Introduction

In 2004, amorphous indium-gallium-zinc-oxide (a-IGZO) was discovered by the group of Prof. Hosono, offering numerous outstanding characteristics such as reasonable field-effect mobility (μFE) > 10 cm2/Vs, extremely low off-current < 10−24 A/μm, steep subthreshold swing (SS) of ~ 0.1 V/dec, and outstanding uniformity even when fabricated at low temperature1,2,3,4,5. Because of these merits, oxide semiconductor (OS) family has been studied intensively and has become a standard channel material of thin-film transistors (TFTs) in high-end active matrix organic light emitting diode (AMOLED) display backplanes. Recently, the OS has gained more attention as a channel candidate for next-generation semiconductor device applications such as 2T0C dynamic random access memory (DRAM)6,7,8,9, because it has potential to overcome the scaling and leakage problems of DRAM technologies due to their ultralow off-current originating from the wide bandgap (EG) nature. Furthermore, their back-end-of-line (BEOL)-compatible low-temperature processibility enables movement of the DRAM peripheral circuitry under the memory array, achieving 3D DRAM technology.

Solid state devices have been scaled down to improve power, performance, area, and cost (PPAC) in the semiconductor industry2. O/S TFTs also must follow this conventional PPAC rule to meet the abovementioned requirements. For this reason, increasing numbers of studies on scaled oxide TFTs have been reported2. Here, it is important to note that the impact of contact resistance in nanoscale semiconductor devices increases considerably with miniaturization. However, the contact resistance of OS TFTs is generally four orders of magnitude higher than that of the current silicon (Si) metal oxide semiconductor field-effect transistors (MOSFETs)10,11,12. This high contact resistance of OS TFTs can be a critical obstacle for miniaturized device applications13. Thus, there is need to improve the electrical contact properties between OS channels and source/drain (S/D) electrodes.

To improve the contact resistance at the channel/electrode interface, both Schottky barrier height and width must be reduced to facilitate carrier injection through the barrier formed at the interface14,15,16. The height is controlled depending on a work function of contact electrode, which can reduce the contact resistance17. The interface-state defects at the channel/electrode interface should be minimized to increase the height control18. The width determined by carrier density (nc) of the channel material can be narrowed by several approaches19,20,21,22,23,24,25,26,27,28,29,30,31,32,33. Increasing nc is the simplest method to reduce barrier width, which was originally devised for the Si MOSFETs, and can be realized through ion implantation and/or plasma treatments using elements such as boron, fluorine, argon, or hydrogen19,20,21,29,30. However, these methods can damage the channel layer below the S/D electrodes, which increases the interface defects and offsets the merits of increasing nc. In addition to these doping techniques, other approaches such as controlling cation composition, metal-induced oxygen scavenging and highly conductive interlayer (IL) insertion have been employed22,23,24,25,26,27,34,35. These methods have an advantage in that they can reduce the barrier width without incurring damage.

In this study, a multi-stack IL using titanium nitride (TiN) and indium-gallium-tin-oxide (IGTO) is inserted between IGZO channel and indium tin oxide (ITO) S/D electrodes. Specific contact resistivity (ρC) of IGZO TFTs with a 3-nm-thick TiN IL is reduced to 5.6 × 10−5 Ω·cm2, which is almost 20-fold lower than that of IGZO TFTs without the IL. More importantly, the electrical contact properties have IGTO thickness (tIGTO)-dependent behavior, decreasing with tIGTO up to 8 nm. It is degraded in the IGZO TFTs with tIGTO greater than 8 nm. Consequently, the device with a TiN/IGTO (3/8 nm) IL has the lowest ρC and width-normalized contact resistance (RCW) of 9.0 × 10−6 Ω·cm2 and 0.7 Ω·cm, which also leads to the largest increase in μFE from 39.9 to 45.0 cm2/Vs compared to the device without the IL. This noticeable improvement in the electrical contact properties could be attributed to enhanced electron injection by reduced Schottky barrier height (SBH) and width through the TiN/IGTO IL insertion, which will be discussed in depth.

Methods

Device fabrication

Transmission line method (TLM) patterns and bottom gate IGZO TFTs were fabricated with top-contact configuration (Fig. 1a,b). 15-nm-thick IGZO thin-films were deposited through plasma-enhanced atomic layer deposition (PEALD) using dimethylbutylamino(trimethylindium) (DATI), trimethylgallium (TMG), diethylzinc (DEZ), and O2 plasma as In, Ga, Zn and oxygen sources, respectively, at 150 °C. Also, 100-nm-thick silicon oxide (SiO2) and heavily doped p+-Si substrates were used as gate dielectric and electrode, respectively. The deposited channel layers were patterned by standard photolithography and wet etching. The TiN and IGTO IL were deposited through direct current (DC) magnetron sputtering at room temperature, followed by in-situ deposition of highly conducting indium tin oxide (ITO) thin-films for S/D electrodes. Then, the deposited S/D structure with the multi-stack IL was patterned using the lift-off method. Finally, the devices were annealed at 500 °C under ambient air for 1 h. The fabricated TLM pattern devices have a channel width (W) of 60 μm and lengths (L) of 20, 30, 40, and 50 μm. The TFTs has W/L of 60/30 μm.

Figure 1
figure 1

Schematic structures of the fabricated devices: (a) The TLM device; (b) the IGZO TFT; (c) device fabrication procedure.

Full size image

Materials and device characterization

The film thickness was determined through spectroscopic ellipsometry (Elli-SE, Ellipso Technology), and nc was obtained through Hall effect measurements (HMS-5000, Ecopia) with a van der Pauw configuration. Surface morphologies of IGTO/IGZO thin-film stacks were characterized using atomic force microscopy (XE-100, Park System Co.). The crystalline structure of the IGTO thin-film was confirmed through grazing incidence X-ray diffraction (Model X’Pert PRO MRD, Malvern Panalytical) and high-resolution transmission electron microscopy (NEOARM, JEOL). Chemical states were determined through X-ray photoelectron spectroscopy (K-Alpha+, Thermo Fisher Scientific Co.) after in-situ surface (~ 3 nm) etching using an accelerated (1 kV) Ar+ ion to avoid confusion by possible contamination and/or oxidation during air exposure.

Electrical characteristics of the fabricated TFTs were measured through a Keithley 2636 source meter at room temperature under dark conditions. The μFE was calculated using the equation \( {\mu_{FE}} = \frac{{L{g_m}}}{{W{C_{OX}}{V_{DS}}}}\)31, where COX, VDS, and gm are the capacitance per unit area, drain-to-source voltage, and transconductance, respectively. In this study, VDS is 0.1 V. Threshold voltage (VTH) was extracted using the constant current method, which was defined as the gate-to-source voltage (VGS) inducing a drain current (ID) of L/W × 10 nA31. Subthreshold swing (SS) was obtained using the equation \( SS = \frac{{d{V_{GS}}}}{{d\log {I_D}}}\) at the subthreshold region of transfer characteristics31.

Results and discussion

To verify the effects of TiN and IGTO IL on electrical contact properties in the IGZO TFT, TLM was conducted. Figure 2 shows total resistance (RT) versus L graphs as a function of gate voltage (VGS) in the IGZO TLM devices with different multi-stack ILs. ρC was calculated using the following equations:14

$$ {R_T} = {R_{ch}} + 2{R_C} = \frac{L - 2\Delta L}{{W{\mu_{FE}}{C_{OX}}\left( {{V_{GS}} - {V_{TH}}} \right)}} + 2{R_C} $$
(1)

where Rch, RC, and ∆L are the channel resistance, the contact resistance, and the change in channel length, respectively. The values of 2RC and 2∆L can be obtained at the intersection of the RT and L curves (Fig. 2)31. More importantly, the RCW can be expressed as36

$$ {R_C}W = \sqrt {{\rho_C}{R_{SH}}} \;\coth \left( {\frac{{L_C}}{{L_T}}} \right) \approx \sqrt {{\rho_C}{R_{SH}}} ,\;\;{\text{if}}\;\;\;{L_C} \gg {L_T} $$
(2)

where RSH, LC, and LT are the sheet resistance of the channel, the physical contact length (30 μm in this study), and the current transfer length, respectively. This equation can be rewritten as

$$ {\rho_C} = \frac{{{{\left( {{R_C}W} \right)}^2}}}{{{R_{sh}}}} $$
(3)
Figure 2
figure 2

VGS-dependent RT variations of the IGZO TLM devices with different multi-stack IL at VDS of 0.1 V: (a) Without IL; (b) 3-nm-thick TiN IL; (c) 3-/5-nm-thick TiN/IGTO IL; (d) 3-/8-nm-thick TiN/IGTO IL; (e) 3-/12-nm-thick TiN/IGTO IL.

Full size image

As such, ρC can be obtained using Eq. (3). It is also important to note that the RCW, an intuitive figure of merit to demonstrate the electrical contact properties, is simultaneously observed along with the ρC throughout this study. The control device without the IL has the largest ρC (RCW) of 8.0 × 10−4 Ω·cm2 (6.9 Ω·cm). Insertion of a 3-nm-thick TiN IL reduces this to 5.6 × 10−5 Ω·cm2 (1.8 Ω·cm), respectively, which indicates that the TiN IL improves the electrical contact properties. This improvement could be attributed to increase in conductivity of the underlying layer through the oxygen scavenging effect of the TiN. Combination of IGTO thin-film with the TiN IL further improves the electrical contact. Of particular interest is tIGTO-dependent behavior, where ρC (RCW) values of 1.4 × 10−5 (0.9), 9.0 × 10−6 (0.7), and 3.0 × 10−5 Ω·cm2 (1.3 Ω·cm) were obtained in the devices using a TiN/IGTO multi-stack IL with tIGTO of 5, 8, and 12 nm, respectively. The device with the TiN/IGTO IL (3/8 nm) has the lowest ρC of 9.0 × 10−6 Ω·cm2, which is almost two orders of magnitude lower than that of the control device (Fig. S1). The RCW also has the same tIGTO dependence which has the smallest value of 0.7 Ω·cm. It is noteworthy that the both of the ρC and RCW are reduced to 1.1 × 10−5 Ω·cm2 and 1.0 Ω·cm in the device featuring an 8-nm-thick IGTO IL (Fig. S2). This outcome suggest that the IGTO IL alone can enhance the electrical contact properties to a certain degree, even without the presence of TiN. However, it simultaneously indicates that the TiN/IGTO multi-stack IL holds greater influence. Additionally, it is worth noting that the device annealed at 400 °C with the TiN/IGTO IL (3/8 nm) demonstrates electrical contact properties comparable to those of the device with the same IL stack annealed at 500 °C (Fig. S3). These findings collectively imply that insertion of a multi-stack IL is not only an effective but also thermally stable method for enhancing the electrical contact properties of IGZO TFT.

Before comprehending the effect of tIGTO on the electrical contact properties, it is quite important to discuss the implications of the intersection in RT versus L graphs. In oxide TFTs, the ∆L generally originates from oxygen vacancy (VO)’s diffusion, which occurs more noticeably in the OS containing high VO concentration. Here, IGZO TFTs with an IL can have higher VO concentration at the contact region than those without an IL due to a synergetic effect of TiN and IGTO layers where the former induces oxygen scavenging effect from the underlying oxide layer and the latter has high VO concentration. It leads to the higher nc, which reduces the Schottky barrier width and improves the electron injection from the source electrode. More importantly, this reduced barrier width makes the electron injection independent on VGS. For this reason, the IGZO TFTs with an IL can have a VGS-independent intersection, i.e., VGS-independent RC, in the RT versus L graphs. Meanwhile, there could not be an intersection, as shown in the RT versus L graph of IGZO TFTs without an IL. In this case, the RC as well as LT becomes dependent on the VGS. It could be because the nc under the S/D electrode is affected by the VGS due to the relatively low VO concentration at the contact region.

To understand the effect of multi-stack IL, the ultraviolet photoelectron spectroscopy (UPS) depth profile and UV/visible spectroscopy was conducted for two cases: (1) the ITO/IGZO thin-film stack; (2) ITO/TiN/IGTO/IGZO (3-/8-nm-thick TiN/IGTO IL) thin-film stack. Work functions and differences between the Fermi-level and the valence band edge (EF–EV) were obtained by the UPS (Fig. S4). EG was extracted through the UV/visible spectroscopy. It was confirmed that both of the IGTO and IGZO possess approximately 3.6 eV of EG. Utilizing the values obtained from these analyses, the energy band diagrams were estimated (Fig. 3). The detailed procedure to depict the band diagrams can be seen in the previous study37. Consequently, the SBH decreases from approximately 0.4 to 0.2 eV by inserting the 3-/8-nm-thick TiN/IGTO IL. These values may deviate slightly. However, it is obvious that the SBH for electron injection is drastically reduced by this contact scheme. This reduction can lead to the improvement in specific contact resistivity by enhancing the electron injection as confirmed in the TLM analyses.

Figure 3
figure 3

Schematic energy band diagrams at the contact region depicted based on UPS and UV/visible analyses: (a) without the IL; (b) 3-/8-nm-thick TiN/IGTO IL.

Full size image

Then, atomic force microscopy (AFM) analysis was conducted. Figure 4 shows tIGTO-dependent surface roughness (Rrms) of the IGTO thin-films deposited on the 15-nm-thick IGZO/SiO2/Si substrates. The Rrms increases to 0.8 nm in the 12-nm-thick IGTO thin-film, and this roughest surface could partially contribute to degradation of the electrical contact properties. The data obtained from the 0-nm-thick IGTO thin-film indicates information of the underlying 15-nm-thick IGZO thin-film. In addition to AFM analysis, Hall effect measurement was performed on the IGTO thin-films to investigate the origin of the tIGTO-dependent the electrical contact. The nc of the IGTO thin-film steadily increases from 7.9 × 1017 to 6.9 × 1019/cm3 with tIGTO (Fig. 5a). Moreover, its bulk resistivity (ρIGTO) simultaneously decreases from 1.1 Ω·cm to 3.7 × 10−3 Ω·cm, a trend originating from percolation conduction2. This result shows that the IGTO IL can act as an n+-layer in IGZO devices but cannot elucidate the tIGTO-dependent the electrical contact. Even, the fact that the 12-nm-thick IGTO thin-film has the lowest ρIGTO conflicts with the trend that the ρC and RCW are the lowest in the tIGTO of 8 nm.

Figure 4
figure 4

Surface images of the IGTO/IGZO thin-film stacks with different tIGTO: (a) 0 nm; (b) 5 nm; (c) 8 nm; (d) 12 nm.

Full size image
Figure 5
figure 5

(a) Hall effect measurements of the IGTO/IGZO thin-film stacks with different tIGTO: 0, 5, 8, and 12 nm. (b) GIXRD patterns of the IGTO thin-films with different tIGTO: 5, 8, and 12 nm.

Full size image

Grazing-incidence X-ray diffraction (GIXRD) was conducted to investigate the crystalline structure of the IGTO thin-film (Fig. 5b). Surprisingly, the 12-nm-thick IGTO thin-film possesses a random polycrystalline In2O3 cubic bixbyite configuration, differing from the other tIGTO conditions. Electron transport can be enhanced by decreased defect scattering after crystallization due to structural ordering and defect confinement, which can contribute to the decrease in bulk ρIGTO observed in the Hall measurement. More importantly, this crystallographic change has a large impact on a metal/semiconductor (MS) interface, a critical factor in the electrical contact properties, and can result in a trend difference between the electrical contact and ρIGTO. However, it is possible that the IGTO thin-films with tIGTO < 12 nm were crystallized but too thin to determine their crystalline structure using GIXRD.

For this reason, cross-sectional high-resolution transmission electron microscopy (HRTEM) was performed on the ITO/TiN/IGTO/IGZO stacks to clearly examine the tIGTO-dependent crystallographic change (Fig. 6). It is noteworthy that electron dispersive spectroscopy (EDS) depth profile was conducted to distinguish the IGTO and IGZO layers (Fig. S5). There is no noticeable nano-/microscale crystal in the 8-nm-thick IGTO thin-film (Fig. 6a and Fig. S6). The fast Fourier transform (FFT) result also shows a diffused hollow ring pattern (Fig. S7). These results indicate that the corresponding IGTO thin-film has an amorphous structure. Meanwhile, the 12-nm-thick IGTO thin-film is crystallized (Fig. S8), and many crystal grains significantly infiltrate from the ITO into the IGTO film (Fig. 6b and Fig. S8). It is worth mentioning that the ITO thin-film has the same In2O3 cubic bixbyite crystalline configuration. Consequently, the MS interface is considerably deteriorated in the thin-film stack with tIGTO of 12 nm. These degradations are not observed in the stack with tIGTO of 8 nm (Fig. S6). This deteriorated interface could be attributed to the worse thermal stability of the polycrystalline structure with grain boundaries than the amorphous structure that make it susceptible to thermal stress during annealing. Thus, the polycrystalline 12-nm-thick IGTO thin-film could be more prone to deformation and cracking by thermal stress compared to the amorphous 8-nm-thick IGTO thin-film, as observed in the HRTEM analyses. Here, it is important to note that the interfacial property can significantly influence electrical contact properties. This is because a rough interface can introduce numerous scattering and localized trapping centers, thereby impeding electron injection at the MS contact region. Furthermore, such roughness can induce variations in the Schottky barrier height, resulting in fluctuations in the contact properties. Consequently, this TEM result implies that the electrical contact can be greatly degraded in a device using the multi-stack IL with tIGTO of 12 nm due to crystallization-induced disruptive interface even if it has the lowest bulk ρIGTO. This underscores the critical importance of the interfacial quality between the S/D electrodes and the channel layer.

Figure 6
figure 6

Cross-sectional HRTEM images of ITO/TiN/IGTO/IGZO stacks with different tIGTO: (a) 8 nm; (b) 12 nm.

Full size image

Figure 7 shows electrical characteristics of IGZO TFTs with different IL structures. The IGZO TFTs without the IL exhibit device performances with μFE of 39.9 ± 1.6 cm2/Vs, current modulation ratio (ION/OFF) > 108, SS of 0.1 ± 0.03 V/dec, and VTH of − 0.5 ± 0.4 V. The μFE is improved to 42.7 ± 1.4 cm2/Vs through insertion of a 3-nm-thick TiN IL, and it further increases using the TiN/IGTO multi-stack IL. Importantly, the tIGTO-dependent behavior is clearly observed, with the IGZO TFTs with tIGTO of 8 nm exhibiting the highest μFE of 45.0 ± 1.6 cm2/Vs. Meanwhile, the devices with tIGTO of 5 and 12 nm showed μFE values of 44.8 ± 1.3 and 42.4 ± 1.8 cm2/Vs, respectively. Figure S9 shows VGS-dependent μFE curves in the IGZO TFTs with different IL conditions. Such improvement by IL insertion is also seen in output characteristics (Fig. 7f–j). All device performances of the IGZO TFTs with different IL stacks are summarized in Table 1. Considering the significant impact of electrical contact on the electrical characteristics of the TFT, it is straightforward to understand the trend where the highest device performances are revealed in the IGZO TFTs with tIGTO of 8 nm. This enhancement can be further pronounced in scaled oxide TFTs where the device performance is dictated by the electrical contact properties. Figure 8 and Table S1 are benchmarking graph and table, respectively. It is important to emphasize that the ρC (RCW) of 9.0 × 10−6 Ω·cm2 (0.7 Ω·cm) achieved in this study surpasses the state-of-the-art results reported in the literature for all types of multicomponent metal oxide transistors, demonstrating the superiority of this study.

Figure 7
figure 7

(a–e) Transfer characteristics of IGZO TFTs with different IL stacks: (a) Without IL; (b) TiN IL; (c) TiN/IGTO IL (tIGTO = 5 nm); (d) TiN/IGTO IL (tIGTO = 8 nm); (e) TiN/IGTO IL (tIGTO = 12 nm). (f–j) Output characteristics: (f) Without IL; (g) TiN IL; (h) TiN/IGTO IL (tIGTO = 5 nm); (i) TiN/IGTO IL (tIGTO = 8 nm); (j) TiN/IGTO IL (tIGTO = 12 nm).

Full size image
Table 1 Summarized electrical figures of merit of IGZO TFTs with different IL stacks.
Full size table
Figure 8
figure 8

Comparison of RCW and μFE in oxide TFTs with different contact approaches.

Full size image

Conclusion

This study demonstrates a significant improvement in electrical contact properties of IGZO TFTs through insertion of a TiN/IGTO IL. The electrical contact properties are dependent on tIGTO, which is attributed to crystallographic change from the amorphous structure to the random polycrystalline structure of IGTO. Despite a decrease in bulk ρIGTO with increasing tIGTO, crystallization at tIGTO of 12 nm aggravates the electrical contact due to side effects such as grain boundary infiltration and degraded MS interfaces. Therefore, the use of an amorphous 8-nm-thick IGTO thin-film is more effective in improving the electrical contact properties. As a result, ρC (RCW) decreases from 8.0 × 10−4 (6.9) to 9.0 × 10−6 Ω·cm2 (0.7 Ω·cm) and μFE increases from 39.9 to 45.0 cm2/Vs, compared to the IGZO devices without the IL. This study not only highlights the effectiveness of the multi-stack IL approach in enhancing electrical contact properties, but also demonstrates that the effect can be maximized using a highly conductive amorphous IL. Importantly, this contact scheme can be applied to various types of oxide TFTs, not just IGZO TFTs.