characterized by good electron transporting properties with field-effect mobilities exceeding 1 cm 2 /Vs when polymer backbones shows directional alignment. [24,31]Moreover, thanks to the non-linear injection properties of P(NDI2OD-T2) with the lateral electric field, the increase of weight of contact resistance decreasing the channel length, at fixed bias, is milder than what expected in TGBC devices. [27,32]However, despite this effect, to retain good electrical characteristics at short channels, some additional strategies are often needed.In our case, to improve the injection properties we exploited the combined effect of a self-assembled monolayer (SAM), to lower the work-function of electrodes, and of mild doping of the semiconductor, to fill deep trap states.In particular, as SAM we used dimethylamino(benzenethiol) (DABT), known to decrease the work-function of noble metals like silver. [33]To dope the semiconductor, we used a derivative of 4-(2,3-dihydro-1,3-dimethyl-1H-benzimidazol-2-yl)-N,Ndimethylbenzenamine (DMBI-H), which is a well-known n-type dopant for P(NDI2OD-T2). [34]In Figure 1c chemical structures of both the molecule adopted for the SAM and the dopant are depicted, along with the adopted polymer semiconductor.The modified benzimidazoline dopant is 4-(1,3,5,6-tetramethyl-2,3-dihydro-1H-benzoimidazol-2-yl)-N,N-diisopropylaniline (N-DiPrBI-Me 2 ); the introduction of a diisopropyl group on the aniline moiety has already been reported to improve the solid-state solubility of the compound in the host polymer, with respect to the standard DMBI-H, leading to a more efficient intercalation in its semicrystalline structure. [35]The presence of two methyl groups on the benzimidazoline core, on the other hand, aims at tuning the electron donating properties of the compound, as similarly presented in the work of Fabiano and co-workers. [36]The combination of these chemical modifications of the DMBI-H structure are expected to enhance the doping efficiency.A more in-depth investigation on this novel dopant molecule and of its performances will be addressed in a future publication.Doping was performed by solution mixing at 1 mol% with respect to the monomeric unit of the polymer.In the Supporting Information, the synthetic procedures and the NMR characterization of the dopant (Figures S2 and S3, Supporting Information) are reported.
The cross-section profile of the fabricated transistors was obtained by FIB-SEM imaging.A typical cross-section of a device is shown in Figure 2a, while in Figure S4 (Supporting Information) images of different electrodes on various samples are collected.In Figure 2b,d magnifications on a single electrode and in the channel, region are performed.From Figure 2b it is possible to notice that electrode thickness varies smoothly along its profile without any abrupt change, which is beneficial to allow the polymer film continuity between contacts and channel regions.The metal nanoparticles are still visible in the dried electrodes, and their granulometry was quantified by measuring the average di-ameter of the particles across various cross-sections images (an example is reported in Figure 2c).A value of ≈ 40 nm was estimated, close to the diameter of the silver nanoparticles composing the starting silver ink (nominal average diameter ≈ 50 nm).While the different organic films cannot be distinguished one from the other, in Figure 2b,d, it is possible to notice a localized thinning of their stack in the contact region with respect to the channel area.The maximum films thinning was quantified as ≈ 36% on average, by considering the average maximum (h max ) and minimum (h min ) heigh of the films equal to 640 and 410 nm, and an average maximum thickness of the electrodes (t max ) equal to 330 nm.
We characterized the electrical properties of our OFETs by first measuring DC characteristic curves.In Figure 3a a comparison of transfer curves between a doped (blue) and undoped (black) device is shown.Upon doping, it is possible to notice a clear increase in the OFF current of four and three order of magnitudes in linear and saturation regimes, respectively, due to the increase of bulk conductivity.Also, the ON current at the same gate voltage is increased of more than one order of magnitude.The minimum OFF and ON resistances of the doped transistor are ≈ 230 kΩ and 2.3 kΩ, respectively, much higher than that of the printed contacts.The threshold voltages in both linear and saturation regimes were extracted, and they are equal to 19 (V d = 5 V) and 15 V (V d = 40 V) for the pristine and 4.2 (V d = 5 V) and 3.9 V (V d = 23 V) for the doped device.Using gradual channel approximation, field-effect mobilities curves were computed in both regimes (Figure 3b).For the pristine device there is a marked gate-voltage dependence of mobility with maximum values ≈ 0.025 cm 2 Vs −1 at V d = 5 V and 0.065 cm 2 Vs −1 at V d = 40 V, respectively.We have not deeply investigated the origin of such non-ideality, but it is a typical fingerprint of the presence of trap states, which can tentatively be ascribed to the marked granularity and roughness of the electrodes, as well as their thickness.Upon molecular doping, the device characteristics in full accumulation are much more ideal, at the expense of an increased OFF current, which is in any case not detrimental for improving a figure of merit like f T .The observed effects are compatible with the introduction of excessive electrons upon doping that fill trap states in the semiconductor and contribute to an increased film electrical conductivity in the OFF state.As a consequence, charge mobility of the doped device is characterized by a markedly reduced bias dependence with respect to the undoped transistor and by an improved reliability factor (r), [5] which goes from 0.7 and 0.6 for the linear and saturation mobility to 0.85 and 0.75 (Figure 3b).In this case, a mobility as high as 0.4 (V d = 5 V) and 1.4 cm 2 Vs −1 (V d = 23 V) is obtained.These values are comparable to what already obtained in optimized short channel P(NDI2OD-T2) OFETs with fs-laser direct-written contacts treated with the same self-assembled monolayer used in this work. [18]n Figure S5 (Supporting Information), the output characteristic curve for the doped device is shown.It is possible to observe that there is no clear saturation of drain current with drain voltage, probably due to a short channel effect.[37] Furthermore, it can be seen that no S-shape at low drain voltages is present, which is an additional indication of good electron injection, in accordance to what deduced before form the mobility curve.
In Figure 4a, the average transfer curve and its standard deviation, obtained by measuring ten doped transistors is presented, indicating good reproducibility.In Figure 4b, the corresponding average mobility curves in both linear and saturation regimes are reported, with mean mobility values ≈ 0.35 and 1 cm 2 Vs −1 respectively, in the V g range between 12 and 23 V. Looking at Figure S6a,b (Supporting Information) it is possible to appreciate a good linear behavior of I d and √I d with gate voltage, indication of good device ideality.By performing a linear fit in the same voltage range used for mobility extraction, it is possible to derive also average threshold voltages equal to 2.3 and 2.7 V in linear and saturation regimes, respectively.
The average quasi-static width-normalized DC transconductance g m /W, an important parameter to have a prediction of the transition frequency of the devices, reaches 0.5 mS cm −1 at V d = V g = 23 V. Furthermore, to preliminary assess the intrinsic shelf-life stability of the devices, they were kept in a nitrogen glovebox after the first measurement, and then re-measured after 2 weeks without any sign of degradation (Figure S7, Supporting Information).
To characterize the semiconductor film on both contacts and channel region, and understand the relation between electrical and morphological properties, we used Atomic Force Microscopy (AFM) on samples prepared following the same procedure for the case of transistors, but without any dielectric and gate electrode.Such characterization is relevant, as the high electrodes thickness may impair the formation of an optimal film for charge transport.As a reference, an AFM image of the bare direct written contacts is shown in Figure 5a characterized by a derived root-meansquare roughness (RMS) of 10 ± 0.5 nm (see details in Figure S8, Supporting Information).Figure 5b shows the AFM topography of the coated polymer films on top of the contacts across the channel area: it is possible to notice that between the elec- trodes the material assumes a highly oriented fibrillar structure, aligned in the perpendicular direction with respect to the channel.The latter is typical of P(NDI2OD-T2) when deposited using directional solution shearing techniques, such as off-centered spin coating and bar-coating from solvents like toluene or mesitylene, which allow strong pre-aggregation in solution. [24,38]Because of the high contrast between the electrode thickness and any feature characterizing the semiconductor morphology, we performed a magnification inside the channel region to clearly distinguish the film morphology.The polymer film topography is very similar to what previously observed in literature for oriented P(ND2OD-T2) and it is characterized by a surface RMS roughness value of 1.8 ± 0.4 nm. [24,31]This proves the successful polymer alignment between the two rather thick direct-written electrodes and explains the good transport properties observed in the fabricated transistors.In Figure 5c a magnification of the contact region coated with the P(NDI2OD-T2) film is reported with a surface RMS roughness value of 20 ± 0.5 nm.The contacts appear only partially covered by the polymer, likely reflecting a different surface interaction.The topography of the polymer is drastically different with respect to the channel, presenting ribbon-like shaped features, apparently interconnected.The non-ideal surface coverage and the different topography could be the reason of an inefficient carrier injection from the contacts, as observed in the pristine OFETs, which we solved by molecular doping of the semiconductor.
To finally assess the dynamic behavior of our devices, we measured the transition frequency f T .Since the undoped transistors suffer severe contact limitations, only devices with doped P(NDI2OD-T2) were considered for dynamic measurements.We adopted the measurement set-up introduced by Perinot et al., which allows to measure separately channel transconductance g m and gate-to-source and gate-to-drain capacitances (C gs , and C gd ). [19]Then, f T value is identified as the crossing point between the total gate admittance, which depends on the total gate capaci-tance C g , and g m .In the frequency range from 1 kHz to 10 MHz, we estimated a mean width-normalized transconductance equal to 0.53 mS cm −1 averaging over ten devices at V d = V g = 25 V and of 0.61 mS cm −1 for the best performing transistor, which are in line with the previous estimations based on quasi-static transfer curves (Figure 6a).Then, average values of C gd , C gs , and C g equal to 0.16, 0.45, and 0.63 pF were obtained by performing a fit in a frequency range between 100 kHz and 10 MHz (Figure 6b).The gate-to-source capacitance is higher since the source is made by three fingers in the interdigitated structure, while the drain is made of two, and since the measurement is performed in saturation, with the source end of the channel showing charge accumulation and the drain end in depletion.As an approximation, by considering a simple parallel plate model, it is possible to estimate theoretically what the total gate capacitance should be, knowing the geometrical overlap (A ov ) and channel (A ch ) area and the dielectric capacitance per unit area C diel .We estimated that C g should nominally fall within a range from 0.54 to 0.76 pF, as experimentally observed.Details about the estimation are reported in the Supporting Information.With such prediction for capacitances and the quasi-static g m , by using Equation ( 1) it is possible to estimate a theoretical transition frequency between 22 and 31 MHz. Figure 6c,d shows the experimental extraction: an average value of f T as high as 25.5 MHz (over ten transistors) and of 28 MHz for the best device are achieved at 25 V, in agreement with the previous estimations.This is the highest transition frequency obtained so far for organic transistors with printed source and drain electrodes.

Conclusion
In this work, the possibility of using high resolution directwritten metallic contacts deposited by an Ultra-Precision Deposition method for the development of fully-solution fabricated organic transistors is demonstrated.In particular, by downscaling the channel length to 1.4 μm and the silver source and drain contacts width to 3.3 μm in TGBC staggered n-type OFETs, based on the model polymer P(NDI2OD-T2), it was possible to achieve an average transition frequency f T of 25.5 MHz at 25 V, the highest f T achieved with OFETs where the critical features is defined by a printing method.A key aspect for the achievement of such AC performance is the combination of a SAM treatment of the silver contacts, a mainstream approach to fine tuning charge injection, with molecular doping of the semiconductor, by means of a benzimidazoline derivative.We highlighted that the printed contacts, characterized by a rounded profile, a maximum height (>300 nm) much thicker than the semiconductor film thickness (40 nm), and a high surface RMS roughness of 10 ± 0.5 nm, do not preclude the formation of an optimal film microstructure in the channel area.Yet, they do not allow ideal coverage by the semiconductor on the contacts, producing non-idealities in the electrical characteristics of OFETs based on the pristine polymer.Such non-idealities disappear upon molecular doping, likely owing to deep trap states filling, allowing the achievement of reproducible devices with excellent performances.
Our results qualify a fully solution-based process for the realization of OFETs where the downscaling of the channel is achieved with a printing, direct-written technique, for applications in the HF bandwidth.Moreover, they pave the way to even higher operational frequencies, which could be achieved by acting on the downscaling of the gate line on the one side, and on the increase of the effective mobility with higher performing semiconductor on the other.Back-on-the-envelope estimations suggest that the UHF band could be achieved in the future with direct-written OFETs, making them candidates for cost-effective and sustainable IoT.

UPD Printing of High-Resolution Electrodes:
The glass substrates used (low alkali 1737F Corning glasses, purchased from Apex Optical Services) were cut into pieces of 1 cm × 1 cm and then cleaned by sonication with acetone and isopropyl alcohol (both purchased from Sigma Aldrich) for 10 and 5 min.The printing of high-resolution electrodes using the UPD method was performed on the XTPL Delta printing system.Silver contacts were printed using the XTPL Ag Nanopaste CL85.The properties of the adopted Ag ink were reported in ref. [23] In order to obtain silver contacts as source and drain electrodes, process parameters had to be optimized.Appropriate nozzle diameter, printing pressure, and printing velocity had a crucial influence on the whole process.The outer diameter of the nozzle was 3.5 μm.In addition, appropriate humidity (40-60%) and temperature (21-24 °C) also had to be kept constant during the fabrication.Sintering of the printed lines was performed in the air on a hotplate (CAT H 17.5D).
Transistor Fabrication: To fabricate our n-type OFETs the glass substrates with the silver electrodes previously printed were cleaned with isopropyl alcohol (Sigma-Aldrich) and then a plasma treatment in argon (100 W for 1 min) was performed.Successively, the substrates were immediately put in a solution of dimethylamino(benzenethiol) (TCI chemicals) and IPA in a concentration of 0.0014 v/v and left there for 25 min.P(NDI2OD-T2) (Polyera Corporation ActivInk N2200) was dissolved in Toluene in concentration of 7 g l −1 by stirring for 3 h.For the doped devices, N-DiPrBI-Me 2 solution was prepared in a glovebox at a concentration of 1 g l −1 and left stirring for 1 h.Then a given amount was added to the P(NDI2OD-T2) solution to have 1 mol% with respect to the monomeric unit of the polymer.The semiconductor solution was deposited by offcentered spin coating in a glovebox and immediately put on a hot plate at 120 °C for the pristine devices or 180 °C for the doped ones for 30 min.
After that, a solution of PMMA (Sigma-Aldrich, averaged M W ≈ 120 00) was dissolved in n-Butyl acetate (Sigma-Aldrich) and deposited by centered spin-coating.Then an annealing at 80 °C for 10 min is followed.PEDOT:PSS (Clevios PJ700 formulation, purchased from Heraeus) gate electrodes were inkjet printed by means of Fujifilm Dimatix DMP2831.At the end, the devices were left in glovebox in annealing for 6 h at 110 °C.
AFM Characterization: The surface topography of the semiconductor films and printed electrodes were characterized with Keysight 5600LS Atomic Force Microscope operating in tapping mode.Gwyddion software was used for image processing and calculation of surface roughness.In particular, to evaluate the RMS.roughness of the electrodes, the image was planarized, removing a polynomial background associated to the curvature of the contact (Figure S5, Supporting Information).
Cross-Section FIB-SEM Imaging: FIB-milled cross sections and SEM imaging of the devices were obtained with a Dual Beam FIB/SEM Helios Nano-Lab 600i (FEI) equipped with a Field Emission source.Samples surface was prepared by depositing a thin Au film (t ≈20 nm) by means of a Q150R sputter coater (Quorum Technologies).FIB-milling and polishing of cross-section was obtained with Ga+ source at 30 kV acceleration voltage and 120 pA of current.SEM images of cross sections were obtained under a sample tilt angle of 52°and with accelerating voltage of 5 kV.Thickness of the device layers and electrode dimensions were evaluated on cross section images by using software ImageJ version 1.53a.
Statistics were performed on 51 images from seven different devices crosssections.Granulometry of Ag nanoparticles composing the electrodes was quantified by measuring the average diameter of the particles on three high-magnification cross-sections images (23 sample measures for each image).
Static and Dynamic Electrical Characterization: OFETs transfer and output characteristic curves were measured with semiconductor parameter analyzer (Agilent B1500A) inside a glovebox with a Wentworth Laboratories probe station.The dynamic characterization of OFETs was carried out in the glovebox using a custom setup that includes an Agilent ENA Vector Network Analyzer and Agilent B2912A Source Meter.To had more information see ref [19].

Figure 1 .
Figure 1.a) Optical images of the bottom source and drain electrodes realized by UPD printing, b) sketch of device architecture, c) chemical structure of materials employed, and d) scheme of the process flow.

Figure 2 .
Figure 2. a) Typical cross-section FIB-SEM image of a device and b) magnification of a single electrode region.c) Detail of a silver electrode to highlight its granulometry.d) Channel region between two direct-written contacts.All images have been acquired under a sample tilt angle of 52°.

Figure 3 .
Figure 3. a) Transfer characteristics in both linear and saturation regime for a pristine (black curve) and doped (blue curve) OFET; the two devices are characterized by the same channel length and width, 1.7 μm and 2 mm respectively, but different dielectric thickness: 350 nm for the pristine and 550 nm for the doped transistor.The difference is due to different stages of optimization of the devices, and the thicker dielectric in the doped case ensures high yield, while reducing by ≈1.6 times the coupling capacitance.b) Field-effect mobility curves computed using gradual channel approximation from transfer characteristics.

Figure 4 .
Figure 4. a) Characteristic transfer and b) field-effect mobility curve averaged over ten devices in linear (V d = 5 V) and saturation regime (V d = 23 V) with standard deviation.

Figure 5 .
Figure 5. a) AFM image of a bare direct written Ag contact printed with UPD technique, b) AFM image of a P(NDI2OD-T2) film deposited on the printed electrodes, with a magnification of the channel region, and c) AFM image of contact region of a P(NDI2OD-T2) film.The scale bar is 1 μm.

Figure 6 .
Figure 6.a) Average channel transconductance in saturation regime (V d = V g = 25 V), b) average total source-to-gate and drain-to-gate admittance in saturation regime; c,d) estimation of average (over ten devices) and maximum transition frequency at 25 V.