Introduction

Amorphous 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 (Ioff), and large optical bandgap1,2. However, there is a serious issue limiting further commercialization applications of AOS TFTs that serious threshold voltage shift (ΔVth) 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 days3,4,5.

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 holes6,7,8,9, the creation of ionized oxygen vacancy (Vo) defects10,11, the formation of metastable peroxides12, the self-trapping of holes by polarons13, 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) doping14,15,16, and the intentional supply of oxygen species, such as high-pressure annealing (HPA)17,18,19, oxygen-containing plasma treatment20, ozone radical treatment21, and ultra-high vacuum sputtering22, would result in a decrease in Vo. However, the improved Vth shift is still too large to drive active-matrix displays without compensation5,22,23. 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:In2O3) 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 In2O3 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 electronegativities24,25,26,27,28,29,30. 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 fn5dm6s2 (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·nH2O solutions (0.4 M) and the corresponding spin-coated thin films annealed at 300 °C. Only Pr(NO)3·nH2O, Nd(NO)3·nH2O, Ho(NO)3·nH2O, and Er(NO)3·nH2O solutions show light colors, which are attributed to the f-f transition of the Pr, Nd, Ho, and Er ions. Figure 2a shows the absorption spectra of the different aqueous Ln(NO)3·nH2O solutions. Typical f-f sharp absorption peaks are seen in Pr(NO)3·nH2O, Nd(NO)3·nH2O, Ho(NO)3·nH2O, and Er(NO)3·nH2O solution (much weaker absorptions peaks are also seen in Sm(NO)3·nH2O, Eu(NO)3·nH2O, Dy(NO)3·nH2O, and Tm(NO)3·nH2O solutions), which are in consistent with the colors of the solutions.

Fig. 1: Ln(NO)3·nH2O aqueous solutions and LnOx films.
figure 1

The photos of the different aqueous Ln(NO)3·nH2O solutions (0.4 M) and the corresponding spin-coated films annealed at 300 °C.

Full size image
Fig. 2: The absorption of Ln(NO)3·nH2O aqueous solutions and LnOx films.
figure 2

a the absorption spectra of the different Ln(NO)3·nH2O and In(NO)3·nH2O aqueous solutions. b UV-visible light absorption of the different LnOx films based on spin-coating Ln(NO)3·nH2O aqueous solutions annealed at 300 °C. c–d UV-visible light absorption of c Ce3 + and CeOx, (d) Pr3 + and PrOx, and (e) Tb3 + and TbOx.

Full size image

Interestingly, after the precursors are spin-coated onto the glass substrates and annealed at 300 °C, all the films are colorless and transparent except PrOx and TbOx. To investigate the color difference between the solution and the corresponding film, UV-visible light absorption was measured for all of the LnOx films, as shown in Fig. 2b. The CeOx, PrOx, and TbOx films exhibit broad-band absorption, while the other films are transparent to the light with the wavelength longer than 300 nm. For the CeOx film, the absorption is strong with a peak centered at ~300 nm and a cut-off edge at ~390 nm. For the PrOx and TbOx 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 CeOx, PrOx, and TbOx 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 Ln4 + ion. Therefore, it is reasonable to deduce that Ce4 + , Pr4 + , and Tb4 + ions exist in the solid films but cannot exist in the aqueous Ln(NO)3·nH2O solutions. The oxidation from Ln3 + to Ln4 + is considered to be taken place during the 300 °C annealing step because all of the as-spin-coated LnOx films without annealing are colorless and transparent. As known, nitrate (NO3−) can releases oxygen-free radicals at a relatively low temperature (~200 °C)34. The oxygen radicals with strong oxidability can help oxidizing the Ce3 + , Pr3 + , and Tb3 + ions to the tetravalent oxidation state. Compared to the work by D.E. Hobart et al.33, the absorption spectra for the PrOx and TbOx 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 LnOx 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 CeOx, PrOx, and TbOx 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 CeOx film are mainly in the tetravalent state (Ce4 + ) with only a small part of trivalent Ce ions (Ce3 + , see V′-W′ peaks in Supplementary Fig. 1), while the number of Pr3 + /Tb3 + ions is comparable to that of Pr4 + /Tb4 + ions in the PrOx/TbOx films (Supplementary Fig. 1).

The performances of the pure In2O3 TFTs

To investigate the effect of different lanthanide doping on the performance of the AOS TFTs, different lanthanide elements are doped into In2O3 by mixing Ln(NO)3·nH2O and In(NO)3·nH2O solutions together to form precursor solutions. The In/Ln molar ratio was optimized to be 19:1 (5 at.% of Ln). Bottom-gate and top-contact TFTs with undoped In2O3 or Ln:In2O3 channel layers were fabricated by spin-coating In2O3 or Ln:In2O3 aqueous precursor solutions and annealing at 300/350 °C in air condition for 1 h (see Fig. 3a). To avoid the influence of the air molecules on the stability of the AOS TFTs, all the TFTs were tested under vacuum condition. Figure 3c, d show output and transfer characteristics of the undoped In2O3 TFT. The mobility is extracted from transfer characteristics at VDS = 15.1 V to be 17.1 cm2 V−1 s−1, which 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 In2O3 TFT exhibits highly conductive with a turn-on voltage (Von, defined as the VGS when IDS 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 In2O3 film35. In addition, the undoped In2O3 TFT shows serious stability problems as discussed later.

Fig. 3: The characteristics of pure In2O3 TFT.
figure 3

a Device structure of TFTs based on solution-processed In2O3 or Ln: In2O3 semiconductors without any passivation. b LED spectrum for the NBIS test. c output and d transfer characteristics of the undoped In2O3 TFT annealed at 300 °C with an aqueous precursor solution. The transfer characteristics of TFTs were measured at VDS = 1.1 and 15.1 V, respectively, and VGS swept from −30 to 30 V, while output characteristics were tested when VGS was held at different voltages between 0 and 30 V in step of 5 V, and VDS swept from 0 to 30 V at each VGS.

Full size image

The transfer characteristic of the AOS TFTs based on Ln:In2O3

Figure 4 shows the transfer characteristic curves of the TFTs based on 14 Ln:In2O3 annealed at 300 °C. All the TFTs exhibit excellent gate-controlled field-effect characteristics except Ce:In2O3 TFT (see Supplementary Fig. 2) that will be discussed later. Compared to the undoped In2O3 TFT, which has a Von of −13.3 V, the Ln:In2O3 TFTs have more positive Von 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.

Fig. 4: Transfer characteristics of Ln:In2O3 TFTs.
figure 4

Transfer characteristics of TFTs based on spin-coating Ln:In2O3 aqueous precursor solutions with annealing at 300 °C in air condition, the VDS was held at 1.1 V (red curve) and 15.1 V (black curve), respectively, and the VGS sweep from −30 to 30 V.

Full size image
Table 1 Summary for the performance of TFTs with Ln:In2O3 channels.
Full size table

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) is much higher than Ga–O. To characterized the effect of Ln doping on the Vo concentration, XPS measurements of O 1 s of undoped In2O3 and 14 Ln:In2O3 films were performed, as shown in Supplementary Fig. 3. The O 1 s spectra can be fitted by three Gaussian distributions with binding energies at 529.5 ± 0.2 eV (Peak 1), 530.0 ± 0.2 eV (Peak 2), and 531.3 ± 0.3 eV (Peak 3), corresponding to the oxygen in oxide lattices (M–O–M), oxygen vacancies (Vo) and metal hydroxide (or loosely bound oxygen) species (M–OH), respectively36. The Vo/(M–O–M + Vo  + M–OH) ratio for undoped In2O3 film is 18.3 %, while those for Ln:In2O3 (except Ce:In2O3) films are in the range of 10.8–4.1 %. The results prove that the Ln doping can effectively suppress Vo generation. It is worth noting that high-intensity M–OH peaks are observed in the XPS spectra of the La:In2O3 and Pr:In2O3 films, which may be attributed to larger radii of La3+ and Pr3+ compared to other trivalent Ln ions (except Ce which is mainly in the tetravalent state). Larger radius difference between In3+ and Ln3+ would cause more serious lattice mismatch and more loosely bound oxygen species. Interestingly, the Ce:In2O3 film has an extremely low Vo ratio of only 7.8% that is ascribed to the strong reducibility of Ce3+, which will further bind with oxygen atoms and oxidized to Ce4+ during annealing. Because the Vo concentration is very low in the Ce:In2O3 film, the free carrier density of it is very low, so almost no field effects are found in the Ce:In2O3 TFTs. It should be noted that decreasing the Ce doping concentration will increase the Ion but will make the Von more negative, as shown in Fig. S2. It is clear that doping Ce into In2O3 (in solution method) will deteriorate the performances of the Ce:In2O3 TFT (even with a small Ce doping amount of 0.8 at.%).

The mobilities of the TFTs based on 13 Ln:In2O3 (except Ce:In2O3) semiconductors are 3.1–6.1 cm2 V−1 s−1 when annealed at 300 °C and 8.2–16.2 cm2 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:In2O3 and Yb:In2O3 TFTs are particularly low that may be due to the special electronic configurations of Eu (4 f 76s2) and Yb (4 f 146 s2), which may lose two 6s electrons and form relatively stable Eu2 + (4 f 7) and Yb2 + (4 f 14). However, there are no direct evidences that Eu2 + and Yb2 + 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:In2O3 (5 at.%) TFT can reach as high as 46.4 cm2 V−1 s−1, while the mobility of the solution-processed Nd:In2O3 (2 at.%) TFT is only 15.6 cm2 V−1 s−1 24,37. The mobility of the TFTs with sputtered Pr/Tb:In2O3 (5 at.%) channel layer can reach as high as 36.2/38.1 cm2 V−1 s−1 (not shown). The crystal structures of the solution-processed In2O3 and Tb:In2O3 films were investigated using X-ray diffraction (XRD), as shown in Supplementary Fig. 5. Both films exhibit clear nanocrystalline nature with In2O3 bixbyite structure, which is consistent with the standard value for bulk cubic-In2O3. Therefore, 5 at.% of Ln doping will not have a remarkable effect on the mobility of the In2O3.

The NBIS stability of the Ln:In2O3 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 In2O3 TFT is rather unstable under the white light illumination, while the Pr:In2O3 and Tb:In2O3 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 VGS 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. Figure 5 shows the evolutions of the transfer curves of different TFTs as a function of the stress time. The Von shift (ΔVon) of the undoped In2O3 TFT is as large as −12.5 V (not shown), and the ΔVon of the 13 Ln:In2O3 (except Ce:In2O3) TFTs are summarized in Table 1 and Supplementary Fig. 8. It is observed that the NBIS of the Pr:In2O3 and Tb:In2O3 TFTs are improved greatly, with the same ΔVon of −3.0 V. The ΔVon of the other 11 Ln:In2O3 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.

Fig. 5: NBIS stability of Ln:In2O3 TFTs.
figure 5

The variations of time-dependent transfer curves of the Ln: In2O3 TFTs when applied a negative gate bias stress of −20 V combining with white LED light illumination of 250 Lux.

Full size image

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 Vo2+. The transition of the Vo ground state to Vo2+ excited states causes spontaneous outward relaxation, which makes the Vo2+ level act as a subgap state below the conduction band minimum (CBM), contributing the delocalized free electrons in the conduction band38,39,40, as illustrated in Fig.6. It has been reported that the formation energy of Vo2+ decreases as the fermi lever (EF) approaching the VBM38. When a negative bias is applied to the gate electrode, the energy band at the gate insulator (GI)/AOS interface bends up, so the EF is closer to the VBM (see Fig. 6). As a result, the formation energy of Vo2+ at the GI/AOS interface decreases. Therefore, more Vo2+ will be formed when a negative VGS is applied during light illumination (NBIS), leading to more electrons in the conduction band, which is the reason for the large negative Von shift under the NBIS. Meanwhile, the ΔVon 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 relaxation of metastable donor states with activation energies of 0.9–1.1 eV3,4.

Fig. 6: The origin of NBIS in stability for AOS TFTs.
figure 6

The schematic diagram for generation of Vo2 + in AOS TFTs under NBIS.

Full size image

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:In2O3 and Tb:In2O3 TFTs cannot be attributed to the suppression of Vo, because there is no evident relationship between the Vo ratios and the ΔVon under the NBIS (see Table 1 and Supplementary Table 1). The VBMs of the Ln:In2O3 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:In2O3films (Supplementary Fig. 10), so the much better NBIS stability of the Pr:In2O3 and Tb:In2O3 TFTs compared to the other Ln:In2O3 TFTs cannot be attributed to the bandgap widening. Therefore, the excellent NBIS stability of the Pr:In2O3 and Tb:In2O3 TFTs is probably due to the downconversion effect of the Pr and Tb ions.

As discussed at the beginning of this section, the CeOx, PrOx, and TbOx 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 CeOx, PrOx, and TbOx 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 Nd4+ and Dy4+ are very unstable because of their large III–IV potentials (E0, see Supplementary Table 1)41. The stability sequence of the tetravalent oxidation states is Ce4+ >Tb4+ ≈Pr4+» Nd4+ >Dy4+. 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 (ECT) is the difference between the ground state of the Ln3 + (Eground) and the energy level of VBM (EV):

$${E}_{{{{\rm{CT}}}}}={E}_{g{round}}\left({{{{\rm{Ln}}}}}^{3+}\right)-{E}_{{{{\rm{V}}}}}$$
(1)

The ECT of Ce (III–IV) can be calculated from the charge transfer absorption peak (300 nm, see Fig. 2c) to be 4.1 eV. The ECT of Pr (III–IV) and Tb (III–IV) should be estimated from the ECT of Ce (III–IV), because the PrOx and TbOx do not display a well-resolved charge transfer absorption band (Figs. 2d, e). L.J. Nugent et al.41 reported that there is a linear unit-slope relationship between the III–IV potentials (E0) and the first charge transfer band energy. Although they did not provide the formulas for the relationship between the ECT and E0, it can be deduced from plots of the E0 versus the first absorption energy that ECT of Pr/Tb can be expressed as

$${E}_{{{{\rm{CT}}}}}\left({{{\rm{Ln}}}}\right)={E}_{{{{\rm{CT}}}}}\left({{{\rm{Ce}}}}\right)+C[{E}^{0}\left({{{\rm{Ce}}}}\right)-{E}^{0}\left({{{\rm{Ln}}}}\right)]$$
(2)

where C is a constant estimated to be −1 eV V−1. It can be calculated from Eq. 2 that ECT 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 PrOx and TbOx enter the blue light regime even without considering the electron cloud expansion effect.

Fig. 7: The schematic diagram of energy levels of the tetravalent lanthanide oxides and In2O3.
figure 7

The valance band (VB) and the conduction band (CB) were formed by bonding orbitals and antibonding orbitals consisted of the Ln5d orbitals and the O2p orbitals in lanthanide oxides.

Full size image

Because Ce4+ is most stable, the energy level of Ce4+ is lowest. It means that, higher energy is required for charge transfer transition from | Ce4f 0 > to | Ce4f 1O2p−1 > , compared to those from | Pr4f 1 > to | Pr4f 2O2p−1 > or from | Tb4f 7 > to | Tb4f 8O2p−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:In2O3, Pr:In2O3, and Tb:In2O3 films stimulated by high-intensity 450-nm light. Only Pr:In2O3 shows a weak red-light peak centered at ~625 nm. So it can be deduced that the absorbed incident light is converted to lattice vibration for the Ce:In2O3 and Tb:In2O3 films, and a small part of the absorbed incident light is converted to red light for the Pr:In2O3 film. It should be noted that the Ln4+ 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 Ln4+, 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 PrOx and TbOx 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 Pr4+ or Tb 4+ that enhances the NBIS stability greatly.

Fig. 8: The schematic diagram of charge transfer transition for Ce4+, Tb4+, and Pr4+.
figure 8

The hole in O2p−1 will pull the transited electron back to its original space by a nonradiative way.

Full size image

It should be noted that the charge transfer transition energies of the Ln3+ → Ln2+ is much higher than those of Ln4+ → Ln3+, as shown in Fig. 9a, so the Ln3+ → Ln2+ transition cannot absorb blue light. As a result, the Ln3+ ions cannot improve the photostability of the TFTs.

Fig. 9: The energy of transition.
figure 9

a, b The energy required for (a) charge transfer and (b) f-d transition of Ln3+ and Ln 4+ ions.

Full size image

Besides tetravalent ions, trivalent ions also exist in the CeOx, PrOx, and TbOx films. That means 4f-5d transition is likely to take place, because the f-d transition energies (Efd) for Ce3+, Pr3+, and Tb3+ ions are much lower than those for other free Ln3+ ions (see Supplementary Table 1). It is known that the Efd 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 106 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 Efd for Ce3+ is lower than that for Pr3+ or Tb3+, and the light absorption spectrum of the CeOx 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 CeOx, PrOx, and TbOx films should be greater than 3.18 eV (<390 nm). It means that the visible light absorption of the PrOx and TbOx 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:In2O3 or Tb:In2O3 semiconductor layers.

It is worth noting that the bivalent Ln ions (Ln2+) have lower Efd than the trivalent Ln ions (Ln3+). And the Efd of some Ln2+ ions are close to the energy of blue light. For example, the Efd for Eu2+, Yb2+, Sm2+, and Tm2+ are 4.30, 4.20, 2.92, and 2.87 eV, respectively (see Fig. 9b). However, only Eu2+ and Yb2+ are stable in the solid state. Moreover, the Ln:In2O3 films must experience high-temperature annealing (>300 °C), so the Ln2+ ions will be oxidized to Ln3+ ions. Thus, it is difficult to achieve stable AOS TFTs doped with Ln2+ ions.

It is worth noting that the Pr:In2O3 and Tb:In2O3 TFTs exhibit worsened stability under positive bias illumination stress (PBIS) with ΔVon of 1.1 and 0.9 V, respectively (see Supplementary Fig. 12). The degradation of the PBIS for Pr:In2O3 and Tb:In2O3 TFTs is ascribed to the light-induced shallow electron traps (possibly Pr3+ and Tb3+ related defects).

Supplementary Figures 13 and 14 show the stability under negative bias temperature stress (NBTS) and positive bias temperature stress (PBTS) of different TFTs at 60 °C without light illumination, respectively. The ΔVon for all the TFTs under NBTS, PBTS, and PBIS are summarized in Supplementary Table 2. The pure In2O3 TFT exhibits poor NBTS stability, while all of the Ln:In2O3 TFTs have good NBTS stability. In contrary to the NBTS stability, the PBTS stability of the pure In2O3 TFT is good (see Supplementary Fig. 14), which may be attributed to the much higher free electron density than the electron traps in the In2O3 film. However, most Ln:In2O3 TFTs show relatively poor PBTS (especially the La, Nd, and Ho incorporated In2O3 TFTs). It reveals that incorporating Ln to In2O3 will induce shallow electron traps in the In2O3 films. The temperature effect of the Ln:In2O3 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. Compared to the undoped In2O3 TFT, which has a turn-on voltage (Von) of −13.3 V, the lanthanide-doped In2O3 (Ln:In2O3) TFTs have more positive Von that range from −1 to 1 V and more steeper SS, revealing that Ln doping with a concentration of only 5 at.% can suppress oxygen vacancy (Vo) generation and decrease free carrier concentration substantially. More interestingly, the NBIS stability of the TFTs are improved greatly when the In2O3 AOS is doped by praseodymium (Pr) or terbium (Tb). The Pr:In2O3 and Tb:In2O3 TFTs exhibit the same Von shift (ΔVon) of −3.0 V, compared to a ΔVon of as high as −12.0 V of the undoped In2O3 TFT. The NBIS stability of the In2O3 TFTs doped with the other lanthanides are not good, with ΔVon at the range of −7.9 to −15.6 V. The mobility of the solution-processed Pr:In2O3 and Tb:In2O3 TFTs are 12.2 and 12.7 cm2 V−1 s−1, respectively. 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 Pr4+ or Tb4+ 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

Methods

Materials

The mate oxide semiconductor films were deposited by spin-coating and thermal decomposition of precursors solution. A 0.2 M In2O3 solution was prepared by dissolving indium nitrate hydrate ((In(NO)3·nH2O), Sigma-Aldrich) in deionized water. The Ln: In2O3 precursors solutions were synthesized by dissolving indium nitrate hydrate ((In(NO)3·nH2O), Sigma-Aldrich) and lanthanides nitrate hydrate ((La(NO)3·nH2O), (Ce(NO)3·nH2O), (Pr(NO)3·nH2O), Nd((NO)3·nH2O), (Sm(NO)3·nH2O), (Eu(NO)3·nH2O), (Gd(NO)3·nH2O), (Tb(NO)3·nH2O), (Dy(NO)3·nH2O), (Ho(NO)3·nH2O), (Er(NO)3·nH2O), (Tm(NO)3·nH2O), (Yb(NO)3·nH2O), (Lu(NO)3·nH2O), aladdin) in deionized water, which was optimized to the total concentration of metal ion of 0.2 M and In/Ln molar ratio of 19:1. All the precursor solutions were stirred vigorously for 12 h at room temperature and filtered through a 0.22 μm syringe filter before spin-coating.

Devices fabrication

A bottom-gate and top-contact source-drain electrode structure was used to fabricate Ln: In2O3 TFT. Firstly a 200-nm thickness Al2O3: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 Al2O3:Nd surface with a stencil shadow mask, for the formation of a hydrophilic surface in the channel area. The Ln: In2O3 precursor films were deposited in the wettable area irradiated by UV by spin-coating Ln: In2O3 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: In2O3 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 In2O3 and Ln: In2O3 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 In2O3 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:

$${I}_{{Ds}}=\frac{W\mu {C}_{i}}{2L}{({V}_{{GS}}-{V}_{{th}})}^{2}$$
(3)

where μ is the field-effect mobility, Ci is the areal capacitance per area of the gate dielectric, Vth is the threshold voltage, W is the channel width, and L is the channel length.