ALD TiO_x as a top-gate dielectric and passivation layer for InGaZnO₁₁₅ ISFETs

To fabricate the TFTs depicted in figure 1, a 25 nm thick Al₂O₃ film was first deposited at 250 °C via ALD on a p⁺⁺ silicon (Si) wafer that also serves as the bottom gate contact. The Al₂O₃ deposition used trimethylaluminum and water vapor as the aluminum and oxygen source, respectively, yielding a growth rate of 0.9 Å/cycle. A 50 nm thick InGaZnO (In:Ga:Zn = 1:1:5) semiconducting film was then deposited by pulsed laser deposition at room temperature and 25 mTorr O₂ using a KrF excimer laser operating at 30 Hz and a 1.8 J cm⁻² energy density. A mesa etch followed to isolate neighboring devices. Subsequently, 25 nm/200 nm thick chrome/gold (Cr/Au) source and drain contacts were formed using electron-beam evaporation and lift-off. The devices consisted of four parallel channels, each with a width (W) of 25 μm and length (L) of 10 μm (see insert in figure 1), thus representing an effective W:L ratio of 10:1. TFTs in this state constitute the bare reference device. Passivation of the devices was accomplished using a two-step process. Initially, the devices were subjected to a N₂O plasma treatment at 50 kHz and 50 W RF power for 2 min. In previous work, this has been found to reduce the effect of hydrogen doping after PECVD SiO₂ passivation steps [20]. The samples in the form of individual chips were then transferred to a Cambridge NanoTech Fiji ALD tool, in which 15 nm thick films of TiO_x were deposited at four temperatures (100 °C, 150 °C, 200 °C and 250 °C). The precursors for TiO_x deposition were water vapor and tetrakis(dimethylamino)titanium. The pulse times for these were 60 ms and 200 ms, respectively, while the purge times were adjusted according to the deposition temperature, ranging from 90 s at 100 °C to 10 s at 250 °C. Following passivation, the TFTs were annealed in an O₂ atmosphere at 300 °C for 30 min.

Zoom In Zoom Out Reset image size

In order to conduct tests in liquid, a microfluidic packaging process was also established. Firstly, a 6 μm thick SU-8 film (SU8-3005) was spun and patterned such that only the TiO_x sensing regions are exposed to the liquid. Thereafter, a PDMS structure with patterned microfluidic channels was attached to the die for liquid analyte delivery. Figure 2 shows a picture of the die after the microfluidic packaging step, where a PMMA clamp has been used to bond the structure together. A motorized syringe pump was used to controllably flow the liquid solutions into the microfluidic cell where the TFTs are located. An off-chip and commercial flow-through Ag/AgCl reference electrode from Microelectrodes, Inc. was used.

Zoom In Zoom Out Reset image size

Figure 3(a) presents a comparison of the deposition-temperature-dependent TFT transfer characteristics (I_D−V_GS) in the linear mode of operation (V_DS = 0.1 V) measured directly after the TiO_x passivation step (prior to O₂ annealing step). The threshold voltage (V_TH) is defined as the value of V_GS where I_D = 1 nA. The on/off ratio (I_ON/I_OFF) is calculated by evaluating the ratio of the maximum on-state current to the minimum off-state current. For the bare device, V_TH = 1.3 V, while the on/off ratio (I_ON/I_OFF) is 2 × 10⁸. It is observed that, in all cases, passivation causes I_OFF to increase, and V_TH to shift negative. Moreover, the degree to which these changes occur depends on the deposition temperature. The 100 °C deposition increases I_OFF to 10⁻⁸ A, while at 250 °C it reaches 10⁻⁴ A. Figure 3(b) depicts the absolute value of I_G versus V_GS behavior at this stage of fabrication. The bare device functions well with ∣I_G∣ ∼ pA, but ∣I_G∣ is found to increase significantly with increased passivation deposition temperature.

Zoom In Zoom Out Reset image size

Each of the samples was then subjected to an O₂ anneal for 30 min at 300 °C. After annealing, the gate current (not shown) decreased to under 100 pA in all of the devices. As shown in figure 3(c), the transfer characteristic also improved, with I_ON/I_OFF increasing to 10⁷ or more. Nonetheless, there remains a performance dependence on the passivation deposition temperature: as before, V_TH shifts more negative as the temperature increases. It was found that the sub-threshold swing (S) tends to increase with increased deposition temperature. The bare device possesses an excellent S of 140 mV dec⁻¹. The device passivated at 100 °C displays a slightly worse value of 185 mV dec⁻¹, while the device passivated at 250 °C increases notably to 363 mV dec⁻¹. These trends indicate that higher passivation temperature increases the trap density either in the InGaZnO bulk or the InGaZnO top surface where the TiO_x is introduced [28]. Passivation clearly has a temperature-related effect on the field-effect mobility (μ_FE,max) too, as shown in figure 3(d). Though the bare device's maximum mobility was 16.5 cm² V⁻¹ s⁻¹, the passivated devices all possess significantly reduced values. At 100 °C, μ_FE,max falls to 2.9 cm² V⁻¹ s⁻¹ and increases with temperature to 6.5 cm² V⁻¹ s⁻¹ at 250 °C. Table 1 summarizes the performance parameters with respect to passivation film deposition temperature. The source of these effects is further investigated in section 3.3.

Positive bias stress (PBS) tests were conducted on the bare, 100 °C and 250 °C samples to contrast device stability. Each device was subjected to a forward bias condition of V_GS = 5 V and V_DS = 0.1 V for 1 h, as shown in figure 4. The bare device experienced a +1.1 V shift in V_TH. In comparison, V_TH of the 100 °C passivated device shifted by 0.2 V. Lastly, the 250 °C passivated device was observed to be the most stable of the devices, showing a nearly negligible 0.05 V positive shift. These results support previous reports [29] showing that the device stability is not solely dependent on the bottom dielectric (Al₂O₃)—semiconductor (InGaZnO) interface, but is also highly dependent on the electric-field induced adsorption of oxygen and desorption of water. Consequently, addition of a TiO_x top-gate dielectric successfully suppresses this source of bias instability, and reduces the shift in V_TH. The sub-threshold swing S was also affected differently following bias stress depending on the temperature of the passivation deposition. S in the bare device increased from 140 to 216 mV dec⁻¹, while in the 100 °C device it only increased from 185 to 190 mV dec⁻¹, and in the 250 °C device it improved from 363 mV dec⁻¹ to 295 mV dec⁻¹. While TiO_x passivation clearly improves the bias stress stability of InGaZnO TFTs, the specific deposition temperature of the passivation film seems to only have a minor impact on these improvements.

Zoom In Zoom Out Reset image size

To better understand the above results, time-of-flight SIMS (TOF-SIMS) was carried out. Past studies using this technique have shown that high temperature ICP-CVD SiO_x passivation can cause diffusion of In, Ga and Zn into the SiO_x passivation layer, in turn reducing μ_FE and increasing S [30]. Furthermore, it has been reported that differences in the type of precursor (O₂ versus H₂O vapor) used for plasma-enhanced versus thermal ALD Al₂O₃ passivation, respectively, of InGaZnO can influence hydrogen doping, which subsequently determines the V_TH shift [31]. Our SIMS analysis (figure 5), however, shows negligible differences in interdiffusion or hydrogen doping due to temperature effects. The latter also serves to support the use of the N₂O plasma treatment prior to TiO_x deposition.

Zoom In Zoom Out Reset image size

It is instead proposed that oxygen gettering by the TiO_x passivation film is responsible for the measured characteristics, which has been observed in organic [23] and ZnO TFTs [32]. In both InGaZnO and InZnO films, oxygen vacancies act as electron donors, while the use of Ga as an oxygen getter enables superior mobility and V_TH stability in InGaZnO [33]. Yen et al reported that e-beam evaporated TiO_x films on InGaZnO₁₁₁ and InZnO₁₁ prior to S/D formation compete for oxygen [34]. Though both TFTs experienced negative shifts in V_TH, S improved for the InGaZnO₁₁₁ TFTs, but deteriorated in the InZnO₁₁ TFTs. This difference was explained by the absence of Ga in InZnO.

Our results indicate that the strength of oxygen gettering is influenced by the TiO_x deposition temperature, with higher temperatures generating more oxygen vacancies to reduce the V_TH. The increase in S with passivation film deposition temperature can likewise be explained by the generation of traps from this phenomenon. Moreover, our results further elucidate the role of the InGaZnO composition in determining the compatibility of passivation strategies. Whereas Yen et al used a 1:1:1 target, our work used a 1:1:5 composition that is more akin to Ga-absent InZnO. Therefore, care must be taken in choosing the composition of InGaZnO, as well as its thin film deposition conditions, when selecting a passivation material and deposition process. The reduction in mobility post TiO_x deposition is attributed to the generation of defects that are symptomatic of the increase in S and can increase the contribution of Coulomb scattering. This effect has been previously reported in TiO_x-InGaZnO TFTs by evaluating the operation temperature-dependent behavior of such devices [35].

The above results underline the promise of ALD TiO_x passivation of InGaZnO TFTs, in particular in terms of enhancing stability for electronic applications that require a dependable baseline. To subsequently validate the suitability of ALD TiO_x as a top-gate dielectric in InGaZnO ISFETs for potentiometric chemical sensing, a 15 nm thick TiO_x film was deposited via ALD at 150 °C on metal electrodes and its pH response was evaluated, as demonstrated in our previous work [36]. This method ties the Ag/AgCl reference electrode to ground (V_REF = 0 V) as is standard for DG-TFT performance, and permits time-based measurements to be performed without the need of an external CHEMFET read-out circuit (e.g., source-follower). Both ISFETs and DG-TFTs are fundamentally limited by the behavior of the top gate dielectric-liquid interface, as demonstrated in the following [6]:

$$\begin{eqnarray}&&{\rm{\Delta }}{V}_{{\rm{T}}{\rm{H}},{\rm{e}}{\rm{f}}{\rm{f}}}=-\displaystyle \frac{{C}_{{\rm{g}},{\rm{t}}{\rm{o}}{\rm{p}}}}{{C}_{{\rm{g}},{\rm{b}}{\rm{t}}{\rm{t}}{\rm{m}}}}{\rm{\Delta }}{{\rm{\Psi }}}_{0},\end{eqnarray} \tag{ 1 }$$

where C_g,top and C_g,bttm are the top and bottom oxide capacitances, respectively, and Ψ₀ represents the top gate-electrolyte interface potential, given by [37]:

$$\begin{eqnarray}&&{\rm{\Delta }}{{\rm{\Psi }}}_{0}=\left(-2.3\displaystyle \frac{kT}{q}\alpha \right){\rm{\Delta }}{\rm{p}}{\rm{H}},\end{eqnarray} \tag{ 2 }$$

where k is the Boltzman constant, T is temperature, q is the charge of a single electron, α is a constant that depends on the quality of the dielectric-analyte interface and ΔpH is the change in pH. At a fixed temperature, therefore, the fundamental sensitivity of an ISFET (ΔΨ₀/ΔpH) is dependent on the interface between the top gate dielectric under investigation and the liquid. In a DG-TFT, this fundamental sensitivity is amplified by the capacitance ratio C_g,top/C_g,bttm (which equals 1 in the case of the ISFET).

Figure 6(a) shows the transient response of ALD TiO_x to buffer solutions with pH = 3, 5 and 7, which confirms the rapid and repeatable change in potential as a function of pH. The data has been normalized with respect to the initial voltage at t = 0 (pH = 7), and baseline drift has been corrected for. The hysteresis, which was calculated as the difference in potential between the two readings at pH = 5, is 10 mV. From the slope in figure 6(b), it is found that the sensitivity of ALD TiO_x is 57.7 mV pH⁻¹ and is therefore very close to the Nernstian Limit (∼59 mV pH⁻¹). Additionally, the linearity of the response is 99.89%. These results are in line with past investigations of TiO_x as a pH sensing layer using sol-gel formation [38] and RF sputtering [8], which recorded 50 mV pH⁻¹ and 59.2 mV pH⁻¹ respectively. The high sensitivity, high linearity and high dielectric permittivity of TiO_x [8, 39] make it a highly attractive candidate as a top-gate dielectric/potentiometric sensing layer in InGaZnO ISFETs and/or DG-TFTs.

Zoom In Zoom Out Reset image size

In addition to sensitivity, it is important to evaluate the stability of InGaZnO ISFETs with an ALD TiO_x top-gate dielectric. From our own tests, we have observed that a 50 nm thick InGaZnO₁₁₅ film is completely etched in acid within 10 min, even when using a 1:150 dilution in water. Such short lifetimes would render these devices impractical for use in the field. To confirm that InGaZnO with a top ALD TiO_x dielectric is suitably protected, the V_GS−I_D characteristic of a device submerged in pH = 5 buffer solution was periodically collected over a 24 h period (figure 7(a)). In between these measurements, the device was unbiased. The TFT continues to operate throughout the entire test, with very little V_TH shift observed. As time passes, the OFF current decreases, thus improving the on/off ratio, but having no effect on the sensor's drift. The sensing baseline drift was monitored by reading out V_GS for a reference drain current (I_REF) of 10 nA, with an average drift of ≤1 mV h⁻¹. These results clearly demonstrate not only that the InGaZnO thin film is protected from the acidic liquid environment, but that the DG-TFT's electrical behavior can also be maintained over an extended period of time. Previous studies confirming the high performance of ALD TiO_x as a barrier layer are therefore consistent with our findings [22, 40].

Zoom In Zoom Out Reset image size