polymer doping enables a two‐dimensional electron gas for...

COMMUNICATION

1805082 (1 of 8) © 2018 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

www.advmat.de

Polymer Doping Enables a Two-Dimensional Electron Gas for High-Performance Homojunction Oxide Thin-Film TransistorsYao Chen, Wei Huang, Vinod K. Sangwan, Binghao Wang, Li Zeng, Gang Wang, Yan Huang, Zhiyun Lu, Michael J. Bedzyk, Mark C. Hersam,* Tobin J. Marks,* and Antonio Facchetti*

DOI: 10.1002/adma.201805082

deposition (PLD), and radiofrequency sputtering.[2,11–13] Nevertheless, solution- processing of MO precursors holds sig-nificant promise for lower cost pro-duction and compatibility with flexible substrates.[2,9,14,15] In this regard, indium oxide (In2O3) is one of the most investi-gated solution-processed oxide semicon-ductors,[7,16–21] however, the carrier density of pristine In2O3 is difficult to control and In2O3 films are usually polycrystalline, limiting their performance uniformity over large areas as well as mechanical flexibility.[22] To address both of these issues, metals such as Zn and/or Ga are added to In2O3, enabling the growth of amorphous MO thin films with enhanced electronic properties and ultimately affording far more stable TFT characteristics.[22–26]

Recently, we reported a new strategy for high performance, near-zero threshold voltage (VTh), bias stress-stable, and amor-phous In2O3 films and TFTs by simply doping the In2O3 with an electrically insulating polymer such as poly(4-vinyl-phenol) (PVP) or polyethylenimine (PEI) to afford amorphous MO:polymer blend semiconducting nanocomposites.[2,17,22] The attraction of electron-rich PEI versus PVP is that the TFT

High-performance solution-processed metal oxide (MO) thin-film transistors (TFTs) are realized by fabricating a homojunction of indium oxide (In2O3) and polyethylenimine (PEI)-doped In2O3 (In2O3:x% PEI, x = 0.5–4.0 wt%) as the channel layer. A two-dimensional electron gas (2DEG) is thereby achieved by creating a band offset between the In2O3 and PEI-In2O3 via work func-tion tuning of the In2O3:x% PEI, from 4.00 to 3.62 eV as the PEI content is increased from 0.0 (pristine In2O3) to 4.0 wt%, respectively. The resulting devices achieve electron mobilities greater than 10 cm2 V−1 s−1 on a 300 nm SiO2 gate dielectric. Importantly, these metrics exceed those of the devices composed of the pristine In2O3 materials, which achieve a maximum mobility of ≈4 cm2 V−1 s−1. Furthermore, a mobility as high as 30 cm2 V−1 s−1 is achieved on a high-k ZrO2 dielectric in the homojunction devices. This is the first demonstration of 2DEG-based homojunction oxide TFTs via band offset achieved by simple polymer doping of the same MO material.

Thin-Film Transistors

The ORCID identification number(s) for the author(s) of this article can be found under https://doi.org/10.1002/adma.201805082.

Dr. Y. Chen, Dr. W. Huang, Dr. B. Wang, Dr. G. Wang, Prof. T. J. Marks, Prof. A. FacchettiDepartment of Chemistry and the Materials Research CenterNorthwestern University2145 Sheridan Road, Evanston, IL 60208, USAE-mail: [email protected]; [email protected]. Y. Chen, Dr. Y. Huang, Prof. Z. LuKey Laboratory of Green Chemistry and Technology (Ministry of Education)College of ChemistrySichuan UniversityChengdu 610064, P. R. ChinaDr. V. K. Sangwan, Prof. M. J. Bedzyk, Prof. M. C. HersamDepartment of Materials Science and EngineeringNorthwestern UniversityEvanston, IL 60208, USAE-mail: [email protected]

Dr. L. Zeng, Prof. M. J. BedzykApplied Physics Program and the Materials Research CenterNorthwestern University2220 Campus Drive, Evanston, IL 60208, USAProf. T. J. MarksDepartment of Materials Science and Engineering and the Argonne Northwestern Solar Energy Research Center (ANSER)Northwestern University2145 Sheridan Road, Evanston, IL 60208, USAProf. A. FacchettiFlexterra Inc.8025 Lamon Avenue, Skokie, IL 60077, USA

As a consequence of their outstanding charge transport properties and excellent optical transparency, metal oxide thin-film transistors (MOTFTs) have been heavily investi-gated for applications in state-of-the-art flat panel display tech-nologies and next-generation electronics.[1–10] To date, several growth techniques have been utilized to fabricate MO films for TFTs including atomic layer deposition (ALD), pulsed-laser

Adv. Mater. 2019, 31, 1805082

Dedicated to Prof. G. A. Pagani on the occasion of his 80th birthday

© 2018 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim1805082 (2 of 8)

www.advmat.dewww.advancedsciencenews.com

electron mobility of In2O3:PEI blends can be, for an optimal PEI content of ≈1.5 wt%, even greater than that of pristine In2O3-based TFTs.[17] Additionally, PEI interfacial layers have been shown to serve as “universal” agents to lower electrode work functions (WFs) of several electrode materials.[27] This result, combined with recent studies[28] demonstrating the generality of PEI doping to enhance MO TFT mobilities beyond In2O3 and to lower the WFs, set the stage for the present communication.[28]

A notably successful strategy to enhance TFT performance, including MO- and organic-based TFTs, utilizes semicon-ducting channel heterojunctions.[29,30] Here, the Fermi energy alignment between two semiconductors in contact leads to band bending at the semiconductor heterojunction, thereby forming a two-dimensional electron gas (2DEG) at this interface.[29,31–35] To date, 2DEG MO transistors have been fabricated using pristine/metal-doped In2O3 and ZnO heterojunctions,[35,36] with ZnO/ZnMgO and In2O3/ZnO being the most investi-gated.[34,35,37] For example, Cheng et al. fabricated MgxZn1−xO/ZnO heterostructures by sputtering[34] and these exhibit a higher field-effect mobility (μ ≈84.2 cm2 V−1 s−1) than ZnO-only TFTs (μ ≈1.5 cm2 V−1 s−1).[34] Recently, Anthopoulos and co-workers[29] reported growing low-dimensional MO heterostruc-tures consisting of solution-processed pristine In2O3 and ZnO layers in which μ increases from <1 cm2 V−1 s−1 for ZnO and ≈15 cm2 V−1 s−1 for In2O3, to ≈45 cm2 V−1 s−1 for In2O3/ZnO heterojunction TFTs. These pioneering studies show that MO heterojunctions having different WFs/Fermi energies can yield 2DEG-enhanced device properties. Inspired by these results and considering the properties of the aforementioned PEI:MO blends,[17,28] an intriguing question arises as to whether PEI doping might tune the In2O3 WF to achieve a 2DEG with pris-tine In2O3 for high-performance TFTs. Furthermore, one could then ask about the role of PEI in tuning performance param-eters, and whether such devices are stable under standard oper-ating conditions. Note that fabrication of PEI-doped MOs is based on aqueous formulations, which should be environmen-tally benign and have minimum waste.[16]

Here, we report the first demonstration of MO homojunc-tion formation where WF tuning is achieved by doping with an electrically insulating polymer. By carefully designing the TFT channel architecture and varying the PEI concentration in the In2O3 layers, electron mobilities surpassing 10 cm2 V−1 s−1, which are >2× that of the pristine In2O3-based TFTs (≈4 cm2 V−1 s−1), are achieved for TFTs processed at 250 °C. Fur-thermore, a much higher average mobility of ≈25 cm2 V−1 s−1 (30.91 cm2 V−1 s−1 max) is achieved by employing a high-k ZrO2 dielectric in the homojunction devices. In contrast to previous

examples of homojunctions, these devices consist of same underlying MO, thus offering a new and efficient route to high performance solution-processed transistors.

Before TFT fabrication, we first evaluated the PEI influ-ence on the In2O3 electronic structure. Thus, ultraviolet pho-toemission spectroscopy (UPS) was performed to measure the WF of the In2O3 films (≈8 nm thick) as a function of the PEI content (Figure 1). Briefly, the x weight % PEI-doped In2O3 blends, indicated here as In2O3:x% PEI; x = 0.0–4.0, were fabricated by spin-coating aqueous PEI-In2O3 precursor solu-tions onto ITO/glass substrates. Next, the spin-coated films were annealed at 250 °C for 20 min, and the whole process of spin-coating and annealing repeated one more time for a total film thickness of ≈8 nm. As shown in Figure 1a, the WFs of the In2O3:x%PEI films strongly depend on the PEI content and shift from 4.00 eV for pristine In2O3 to 3.89 eV for 0.5 wt% PEI to as low as 3.62 eV for 4 wt% PEI. This substantial WF shift, approaching 0.4 eV, is attributed to the electron-doping ability of PEI.[17,27]

Next, homojunction TFTs based on two layers of In2O3 with varying PEI contents were fabricated. Figure 2 shows the device structures, indicated here as TFT-n (n = 1–4), where the Si/SiO2 (300 nm) substrate acts as the gate electrode/gate insulator, a bilayer In2O3-based film acts as the semiconducting layer, and thermally evaporated Al lines (40 nm thick) act as the source-drain electrodes (see Supporting Information for details). The first MO layer (In2O3 precursor or PEI-doped In2O3 precursor) was spin-coated on the substrate at 3000 rpm for 30 s and then annealed for 20 min at 250 °C in ambient conditions (relative humidity ≈30%). Then, a second layer was deposited by the same procedure. The thicknesses of both pristine In2O3 and In2O3: x% PEI layers are ≈4 nm, so that the total semicon-ductor channel thickness is ≈8 nm. The channel width and length for all devices are 1000 and 100 μm, respectively. These bilayer In2O3-based devices allow assessing the impact of the PEI doping level and compositional position of the layers on the homojunction charge transport characteristics.

As shown in Figures 2b, and S1–S3 in the Supporting Information, all TFT-n structures exhibit a low off-current (Ioff) of 10−8–10−10 A and an on/off current ratio (Ion/Ioff) of >105–106, which are reasonable considering that the semiconductor is unpatterned. Interestingly, the on-current of TFTs-2–4 (Ion, measured at VDS = VGS = +80 V) follows a different trend with respect to PEI content versus that of TFT-1 where Ion first increases for PEI contents of 0.5–1 wt% and then decreases at greater PEI concentrations (1.5–4 wt%). Moreover, the variation of Ion also strongly depends on the device structure. Specifically,

Adv. Mater. 2019, 31, 1805082

Figure 1. a) UPS binding energy plots for In2O3 with different PEI contents. b) Schematic energy band diagram illustrating the shift of EWF.



as shown in Figures 2b, and S4 in the Supporting Informa-tion, Ion of TFT-2 increases from 9.66 × 10−4 A (TFT-1, no PEI) to 1.78 × 10−3

A (0.5wt% PEI), stabilizes at ≈1.6 × 10−3 A for 1–1.5 wt% PEI contents, and next falls to 5.21 × 10−4 A (2 wt% PEI) and 1.91 × 10−4 A (4 wt% PEI). This trend is con-sistent with our previous results for PEI-doped In2O3 TFTs.[17] Similar Ion trends are noted for the TFT-3 [9.66 × 10−4 A (0 wt% PEI)→1.41 × 10−3 A (0.5 wt% PEI)→1.43 × 10−3 A (1 wt% PEI)→1.53 × 10−3 A (1.5 wt% PEI)→1.14 × 10−3 A (2 wt% PEI)→6.32 × 10−4 A (4 wt% PEI)] and TFT-4 [9.66 × 10−4 A (0 wt% PEI)→1.64 × 10−3 A (0.5 wt% PEI)→3.30 × 10−3 A (1 wt% PEI)→2.06 × 10−3 (1.5 wt% PEI)→2.06 × 10−3 A (2 wt% PEI)→1.14 × 10−3 A (4 wt% PEI)], although the currents of the latter devices are considerably larger.

TFT metrics including field-effect electron mobility (μe) and threshold voltage (VTh) were extracted from the saturation region using conventional MOSFET equations; data are sum-marized in Table 1. Pristine bilayer In2O3 TFTs (TFT-1) exhibit good n-channel characteristics with negligible hysteresis, an average μe of 4.16 cm2 V−1 s−1, VTh = +4.8 V, and Ion/Ioff = 106. From these transport metrics, it is clear that PEI doping induces a dramatic change in the transport. Thus, compared to TFT-1, μe of TFT-2 first increases to 6.03 cm2 V−1 s−1 for a 0.5 wt% PEI content, next gradually declines with increasing PEI content [5.61 cm2 V−1 s−1 (1 wt% PEI), 5.63 cm2 V−1 s−1 (1.5 wt% PEI), 2.82 cm2 V−1 s−1 (2 wt% PEI), and 1.13 cm2 V−1 s−1 (4 wt% PEI)] as the PEI concentration increases. Simultaneously, VTh shifts to more positive values (+4.8 V in TFT-1) increasing from

Adv. Mater. 2019, 31, 1805082

Figure 2. a) The four different TFT structures used in this study. b) Representative transfer characteristics, c) mobility versus PEI content (wt%), and d) threshold voltage versus PEI content (wt%) for the indicated devices.



+5.2 to +10.2 V as the PEI content increases from 0.5 to 4 wt%, in accordance with previous data for In2O3:PEI-only TFTs.[17]

For TFT-3 and TFT-4, recall that one layer is pristine In2O3 and the other is In2O3:x wt% PEI (x = 0.5, 1, 1.5, 2, 4). As shown in Figure 2 and Table 1, for the TFT-3 devices μe first increases from 4.16 cm2 V−1 s−1 (0 wt% PEI) to 6.29 cm2 V−1 s−1 (0.5 wt % PEI) then falls to 5.50 cm2 V−1 s−1 (1 wt % PEI), 5.22 cm2 V−1 s−1 (1.5 wt% PEI), 4.42 cm2 V−1 s−1 (2 wt% PEI), and 3.64 cm2 V−1 s−1 (4 wt% PEI). For these devices VTh remains in a narrow range (+7 to +8 V) except increasing mark-edly to ≈+16 V for the largest PEI content of 4 wt%. Clearly, the mobility of TFT-3 is either similar or higher than those of both TFT-1 and TFT-2. For the TFT-4 device series, having the x wt% PEI-doped In2O3 as a top layer and the pristine In2O3 as bottom layer, μe reaches 6.51 cm2 V−1 s−1 for 0.5 wt% PEI and further increases to a remarkable value of 10.05 cm2 V−1 s−1 for 1 wt% PEI. For higher PEI concentrations, μe gradually declines to 9.21 cm2 V−1 s−1, 6.87 cm2 V−1 s−1, and 6.59 cm2 V−1 s−1 for 1.5 wt%, 2 wt%, and 4 wt% PEI content, respectively. For these devices VTh minimizes at ≈+3 V for the device having the greatest mobility (1 wt% PEI) and, consistent with the other TFT-n devices, increases greatly (≈+15 V) for the largest PEI con-tent. Interestingly, the TFT-4 electron mobilities are the highest of the TFT-n structures (Table 1). We attribute this significant mobility enhancement to the homojunction structure forming a 2DEG at the semiconductor–semiconductor interface due to the Fermi level alignment.[35] This is consistent with the nega-tive VTh shift, high μe, and higher Ioffs for the optimized TFT-4 devices, characteristics similar to other oxide–oxide heterojunc-tion/homojunction TFTs.[29,32] Note, in conventional TFTs such

as TFT-1 and TFT-2 (Figure 2), carrier density modulation by VGS occurs via typical field-effect and the channel is formed in the semi-conductor layer near the dielectric surface. For TFT-4, also for VGS = 0 V, charge car-riers (electrons) accumulate in the high WF region of the homojunction near the inter-face between the In2O3 and the In2O3:PEI layers, thus in the In2O3 layer. Since both MO layers are ≈4 nm thick, this carrier den-sity must be located at ≈3-4 nm away from the dielectric surface. Thus, application of a VGS > 0 V will accumulate additional carriers which, depending on the VGS strength, can spatially overlap with the 2DEG layer since it is known that the gate field can tune the carrier concentrations in the TFT channel region up to 10 nm.[38] Application of a VGS < 0 V can shut down the channel by an identical mechanism. A similar phenomenon has been reported for other 2DEG TFTs.[32,39]

We also investigated TFF-n bias stress sta-bility by subjecting devices to a constant VGS bias of +20 V for 1200 s in ambient condi-tions, without encapsulation. As shown in Figure S5 in the Supporting Information, compared to TFT-1 and TFT-2-based devices where VTh shifts by 12.1 and 12.4 V, respec-tively, TFT-3, and particularly TFT-4 exhibit

more stable VTh (VTh shift = 11.4 and 7.1 V, respectively) and more stable mobility characteristics.

Next, variable-temperature I–V measurements were carried out to further probe the electron transport mechanism in these homojunction devices (Figures 3; and S6, Supporting Informa-tion). Specifically, we compare control device TFT-1 to homojunc-tion devices TFT-3 and TFT-4. For these measurements, TFTs were photolithographically patterned (see Experimental Section). Figure 3a shows representative transfer characteristics of TFT-4 (1 wt% and 4 wt% PEI) and TFT-1 measured at VDS = 80 V over 290–50 K. Overall, TFT-1 shows a larger fall in conductivity at all VGS values with falling temperature versus TFT-4. Further-more, TFT-4 with 4 wt% PEI shows a smaller decrease in the conductivity compared to TFT-4 with 1 wt% PEI. Figure 3b shows Arrhenius plots of saturation mobility over 50–290 K for the TFT-1, TFT-3 (1 wt% and 4 wt% PEI), and TFT-4 (1 wt% and 4 wt% PEI) devices. Overall, the mobility of TFT-1 falls more rap-idly than that of TFT-4 with falling temperature, and the TFT-1 device clearly exhibits thermally-activated behavior (Figure S6b, Supporting Information); a typical characteristic of trap- limited conduction (TLC) in MOTFTs.[29,40] Transistors based on TFT-3 with 4 wt% PEI also show similar thermally activated transport behavior at all temperatures (Figure S6b, Supporting Information) likely due to the rougher interface (vide infra). However, TFT-4 (1 or 4 wt% PEI) and TFT-3 (1 wt% PEI) show far weaker temperature dependence, with μe saturating at ≈4 cm4 Vs−1 at low temperatures, consistent with percola-tion conduction (PC; Figure S6b, Supporting Information).[41,42] This trend is similar to that of previously reported In2O3/ZnO heterojunction devices.[40] Thus, charge conduction in TFT-3

Adv. Mater. 2019, 31, 1805082

Table 1. Performance metrics of the indicated TFT devices on Si/SiO2 (300 nm) substrates with Al source-drain electrodes.

Device structurea) Semiconductor PEI [wt%] Mobility [cm2 V−1 s−1] VTh [V] Von [V] Ion/Ioff

TFT-1 In2O3/In2O3 0 4.16 ± 0.23 4.8 ± 2.5 −9.7 ± 3.6 106

TFT-2 In2O3:x% PEI/

In2O3:x% PEI

0.5 6.03 ± 0.27 5.2 ± 1.1 −8.1 ± 2.0 106

1 5.61 ± 0.56 6.6 ± 2.4 −8.9 ± 1.6 105

1.5 5.63 ± 0.56 7.1 ± 1.5 −15.5 ± 3.3 105

2 2.82 ± 0.24 9.8 ± 1.8 −5.7 ± 1.2 105

4 1.13 ± 0.13 10.2 ± 2.2 −5.9 ± 3.6 105

TFT-3 In2O3:x% PEI/

In2O3

0.5 6.29 ± 0.18 7.8 ± 1.9 −7.8 ± 1.2 106

1 5.50 ± 0.14 7.5 ± 2.1 −6.0 ± 2.1 106

1.5 5.22 ± 0.30 6.7 ± 1.0 −4.2 ± 0.2 106

2 4.42 ± 0.15 7.4 ± 2.6 −4.1 ± 1.9 106

4 3.64 ± 0.18 15.8 ± 2.7 −2.8 ± 1.5 106

TFT-4 In2O3/

In2O3:x% PEI

0.5 6.51 ± 0.48 6.4 ± 1.6 −2.2 ± 1.8 106

1 10.05 ± 0.72 3.2 ± 1.9 −7.3 ± 4.2 105

1.5 9.21 ± 0.52 5.2 ± 2.6 −7.9 ± 3.8 105

2 6.87 ± 0.42 13.6 ± 1.8 −2.2 ± 1.1 106

4 6.59 ± 0.31 14.7 ± 1.5 −2.1 ± 1.5 106

a)Average from ≥10 devices.



with 1 wt% optimized PEI doping or TFT-4 occurs via formation of a 2DEG at the interface between the intrinsic and doped In2O3 layers. The physical distance between this plane of a high carrier density and the gate dielectric surface is expected to suppress both long-range Coulomb scattering by trapped charges and short-range scattering from dielectric surface topological imper-fections and chemical defects. The term 2DEG is used here since charge conduction in these homojunctions is still ideal as seen in phonon-limited ZnMgO/ZnO heterojunctions grown by molecular beam epitaxy (MBE)[43,44] and in the temperature-inde-pendent regime as seen in sputtered ZnMgO/ZnO heterojunc-tions.[45] Further optimization of film solution-processing and dopant chemistry would likely improve the performance further.

Two additional experiments were performed to test the model of a homojunction 2DEG as the origin of enhanced TFT performance. First, the WFs of the semiconductor bilayers, TFT-3 [In2O3:x% PEI(bottom)/In2O3(top)] and TFT-4 [In2O3(bottom)/In2O3:x% PEI(top)], were measured to exclude the possibility of WF equilibration via PEI diffusion. As shown in Figure S7 in the Supporting Information, the WFs of the TFT-3 and TFT-4 channels with x% PEI = 4 wt% are 3.96 and 3.61 eV, respectively, thus, close to those of TFT-1 (4.00 eV) and TFT-2 (3.62 eV), respectively, indicating that the WFs remain unchanged relatively distant from the interface. Additionally, since variation of contact resistance can affect TFT performance by varying charge injection parameters,[46] we measured the con-tact resistance (Rc) for TFT-1 (pristine In2O3) and all the other TFT-n with the largest PEI content of 4 wt% and found that Rc remains in a narrow range in all cases (4–6 kΩ, Figure S8, Supporting Information). Thus, the mobility variations cannot be ascribed to variations in charge injection characteristics by PEI changing the source/drain electrode WFs.

Next, MO film composition, morphology, and microstructure were analyzed. X-ray photoelectron spectroscopy (XPS) meas-urements assessed how PEI influences the chemical properties and stoichiometry of these bilayer films. The O1s spectra of all the films show a similar peak evolution, and both the PEI-doped In2O3 and pristine In2O3 films show a similar M-O-M content of 74-75% (Figures S9–S12, Supporting Information), in agree-ment with previous In2O3:PEI results.[17] Grazing incidence X-ray diffraction (GIXRD) measurements were also carried out to examine the impact of PEI doping on the bilayer film

microstructure. From Figure 4a, the pristine In2O3 films in TFT-1 are polycrystalline exhibiting a sharp, strong (222) reflec-tion at 2θ = 31.1° with a degree of crystallinity (χc) of 23.2%. Note that the broad high intensity peak located at 2θ = 21° is attributed to the bottom ≈300 nm SiO2 on Si substrate. Upon PEI doping of both layers (TFT-2), all scans reveal the presence of a broad reflection at 2θ = 30°–35°, assigned to the amorphous phase. These results are in accordance with previous findings that PEI doping strongly frustrates In2O3 crystallization.[17] The degree of crystallinity χc of all films were also extracted following the meth-odology in previous work (Figure 4b).[17] Interestingly, for the semiconductor bilayers of the TFT-3 series, where the top layer is pristine In2O3, the degree of crystallinity χc decreases (χc = 23.2% → 12.7% → 5.3% → 2.5%) as the bottom layer PEI content increases (x wt% PEI = 0 wt% → 0.5 wt% → 1 wt% → 1.5 wt%). For PEI contents greater than 2 wt%, the films are amorphous. Interestingly, compared to TFT-3 (Figures 4a,b; and S13, Sup-porting Information), TFT-4 devices having the same PEI con-tent are all amorphous except for that having the minimum PEI content (0.5 wt%), which exhibits χc of only 3.7%. Thus, placing the In2O3:PEI layer uppermost on the bilayer is more effective at promoting amorphous bilayer films. Irregardless, TFT-2, TFT-3, and TFT-4 exhibit lower crystallinity than TFT-1, thus favoring mobility due to the absence of carrier-scattering grain bounda-ries.[8] Thus, the GIXRD data are consistent with the TFT metrics.

The film surface topology of the MO bilayers was also assessed by atomic force microscopy (AFM), and representative images are shown in Figures 4, and S14–S19 in the Supporting Information. Pristine bilayer In2O3 films in TFT-1 are smooth with an RMS roughness (σrms) of 0.34 nm. The TFT-2 semicon-ductor films all exhibit σrms values of 0.31–0.41 nm, similar to pristine In2O3. Interestingly, the bilayer films corresponding to TFT-3 are considerably rougher (σrms = 0.48–0.63 nm) and TFT-4 is smoother (σrms = 0.21–0.27 nm) than both TFT-1 and TFT-2. Note, the maximum height features are over 4 nm high in TFT-3 (Figure S17, Supporting Information).[47,48] Since surface rough-ness should track roughness at the homojunction in these ultra-thin films, the poor morphology of TFT-3 devices likely accounts for the lower performance versus the TFT-4 series (see Table 1).

X-ray reflectivity (XRR) measurements were next conducted to extract the electron density depth profiles of the semiconducting channel layers, thicknesses of the different layers, and the

Adv. Mater. 2019, 31, 1805082

Figure 3. a) Transfer characteristics of devices TFT-1 and TFT-4 (1.0 wt% and 4.0 wt% PEI) at VDS = 80 V for temperatures ranging from 290 to 50 K. b) Plot of the electron mobility (μ) of TFT-1, TFT-3, and TFT-4(1.0 wt% and 4.0 wt% PEI) transistors as a function of inverse temperature measured between 290 and 50 K.



roughness of surfaces and buried interfaces (Figure S20, Table S1, Supporting Information). From Figure S20b in the Sup-porting Information, the TFT-2 upper layer has lower electron density than the bottom layer. In TFT-3, the interface between the two layers is not well-defined, but instead shows increased electron density in the vertical direction, indicating that the upper layer has a higher electron density than the bottom layer.

In contrast, for the TFT-4 channel, greater electron density is found at the homojunction interface that increases with increasing PEI concentration, which is completely different from the other films. Note, for TFT-4, the greater electron den-sity is found for a PEI content of 4% while maximum trans-port is measured for the 1% PEI blend, however, we previously reported for these MO blends that larger electron density does not always translate into higher mobilities since excessive PEI can also disrupt film microstructure and increase trap density.[17] The film thicknesses of the MO films slightly increase after PEI doping from 8.2 nm (TFT-1) to 8.7–9.8 nm (TFT-2, TFT-3, and TFT-4). Surface roughness σrms values were also extracted from the XRR results (Table S1, Supporting Information). The σrmss of TFT-1, TFT-2, and TFT-3 are ≈0.42, ≈0.50–0.55, and ≈0.52–0.71 nm, respectively. Interestingly, the RMS roughness of the TFT-4 series (σrms = 0.43–0.47 nm) is lower and close to that of TFT-1. Overall, these XRR data are consistent with the RMS roughness from the AFM analysis.

Finally, cross-sectional TEM was performed to examine the homojunctions in TFT-3 and TFT-4. Note, the PEI doping con-

centration here is 4%. As shown in Figure 5a, the two In2O3 layers are difficult to distinguish in the TFT-3 channel film, although the channel in proximity to the SiO2 layer exhibits a lighter contrast, thus reduced density, than the upper layer. For the TFT-4 channel, two distinct In2O3/In2O3:4% PEI layers and the presence of a clear interface are readily identified. Further-more, the layer in contact with the SiO2 is darker, thus denser that the upper layer. These images are consistent with the XRR reflectivity data electron density profile and corroborate the superior TFT-4 device performance.

Furthermore, to demonstrate high-performance TFTs using these polymer-doping-induced homojunction devices, a high-quality ZrO2 film was also employed as the gate dielec-tric (Figure S21, Supporting Information).[49] It is known that high-k dielectrics can greatly enhance the mobility of inorganic TFTs.[50–53] The device fabrication process details can be found in the Experimental Section and representative transfer charac-teristics are shown in Figure 5b. All of the present In2O3 devices with ZrO2 as the dielectric layer (Table S2, Supporting Infor-mation) exhibit much higher μe values than those using SiO2. Most importantly, the homojunction devices based on the TFT-4 architecture with 1% of PEI doping show a high average μe of 24.78 cm2 V−1 s−1 (μe,max = 30.91 cm2 V−1 s−1), which is ≈2× higher than the control TFT-1 device (μe/μe,max = 14.44/18.13 cm2 V−1 s−1). These results indicate that combining MO pol-ymer doping with a high-k dielectric is an effective strategy to realize high-performance homojunction TFTs.

Adv. Mater. 2019, 31, 1805082

Figure 5. a) High-resolution cross-sectional TEM images of indicated device structures. b) Representative transfer characteristics of TFT-1 and TFT-4 (1 wt% PEI) using ZrO2 as the gate dielectric layer (VDS = 2 V).

Figure 4. a) GIXRD plots for the semiconductor layer of TFT-n. Note, TFT-1 corresponds to the black line reported in TFT-2. b) Contour color-filled comparative maps of crystallinity. c) AFM height images of the indicated semiconductor layers. d) Corresponding film RMS roughness for the TFT-n measured by AFM.



In conclusion, the first MO homojunction TFT devices composed of In2O3 bilayers are demonstrated using a polymer (PEI) to selectively dope one of the two MO layers. The greater performance for the TFT architecture having a top PEI-doped In2O3 layer and a bottom pristine In2O3 layer (TFT-4) reveals that 1) PEI-doping can effectively tune the WF (Fermi level) of In2O3, thus forming a 2DEG via Fermi level alignment between the two In2O3 layers; 2) PEI-doped In2O3 provides a superior morphology as the uppermost layer of the bilayer – a critical factor for the TFT architecture; 3) PEI-doped In2O3 has a more favorable electron density as the uppermost layer of the bilayer, and thus, In2O3 WF modulation is achieved by polymer doping; 4) solution-processed homojunction devices of the type In2O3/In2O3:x% PEI are fabricated by an aqueous route. A high average electron mobility of up to 10.05 cm2 V−1 s−1 is achieved on a 300 nm SiO2 dielectric. Furthermore, by employing a high-k (ZrO2) dielectric layer, a much higher mobility >30 cm2 V−1 s−1 has been demonstrated using the homojunction devices. This work paves a new road for fabricating high-performance TFTs, including MO–organic hybrid devices.

Experimental SectionPrecursor Preparation: In(NO3)3 (99.999%) and PEI

(average MW ≈25 000 by LS; average Mn ≈10 000 by GPC, branched) were purchased from Sigma-Aldrich. For In2O3 TFT precursors, appropriate amounts of the metal salt In(NO3)3 · xH2O were dissolved in deionized-water to achieve a 0.1 m In concentration. Then, 20 mg mL−1 of the PEI stock solution was prepared by dissolving the PEI in the di-water. These solutions were stirred for at least 6 h under ambient conditions. The PEI weight concentration (In2O3:x% PEI; x = 0.5, 1, 1.5, 2, and 4 wt %) was controlled by adding the polymer solution to the MO precursor solution. After addition, the precursor solutions were stirred for 8 h before use. For ZrO2 dielectric layer, appropriate amounts of ZrCl4 was dissolved in 2 mL 2-methoxyethanol with 120 μL HNO3 at a concentration of 0.2 m, and the solution was stirred more than 6 h before use.

Thin-Film Fabrication and Electrical Characterization: All solutions were filtered through 0.2 μm PFET syringe filters prior to device fabrication. Highly doped (n++) silicon and 300 nm thick thermally grown SiO2 (WRS Materials) were used as gate electrode and gate dielectric layer, respectively. The substrates were first cleaned by ultrasonication in acetone and isopropyl alcohol and then subjected to an O2 plasma for 20 min. For ZrO2 dielectric layers, the ZrO2 precursor solution was filtered through a 0.2 μm poly(tetrafluoroethylene) (PTFE) filter and spin-coated on a n++ Si wafer at 3000 rpm for 30 s. The substrate was then annealed at 300 °C for 10 min. This spin-coating and annealing process was repeated four times. Then, the dielectric film was annealed at 500 °C for 1 h. The ZrO2 film thickness is 64 nm and the areal capacitance on Si substrates having a ≈2 nm thick native oxide layer is 300 nF cm−2 (k = 21.4). Before the spin-coating of In2O3 on ZrO2, the dielectric surface was treated with an oxygen plasma for 5 min. For all devices, the first MO layer (In2O3 precursors or PEI-doped In2O3) was spin-coated on substrates at 3000 rpm for 30 s and subsequently annealed for 20 min at 250 °C (relative humidity ≈30%) from either the pristine In2O3 precursor or PEI-doped In2O3 precursor solution. The second layer was then formed by repeating the process to obtain the desired device structure and film thickness. To complete the TFT, Al source and drain (S/D) electrodes (thickness = 40 nm) were deposited by thermal evaporation through metal shadow masks. The channel width and length for all devices were 1000 and 100 μm, respectively. TFT characterization was performed under ambient conditions using an Agilent B1500A semiconductor parameter analyzer.

For variable temperature transport studies, TFTs were fabricated by photolithography and the semiconductor layers were etched into

well-defined channels. Semiconductor layer patterning was achieved by spin-coating one layer of S1813 photoresist (5000 rpm for 30 s, 1.2 μm) on the surface, which was thermally annealed at 115 °C for 1 min, and then exposed to UV light (Inpro Tech., F300S) for 2 s through a photo mask. Next, the S1813 film was removed by soaking in MF319 for 30 s, and the films were cleaned with water and dried with an N2 flow. The exposed oxide films were next removed in aqueous oxalic acid (10% w/v) for 30 s, and rinsed with water. The remaining S1813 film was then removed with acetone. Finally, the samples were further thermally annealed at 250 °C for 5 min to remove any residual solvent. Al S/D electrodes (thickness = 40 nm) were deposited by thermal evaporation through metal shadow masks. The channel width and length for all devices were 1500 and 150 μm, respectively. Temperature-dependent experiments were carried out under high vacuum (≈10−6 Torr) at temperatures ranging from 50 to 290 K using a cryogenic probe station (LakeShore CRX 4K) connected to Keithley 2400 source-meters that were controlled via home-made LabVIEW programs.

The carrier mobility (μ) was evaluated in the saturation region with the conventional metal-oxide-semiconductor field-effect transistor model as follows:[8]

2DSi

GS Th2I

C WL

V Vµ ( )= −

(1)

where IDS is the drain–source current, Ci is the dielectric capacitance per unit area (the Ci of 300 nm SiO2 is 11 nF cm−2), W and L are the channel width and length, respectively, VGS is the gate-source voltage, and VTh is the threshold voltage.

Characterization: The AFM images were recorded on a Bruker Dimensional Icon system in the tapping mode. GIXRD and XRR measurements were acquired with a Rigaku SmartLab diffraction workstation using Cu Kα (1.54 Å) radiation. UPS and XPS were performed on a Thermo Scientific ESCALAB 250 Xi spectrometer. Note, the film surfaces were etched about 2 nm before collecting the spectra. Cross-sectional TEM measurements were performed using a JEOL ARM 300F instrument, with samples prepared on Si/SiOx using fast ion bombardment (FIB) techniques (FEI Helios NanoLab 600). A thin Au layer was deposited on the sample surface to protect from damage during the FIB processing.

Supporting InformationSupporting Information is available from the Wiley Online Library or from the author.

AcknowledgementsY.C. and W.H. contributed equally to this work. The authors thank AFOSR (FA9550-18-1-0320), the Northwestern University MRSEC (NSF DMR-1720139), and Flexterra Corp. for support of this research. This work made use of the J. B. Cohen X-ray Diffraction Facility, EPIC facility, Keck-II facility, and SPID facility of the NUANCE Center at Northwestern University, which received support from the MRSEC program (NSF DMR-1720139); the International Institute for Nanotechnology (IIN); the Keck Foundation; and the State of Illinois. A.F. thanks the Shenzhen Peacock Plan project (KQTD20140630110339343) for support. Y.C. thanks the joint-Ph.D. program supported by China Scholarship Council for a fellowship.

Conflict of InterestThe authors declare no conflict of interest.

Adv. Mater. 2019, 31, 1805082



Adv. Mater. 2019, 31, 1805082

Keywords2D electron gases, homojunctions, oxide electronics, PEI-doped In2O3

Received: August 4, 2018Revised: October 10, 2018

Published online: November 29, 2018

[1] A. Liu, H. Zhu, H. Sun, Y. Xu, Y. Y. Noh, Adv. Mater. 2018, 30, 1706364.

[2] X. Yu, T. J. Marks, A. Facchetti, Nat. Mater. 2016, 15, 383.[3] W. J. Lee, W. T. Park, S. Park, S. Sung, Y. Y. Noh, M. H. Yoon,

Adv. Mater. 2015, 27, 5043.[4] K. K. Banger, Y. Yamashita, K. Mori, R. L. Peterson, T. Leedham,

J. Rickard, H. Sirringhaus, Nat. Mater. 2011, 10, 45.[5] W. T. Park, I. Son, H. W. Park, K. B. Chung, Y. Xu, T. Lee, Y. Y. Noh,

ACS Appl. Mater. Interfaces 2015, 7, 13289.[6] X. Zhao, D. Madan, Y. Cheng, J. Zhou, H. Li, S. M. Thon,

A. E. Bragg, M. E. DeCoster, P. E. Hopkins, H. E. Katz, Adv. Mater. 2017, 29, 1606928.

[7] J. Socratous, K. K. Banger, Y. Vaynzof, A. Sadhanala, A. D. Brown, A. Sepe, U. Steiner, H. Sirringhaus, Adv. Funct. Mater. 2015, 25, 1873.

[8] E. Fortunato, P. Barquinha, R. Martins, Adv. Mater. 2012, 24, 2945.[9] Y. Chen, B. Wang, W. Huang, X. Zhang, G. Wang, M. J. Leonardi,

Y. Huang, Z. Lu, T. J. Marks, A. Facchetti, Chem. Mater. 2018, 30, 3323.

[10] R. Branquinho, D. Salgueiro, L. Santos, P. Barquinha, L. Pereira, R. Martins, E. Fortunato, ACS Appl. Mater. Interfaces 2014, 6, 19592.

[11] K. Nomura, H. Ohta, A. Takagi, T. Kamiya, M. Hirano, H. Hosono, Nature 2004, 432, 488.

[12] P. F. Carcia, R. S. McLean, M. H. Reilly, G. Nunes, Appl. Phys. Lett. 2003, 82, 1117.

[13] Y. Y. Lin, C. C. Hsu, M. H. Tseng, J. J. Shyue, F. Y. Tsai, ACS Appl. Mater. Interfaces 2015, 7, 22610.

[14] S. Jeong, J. Moon, J. Mater. Chem. 2012, 22, 1243.[15] K. K. Banger, R. L. Peterson, K. Mori, Y. Yamashita, T. Leedham,

H. Sirringhaus, Chem. Mater. 2014, 26, 1195.[16] W. J. Scheideler, R. Kumar, A. R. Zeumault, V. Subramanian,

Adv. Funct. Mater. 2017, 27, 1606062.[17] W. Huang, L. Zeng, X. G. Yu, P. J. Guo, B. H. Wang, Q. Ma,

R. P. H. Chang, J. S. Yu, M. J. Bedzyk, T. J. Marks, A. Facchetti, Adv. Funct. Mater. 2016, 26, 6179.

[18] H. Faber, Y. H. Lin, S. R. Thomas, K. Zhao, N. Pliatsikas, M. A. McLachlan, A. Amassian, P. A. Patsalas, T. D. Anthopoulos, ACS Appl. Mater. Interfaces 2015, 7, 782.

[19] J. Leppaniemi, O. H. Huttunen, H. Majumdar, A. Alastalo, Adv. Mater. 2015, 27, 7168.

[20] H. S. Kim, P. D. Byrne, A. Facchetti, T. J. Marks, J. Am. Chem. Soc. 2008, 130, 12580.

[21] Y. S. Rim, H. J. Chen, T. B. Song, S. H. Bae, Y. Yang, Chem. Mater. 2015, 27, 5808.

[22] X. Yu, L. Zeng, N. Zhou, P. Guo, F. Shi, D. B. Buchholz, Q. Ma, J. Yu, V. P. Dravid, R. P. Chang, M. Bedzyk, T. J. Marks, A. Facchetti, Adv. Mater. 2015, 27, 2390.

[23] S. Jeong, Y. G. Ha, J. Moon, A. Facchetti, T. J. Marks, Adv. Mater. 2010, 22, 1346.

[24] J. W. Hennek, M. G. Kim, M. G. Kanatzidis, A. Facchetti, T. J. Marks, J. Am. Chem. Soc. 2012, 134, 9593.

[25] J. W. Hennek, J. Smith, A. Yan, M. G. Kim, W. Zhao, V. P. Dravid, A. Facchetti, T. J. Marks, J. Am. Chem. Soc. 2013, 135, 10729.

[26] B. K. Sharma, B. Jang, J. E. Lee, S. H. Bae, T. W. Kim, H. J. Lee, J. H. Kim, J. H. Ahn, Adv. Funct. Mater. 2013, 23, 2024.

[27] Y. Zhou, C. Fuentes-Hernandez, J. Shim, J. Meyer, A. J. Giordano, H. Li, P. Winget, T. Papadopoulos, H. Cheun, J. Kim, M. Fenoll, A. Dindar, W. Haske, E. Najafabadi, T. M. Khan, H. Sojoudi, S. Barlow, S. Graham, J. L. Bredas, S. R. Marder, A. Kahn, B. Kippelen, Science 2012, 336, 327.

[28] W. Huang, P. Guo, L. Zeng, R. Li, B. Wang, G. Wang, X. Zhang, R. P. H. Chang, J. Yu, M. J. Bedzyk, T. J. Marks, A. Facchetti, J. Am. Chem. Soc. 2018, 140, 5457.

[29] H. Faber, S. Das, Y. H. Lin, N. Pliatsikas, K. Zhao, T. Kehagias, G. Dimitrakopulos, A. Amassian, P. A. Patsalas, T. D. Anthopoulos, Sci. Adv. 2017, 3, e1602640.

[30] P. Zhang, H. Wang, D. Yan, Adv. Mater. 2017, 29, 1702427.[31] K. Tetzner, I. Isakov, A. Regoutz, D. J. Payne, T. D. Anthopoulos,

J. Mater. Chem. C 2017, 5, 59.[32] D. Khim, Y. H. Lin, S. Nam, H. Faber, K. Tetzner, R. Li, Q. Zhang,

J. Li, X. Zhang, T. D. Anthopoulos, Adv. Mater. 2017, 29, 1605837.[33] S. Nam, J. H. Yang, S. H. Cho, J. H. Choi, O. S. Kwon, E. S. Park,

S. J. Lee, K. I. Cho, J. Jang, C. S. Hwang, J. Mater. Chem. C 2016, 4, 11298.

[34] I. C. Cheng, B. S. Wang, H. H. Hou, J. Z. Chen, ECS Trans. 2013, 50, 83.

[35] J. G. Labram, Y. H. Lin, T. D. Anthopoulos, Small 2015, 11, 5472.[36] X. Yu, N. Zhou, J. Smith, H. Lin, K. Stallings, J. Yu, T. J. Marks,

A. Facchetti, ACS Appl. Mater. Interfaces 2013, 5, 7983.[37] H. Tampo, H. Shibata, K. Maejima, T. W. Chiu, H. Itoh, A. Yamada,

K. Matsubara, P. Fons, Y. Chiba, T. Wakamatsu, Y. Takeshita, H. Kanie, S. Niki, Appl. Phys. Lett. 2009, 94, 242107.

[38] M. Nakata, H. Tsuji, H. Sato, Y. Nakajima, Y. Fujisaki, T. Takei, T. Yamamoto, H. Fujikake, Jpn. J. Appl. Phys. 2013, 52, 03BB04.

[39] T. J. Seok, Y. Liu, H. J. Jung, S. B. Kim, D. H. Kim, S. M. Kim, J. H. Jang, D. Y. Cho, S. W. Lee, T. J. Park, ACS Nano 2018, 12, 10403.

[40] Y. H. Lin, H. Faber, J. G. Labram, E. Stratakis, L. Sygellou, E. Kymakis, N. A. Hastas, R. Li, K. Zhao, A. Amassian, N. D. Treat, M. McLachlan, T. D. Anthopoulos, Adv. Sci. 2015, 2, 1500058.

[41] S. Lee, K. Ghaffarzadeh, A. Nathan, J. Robertson, S. Jeon, C. Kim, I. H. Song, U. I. Chung, Appl. Phys. Lett. 2011, 98, 203508.

[42] T. Kamiya, K. Nomura, H. Hosono, J. Disp. Technol. 2009, 5, 462.[43] H. Tampo, H. Shibata, K. Maejima, A. Yamada, K. Matsubara,

P. Fons, S. Kashiwaya, S. Niki, Y. Chiba, T. Wakamatsu, H. Kanie, Appl. Phys. Lett. 2008, 93, 202104.

[44] J. Falson, D. Maryenko, Y. Kozuka, A. Tsukazaki, M. Kawasaki, Appl. Phys. Express 2011, 4, 091101.

[45] H.-A. Chin, I. C. Cheng, C.-I. Huang, Y.-R. Wu, W.-S. Lu, W.-L. Lee, J. Z. Chen, K.-C. Chiu, T.-S. Lin, J. Appl. Phys. 2010, 108, 054503.

[46] J. W. Hennek, Y. Xia, K. Everaerts, M. C. Hersam, A. Facchetti, T. J. Marks, ACS Appl. Mater. Interfaces 2012, 4, 1614.

[47] Y. Wang, X. W. Sun, G. K. L. Goh, H. V. Demir, H. Y. Yu, IEEE Trans. Electron Devices 2011, 58, 480.

[48] L. Li, T. Xue, Z. Song, C. Liu, F. Ma, Rare Met. Mater. Eng. 2016, 45, 1992.

[49] B. Wang, X. Yu, P. Guo, W. Huang, L. Zeng, N. Zhou, L. Chi, M. J. Bedzyk, R. P. H. Chang, T. J. Marks, A. Facchetti, Adv. Electron. Mater. 2016, 2, 1500427.

[50] B. Wang, W. Huang, L. Chi, M. Al-Hashimi, T. J. Marks, A. Facchetti, Chem. Rev. 2018, 118, 5690.

[51] A. Zeumault, V. Subramanian, Adv. Funct. Mater. 2016, 26, 955.[52] Y. H. Kim, J. S. Heo, T. H. Kim, S. Park, M. H. Yoon, J. Kim,

M. S. Oh, G. R. Yi, Y. Y. Noh, S. K. Park, Nature 2012, 489, 128.[53] M. G. Kim, M. G. Kanatzidis, A. Facchetti, T. J. Marks, Nat. Mater.

2011, 10, 382.

polymer doping enables a two‐dimensional electron gas for...

Documents