The effect of charge transfer transition on the photostability of lanthanide-doped indium oxide thin-film transistors

Amorphous oxide semiconductors are promising for their use in thin-film transistor (TFT) devices due to their high carrier mobility and large-area uniformity. However, their commercialization is limited by the negative gate bias stress experienced under continuous light illumination. Here, we report an approach to improve the negative bias illumination stress (NBIS) stability of amorphous oxide semiconductors TFTs by using lanthanide-doped indium oxide semiconductors as the channel layer. The effect of different lanthanide dopants on performances of solution-processed Ln:In2O3 TFTs are investigated. All lanthanides exhibit strong suppression of oxygen vacancy, which shift the Von from −13.5 V of pure In2O3 TFT to −1~1 V of Ln:In2O3 TFTs (except Ce). However, only Pr:In2O3 and Tb:In2O3 TFTs exhibit much better NBIS stability with same ΔVon of −3.0 V, compared to much higher ΔVon of −7.9~−15.6 V for other Ln:In2O3 TFTs. Our comprehensive study reveals that praseodymium and terbium act as a blue light down-conversion medium with low charge transfer transition energy for lowing photosensitivity of oxide semiconductors. Thin-film transistors based on amorphous oxide semiconductors have promising applications, but their stability is hampered by negative bias illumination stress. Here, a systematic study of lanthanide-doped indium oxide semiconductors reveals that Pr and Tb are most efficient in improving the photostability of devices.

A morphous oxide semiconductors (AOSs) have attracted considerable attention, owing to their outstanding properties, such as high carrier mobility with large-area uniformity, low off-state current (I off ), and large optical bandgap 1,2 . However, there is a serious issue limiting further commercialization applications of AOS TFTs that serious threshold voltage shift (ΔV th ) is observed when the TFTs experience a negative gate bias stress combined with continuous light illumination even in the visible spectrum (NBIS), which cannot be fully recovered even after removing the stress for days [3][4][5] .
It is widely accepted that the NBIS instability is attributed to the intrinsic states of AOS materials regardless of the device structures. Several degradation models are proposed to reveal the mechanism of the NBIS instability of the AOS TFTs, including the trapping of photogenerated holes [6][7][8][9] , the creation of ionized oxygen vacancy (Vo) defects 10,11 , the formation of metastable peroxides 12 , the self-trapping of holes by polarons 13 , etc.
To improve the NBIS stability of AOS TFTs, reducing concentration of preexisting Vo defects (subgap states) is the most commonly used method. The cation (oxygen binder) doping [14][15][16] , and the intentional supply of oxygen species, such as highpressure annealing (HPA) [17][18][19] , oxygen-containing plasma treatment 20 , ozone radical treatment 21 , and ultra-high vacuum sputtering 22 , would result in a decrease in Vo. However, the improved V th shift is still too large to drive active-matrix displays without compensation 5,22,23 . J. Kim and H. Hosono et al. 22 developed a wide-bandgap AOS (ZnGaO) to improve the NBIS stability of the AOS TFTs by widening the optical bandgap to keep the photoexcitation energy between the subgap states and the conduction band minimum (CBM) larger than ∼3 eV. However, this approach will decrease the mobility seriously, because the 4 s orbitals of adjacent metal ions (both Ga and Zn) do not overlap that would destroy the electron transport paths in the amorphous state. Therefore, there is a tradeoff between the mobility and NBIS stability of the AOS TFTs.
Here, we report an approach to improve the NBIS stability of AOS TFTs by using lanthanide-doped indium oxide (Ln:In 2 O 3 ) semiconductors as the channel layer. It is found that only praseodymium (Pr) and terbium (Tb) can improve the NBIS stability of the AOS TFTs greatly, and doping Pr/Tb into In 2 O 3 would not affect the mobility much, so the tradeoff between the mobility and NBIS stability of the AOS TFTs can be broken. Comprehensive studies reveal that different lanthanides (Lns) have a different effect on the NBIS stability of AOS TFTs, and Pr/Tb acts not only a Vo suppressant but also a blue light downconversion medium with low charge transfer transition energy. The result is different from those reported elsewhere, where the lanthanides are merely regarded as free carrier suppressants of the AOSs due to ultra-low electronegativities [24][25][26][27][28][29][30] . T. Kamiya et al. 31,32 gave a comprehensive study on the effect of the Ln doping on the photoluminescence characteristics of the AOS films but the effect of Ln doping on the TFT characteristics has not been investigated. It is should be noted that promethium (Pm) is not included in this work because of its radioactivity.

Results and discussion
The absorption characteristics of the different Ln ions. As known, lanthanides have electron configurations of [Xe] 4 f n 5d m 6s 2 (n = 0-14, m = 0-10). The unfilled 4f orbitals form complicated energy level structures due to Coulomb interactions, spin-orbit coupling, and external field interactions. Lanthanides exhibit various optical properties which are attributed to f-f transition, f-d transition, or charge transfer transition (electrons transit between the ligands and Ln ions). Figure 1 shows the photos of the different aqueous Ln(NO) 3  Interestingly, after the precursors are spin-coated onto the glass substrates and annealed at 300°C, all the films are colorless and transparent except PrO x and TbO x . To investigate the color difference between the solution and the corresponding film, UVvisible light absorption was measured for all of the LnO x films, as shown in Fig. 2b. The CeO x , PrO x , and TbO x films exhibit broadband absorption, while the other films are transparent to the light with the wavelength longer than 300 nm. For the CeO x film, the absorption is strong with a peak centered at~300 nm and a cutoff edge at~390 nm. For the PrO x and TbO x films, the absorption peaks are broadened into the visible light without distinct cut-off edges. The broad-band and strong absorption implies that the absorption cannot be caused by f-f transition, because f-f transition has some features, including i) f-f transition is forbidden due to the same parity, so the f-f transition absorption is relatively weak; ii) f-f transition spectrum is line spectrum with sharp peaks because the intimal 4 f orbitals are screened by the outer 5d and 6 s orbitals and hardly influenced by the chemical environment; iii) f-f transition can be observed in most of the Ln ions. The absorption spectra of CeO x , PrO x , and TbO x films is similar to that reported by D.E. Hobart et al. 33 in 1980 (see Fig. 2c-e), who ascribed the broad-band and absorption to the charge transfer transition between the delocalized ligand molecular orbital and the Ln 4 + ion. Therefore, it is reasonable to deduce that Ce 4 + , Pr 4 + , and Tb 4 + ions exist in the solid films but cannot exist in the aqueous Ln(NO) 3 ·nH 2 O solutions. The oxidation from Ln 3 + to Ln 4 + is considered to be taken place during the 300°C annealing step because all of the as-spin-coated LnO x films without annealing are colorless and transparent. As known, nitrate (NO 3 − ) can releases oxygen-free radicals at a relatively low temperature (~200°C) 34 . The oxygen radicals with strong oxidability can help oxidizing the Ce 3 + , Pr 3 + , and Tb 3 + ions to the tetravalent oxidation state. Compared to the work by D.E. Hobart et al. 33 , the absorption spectra for the PrO x and TbO x films in this work span a much wider range (extend to the near-infrared region), which may be attributed to the larger metal-ligand electron cloud overlap in the solid state. Moreover, the oxygen vacancies and structural defects of the amorphous LnO x films will give rise to the electron cloud expansion effect and further broaden the absorption spectra to the near-infrared region, as discussed later.
To further analyze, the valence state of the CeO x , PrO x , and TbO x films, X-ray photoelectron spectroscopy (XPS) experiments were carried out, as shown in Supplementary Fig. 1. It can be found that the Ce ions in the CeO x film are mainly in the tetravalent state (Ce 4 + ) with only a small part of trivalent Ce ions (Ce 3 + , see V′-W′ peaks in Supplementary Fig. 1), while the number of Pr 3 + /Tb 3 + ions is comparable to that of Pr 4 + / Tb 4 + ions in the PrO x /TbO x films ( Supplementary Fig. 1).   indicates that the indium nitrate aqueous precursor can form a continuous mate oxide skeleton by thermal decomposition even at a low temperature of 300°C. However, the undoped In 2 O 3 TFT exhibits highly conductive with a turn-on voltage (V on , defined as the V GS when I DS is increase to 1 nA) of as negative as −13.3 V, which is due to the large amounts of free carriers resulted from the high-density oxygen vacancies in the undoped In 2 O 3 film 35 . In addition, the undoped In 2 O 3 TFT shows serious stability problems as discussed later.
The transfer characteristic of the AOS TFTs based on Ln:In 2 O 3 . Figure 4 shows the transfer characteristic curves of the TFTs based on 14 Ln:In 2 O 3 annealed at 300°C. All the TFTs exhibit excellent gate-controlled field-effect characteristics except Ce:In 2 O 3 TFT (see Supplementary Fig. 2) that will be discussed later. Compared to the undoped In 2 O 3 TFT, which has a V on of −13.3 V, the Ln:In 2 O 3 TFTs have more positive V on that range from −1 to 1 V (see Table 1) and steeper SS, revealing that Ln doping with a concentration of only 5 at.% can suppress Vo generation and decrease free carrier concentration substantially.
In conventional AOS (IGZO), the concentration of the free carrier suppressant (Ga) should be very high (33.3 at.%), because Ga is not a strong free carrier suppressant. The electronegativity of Ga is 1.6 with Ga-O dissociation energy of 374 kJ mo1 −1 , no much difference compared to In (the electronegativity of In is 1.7 with In-O dissociation energy of 346 kJ mo1 −1 ). High Ga concentration will destroy overlaying of the In 5 s orbitals (especially in the amorphous state), and result in electron mobility decrease. Instead, the lanthanides are a strong carrier suppressant, because their electronegativity (1.10-1.27, Supplementary Table 1) is much lower than Ga, and the Ln-O dissociation energy (397-799 kJ mo1 −1 , Supplementary Table 1 The results prove that the Ln doping can effectively suppress Vo generation. It is worth noting that highintensity M-OH peaks are observed in the XPS spectra of the La:In 2 O 3 and Pr:In 2 O 3 films, which may be attributed to larger radii of La 3+ and Pr 3+ compared to other trivalent Ln ions (except Ce which is mainly in the tetravalent state). Larger radius difference between In 3+ and Ln 3+ would cause more serious lattice mismatch and more loosely bound oxygen species. Interestingly, the Ce:In 2 O 3 film has an extremely low Vo ratio of only 7.8% that is ascribed to the strong reducibility of Ce 3+ , which will further bind with oxygen atoms and oxidized to Ce 4+ during annealing. Because the Vo concentration is very low in the Ce:In 2 O 3 film, the free carrier density of it is very low, so almost no field effects are found in the Ce:In 2 O 3 TFTs. It should be noted that decreasing the Ce doping concentration will increase the I on but will make the V on more negative, as shown in Fig. S2. It is clear that doping Ce into In 2 O 3 (in solution method) will deteriorate the performances of the Ce:In 2 O 3 TFT (even with a small Ce doping amount of 0.8 at.%).
The mobilities of the TFTs based on 13 Ln:In 2 O 3 (except Ce:In 2 O 3 ) semiconductors are 3.1-6.1 cm 2 V −1 s −1 when annealed at 300°C and 8.2-16.2 cm 2 V −1 s −1 when annealed at 350°C (see Table 1 and Supplementary Fig. 4). The higher mobility with higher annealing temperature is attributed fewer impurities and a higher degree of order. Further increase the annealing temperature (>350°C) has not tried, because the electrodes will be oxidized at higher annealing temperature. It can be seen from Table 1 that the mobility of the Eu:In 2 O 3 and Yb:In 2 O 3 TFTs are particularly low that may be due to the special electronic configurations of Eu (4 f 7 6s 2 ) and Yb (4 f 14 6 s 2 ), which may lose two 6s electrons and form relatively stable Eu 2 + (4 f 7 ) and Yb 2 + (4 f 14 ). However, there are no direct evidences that Eu 2 + and Yb 2 + ions exist in the annealed solid films.
It is worth noting that the mobility of solution-processed AOS TFTs is generally lower than that of vacuum-based AOS TFTs because of the lower film quality and the impurities induced by the precursors. In our previous studies, the mobility of the vacuum-based Nd:  The NBIS stability of the Ln:In 2 O 3 TFTs. Supplementary Figure 6 shows the evolutions of transfer curves of the TFTs illuminated with different intensity of white light (without bias stress). The schematic diagram and the photo of the photostability testing equipment are shown in Supplementary Fig. 6. During illumination, the TFT channel was directly exposed to white LED light whose spectrum is depicted in Fig. 3b, and the intensity of the white light ranges from 500 to 5000 Lux. The undoped In 2 O 3 TFT is rather unstable under the white light illumination, while the Pr:In 2 O 3 and Tb:In 2 O 3 TFTs are hardly affected by the white light illumination (highly stable even under 5000-Lux-light illumination, see Supplementary Fig. 7). For further investigating the effect of Ln doping on the photostability of the AOS TFTs, the devices were tested under the NBIS. During the NBIS, the TFTs were bias with a V GS of −20 V under continuous white LED irradiation (250 Lux), and the transfer curves were recorded at 0, 100, 600, 1200, and 3600 s, respectively.  Table 1 and Supplementary Fig. 8. It is observed that the NBIS of the Pr:In 2 O 3 and Tb:In 2 O 3 TFTs are improved greatly, with the same ΔV on of −3.0 V. The ΔV on of the other 11 Ln:In 2 O 3 TFTs are at the range of −7.9 to −15.6 V (see Table 1). The result shows that only the Pr and Tb are effective dopants for resolving the NBIS instability problem of the AOS TFTs.
It is known that high-density subgap states are formed by Vo existed in AOSs near valance band maximum (VBM). Under light illumination, some of the Vo will be thermal excited and lose two electrons and become Vo 2+ . The transition of the Vo ground state to Vo 2+ excited states causes spontaneous outward relaxation, which makes the Vo 2+ level act as a subgap state below the conduction band minimum (CBM), contributing the delocalized free electrons in the conduction band 38-40 , as illustrated in Fig.6. It has been reported that the formation energy of Vo 2+ decreases as the fermi lever (E F ) approaching the VBM 38 . When a negative bias is applied to the gate electrode, the energy band at the gate insulator (GI)/AOS interface bends up, so the E F is closer to the VBM (see Fig. 6). As a result, the formation energy of Vo 2+ at the GI/AOS interface decreases. Therefore, more Vo 2+ will be formed when a negative V GS is applied during light illumination (NBIS), leading to more electrons in the conduction band, which is the reason for the large negative V on shift under the NBIS. Meanwhile, the ΔV on is increase with illumination intensity (as seen in Supplementary Fig. 9). The photoresponse is reversible very slowly with a time constant exceeding thousands of seconds. This is attributed to the  The effect of Pr and Tb in NBIS stability. It can be deduced from the analysis above that there are mainly three ways to improve the NBIS stability of the AOS TFTs. i) to decrease the Vo concentration; ii) to widen the bandgap of the AOSs and deepen the VBM and the Vo subgap states, so that the Vo cannot be activated by the visible light; iii) to downconvert the incident light by doping some lanthanide ions into the AOSs, so that the energy of the output light is not enough to activate the Vo. Here, the excellent NBIS stability of the Pr:In 2 O 3 and Tb:In 2 O 3 TFTs cannot be attributed to the suppression of Vo, because there is no evident relationship between the Vo ratios and the ΔV on under the NBIS (see Table 1 and Supplementary Table 1). The VBMs of the Ln:In 2 O 3 thin films were characterized by ultraviolet photoelectron spectroscopy (UPS) measurements. It is found that there is no much difference on the VBMs for all of the 14 Ln:In 2 O 3 films ( Supplementary Fig. 10 As discussed at the beginning of this section, the CeO x , PrO x , and TbO x films exhibit broad-band absorption (see Fig. 2), while the other films are transparent to the light with the wavelength longer than 300 nm; and the broad-band and strong absorption the CeO x , PrO x , and TbO x films are mainly resulted from the charge transfer transition from ligands to Ln 4 f. Among all of the lanthanides, only Ce, Pr, Tb, Nd, and Dy have tetravalent oxidation states, but Nd 4+ and Dy 4+ are very unstable because of their large III-IV potentials (E 0 , see Supplementary Table 1) 41 . The stability sequence of the tetravalent oxidation states is Ce 4+ >Tb 4+ ≈Pr 4+ » Nd 4+ >Dy 4+ . In lanthanide oxides, the Ln5d orbitals and the O2p orbitals make up the chemical bonds; the bonding orbitals form the valance band (VB), and antibonding orbitals form the conduction band (CB), as illustrated in Fig. 7. So, the charge transfer energy (E CT ) is the difference between the ground state of the Ln 3 + (E ground ) and the energy level of VBM (E V ): The E CT of Ce (III-IV) can be calculated from the charge transfer absorption peak (300 nm, see    41 reported that there is a linear unit-slope relationship between the III-IV potentials (E 0 ) and the first charge transfer band energy. Although they did not provide the formulas for the relationship between the E CT and E 0 , it can be deduced from plots of the E 0 versus the first absorption energy that E CT of Pr/Tb can be expressed as where C is a constant estimated to be −1 eV V −1 . It can be calculated from Eq. 2 that E CT of Pr and Tb (III-IV) are 2.5 eV (496 nm) and 2.6 eV (477 nm), respectively. Therefore, the charge transfer absorption peaks of PrO x and TbO x enter the blue light regime even without considering the electron cloud expansion effect. Because Ce 4+ is most stable, the energy level of Ce 4+ is lowest. It means that, higher energy is required for charge transfer transition from | Ce4f 0 > to | Ce4f 1 O2p −1 > , compared to those from | Pr4f 1 > to | Pr4f 2 O2p −1 > or from | Tb4f 7 > to | Tb4f 8 O2p −1 > , as illustrated in Fig. 8. After charge transfer transition, the hole in O2p −1 will pull the transited electron back to its original space by a nonradiative way (or downconverted to red light). Supplementary Fig. 11 shows the enhanced photoluminescence spectra of the Ce:In 2 O 3 , Pr:In 2 O 3 , and Tb:In 2 O 3 films stimulated by high-intensity 450-nm light. Only Pr:In 2 O 3 shows a weak redlight peak centered at~625 nm. So it can be deduced that the absorbed incident light is converted to lattice vibration for the Ce:In 2 O 3 and Tb:In 2 O 3 films, and a small part of the absorbed incident light is converted to red light for the Pr:In 2 O 3 film. It should be noted that the Ln 4+ ions tend to move to Vo, because of the weak electrostatic attraction, and the charge transfer transition is more easily to take place between Vo and Ln 4+ , so it can absorb light with lower energy (longer wavelength), as illustrated in Fig. 8. T. Kamiya et al. 2 reported that amorphous IGZO TFTs respond to photon energies above 2.3 eV (corresponding to the wave lengths below 540 nm), which is lower than the bandgap (3.1 eV). It means that the blue light and parts of the green light of the LED light are responsible for the photoinstability of the AOS TFTs. It can be observed from Figs. 2d, e that the absorption spectra for both of the PrO x and TbO x films extend to the near-infrared region, so they can absorb almost the whole range of the incident white LED light (see Fig. 3b). As a result, the incident white LED light is downconverted to the nonradiative transition by Pr 4+ or Tb 4+ that enhances the NBIS stability greatly.
It should be noted that the charge transfer transition energies of the Ln 3+ → Ln 2+ is much higher than those of Ln 4+ → Ln 3+ , as shown in Fig. 9a, so the Ln 3+ → Ln 2+ transition cannot absorb blue light. As a result, the Ln 3+ ions cannot improve the photostability of the TFTs.
Besides tetravalent ions, trivalent ions also exist in the CeO x , PrO x , and TbO x films. That means 4f-5d transition is likely to take place, because the f-d transition energies (E fd ) for Ce 3+ , Pr 3+ , and Tb 3+ ions are much lower than those for other free Ln 3+ ions (see Supplementary Table 1). It is known that the E fd can be lower greatly by choosing ligands with lower electronegativity due to the larger metal-ligand electron cloud overlap (electron cloud expansion effect). Different from the f-f transition, the f-d transition is allowed by the parity selection rule, so it has a fairly high transition intensity that is 10 6 times higher than f-f transition. And f-d transition has a wide absorption band because the 5d energy level is widened by ligand ions. Since the E fd for Ce 3+ is lower than that for Pr 3+ or Tb 3+ , and the light absorption spectrum of the CeO x film is cut-off at~390 nm (Fig. 1e), it can be deduced that the light energy for the f-d transition for all of the CeO x , PrO x , and TbO x films should be greater than 3.18 eV (<390 nm). It means that the visible light absorption of the PrO x and TbO x films cannot be attributed to the f-d transition. Therefore, the charge transfer transition is the only reason for the greatly improved NBIS stability of the TFTs with the Pr:In 2 O 3 or Tb:In 2 O 3 semiconductor layers.
It is worth noting that the bivalent Ln ions (Ln 2+ ) have lower E fd than the trivalent Ln ions (Ln 3+ ). And the E fd of some Ln 2+ ions are close to the energy of blue light. For example, the E fd for Eu 2+ , Yb 2+ , Sm 2+ , and Tm 2+ are 4.30, 4.20, 2.92, and 2.87 eV, respectively (see Fig. 9b). However, only Eu 2+ and Yb 2+ are stable in the solid state. Moreover, the Ln:In 2 O 3 films must experience high-temperature annealing (>300°C), so the Ln 2+ ions will be oxidized to Ln 3+ ions. Thus, it is difficult to achieve stable AOS TFTs doped with Ln 2+ ions.
It is worth noting that the Pr:In 2 O 3 and Tb:In 2 O 3 TFTs exhibit worsened stability under positive bias illumination stress (PBIS) with ΔV on of 1.1 and 0.9 V, respectively (see Supplementary  Fig. 12). The degradation of the PBIS for Pr:In 2 O 3 and Tb:In 2 O 3 TFTs is ascribed to the light-induced shallow electron traps (possibly Pr 3+ and Tb 3+ related defects).   Supplementary Fig. 14), which may be attributed to the much higher free electron density than the electron traps in the In 2 O 3 film. However, most Ln:In 2 O 3 TFTs show relatively poor PBTS (especially the La, Nd, and Ho incorporated In 2 O 3 TFTs). It reveals that incorporating Ln to In 2 O 3 will induce shallow electron traps in the In 2 O 3 films. The temperature effect of the Ln:In 2 O 3 TFTs is still under investigated, and will be published in the future.

Conclusions
In conclusion, the effect of different lanthanide dopants on the NBIS stability of solution-processed AOS TFTs are investigated. Comprehensive studies reveal that Pr/Tb acts as not only a Vo suppressant but also a blue light downconversion medium with low charge transfer transition energy. Most of the incident white light can be absorbed by Pr 4+ or Tb 4+ ions by charge transfer transition, and downconverted to nonradiative transition or red light. As a result, the NBIS stability of the AOS TFTs is improved greatly. The result is different from those reported elsewhere, where the lanthanides are merely regarded as free carrier suppressants of the AOSs due to ultra-low electronegativities Devices fabrication. A bottom-gate and top-contact source-drain electrode structure was used to fabricate Ln: In 2 O 3 TFT. Firstly a 200-nm thickness Al 2 O 3 :Nd gate dielectric layer with a capacitance density of 38 nF cm −2 was formed by anodization on the surface of a 300-nm thick Al:Nd alloy film deposited on glass by sputtering and patterned by wet etch, which the details process have been reported in our previous report. Then, the substrates were cleaned ultrasonically in deionized water and isopropanol for 10 min, respectively, and dried in an air oven of 80°C. Next, an ultraviolet light irradiating of a long time was used to treat a part (channel area) of the Al 2 O 3 :Nd surface with a stencil shadow mask, for the formation of a hydrophilic surface in the channel area. The Ln: In 2 O 3 precursor films were deposited in the wettable area irradiated by UV by spin-coating Ln: In 2 O 3 precursor solutions at 2000 rpm for 5 s and 6500 rpm for 40 s, followed by drying at 40°C and thermal annealing at 300/350°C for 1 h in an air condition. Finally, the Al source and drain electrodes were deposited on the Ln: In 2 O 3 layer by thermal evaporation, defined the channel area with 1000 μm width and 300 μm length by using a stencil shadow mask.
Devices characterization. The electrical characteristics of undoped In 2 O 3 and Ln: In 2 O 3 TFTs were measured using a semiconductor parameter analyzer system (Agilent B1500A) in conjunction with a probe station (Lakeshore CRX-VF) at room temperature and air condition. The NBIS stability was tested by monitoring evolutions of the transfer curves of MO TFTs based on a series of Ln incorporated In 2 O 3 semiconductors as a function of the stress time under gate bias stresses of −20 V combining with white LED irradiation of optical density of 250 Lux, respectively. The energy levels of the functional layer were measured with Ultraviolet Photoelectron Spectrometer (UPS) of K-ALPHA + of Thermo Fisher Scientific. The absorption spectrum was measured by using UV-2600. The enhanced photoluminescence spectra was measured by using FLS1000. The X-ray Photoelectron Spectroscopy (XPS) was measured by using K-ALPHA + of Thermo Fisher Scientific with Mono AlKα. The carrier mobility for the saturation regime was calculated using: where μ is the field-effect mobility, C i is the areal capacitance per area of the gate dielectric, V th is the threshold voltage, W is the channel width, and L is the channel length.

Data availability
The data that support the plots within this paper and other findings of this study are available from the corresponding authors upon reasonable request.