High performance carbon nanotubes thin film transistors by selective ferric chloride doping

Single wall carbon nanotubes (SWNT) have been a significant research topic as active layers for thin film transistors (TFTs) due to their high charge carrier mobility beyond that of crystalline silicon. In this study, we report an effective approach to achieve a very high field-effect mobility and on/off ratio for solution processed semiconducting SWNT TFTs, by selective doping through contact with a thin ferric chloride (FeCl3) dopant layer. The semiconducting layer is formed by a double spin coating of the highly purified (>99%) high pressure carbon mono oxide (HiPCO) SWNT sorted by wrapping of poly (3-dodecylthiophene-2,5-diyl) (P3DDT). In order to achieve effective hole injection from the top Au source electrode without increasing the off-state drain current, less purified (98-99%) SWNTs produced by the plasma discharge process sorted by wrapping of poly (9,9-di-n-dodecylfluorene) (PFDD) are formed on the top of HiPCO film. Significantly improved TFT performance is achieved by the insertion of a few nanometers of a FeCl3 dopant layer at the semiconductor-contact interface. A significant high hole field-effect of 48.35 ± 3.11 cm2V−1s−1 (bare: 6.18 ± 0.87 cm2V−1s−1) with a reasonable on/off current ratio of 105, and low off current of ∼80 pA, are obtained by controlling the concentration of FeCl3 dopant (thickness = 1.5 nm) at the contact. Mobility is improved further at 2.5 nm thickness of the FeCl3 dopant layer resulting in a hole mobility of 177 ± 13.2 cm2 V−1s−1, an on/off ratio of 7.4 × 103, and off state current of 1.2 × 10−9 A.


Introduction
Carbon nanotubes (CNTs) offer considerable potential as semiconducting materials replacing silicon microelectronics, which have flourished according to the scaling law proposed by G. Moore [1][2][3][4][5][6][7]. The inherent superior charge carrier mobility of CNTs beyond single crystalline silicon enables hyperscaling of modern digital circuits, which can remarkably increase computing power over the downscaling of circuits. The rapid progress in CNT based high speed thin-film transistors (TFTs) has recently led to the demonstration of high speed analog and digital circuits, and even modern microprocessors [8][9][10][11]. However, this has mainly been performed using high cost, high vacuum based deposition techniques of the CNT film, such as chemical vapor deposition, while the solution processed CNT networked films suffer from a relatively low charge carrier mobility and less device uniformity [12]. Considering the above mentioned limitations, solution processed high-performance and large area uniform CNT network based TFTs should be developed [13]. To achieve high performance printed CNT TFTs, the morphology of the semiconducting CNT (semi-CNT) networked film, contact, and semiconductor-dielectric interface should be optimized [14,15]. Basically, random networked semi-CNT films are obtained by the coating of CNT dispersed solution through various solution based deposition methods. The ideal morphology of a semi-CNT networked film to facilitate charge transport is highly dense and aligned CNT networks to the channel direction because the charge transport is primarily through extended π -orbitals on the CNT growth direction [16]. In an aligned CNT networked film, the charge carriers can transport mainly from the source to the drain electrode through a single wall CNT (SWNT) chain effectively, hopping between a few adjunct semi-CNTs [17]. In particular, if the length of a CNT is longer than the channel length of TFTs and a single CNT can directly connect the source and drain, a ballistic charge transport can ideally occur, resulting in extremely high carrier mobility of over 1000 cm 2 /Vs [18,19]. Many attempts have been made to align CNTs with TFT channel, including Langmuir-Blodgett methods and applying the high electrical field assisted alignment method [20]. However, those methods worked effectively only in a small area with low reproducibility, and it is very difficult to achieve a highly aligned and morphologically uniform CNT film at the wafer scale. Thus, a more realistic alternative is currently implemented by CNT TFTs through randomly networked CNT films that exhibit high performance uniformity with a reasonable field-effect mobility.
In this study, we report a simple and effective method to obtain high performance semi-CNT TFTs by using selective ferric chloride (FeCl 3 ) doping at the contact. The novel feature of this work is obtaining a high drain current density at the channel of CNT TFTs by injecting large amounts of charge carrier from the top Au source electrode through the insertion of a (1) less purified semiconducting CNT layer and (2) thin FeCl 3 doping layer at the semiconductor-contact interface. The bottom gate/top contacts CNT TFTs with randomly networked semi-CNT films composed of highly purified ( > 99%) poly (3-dodecylthiophene-2,5-diyl) (P3DDT) wrapped high pressure carbon mono oxide (HiPCO) single well SWNT (P3DDT-HiPCO) at the bottom. Less purified (98-99%) poly (9,9-di-n-dodecylfluorene) (PFDD) wrapped plasma discharge (PD) SWNT (PFDD-PD) at the top exhibited a significantly improved hole fieldeffect of 48.35 ± 3.11 cm 2 V −1 s −1 , with a reasonably high on/off current ratio of 10 5 (off current of ∼ 80 pA) from the bare devices of 6.18 ± 0.87 cm 2 V −1 s −1 by inserting a 1.5 nm thick FeCl 3 layer at the contact. The hole mobility was dramatically improved, up to 177 ± 13.2 cm 2 V −1 s −1 , with an on/off ratio of 7.4 × 10 3 and an off state current of 1.2 × 10 −9 A for the devices with a 2.5 nm thick FeCl 3 dopant layer.

Results and discussion
To achieve high-performance printed CNT TFTs, it is essential to sort semi-CNTs with electronic grade purity from bare mixtures of metallic and semi-CNTs on a mass production scale. For this purpose, various separation methods including density gradient ultracentrifugation, DNA based wrapping separation, a series of column chromatography, and selective wrapping of conjugated polymers have been proposed [21][22][23][24]. Among these, sorting semi-CNTs through conjugated polymer wrapping is considered the most effective method owing to several advantages. A highly stable and long shelf-life semi-CNT formulation can be produced at a relatively small energy cost and short processing time because wrapping conjugated polymers can work not only as a sorting medium, but also as a surfactant of the semi-CNT to provide enough solubility in various solvents [25]. In addition, the subsequent purification process for removing excess conjugated polymers and metallic CNT is not complicated. For this study, we obtained highly purified ( > 99%) semi-CNT by wrapping P3DDT on HiPCO SWNTs, and a less purified (98-99%) PFDD wrapped plasma discharge (PD) SWNTs (PFDD-PD) [26]. The polymer-CNT composite semiconductors were obtained by a selective dispersion method using P3DDT-HiPCO semi-CNT and PFDD-PD semi-CNT, as reported previously [26]. The chemical structure of P3DDT, PFDD, SWNT, and the schematic of the semiconducting-CNT ink fabrication procedure are provided in Figure S1 and Figure 1(a). Figure 1(b) shows the UV-vis-NIR absorption spectra of P3DDT-HiPCO semi-CNT composite solutions. The absorption peaks appear in the 1000-1500 nm, and 600-800 nm bands corresponding S 11 and S 22 , respectively [27][28][29]. S n indicates the electronic transition energy between corresponding levels in the valance and conduction band [30]. Due to overlapping between the absorption of the P3DDT polymer (440-600 nm) and the absorption peak of metallic HiPCO CNTs in UVvis-NIR absorption spectra, the P3DDT-HiPCO semi-CNT composite film was also analyzed using the Raman spectra. Figure 1(c) displays that the radial breathing mode (RBM) of the Raman spectra is excited at 633 nm. The P3DDT-HiPCO semi-CNT composite film has a flat baseline in the metallic region (180-235 cm −1 ), which confirms a high-purity semiconducting material [31]. The purity of the P3DDT-HiPCO semi-CNT composite sample was estimated to be > 99% based on the absorption spectra, Raman spectra, and an off state current of more than 10 P3DDT-HiPCO TFTs [27,29,32]. Absorption peaks of the PFDD-PD semi-CNT composite solution appeared in the 1400-1900, 700-1100, and 450-550 nm bands for S 11 , S 22 , and S 33 , respectively, as shown in Figure 1(e). The purity of semi-CNT was evaluated from the absorption peak ratio ( = A CNT /(A CNT + A B ), where A CNT is the surrounding area of the M 11 and S 22 bands surrounded by the linear baseline in the 615-1190 nm region, and A B is the area covered by the linear baseline over the same region) as proposed in the previous reports [29,33,34]. The absorption peak ratio for the PFDD-PD semi-CNT composite solution was 0.396, and consequently the estimated purity of the PFDD-PD semi-CNT composite solution was 98-99% [29]. Figure 1(f) displays a typical RBM of the PFDD-PD semi-CNT composite spectra excited at 633 nm [29]. Figure 1(d) and (g) display the TEM image of the P3DDT-HiPCO semi-CNT, and PFDD-PD semi-CNT composite. The pristine SWNTs were formed as aggregated bundles [35,36], but the polymer-CNT composites were mostly isolated in toluene solution [37].
Bottom gate-top contact (BGTC) TFTs were fabricated using conjugated polymer wrapped semi-CNT composite ink as a channel on a photo-patternable polyimide (PI) film [38]. Heavily doped, thermally grown silicon wafer substrates 300 nm thick were used as the gate electrode. The structure of PI and the schematic structure of the BGTC are provided in Figure 2(a). The CNT film was formed by consecutive spin coating of P3DDT-HiPCO semi-CNT and PFDD-PD semi-CNT. Finally, the FeCl 3 dopant layer and Au electrode was deposited by thermal evaporation through a shadow mask for electrode patterning. Details of fabrication procedures are described in the experiment section. X-ray photoelectron spectroscopy (XPS) was used to confirm formation of the FeCl 3 layer on the semi-CNT composite film after deposition, as shown in Figure 2(b). XPS samples for the FeCl 3 doped semi-CNT composite were prepared using the same conditions for the transistor fabrication on Au. The main peaks of XPS spectra for the composite films appeared at 284 eV for C 1s peaks of both, semi-CNT and the conjugated polymers, and at 333, 355, and 641 eV for the Au substrate. The XPS spectra showed Fe 2p3 (715 eV) and Cl 2p (199 eV) peaks corresponding to the stable formation of the FeCl 3 dopant layer on the conjugated polymer wrapped semi-CNTs. Figure 2(c) to (f) show the AFM and SEM image of the undoped semi-CNT composite film (pristine), and the FeCl 3 deposited semi-CNT film. The spin coated binary semi-CNT films showed random networked morphology. The highly dense semi-CNT network film was achieved by two consecutive spin coatings of P3DDT-HiPCO semi-CNTs, and two consecutive spin coatings of PFDD-PD semi-CNTs, respectively. After each spin coating process, the film was annealed at 120°C for 90 sec to remove the residual solvent. The area density of CNT increased by the number of spin coatings, and saturated after four spin coatings (estimated CNT area density: 85%, Figure S2). Interestingly, FeCl 3 deposited films showed that FeCl 3 was well adsorbed on the semi-CNT surfaces. It is expected to effectively induce electron transport from the CNT surface to the FeCl 3 stably adhered to the semi-CNT surface because the efficiency of molecular doping strongly depends on the distance between the host and the dopant molecule [39,40]. We chose the FeCl 3 as a p-type dopant for semi-CNT based composites because it shows efficient p-type doping to various organic semiconductors [41][42][43]. In addition, FeCl 3 is well known as a catalyst for CNT growth and has been used as an anchoring source to functionalize CNT surfaces with various chemicals [44][45][46][47].  semiconducting layer, and the SiO 2 gate dielectric showed ambipolar characteristics with slightly better hole transport ( Figure S3). The ambipolar characteristics change to unipolar p-type by applying the polyimide gate dielectrics as an effect of the fluorine atom on polyimide [26,[48][49][50][51][52][53][54][55][56][57]. All transfer curves with the P3DDT-HiPCO and PFDD-PD double layer with polyimide gate dielectrics showed unipolar p-type characteristics. The printed semi-CNT TFTs with the P3DDT-HiPCO and PFDD-PD double layer exhibited a reasonably high hole mobility of 6.18 ± 0.87 cm 2 V −1 s −1 , an on/off ratio of 10 6 , and an off state current of 1 × 10 −12 A. The mobility of printed semi-CNT TFTs with double CNT layers is more than an order of magnitude higher than those of TFTs wrapped conjugated polymers (PFDD and P3DDT) as the active layer. This indicates that charge carriers mainly transport through the semi-CNT network instead of the wrapped polymers. The high fieldeffect mobility and current on/off ratio of semi-CNT TFTs is mainly due to the top PFDD-PD double layer for low contact resistance (R c ), and bottom P3DDT-HiPCO for low off current. The R c for hole injection decreased to 1.5 kΩ·cm by insertion of a PFDD-PD CNT networked film between the P3DDT-HiPCO and Au electrode. The R c for a P3DDT-HiPCO single CNT TFT was 20 kΩ·cm. The less purified (98-99%) semiconducting PFDD-PD CNT networked film contained a small amount of metallic CNTs leading to a high sheet conductivity, and better hole injection. In the bi-layered films, however, the channel is only formed in the highly purified P3DDT-HiPCO semi-CNT, and can maintain a low off current.
Considering the increased field-effect mobility by improving the charge injection properties through the PFDD-PD CNT networked film, we insert a FeCl 3 dopant layer that is a few nanometers thick between the PFDD-PD film and the Au source/drain electrode for a selective contact doping technique. Figure 3(b) shows typical transfer curves of the printed bilayered semi-CNT TFTs with the FeCl 3 dopant. The semi-CNT TFTs with a 1.5 nm thick FeCl 3 layer showed a significantly improved field-effect mobility of 48.35 ± 3.11 cm 2 V −1 s −1 (high uniformity ≥ 90.2%, see Figure S4 in supporting information), an on/off ratio of 1 × 10 5 , and off state current of 8 × 10 −11 A. The doping concentration can be controlled by the thickness of the FeCl 3 layer. The basic electrical parameters of semi-CNT TFTs are summarized in Table 1. As the thickness (doping concentration) of the FeCl 3 layer increased, the field-effect mobility and off-state drain current increased ( Figure S5). Figure 3(c) shows the field-effect mobility and on/off ratio of semi-CNT TFTs with five different FeCl 3 doping concentrations. The semi-CNT TFTs with a 1 nm thick FeCl 3 layer exhibited a slightly improved hole mobility of 10.8 ± 3.7 cm 2 V −1 s −1 , on/off ratio of 4 × 10 5 , and off state current of 1.7 × 10 −12 A. By increasing the thickness of the FeCl 3 layer to 2 nm, a surprisingly high hole mobility of 116 ± 15.16 cm 2 V −1 s −1 , on/off ratio of 1 × 10 5 , and reasonably low off state current of 1.01 × 10 −10 A. The 2.5 nm-FeCl 3 doped SWNT transistors exhibit a hole mobility of 177 ± 13.2 cm 2 V −1 s −1 , on/off ratio of 7.4 × 10 3 , and off state current of 1.2 × 10 −9 A. To the best of our knowledge, this is one of the highest field effect mobilities and on/off current ratios obtained with printed semi-CNT TFTs reported thus far. Such an impressive high field-effect mobility is attributed to the efficient injection of a large amount of charge carrier from the FeCl 3 /Au source electrode by a very low contact resistance (R c = 0.17 kΩ·cm). As the FeCl 3 thickness increased, the threshold voltage (V th ) gradually shifted to a positive value as the carrier density increased, and the TFTs operated in near depletion mode for FeCl 3 thicker than 2.5 nm (see Figure S5). Bias stress stability was checked for the bias stress test of FeCl 3 doped semi-CNT TFTs. The bias stress response of the FeCl 3 device was conducted at V d = −5 V, and V G ranged from 15 V to −20 V (see Figure 3(d)). After continuous bias stressing, the on and off currents stayed nearly the same with a slight shift in V on ( < 1 V).
In order to confirm the availability of FeCl 3 doping on the semi-CNT composite film, the Fermi level (E F ) of the undoped P3DDT-HiPCO, PFDD-PD double layer (pristine), and the FeCl 3 doped film was measured using ultraviolet photoelectron spectroscopy (UPS) (Figure 4(a) and Figure S6). Figure 4(b) shows the related energy level of bare and FeCl3 doped semi-CNT composite. UPS samples were prepared on a Au substrate, the same as the XPS samples. In the UPS spectra, the cut-off binding energy of the FeCl 3 doped semi-CNT film was reduced by 0.23 eV compared to the undoped ones. Thus, the work function of the FeCl 3 doped film increased from 4.22 to 4.45 eV of the bare film. The doped film of E f is 4.45 eV with respect to E vac [59]. No distinct difference was found in samples with the FeCl 3 film over 1 nm thick. This is direct evidence of FeCl 3 p-doping of the P3DDT-HiPCO and PFDD-PD double layer, because E F shifted towards the HOMO level due to the increase in the hole concentration. The improved charge injection through the tunneling across the Schottky barrier was achieved by selective FeCl 3 doping at the contact to reduce R c in printed semi-CNT TFTs [42,60].

Conclusion
In conclusion, we reported a solution for processed high mobility and on/off ratio semi-CNT TFTs by improving charge injection properties. Compared to bare semi-CNT TFTs with a P3DDT-HIPCO single semiconducting layer, the field-effect mobility of semi-CNTs was significantly improved over 100 cm 2 V −1 s −1 with high on/off current ratio of 10 5 by applying less purified PFDD-PD semi-CNT networked films and patterned FeCl 3 dopant layer at Au contact. We also demonstrated the record high and outstanding performance semi-CNT TFTs of 177 ± 13.2 cm 2 V −1 s −1 , on/off ratio of 7.4 × 10 3 and off state current of 1.2 × 10 −9 A device with 2.5 nm thick Table 1. Electrical parameters of semi-CNT TFTs.

Saturation Linear
Structure  FeCl 3 dopant layer. This selective doping method offers a promising prospect for high-performance printed random networked semi-CNT TFTs and integrated circuits, without complicated, large area, and uneven CNT alignment technology.

Conjugated polymer-carbon nanotube composite
P3DDT (20,000 ≤ Mn ≤ 50,000 gmol −1 , Leika metal) and PFDD (Mw ≈ 15,000-200,000, Lumtec). HiPCO SWNTs and PD SWNTs were purchased from Nanointegris Inc. Toluene solutions of P3DDT (2 mg/ml) and PFDD (2 mg/ml) were prepared and heated at 80°C for 2 h for complete dissolution. After cooling, 1 mg of the HiPCO and PD based SWNT powders were added to P3DDT and PFDD solutions, respectively. The solutions were combined in an ultrasonic bath (Branson 5510) for 1 h, and then centrifuged at 85,000 g for 1 h to separate semi-CNT. The supernatant fluid was further centrifuged first at 199,000 g for 1 h, and then ultra-centrifuged at 320,000 g for 12 h, to remove all the residual polymer (Vision scientific Inc VS-65 ultracentrifuge, V1308Ti fixed rotor). The produced pellets were washed 5 times to remove all the polymer and then collected. Finally, these enriched semi-CNTs were re-dispersed in toluene using an ultrasonic bath.

Device fabrication
The substrates were sequentially washed in an ultrasonic bath with deionized water, acetone, and isopropanol for 10 min each. Semi-CNTs were deposited on SiO 2 (60 mg ml −1 in cyclohexanone) by spin coating at 1000 rpm for 60 s. The spin-coated films were then baked at 90°C for 120 s and exposed to 365 nm UV light at a dose of more than 1.5 J cm −2 for cross-linking and patterning the semiconducting active layer. The films were subsequently annealed on a hot plate at 90°C for 10 min and 110°C for 30 min. Semi-CNTs were then deposited by spin coating twice at 700 rpm for 60 s, and the resulting films were annealed in a glove box on a hot plate at 120°C for 30 min in N 2 filled glove box. Exposed PI was developed with cyclohexanone, acetone, and isopropyl alcohol for 10 s each, and then annealed on a hot plate at 120°C for 30 min to remove the residual solvents and moisture. The development process removed unnecessary organic-CNT composite films to pattern the semiconducting layer. The FeCl 3 was evaporated through a metal mask source and drain electrodes were deposited on the hybrid films by thermal evaporation under 5.0 × 10 −6 torr with a 1.0 Å s −1 deposition rate.

Measurement
To quantify the purities of semi-CNT, UV-Vis spectra was measured using a Carry 5000 (Varian Inc.) spectrophotometer. Raman measurements were obtained using a LabRAM HV Evolution (HORIBA), while transmission electron microscopy (TEM) images were obtained using a JEM-2100f (JEOL). Atomic force microscope (AFM) topographic images were obtained using a XE-100 (PSIA) scanning probe microscope. A Keithley 4200 SCS instrument in a nitrogen filled glove box was used to perform electrical measurements. Capacitance-voltage (CV) measurements were performed using a Keithley 4200 connected to an Agilent 4284 LCR meter.
X-ray photoelectron spectroscopy (XPS) and ultraviolet photoelectron spectroscopy (UPS) were measured at the Korea Basic Science Institute (KBSI).

Disclosure statement
No potential conflict of interest was reported by the author(s).