Ion-gated transistors based on porous and compact TiO 2 films: Effect of Li ions in the gating medium

Ion-gated transistors (IGTs) are attractive for chemo- and bio-sensing, wearable electronics, and bioelectronics, because of their ability to act as ion/electron converters and their low operating voltages (e.g., below 1 V). Metal oxides are of special interest as transistor channel materials in IGTs due to their high mobility, chemical stability, and the ease of processing in air at relatively low temperatures ( < 350 ○ C). Titanium dioxide is an abundant material that can be used as a channel material in n -type IGTs. In this work, we investigate the role of the morphology of the TiO 2 channel (porous vs compact films) and the size of the cations in the gating media ([EMIM][TFSI] and

Metal oxide semiconductors are exploited in a variety of applications, such as sensing, energy conversion and storage, and display technologies. [1][2][3][4] They are widely studied as active channel materials for thin-film transistors (TFTs) and ion-gated transistors (IGTs) due to their abundance, chemical stability, transparency, and the ease of processing at low temperatures under ambient conditions. 5,6 Amorphous indium-gallium-zinc-oxide (a-IGZO) is currently used in n-type TFT backplanes of flat panel and organic light-emitting diode (OLED) displays, due to its high electrical conductivity, film uniformity, and transparency, as well as its mechanical and chemical stability. 3,[7][8][9][10][11] Indium-based oxides also show excellent performance as channel materials for IGTs (charge carrier mobility up to ∼120 cm 2 V −1 s −1 ). 12 However, as the availability of indium in the earth's crust is limited, there is an increasing interest in indium-free metal oxides, such as ZnO, SnO 2 , CuxO, NiO, and TiO 2, both for TFTs and IGTs. 5,13-28 Among indium-free metal oxides, TiO 2 (n-type) is sought for its availability, cost-effective processability, high chemical stability, and biocompatibility. 29,30 Our groups reported on TiO 2 -based IGTs using 1-ethyl-3-methylimidazolium bis(trifluoromethylsulfonyl)imide ([EMIM][TFSI]) as the gating medium and found that the patterning of the channel leads to enhanced transistor performance. 19 IGTs use ion-gating media, such as aqueous saline solutions, polymer electrolytes, ionic liquids, and ion gels. A high specific capacitance (∼10-100 μF/cm 2 ), which permits operation at low voltages (<2 V), is achieved by the formation of a thin electrical double layer at the semiconductor/gating medium interface. Besides pure electrostatic or electrochemical doping, different mechanisms have been considered to describe the operation of IGTs, such as interface-confined electrochemical doping. 6,21,[31][32][33][34][35][36][37]

ARTICLE scitation.org/journal/adv
In this work, we explored IGTs based on TiO 2 active layers deposited using thermal evaporation to produce compact films, and solution casting to produce porous films. Transistors based on compact films show electron mobilities about two orders of magnitude higher than those based on porous ones. In order to gain further insight into the doping mechanism, we employed two gating media: [EMIM][TFSI] and lithium bis(trifluoromethanesulfonyl)imide [TFSI]. The characteristics of transistors based on compact and porous films depend on the gating media, which points to different working mechanisms. While [EMIM] + ions (ionic radius ∼ 3.5 Å) are expected to dope TiO 2 by electrostatic and interface-confined electrochemical doping, 19 [Li] + ions, having a small radius (∼0.7 Å), are expected to dope TiO 2 electrochemically via lattice intercalation. [38][39][40] Thermal (e-beam) evaporation was carried out at a deposition rate of 0.7 Å/s and a pressure of 3 × 10 −5 Torr using a Thermionics e-beam evaporator. Solution processed TiO 2 films were obtained by drop casting a suspension containing 2 g of TiO 2 (a mixture of rutile and anatase in xylene, MilliporeSigma, No. 700339) and 0.4 g of polyvinylidene fluoride (PVDF, MilliporeSigma) binder in 20 ml of dimethylformamide (DMF, Caledon Laboratory Chemicals, purity 98.5%). The suspension was mixed in a planetary mixer (Thinky ARM-310) for 30 min at 2000 rpm and then filtered through a 0.2 μm nylon filter. Finally, it was drop cast on the substrates that were first heated at 60 ○ C for 30 min to remove excess solvent, and subsequently annealed at 450 ○ C in a tube furnace for 1 h in ambient air. 19 The thicknesses of the films, measured using a profilometer (Dektak 150) were ∼60 nm for the evaporated films and ∼3 μm for the dropcast films. A high thickness was required for the drop-cast films in order to achieve a complete coverage of the portion of the substrate containing the devices (Fig. S1).
The scanning electron microscopy (SEM) analysis was performed using JEOL FEG-SEM, JSM 7600F. The x-ray diffraction (XRD) study was performed with a Bruker D8 diffractometer using a CuKα beam, at every 2θ = 0.01 ○ . The atomic force microscopy (AFM) study was performed using a Multimode Nanoscope (Bruker) in tapping mode, in air, using Bruker NCHV probes with the resonance frequency between 270 KHz and 400 KHz, k = 40 N/m, and a nominal tip radius of 8 nm.
Ion-gated transistors were fabricated according to a procedure described in previous publications. 19,20 The TiO 2 films were deposited, as described above, on SiO 2 /Si substrates prepatterned with Ti/Au (5 nm/40 nm) source-drain electrodes (W/L = 4000 μm/10 μm) by vacuum sublimation and drop casting. As gating media, we brane. Activated carbon (PICACTIF SUPERCAP BP10) on carbon paper (Spectracarb 2050) was used as the gate electrode. The high surface area (∼1000 m 2 g −1 to 2000 m 2 g −1 ) of this electrode is amenable to high current modulation, while its inertness toward undesired electrochemical reactions is advantageous for device stability. 19,41 Transistor characteristics were measured in a N 2 -purged glove box (O 2 and H 2 O < 3 ppm) with a semiconductor parameter analyzer (Agilent B1500A). We prepared four types of transistors, based on evaporated and solution-processed TiO 2 channels and two different gating media, i.e., [ Fig. S2 for the device configuration during cyclic voltammetry). Device characterization was performed in the following order: cyclic voltammetry (three cycles were performed, the third cycle is reported), output characteristics, transfer characteristics in the linear regime (three cycles were performed, the second cycle is reported), and transfer characteristics in the saturation regime (three cycles were performed, the second cycle is reported).
SEM and AFM images reveal that the evaporated films are compact, with a root mean square (rms) roughness of 1.0 ± 0.1 nm  Table S1 for the values of hysteresis loop area of evaporated TiO 2 IGTs). 21,42 The threshold voltages, extracted from the linear transfer characteristics, are in the range of 0.4 V-0.5 V for both gating media (Table S2). The saturation (V ds = 1 V) transfer characteristics [Figs. 2(c) and 2(f)] show that the ON/OFF ratios are rather low (Table S2) due to high OFF currents (about 10 −4 -10 −3 A), which might be attributed to the higher intrinsic conductivity of the evaporated TiO 2 films. Figure 3 shows the output and transfer characteristics of solution-processed   Table S4). The effect of the gating medium on the transistor current can be explained by a change in the doping mechanism; while the large [EMIM] + ions likely lead to a combination of electrostatic and interface-confined electrochemical doping, 19 38,39 The number of Li intercalation events is favored by the porous morphology of the solution processed films, which explains the much larger difference in the saturation current and ON/OFF ratio with respect to [EMIM] + gating. Moreover, the small size of Li + ions (∼0.7 Å) 38 also facilitates a higher packing density at the gating media/TiO 2 interface, which might lead to more efficient electrostatic and interface-confined electrochemical doping.
The charge carrier mobility was extracted from the transfer characteristics in the linear regime according to the formula μ lin = L I d,lin W e n V d , where L indicates the interelectrode distance (10 μm), W the electrode width (4000 μm), V d the drain-source voltage (0.1 V), n the charge carrier density, e the elementary charge (1.6 × 10 −19 C), and I d,lin the drain-source current at 1.5 V. 19 The charge carrier density is calculated from the linear transfer characteristics, using the formula n = Q eA = ∫ I g dV g r v eA , where Q is the doping charge, Ig is the gate-source current in the forward scan [Figs. 2(b), 2(e), 3(b), and 3(e)], rv is the sweep rate of the gate voltage, and A is the surface area of the semiconductor in contact with the membrane soaked with the gating media (in our case 0.36 ± 0.04 cm 2 ). 19 For the porous films, we also calculated the mobility using the surface area of the same region measured from Brunauer-Emmett-Teller (BET) analysis (see the supplementary material). 19,42 The calculated charge carrier densities and mobilities are summarized in Tables S4  and S5. For evaporated TiO 2 IGTs, a charge carrier mobility of ∼0.5 cm 2 /V s was extracted for both gating media (Table S2). The obtained mobility value is comparable with values reported in the previous articles based on metal oxide TFTs and IGTs. [17][18][19]25,27,[43][44][45][46] Solution-processed TiO 2 IGTs showed a lower mobility with a significant dependence on the gating media.    Figure 4 shows the cyclic voltammetry (CV) of the evaporated and solution-processed TiO 2

based IGTs with [EMIM][TFSI] and [Li][TFSI] + [EMIM][TFSI]
, performed in a two-electrode configuration, with the TiO 2 channel acting as the working electrode and the activated carbon gate (specific capacitance of ∼100 F/g) acting simultaneously as the counter and quasi-reference electrode. 41 The currents obtained with solution-processed TiO 2 are higher than those measured with the evaporated TiO 2 films due to their higher thickness.
The cyclic voltammetries show wide cathodic and anodic waves related to the reversible reduction-oxidation process of TiO 2 . The cathodic signal represents the n-doping of TiO 2 and the anodic one represents the dedoping process.   Tables S6 and S7 (for the porous solution-processed TiO 2 films, we also included values calculated using the BET surface area). The porous films have higher pseudocapacitance than the compact films; hence, the former can store more charge. All the devices show a high pseudocapacitance at low sweep rates.
[Li] + based gating media lead to a higher pseudocapacitance than do the pure [EMIM] + based gating media.
In summary, we investigated the electrical and electrochemical properties of n-type IGTs based on evaporated (predominantly anatase) and solution-processed (mixture of rutile and anatase) polycrystalline TiO 2 48 The transistor characteristics of the evaporated films are not significantly affected by the presence of Li + in the gating medium. For the solution-processed films, the presence of Li + in the gating medium leads to a significant increase in the electron mobility. We hypothesize that the presence of Li + facilitates charge transport at grain boundaries in the solution-processed films. The evaporated TiO 2 films show a higher mobility of 0.5 cm 2 V −1 s −1 compared to the solution-processed films. . This points to the complex arrangements of the ions in the electrical double layer. Work is in progress to shed light on such complexity (originated from the number of ionic interactions, from electrostatic to van der Waals) by atomic force microscopy force-distance profiling performed in highly viscous ionic liquids. 49 See the supplementary material for the schematic representation of the coverage of the TiO 2 film on the substrate, device configuration during the electrical and electrochemical characterizations, AFM images of the TiO 2 films, explanation of the hysteresis loop area calculations, values of hysteresis loop area, charge carrier density, mobility, ON/OFF ratio, threshold voltage of IGTs, and the pseudocapacitance values of the TiO 2 films.
The authors are grateful to Yves Drolet for technical assistance. Funding for this project was provided by NSERC (Discovery grants, F.C. and C.S.) and the Québec Ministry of Economy Science and Innovation (Project No. PSR-SIIRI-810). I.V. and A.S. are grateful to FRQNT for financial support through a doctoral scholarship, A.S. is grateful to the Trottier Energy Institute for a doctoral scholarship, and B.G. is grateful to FRQNT for a Master's scholarship. F.S. acknowledges financial support from Alma Mater Studiorum Università di Bologna (Researcher Mobility Program, Italian-Canadian cooperation agreement). This work was supported by CMC Microsystems through the MNT program. This work also benefited from the financial support of FRQNT through a grant awarded to RQMP.

DATA AVAILABILITY
The data that support the findings of this study are available from the corresponding author upon reasonable request.