Rapid photo-assisted activation and enhancement of solution-processed InZnO thin-film transistors

Solution process fabrication of amorphous oxide semiconductors (AOS) offers attractive advantages such as simplicity and versatility over traditional vacuum-based fabrication methods. In particular, solution-processed AOS channel materials in thin-film transistor (TFT) will have important applications in ubiquitous and mass-produced technologies while not being cost-prohibitive [1–6]. For truly ubiquitous devices, deposition on any surface is especially appealing. However, this becomes challenging when depositing on large area and flexible substrates since not all deposition methods can scale well and most flexible substrates are sensitive to high temperatures. Although AOS can be deposited for large-area applications through solution process, several major issues still persist for solution-processed oxide TFTs such as having typically poor characteristics (<1 cm² V⁻¹ s⁻¹) compared with their vacuum-processed counterparts which necessitates additional annealing/activation processes to boost the performance. Furthermore, these processes are usually lengthy (>1 h) and performed at high temperatures (>400 °C) [4–6]. Thus, many researchers have been aggressively lowering the post-annealing temperature of solution AOS through microwave annealing (<300 °C), plasma treatment (<300 °C), ultraviolet (UV) with ozone treatment (<300 °C), combustion synthesis (<250 °C), high pressure annealing (<250 °C), water-based/hydrolysis route (<250 °C), low temperature steam annealing (<200 °C), photochemical activation (<150 °C), among others [4–15].

Nevertheless, there are still insufficient studies focused on drastically shortening process time to instantaneous periods (<1 s) while maintaining low temperatures for post-annealing activation of solution AOS channel. A handful of studies have demonstrated large process time reduction by employing laser irradiation and combining microwave annealing with photoirradiation [16–18]. Though successful, performance improvement is limited to 1.5 cm² V⁻¹ s⁻¹ and process time is still in the order of few minutes—requiring multiple shots especially for laser irradiation. Ultimately, aside from decreasing the activation process time, understanding the mechanism why channel activation is successful even when employing an instantaneous process is also critical.

In this work, we used a rapid (<100 ns) and low temperature (room temperature, RT) activation method, a single shot of KrF excimer laser annealing (ELA), to enhance the electrical characteristics of solution-processed AOS TFT—amorphous InZnO (a-IZO) TFT. A single laser shot ensures a rapid process (<100 ns) and that high temperatures will be localized near the channel. Thus, substrate temperature will be nearly RT, and performance improvement is likely as we have previously reported for vacuum-processed AOS TFTs especially IZO TFTs subjected to a higher wavelength XeCl ELA [19–22]. However, improving the performance of solution-processed films is not as straightforward compared to vacuum-processed films because as-deposited solution films generally contain more voids and impurities. We show that by irradiating a-IZO TFTs with a single shot of a shorter wavelength KrF ELA, the mobility (µ) is improved by up to ~20 times its original value, the on-current is increased, and subthreshold degradation is suppressed. Through surface and bulk characterization techniques, we elucidated the improvement mechanism of solution a-IZO TFTs after ELA. The analysis and discussion presented here is vital for developing rapid and low temperature fabrication of solution-processed oxide semiconductor TFTs.

A 50 nm-thick a-IZO (In:Zn  =  77:23) channel was deposited by solution process via subsequent spin-coating of five layers of a-IZO precursor at a main spin of 3000 rpm for 30 s. The spin-coating deposition was done on Si substrate with a 100 nm thermal oxide SiO₂. The Si (resistivity  <  0.002 Ω·cm) and SiO₂ were used as gate and gate insulator, respectively. UV ozone treatment at 115 °C for 10 min was performed on the SiO₂/substrate prior to the IZO channel deposition. Each IZO layer was prebaked at 150 °C for 5 min and post-baked at 300 °C for 1 h after the 5th IZO layer deposition. The channel was then patterned through photolithography and wet etched by 0.02 mol l⁻¹ HCl solution. A stack of Mo/Pt (80/20 nm) source/drain electrodes were deposited by radio frequency magnetron sputtering and patterned using a lift-off technique. The final TFT structure is a bottom-gate top-contact structure shown in figure 1(a).

Zoom In Zoom Out Reset image size

Instead of a customary 2 h of 300 °C post-annealing treatment [23], TFTs were irradiated with a single shot of KrF excimer laser (λ  =  248 nm; with optical pulse stretcher; temporal profile shown in figure 1(b)) in either ambient or vacuum atmosphere at room temperature and pressure. For the vacuum condition, TFTs were placed in a vacuum-sealed chamber and the pressure was reduced to  <10⁻³ Pa before being subsequently irradiated. Electrical properties were analyzed by measuring transfer characteristics before and after ELA using a semiconductor parameter analyzer (Agilent 4156C). Additional film samples, IZO/SiO₂/Si were also fabricated and exposed to ELA to analyze the effect of ELA on the optical properties, elemental concentration, physical properties, and chemical bonding. Characterizations such as spectroscopic ellipsometry (Horiba UVISEL), secondary ion mass spectrometry (SIMS, ULVAC-PHI ADEPT-1010), x-ray diffraction (XRD, Rigaku RINT-TTRIII/NM), and x-ray photoelectron spectroscopy (XPS, ULVAC-PHI Versa Probe II) were performed to elucidate the TFT performance enhancement mechanism.

Figure 2 compares the transfer characteristics of solution-processed IZO TFT before and after KrF ELA at 60 mJ cm⁻² and 80 mJ cm⁻² in air and vacuum atmosphere. Before KrF ELA, smaller on-current, significant subthreshold degradation, and a large hump can be observed in the transfer characteristics. These TFTs exhibit inferior electrical performance with µ of 0.15  ±  0.06 cm² V⁻¹ s⁻¹ at a drain voltage (V_ds) of 0.1 V. The average and standard deviation values were obtained by measuring multiple TFTs at different locations on the same sample. Other average electrical parameters were also mediocre including a large negative turn-on voltage (V_on) of  −6.2 V, large subthreshold swing (S) of 0.82 V/dec, and low on-current of ~1.4  ×  10⁻⁸ A (see table 1 for the summary of the electrical characteristics at V_ds  =  0.1 V). Poor characteristics are expected for IZO TFTs before ELA since additional post-annealing treatment was not performed. This post-annealing treatment is vital especially for solution-processed TFTs since solution-processed channel materials generally have higher amount of inherent defects and residual impurities originating from the precursor. The typical post-annealing treatment is performed at high temperatures (>400 °C) and long time (>2 h) to create defect/impurity-free films and improve the interface with source/drain electrodes.

Zoom In Zoom Out Reset image size

Table 1. Summary of electrical characteristics at V_ds  =  0.1 V before, after ELA at 60 mJ cm⁻², 80 mJ cm⁻² in air, 80 mJ cm⁻² in vacuum and after 300 °C 2 h of conventional furnace annealing.

Sample Vds  =  0.1 V	µ (cm2 V−1 s−1)	Von (V)	S (V dec−1)	On-current (A)	On/off ratio
Before ELA	0.15  ±  0.06	−6.2  ±  2.3	0.81  ±  0.17	1.4  ±  1.6  ×  10−8	~1.4  ×  10−4
60 mJ cm−2 air	0.98  ±  0.27	−2.6  ±  1.3	0.38  ±  0.07	3.8  ±  0.9  ×  10−7	~3.8  ×  10−4
80 mJ cm−2 air	1.66  ±  0.96	−2.9  ±  1.0	0.32  ±  0.06	6.7  ±  3.3  ×  10−7	~6.7  ×  10−4
80 mJ cm−2 vacuum	3.30  ±  0.55	−3.1  ±  1.2	0.37  ±  0.05	1.3  ±  0.3  ×  10−6	~1.3  ×  10−5
300 °C annealing	2.98  ±  0.14	1.6  ±  0.3	0.33  ±  0.02	1.2  ±  0.1  ×  10−6	~1.2  ×  10−5

To improve the performance of solution-processed IZO TFTs, KrF ELA was performed at a fluence (F) of 60 mJ cm⁻² and 80 mJ cm⁻² instead of the conventional high temperature and lengthy post-annealing process. Transfer characteristics shown in figures 2(b)–(d) demonstrate how ELA can induce superior performance, such as greatly increased on-current and suppressed subthreshold degradation. The enhancement was also dependent on F. At 60 mJ cm⁻², the µ improved from 0.15  ±  0.06 cm² V⁻¹ s⁻¹ to 0.98  ±  0.27 cm² V⁻¹ s⁻¹. Other electrical parameters such as V_on, S, and on-current were also enhanced to  −2.6 V, 0.38 V dec⁻¹, and 3.8  ×  10⁻⁷ A, respectively. Despite the µ increase at 60 mJ cm⁻², the performance is still non-ideal since higher µ greater than 1 cm² V⁻¹ s⁻¹ is preferred. To achieve this, F was increased to 80 mJ cm⁻² which yielded better µ enhancement to 1.66  ±  0.96 cm² V⁻¹ s⁻¹. The elevated F also induced a higher on-current of 6.7  ×  10⁻⁷ A, V_on of  −2.9 V, and an improved S of 0.32 V dec⁻¹. When performing ELA at 80 mJ cm⁻² in vacuum (<10⁻³ Pa) instead of ambient atmosphere, both the µ and on-current were further boosted to 3.30  ±  0.55 cm² V⁻¹ s⁻¹ and 1.3  ×  10⁻⁶ A, respectively, with the on-current being two orders of magnitude higher than the TFT on-current before ELA. Consistent with other irradiated samples, both the V_on (−2.9 V) and S (0.37 V dec⁻¹) were also superior to the non-irradiated TFTs. Irradiation in vacuum helped inhibit incorporation and adsorption of additional moisture and excess oxygen from the atmosphere.

The massive improvement of a-IZO TFTs subjected to ELA is highlighted in the enhancement of the output characteristics shown in figure S2 (supplementary section (stacks.iop.org/JPhysD/53/045102/mmedia)). Before ELA, minimal drain currents of  <300 pA are achieved when varying gate voltage from 0 to 10 V. Performing ELA at 60 mJ cm⁻² and 80 mJ cm⁻² enhanced the drain currents to ~1 µA and ~5 µA, respectively. Clear transition from linear to saturation behavior in the output characteristics is also observed after ELA. In addition, no contact resistance from the output characteristics is observed after ELA indicating that laser-induced heating did not induce damage at the source/drain-channel interface.

For comparison, the transfer and output characteristics of a-IZO TFT subjected to a standard 300 °C 2 h furnace annealing is shown in figure S3 (supplementary section) and the average electrical characteristics at V_ds  =  0.1 V are also included in table 1. The 80 mJ cm⁻² vacuum sample still showed a higher µ (3.3 cm² V⁻¹ s⁻¹), slightly better on-current (1.3  ±  0.3  ×  10⁻⁶ A) and on/off ratio (~1.3  ×  10⁻⁵) compared to a-IZO TFT activated through furnace annealing at 300 °C for 2 h which had µ, on-current, and on/off ratio of 2.98 cm² V⁻¹ s⁻¹, 1.2  ±  0.1  ×  10⁻⁶ A, ~1.2  ×  10⁻⁵. The ELA sample achieved better electrical characteristics even when using an instantaneous process that takes  <100 ns compared to the lengthy 2 h annealing. The drastic reduction in process time is not the sole advantage of ELA. In this photo-activated process, high temperatures are localized on the layer that requires heating—the channel. Thus, the substrate temperature can be lower enabling the usage of flexible substrates. On the opposite end, heating during furnace annealing is not selective and all layers, even layers which do not need high temperature annealing, are subjected to high temperatures which limits the use of flexible substrates.

For ELA, utilizing an optical pulse stretcher (OPS) is also an important factor since performing ELA without OPS yielded µ improvement but the resulting µ was insufficient (<1 cm² V⁻¹ s⁻¹). The OPS extends the excimer laser pulse duration by stretching the temporal profile to prevent ablation and generate longer and more efficient laser-induced heating [24]. Employing a single ELA shot is also vital to not only reduce process complexity and process time but also to minimize laser-induced damage and easily control the total laser energy imparted to the film. With a single shot, irradiating at even higher F (⩾100 mJ cm⁻²) yields conductive films not suitable for channel applications but through clever exploitation of the laser's precise location control will have very interesting applications in all solution-processed TFTs [25, 26].

Figure 3 shows the 2D COMSOL multiphysics simulation we performed to confirm high temperature localization, laser-induced heating duration, and determine temperatures at different interfaces. It is evident from figure 3(a) that high temperature heating is solely localized in the upper layers with the highest heating occurring near the source/drain regions because of their higher thermal conductivity compared with the channel and gate insulator. Moreover, the simulation shows that laser-induced heating penetrates throughout the IZO layer. Figure 3(b) plots the maximum temperature over the entire pulse duration and reveals that majority of the high temperature heating occurs in  <100 ns. More interestingly, this indicates that substrate temperature barely increases from RT over the laser duration. This bodes well for utilizing KrF ELA on TFTs employing any flexible/stretchable/rigid substrate. As expected, the comparison in figure 3(c) shows that laser-induced heating is higher at 80 mJ cm⁻². At the IZO near the source/drain region, ELA at 60 mJ cm⁻² induces a temperature of ~358 °C while it is ~476 °C for 80 mJ cm⁻² which implies a difference of 118°. The induced temperature difference is 63° at the IZO center region. The lower induced temperatures likely explains why 60 mJ cm⁻² caused a more modest µ improvement compared to 80 mJ cm⁻². Nevertheless, both cases enhanced the channel's contact with the source/drain by heating the channel/electrode interface to at least ~631 K (~358 °C).

Zoom In Zoom Out Reset image size

To check if crystallization occurred after ELA, we performed XRD measurements on a-IZO film before and after ELA as shown in figure S4 (supplementary section). We performed the XRD measurements at a grazing incidence angle of 0.4 to enhance the signal from the thin 50 nm IZO film and minimize the peaks from the Si substrate that can be found at 50–52°. From figure S4, a-IZO films before ELA is amorphous as shown by the broad peak at ~32°. All samples subjected to ELA at 60 mJ cm⁻² and 80 mJ cm⁻² also show amorphous nature as shown by similar broad peaks with samples before ELA. However, there is observed peak narrowing at ~32° for the 80 mJ cm⁻² air samples which likely signifies a transition between amorphous and crystalline state that will proceed to complete crystallization if fluence is further increased. We found that the solution-processed a-IZO film is fully crystallized after ELA at 120 mJ cm⁻².

We performed additional characterization to determine the improvement mechanism by investigating the changes in elemental concentration, optical properties, and chemical bonding in solution-processed IZO after ELA. From the SIMS measurement analysis in figure 4, a higher H amount is present in the IZO bulk of irradiated samples. An additional amount of H—considered a shallow donor in AOS [27] which increases their carrier concentration can induce a better µ and on-current for irradiated samples. Furthermore, H when incorporated in AOS such as IZO can also passivate traps such as oxygen vacancies (V_O) and form H_O (H at a V_O site) which further improved the device performance of irradiated TFTs [28–31]. Nevertheless, the incremental rise in the H amount solely on its own cannot sufficiently account for the performance enhancement after ELA. We also checked the elemental concentration of other elements such as carbon which illustrates that there is a slight reduction of carbon impurities in the IZO bulk for ELA samples as shown in figure 4(b). Impurity reduction is paramount in solution-processed channel films since these impurities which is usually precursor-derived act as traps and hinder the efficient formation of the metal oxide framework which function as carrier pathways. This problem is not as pronounced in vacuum-processed films since these do not typically need additional precursors to facilitate efficient oxide film formation.

Zoom In Zoom Out Reset image size

We also performed an investigation of the optical properties such as the band gap (E_g) and Urbach energy (E_u) to further elucidate the underlying mechanism. Figure S6 (supplementary section) shows that irradiation at 60 mJ cm⁻² barely altered the E_g of solution-processed IZO film. However, depending on the irradiation atmosphere, E_g can be lowered by up to 0.7 eV, from 3.2 eV before ELA to 2.5 eV after irradiation at a higher F of 80 mJ cm⁻² in vacuum. More importantly, we also examined the changes in E_u after irradiation since the subgap disorder in IZO, which is a combination of the tail states of both conduction and valence bands, can be estimated by E_u. In AOS, the contribution of the valence band tail states is more dominant than conduction band tail states—making E_u a better indicator of valence band tail states [32, 33]. These tail states heavily influence the electrical characteristics especially the on-current and subthreshold region [34]. From figure S7 (supplementary section), there is a significant reduction in E_u after KrF ELA especially when performed at a higher F of 80 mJ cm⁻² compared with both the before ELA and 60 mJ cm⁻² cases. Considering that samples with lower E_u had higher on currents and suppressed subthreshold degradation, these results confirm that ELA especially at higher F is effective in decreasing the amount of subgap trap states. The S values which also describes the quality of the semiconductor channel and indicates the amount of subgap traps in the channel also agree well with the E_u results. Before ELA, the IZO TFT had a large S of 0.81 V dec⁻¹ which were more than halved up to 0.32 V dec⁻¹ when performing ELA at 80 mJ cm⁻² indicating a drastic reduction of subgap traps. Carrier transport is another aspect where tail states perform an important role since these represent the structural order and randomness that contribute to potential barriers inhibiting efficient carrier transport [35]. Therefore, the E_u decrease after ELA, which implies reduced tail states and better structural organization, will not only lessen subgap traps but also diminish potential barriers. This reduction leads to an enhanced carrier transport as reflected by the better performance, higher µ, and increased on-current exhibited by solution IZO TFTs after ELA.

To further corroborate the mechanisms that were earlier discussed, we compared the bonding changes in IZO before and after ELA through XPS. Figure 5 compares the O1s core level spectra for IZO before and after ELA at 80 mJ cm⁻² in both the film's surface and bulk. The O1s spectra can be further deconvoluted to oxygen in the metal oxide lattice (O_M), oxygen-deficient regions (O_def), hydroxyl bonds/carbon-related impurities (O_H), and precursor-related impurities (O_imp) with peak assignments at 529.5  ±  0.1, 531.1  ±  0.1, 532.1  ±  0.1, and 533.0  ±  0.1 eV, respectively [13, 36, 37]. At the surface, both IZO films have similar metal oxide bonding with IZO subjected to ELA having slightly lower V_O amount. However, the IZO film before ELA contains some precursor-related impurities which degraded the electrical characteristics as indicated by the sub-peak at ~533 eV. Usually, a high temperature post-annealing is performed to reduce these precursor-related impurities which is why IZO films after ELA have reduced impurities as shown by the absence of the 533 eV subpeak. Comparing the O1s XPS in the bulk, IZO after ELA comprise a lower amount of V_O traps and hydroxyl/carbon-related impurities which are both supported by the SIMS profiles of H and C. The higher H amount in the bulk helped reduce V_O and the carbon-related impurities were lessened by KrF ELA. Thus, ELA promotes the formation of metal oxide framework as seen by the higher amount of O_M in after ELA samples. A higher amount of metal oxide bonds are preferred for the channel since a sizeable amount of carrier pathways are essential to achieve high performance. These mechanisms are also observed in non-laser photo-assisted methods where high energy photons induce photochemical cleavage of metal alkoxy groups and disordered networks and promote subsequent reorganization of the metal oxide framework [15]. The difference of ELA with other photo-assisted methods that take longer is that ELA combines these photochemical effects with laser-induced high temperature heating which provides additional localized thermal energy that efficiently decomposes precursor-related impurities, remove metal oxide defects, and simultaneously reorders the metal oxide structure at a lower substrate temperature (RT) in an instant (<100 ns). However, an important yet challenging next step is to use rapid photo-assisted methods to jointly enhance both the performance and stability of solution-processed oxide TFTs.

Zoom In Zoom Out Reset image size

Figure 6 summarizes the difference between IZO films before and after ELA. Before irradiation, the film includes many precursor-related impurities (represented as R and H₂O in figure 6) and V_O traps which hinder the formation of an efficient metal oxide framework. Consequently, the IZO TFT before ELA exhibited a lower on-current, poor performance, and had subthreshold degradation. After ELA, subgap traps, V_O, and precursor-related impurities are decreased that facilitated the subthreshold degradation suppression and performance increase. In addition, more efficient carrier transport is achieved due to greater number of carrier pathways, as illustrated by having more metal-oxide bonds, and diminished potential barriers contributing to higher µ and on-current which are indispensable for achieving high performance.

Zoom In Zoom Out Reset image size

In this work, we demonstrated a rapid (<100 ns) and low temperature (~RT at substrate) post-annealing process, KrF ELA, to significantly enhance the performance of solution-processed IZO TFT. Through this method, subthreshold degradation suppression, higher on-current, and mobility enhancement of up to 20 times can be achieved. Comprehensive analysis shows that ELA induces performance improvement due to a combination of reduction of precursor-related impurities, subgap traps, and tail states, and the efficient formation of a robust metal oxide framework. KrF ELA is very promising for high throughput fabrication of solution-processed devices on flexible substrates.

The authors thank Nissan Chemical Corporation for providing the IZO precursors.