Organization of this Thesis

In Chapter 2, we present the fabrication of the suspended-NW-channel TFT and SG VMOSFET. Basic process flows are briefly described. The measurement setups are also presented in this chapter.

In Chapter 3, we present and discuss the electrical characteristics of the fabricated devices tested in different environments, including atmosphere, vacuum and ammonia. The results of the suspended-NW-channel TFT and SG VMOSFET

will be discussed respectively.

Finally, we summarize the major conclusions from our experimental results in Chapter 4.

Table Captions

Table 1.1 Typical examples of mixed-potential-type gas sensors [8].

Chapter 2

Device Structure and Fabrication

2.1 Device Structure and Process Flow of Suspended-NW-Channel TFTs

Fig. 2.1 shows the structure and process flow of the suspended-NW-channel TFT. Fig. 2.1(a) shows the top view of the device and the device key fabrication process flow is shown in Figs. 2.1(b)~(f) which illustrate the cross-sectional views at various steps along the dotted line AB in Fig. 2.1(a).

The fabrication began on a 6-in Si substrate capped with a 1000-nm thermal oxide layer. First, a 150nm-thick in-situ doped n⁺ poly-Si was deposited by low pressure chemical vapor deposition (LPCVD) at 550℃ and then patterned to form the gate electrode as show in Fig. 2.1(b). After the pre-furnace standard clean process, a 20-nm silicon nitride dielectric, sacrificial TEOS oxide layer and 150-nm amorphous Si were deposited in order. All these films were deposited by LPCVD at 780℃, 700℃

and 550℃, respectively. Besides, the sacrificial layer was split into 20-nm, 50-nm, and 80-nm in thickness to achieve different air gap thickness, as would become clear later. The amorphous Si was then re-crystallized to become polycrystalline in a furnace at 600℃ in N2 ambient for 24 hours, as shown in Fig. 2.1(c). The source/drain (S/D) photoresist patterns were subsequently defined and an anisotropic plasma etching was performed to form the S/D simultaneously with the NW channels at the sidewall of the gates as shown in Fig. 2.1(d). Note that the dimension of the NW

channels can be controlled by the over etching time of the plasma etching.

Subsequently, using another photoresist pattern to cap the channel region, an n-type S/D implant was conducted using P³¹⁺ at a dose of 5×10¹⁵ cm^-2, as shown in Fig.

2.1(e). The dopant activation was simultaneously done with the deposition of a 500-nm TEOS oxide passivation layer deposited by LPCVD at 700℃. Finally, the contact holes were open by buffered oxide etchant (BOE) and hot 160℃ H3PO4 wet etching process to remove the capping oxide and nitride over the electrodes, respectively. Here, a typical NW-channel TFT was completed.

To suspend the NW-channel, a photoresist pattern was formed to cap the device except the central channel region, as shown in Fig. 2.1(a). A BOE wet etching was used to remove the sacrificial TEOS oxide layer between the NW channel and the silicon nitride layer. After removing the photoresist, we put the wafer on a hot plate of 115℃ to desiccate the devices. The NW channels became suspended and exposed to the air after this step, as shown in Fig. 2.1(f).

2.2 Device Structure and Process Flow of SG Vertical-MOSFETs

Fig. 2.2(a) shows the top view of the device and Figs. 2.2(b) to (f) show the device key fabrication process flow with cross-sectional views along the dotted line

AB in Fig. 2.2(a).

Briefly, 6-in n-type Si wafers, with resistivity between 2-7 Ω^.cm, were used as the starting substrates of p-type SG vertical-MOSFETs. First, S/D photoresist patterns were formed with a photolithography process, and then an anisotropic plasma etching step was applied to define the vertical channel region. In this thesis, the vertical

channel length was split to 300-nm, 500-nm and 1000-nm by controlling the etching-time. To remove the damage resulted during the plasma etch, a 35-nm sacrificial thermal oxide was grown on the surface of the Si wafers. This oxide layer also serves as a screen oxide to avoid channeling effect in the next channel implantation step. A phosphorous ion implantation at 170 keV with a dose of 4×10¹² cm^-2 was used to dope the sidewall channel regions, and was carried out at a tilt angle of 45°, twist angle of 72° with the wafer rotated 4 times during implantation, as shown in Fig. 2.2(b). Afterwards, the sacrificial oxide was etched by a BOE wet etching. The RCA clean was applied before the sequential deposition of four films, as shown in Fig. 2.2(c). First, a 3-nm TEOS oxide was deposited by LPCVD at 700℃ to serve as a buffer layer. Second, a 20-nm silicon nitride was deposited by LPCVD at 780℃ as gate oxide layer. Third, a sacrificial TEOS oxide layer of 50-nm or 80-nm as a sacrificial layer. Finally, a 150nm-thick in-situ doped n⁺ poly-Si was deposited by LPCVD at 550℃ to serve as the material of gate electrode. Afterwards, a second mask was used to define the pad area of the gate. The gate pad and the sidewall spacer gate electrode were formed simultaneously with an anisotropic plasma etching step, as shown in Fig. 2.2(d). Next, a diluted hydrofluoric acid (DHF) wet etching was performed to etch the TEOS oxide layer capping on S/D area. Then BF249+ ions was implanted to dope the S/D area at 25 keV with a dose of 5×10¹⁵ cm^-2 and activated by rapid thermal annealing (RTA) at 1000℃, as shown in Fig. 2.2(e). And then a 400-nm passivation TEOS oxide layer was deposited to cap the entire wafer by plasma enhanced chemical vapor deposition (PECVD). Finally, we removed the top passivation layer, silicon nitride gate dielectric and TEOS oxide buffer layer by BOE, hot 160℃ H3PO4 and DHF, respectively, in order to open the contact holes. Here, a typical SG vertical-MOSFET was completed. Figs. 2.3(a) and (b) show the

Fig. 2.2(a).

The steps to suspend the sidewall spacer gate electrode are just like that described in Sec. 2.1. First, a photoresist mask was used to define the open region.

Next, BOE and hot 160℃ H3PO4 wet etching were used to remove the sacrificial layer and photoresist mask in order. Finally, the wet humidity remained on the wafer surface was removed by a hot plate set at 115℃. The resultant SG vertical-MOSFET is shows in Fig. 2.2(f).

2.3 Measurement Setup

2.3.1 Measurement Setup of Basic Electrical Characteristics

The basic I-V characteristics were measured by a precise measurement system including an HP 4156A precision semiconductor parameter analyzer, an Agilent TM 5250A switch and the Interactive Characterization Software (ICS) operating software at a stable environment at room temperature.

The basic electrical parameters of our device were extracted from the basic I-V characteristics. Due to the different operation behaviors exhibited by the two types of devices studied in this work, different definitions of threshold voltage (Vth) are necessary. Vth was defined as a value of gate voltage (VG) when the drain current (ID) equals A under drain voltage (VD) was 0.1V for the suspended- l TFTs, and

L vertical-MOSFETs, where W is the channel width and L is the channel length. The subthreshold swing (SS) was calculated by the following equation:

log ⁻1

For both suspended-NW-channel TFT and SG vertical-MOSFET device, the average value of SS is considered.

2.3.2 Measurement Setup of Gas Sensor

In this study, the gas sensor measurements of all devices were evaluated by an airtight measurement system including a Keithley 4200 semiconductor characterization system, an ammonia source, a gas throttle, a flow controller, a vacuum pump and a closed chamber, as shown in Fig. 2.4.

The measurement system is capable of adjusting the ammonia concentration contained in the environments. In normal conditions, the environment is filled with fresh air in the closed chamber and the fresh-state I-V characteristics are recorded with the Keithley 4200 semiconductor characterization system. For gas sensing measurements, pure ammonia gas is introduced into the closed chamber. The ammonia concentration is controlled by the gas throttle and flow controller. According to a previous research, the diffusion coefficient of ammonia is 0.241 cm/s in nitrogen environment at atmospheric pressure and 20℃ [16]. So the measurements are conducted two minutes after the gas injection so that the ammonia could be uniformly distributed in the closed chamber. This scheme allows us to evaluate the response of the test device to the variation of ammonia content in the environment. The ammonia could be further injected to increase its concentration and repeated the measurements.

The injection gases can be evacuated via a vacuum pump.

Chapter 3

Results and Discussion

3.1 Basic Characteristics of Suspended-NW-Channel TFTs

Figs. 3.1(a) and (b) show the typical transfer characteristics of the conventional NW-channel TFTs and suspended-NW-channel TFTs, respectively. The difference in structure between the conventional and suspended-NW-channel TFTs is the existence of the air gap in the latter. The suspended-NW-channel TFT depicts a better subthreshold swing (SS) of 221 mV/dec as compared with a poorer SS of 627 mV/dec for the conventional one. Owing to the low dielectric constant of air, a nominal EOT of 341.73 nm for suspended-NW-channel TFT is actually much thicker than the EOT of 95.35 nm for the conventional NW-channel TFT. According to the relationship between SS and EOT:

⎟⎟⎠

where Cdep is depletion layer capacitance and Cox is the effective oxide capacitance, i.e., Cox∝ 1/EOT, the suspended-NW-channel TFT should has poorer SS, which is contrary to the observed trend. It can be attributed to the swaying of the NW channels.

A detailed discussion about the operation mechanism is given later in Section 3.1.1.

3.1.1 Hysteresis Phenomenon

Fig. 3.2 shows the fresh transfer characteristics of a suspended-NW-channel TFT. The device exhibits a hysteresis phenomenon under forward and reverse sweeping measurments [11]. It is attributed to the motion of the NW being swayed by the electrosatic force (Fe) and elastic recovery force (Fk). Figs. 3.3(a) to (e) show the motion of the NW through forward and reverse sweeping measurments. First, Fe is very small when the device is in the off-state. The NW is suspended as shown in Fig.

3.3(a). At some momnet during the forward sweeping, Fe would become greater than Fk and the NW is pulled toward and eventually touches the gate, as shown in Fig.

3.3(b). As the gate bias increases further, more channel regions are in contact with the gate, as shown in Fig. 3.3(c). On the other hand, during reverse sweeping the attached NW is released from the gate as the gate bias is sufficiently low. When the NW channels are in conatct with the gate nitride, the electrostic electric force is stronger than when they are apart due to the thinner EOT, while additional van der Waal force (Fv) is exerted (Fig. 3.3(d)). As a result, the pull-out action would occur at a gate bias lower than that of pull-in, leading to the hystersis characteristics. Finally, Fig. 3.3(e) shows the NW is pulled out from gate nitride and returns to suspensory status. Such mechanical pull-in and pull-out phenomena are depicted in Fig. 3.2.

Owing to the fact that the NW-channel diameter is smaller than 100 nm, it is very soft.

Sudden increase in drain current at VG=Vpi is not significant.

3.1.2 Hydrogen Passivation Effects

Fig. 3.4 shows the transfer characteristics of a suspended-NW-channel TFT

vacuum, and 80 minutes and 21 hours after returning to atmosphere. The air pressure is 0.2 torr for the vacuum state. It can be seen that SS dramatically increases from 223 mV/dec to infinite as the environment is switched from atmospheric to vacuum state.

After returning to the atmosphere, the characteristics recover gradually and the SS is back to 235 mV/dec after 21 hours. This phenomenon can be ascribed to the passivation effect of water vapor [17][18]. Polycrystalline silicon is known to be granular in structure and contains a lot of grain boundaries at which many defects are located. Those defects are the root cause for mobility and SS degradation. Figs. 3.5(a) to (c) show the schematic diagram of poly-silicon grain boundaries. Generally, there exists an annealing procedure which introduces hydrogen-related species to decrease the amount of active defects and increase mobility of the polycrystalline silicon films.

During the treatment, hydrogen species diffuse into polycrystalline silicon film and passivate the dangling bonds [18], as shown in Fig. 3.5(a). In vacuum state, the passivated hydrogen species are desorbed from polycrystalline silicon film and released into the environment, and leave the previously-passivated sites with the dangling bonds again. It is attributed to the reduction of chemical potential due to the decreasing pressure. The relationship between chemical potential and pressure of a perfect gas is [19] where μ^o and p are the chemical potential and pressure, respectively, at the ^o standard state. The chemical potential in the gas is reduced when the pressure decreases to 0.2 torr. As a result, the hydrogen species tends to get expelled from grain boundaries to the vacuum. When the device is turned on, a large number of electrons are trapped in those defects, causing mobility degradation and reducing carrier concentration. Fig. 3.5(b) shows the schematic diagrams of hydrogen desorption from

the grain boundary. This explains why in Fig. 3.4 the drain current is too low to detect as the device is measured in vacuum. Fig. 3.6 compares the transfer characteristics of the NW devices with or without the air gap. Clearly the suspended-NW device has a much lower current. This is attributed to its larger area/volume ratio as compared with the conventional-NW device which has no air gap. In other words, desorption of the hydrogen species from the grain boundaries is expected to be more efficient in the suspended-NW device. Therefore, the hydrogen de-passivation effect of the suspended-NW is more evident than the conventional-NW device. As the environment returns to the atmosphere, hydrogen species contained in the environment may diffuse slowly back into the grain boundaries. Dangling bonds may be re-passivated by these returning hydrogen species, especially for those located closer to the NW surface.

Nonetheless, the dangling bonds located in the NW core region would require a longer time to get re-passivated owing to the low environment temperature, as schematically shown in Fig. 3.5(c). Therefore, it may take a very long time to fully recover to the fresh state at room temperature. Fig. 3.7 shows the transfer characteristics of a suspended-NW-channel TFT measured at different environments or conditions, including the one with a forming gas (5% H2 in N2) sintering treatment in a furnace at 400℃ for 30 minute. After the sintering treatment with a forming gas, the SS is 246 mV/dec, smaller than 338 mV/dec of the fresh state. This indicates the effectiveness of hydrogen passivation of the defects, and provides an indirect evidence for the above de-passivation effect used for explaining the experimental observations.

3.2 Suspended-NW-Channel TFTs for Gas Sensor Applications

At first, the NH3 concentration in the measurement system is calculated. We have introduced this system in Section 2.3.2 and the system configuration is shown in Fig. 2.4. The ideal gas law will be considered as:

nRT

PV = , (Eq.3-3) where P is pressure, V is volume, n is the number (usually in moles) of gas molecules presenting in the system, R is the universal gas constant and T is the temperature [20].

In this study, in normal situations the pressure is controlled at 1 atmosphere, the volume of closed chamber is 52 liters, the temperature is controlled at room temperature and the value of universal gas constant is 0.08206 L⋅atm K⋅mol [20].

According to a standard of BIPM (Bureau international des poids et mesures), the moist air density can be calculated by the CIPM-2007 revised formula [21][22]. So we can calculate the weight of moist air in the closed chamber from the moist air density. In addition, the atom mass of NH3 has a value of 17.034 Da, meaning that 6.022×10²³ atoms of NH3

collectively have a mass of 17.034 gram. For example, moist air density is 1.1938 kg/cm³ at pressure, temperature and relative humidity (RH) of 101325 Pa, 20℃ and 100%, respectively [22]. We can calculate that the NH3 concentration of1 ppm in the chamber is about 4.13×10¹³ cm^-3.

Transfer characteristics of a suspended NW-channel TFT with an air gap of 20 nm measured in environments with various NH3 concentrations are comapred in Fig.

3.8. In the measurements, NH3 was injected into the closed chamber where the test device is located. As can be seen in the figure, the threshold voltage (Vth) decreases while the subthreshold swing (SS) increases under forward sweeping with increasing

NH₃ concentration. Fig. 3.9 and Fig. 3.10 show the Vth shift and SS as a function of NH3 concentration in the test environment. The Vth shift and SS variation increases with increasing the NH3 concentration. Because the cross-sectional dimensions of our NW structure are smaller than 100 nm, the device appears to be sensitive to the variation of the NH3 concentration.

3.2.1 NH₃ Passivation Effect

In Sec. 3.1.2, we have discussed the hydrogen passivation effect that affects the basic characteristics of the devices such as SS. Similarly, when the NH3 is introduced, the SS shows response to the NH3 concentration. According to a previous work [23], ammonium (NH4+) produced through the reaction of NH3 and water vapor can be expressed by

3( )g 2 ( )g 4 ( )g ( )

NH +H O →NH⁺ +OH⁻_g , (Eq.3-4) and the NH4+ tends to be adsorbed and accumulated on the NW surface. Fig. 3.10 shows the SS under forward and reverse sweeping measurements as a function of NH3

concentration in the test environment. As can be seen in the figure, the SS decreases with increasing NH3 concentration. According to the discussion made previously, the improvement in SS is attributed to the recombination of the dangling bonds with the NH₄⁺ molecule at the NW surface. Enhanced passivation of the defects conatined in the poly-Si NW channels in the NH3-containing environment well explains the results.

And then the Vth shift takes place due to the improved SS with NH3 passivation as shown in Fig. 3.9. As a result, the NH3 concentration can be detected through the Vth

shift and SS improvement.

3.2.2 Sensing Limit

Fig. 3.12 shows the transfer characteristics of a suspended-NW-channel TFT with an air gap of 85 nm, measured in environments with various NH3 concentration to test the sensing limit of our devices. As shown in Fig. 3.13, the Vth shift tends to saturate when the NH3 concentration is high. It is attributed to the limited suface sites on the NW surface for NH4+ to adsorb. Fig. 3.14 shows the real-time sensing characteristics. A response in the drain current is clearly detected as the NH3

concentration is introduced, though the signal is weak as the NH3 concentration is higher than 10¹³ cm^-3.

3.2.3 Relationship between Air Gap and Sensibility

The air gap is an important feature for the suspended-NW-channel TFTs. Air gap thickness directly influences the basic electrical characteristics such as Vth, SS and hysteresis window [15], as can be understood by comparing the results shown in Figs.

3.8 and 3.11. Therefore, the air gap is one of the major factors in affecting the gas sensing capability of the suspended-NW-channel TFTs. In this part, we consider the detection of NH3 with concentration lower than 10¹³ cm^-3 which is smaller than the detection limited mentioned above. Fig. 3.14 compares the relative conductance versus NH3 concentration extracted from devices with air gap thickness of 20 nm or 85 nm. The relative conductance means the conductance normalized to the value measured as no NH3 is introduced. As can be seen in the figure, the slope of fitting line for the 20 nm air gap is larger than that of the device with 85 nm air gap. It is attributed to the different EOT values of the devices. Figs. 3.15(a) and (b) show the

3.8 and 3.11. Therefore, the air gap is one of the major factors in affecting the gas sensing capability of the suspended-NW-channel TFTs. In this part, we consider the detection of NH3 with concentration lower than 10¹³ cm^-3 which is smaller than the detection limited mentioned above. Fig. 3.14 compares the relative conductance versus NH3 concentration extracted from devices with air gap thickness of 20 nm or 85 nm. The relative conductance means the conductance normalized to the value measured as no NH3 is introduced. As can be seen in the figure, the slope of fitting line for the 20 nm air gap is larger than that of the device with 85 nm air gap. It is attributed to the different EOT values of the devices. Figs. 3.15(a) and (b) show the