Thesis organization

This thesis is divided into six chapters. Chapter 1 introduces general background of nanowire-based sensors, motivation of this study, and Ge condensation technology. Chapter 2 describes process flow of nanowire fabrication, device structure, and the method of electrical characteristics measurement. In Chapter 3, experimental results of SiGe and Si nanowire sensors are presented and discussed. In Chapter 4, fabrication and electrical characteristics of

non-homogeneous SiGe nanowire are investigated. Conclusion is made in Chapter 5, and some proposals for future work are given in Chapter 6.

Chapter 2

Experiment

Top-down approach was used for SiGe and Si nanowires fabrication in this study. Instead of commonly used e-beam lithography, we made use of spacer formation to obtain SiGe and Si spacers on nanometer scale, namely, SiGe and Si nanowires. Ge condensation technique was utilized to increase the Ge concentration in SiGe nanowires and reduce the size simultaneously.

2.1 Process flow

A p-type (Boron doped) Si substrate (100) was used in this study. The resistivity of the silicon substrate was about 1~10 Ω-cm. Samples were prepared by following processes:

1. Standard RCA clean and wet oxidation for about 6 hours at 980°C to grow 5000Ǻ thick bottom oxide by ASM/LB45 furnance system. The structure is shown in Fig. 2-1.

2. Mask #1: Define active area. TEL CLEAN TRACK MK-8 and G-line lithography system were employed to transfer pattern onto oxide layer. Then, dry etching was carried out with TEL5000 R.I.E. system to form oxide step, which is 3000 Ǻ high and ready for sidewall spacer formation. The structure is shown in Fig. 2-2.

3. After standard RCA clean, 150Å α-Si layer was deposited on bottom oxide at 650°C for

~2 hours by ASM/LB45 furnance system. This α-Si layer served as seed layer for SiGe film deposition. The structure is shown in Fig. 2-3. This step was skipped for Si

nanowire fabrication.

4. After standard RCA clean, either poly-crystalline SiGe film was deposited by ANELAVA SiGe UHV-CVD at 665°C, or poly-crystalline Si film was deposited by ASM/LB45 furnance system. The structure is shown in Fig. 2-4.

5. Mask #2: Define S/D region and form spacer. Dry etching was carried out with TCP9400 SE poly etcher to remove unwanted part of active layer. Only the S/D region and SiGe or

Si nanowires were retained. The structure is shown in Fig. 2-5.

6. Mask #3: Remove unwanted sidewall spacer. TCP poly etcher was employed to remove unwanted spacer, which would have resulted in short circuit between two nanowire devices if not removed. The structure is shown in Fig. 2-6.

7. Phosphorus (P) or boron-fluoride (BF2) ion implantation. The implantation dose is varied from 1×10¹⁵ to 3×10¹⁵ ions/cm². Implantation energy was from 15keV to 30keV for P⁺ implantation, 50keV for BF2+ implantation.

8. Annealing in furnace at 900°C for 30min to activate dopants.

9. 3000Ǻ Aluminum deposition by thermal coater

10. Mask #4: Define aluminum contact pad. Al pads are formed by wet etching (HNO3:CH3COOH:H3PO4:H2O=2:9:50:10). The structure is shown in Fig. 2-7.

11. Al sintering at 400℃ in N2 ambient for 30 minutes.

2.2 Functionalization

First, we used amino-propyl-trimethoxy-silane (APTMS) to modify the surface of native

oxide layer around nanowires. Hydroxyl functional groups on the surface of native oxide layer were replaced by the methoxy groups of APTMS molecules. After APTMS modification, the surface of nanowire was terminated by amine groups. In our experimental environment, amine groups were prone to be positively charged; that is, the surface potential nanowire increased, and the conductance of n-type (p-type) nanowires increased (decreased). Next, we used bis-sulfo-succinimidyl suberate (BS3) as linker between APTMS and IgG antibody. BS3 treatment resulted in negative charges. Hence, the conductance of n-type (p-type) nanowires decreased (increased). After APTMS and BS3 modification, nanowires were capable of capturing IgG antibody. Our group has already demonstrated IgG sensing capability of SiGe nanowire sensor. In this study, we focused on investigating the change of conductance (ΔG) and sensitivity (S) of different samples under identical surface condition. Comparison was made by using ΔG and S corresponding to APTMS treatment since the situation became more complex, and the variation of experimental data was larger after BS3 treatment.

2.3 Measurement of electric characteristics

HP4156A was used in this study to measure the electric characteristics of nanowire

sensors. Drain voltage (VD) was varied from -8V to 8V and 200mV a step, and back gate voltage (VG) was 0V. The measurement of electric characteristics was performed at every stage of surface modification, and the average conductance was then extracted from ID-VD

characteristics with VD=3~6V.

Chapter 3

SiGe and Si Nanowire Sensors

3.1 Cross-section view of poly-crystalline SiGe and Si nanowires

Si0.8Ge0.2 and Si nanowires were successfully fabricated by sidewall spacer formation.

Fig. 3-1 and Fig. 3-2 are the SEM images of Si0.8Ge0.2 and Si nanowires respectively. The average height of Si0.8Ge0.2 nanowire was 175nm, and the average width was about 60nm.

The average height and width of Si nanowire were about 160nm and 90nm respectively.

According to SEM images, the surface area of Si0.8Ge0.2 nanowire was close to that of Si nanowire, and the volume of Si0.8Ge0.2 was smaller than that of Si nanowire.

3.2 Electrical characteristics of SiGe and Si nanowire sensors before and after APTMS

and BS3 modification

Both p-type and n-type nanowires were fabricated. Fig. 3-3 is the ID-VD characteristics of a p-type Si0.8Ge0.2 nanowire device, which is 13μm in length. Normal denotes the data recorded from the blank nanowire. Conductance extracted from ID-VD characteristics is shown in Fig. 3-4.

After APTMS modification, the conductance of this p-type Si0.8Ge0.2 nanowire device decreased because APTMS molecules were positively charged. Then the p-type Si0.8Ge0.2 nanowire underwent BS3 modification, and the conductance increased simultaneously since BS3 molecules possessed negative charges. The ID-VD characteristics and conductance of a n-type Si0.8Ge0.2 nanowire device are shown in Fig. 3-5 and Fig. 3-6 respectively. The length of this

nanowire was 8μm. The n-type Si0.8Ge0.2 nanowire exhibited opposite behavior to p-type nanowire. The conductance of n-type Si0.8Ge0.2 nanowire increased after APTMS modification and decreased after BS3 treatment. This complementary sensing behavior was consistent with the fact that APTMS molecules were positively charged, and BS3 molecules were negatively charged. Moreover, complementary conductance changes verified that the response of Si0.8Ge0.2 nanowire sensor was indeed the result of specific binding of APTMS molecules and BS3 molecules. Besides, the capability of complementary sensing is important because signals due to electrical noise and non-specific binding could be filtered off if we fabricated both n- and p-type SiGe nanowires on the same chip.

The ID-VD curves and conductance of a 9μm-long p-type Si nanowire are shown in Fig.

3-7 and Fig. 3-8 respectively, and the electrical characteristics of a 17μm-long n-type Si nanowire are shown in Fig. 3-9 and Fig. 3-10. The ID-VD curves of these devices were all linear, which indicated that it was ohmic contact between Al electrodes and Si0.8Ge0.2 or Si.

Sensing behavior of Si nanowires was similar to that of SiGe nanowires.

To clarify the difference between SiGe and Si nanowires, the change of conductance (ΔG ) of several nanowires on each chip after APTMS modification were shown in Fig. 3-11. In our experiment, nanowires with length in a range from 3μm to 50μm were fabricated in a chip. In order to compare ΔG between nanowires, we normalized ΔG by multiplying ΔG and the length of nanowire (LNW) together. According to Nair’s simulation results[66], the length of nanowire was reported as a factor of sensitivity; however, the impact of length variation on sensitivity should be neglectable in our experiment because a large amount of positive charges

were induced by APTMS modification and covered entire nanowire.

Complementary sensing is clearly seen in Fig. 3-11. P-type Si0.8Ge0.2 and Si nanowires exhibited negative conductance change while the conductance change of n-type nanowires was positive. Average ΔG of p-type Si0.8Ge0.2 nanowires was the greatest in magnitude among all samples. It was 74% larger than the ΔG of p-type Si. As for n-type Si0.8Ge0.2 and Si nanowires, ΔG of n-type Si0.8Ge0.2 was 36% larger than that of n-type Si. Both p-type and n-type Si0.8Ge0.2 nanowires provided larger ΔG than Si nanowires. According to Equation (1.2), since the surface areas of Si0.8Ge0.2 and Si nanowires were very close, there were two possible reasons for this phenomenom: higher surface charge density or higher carrier mobility in Si0.8Ge0.2 nanowire. Surface charge density is related to the difference of surface chemistry between SiGe and Si nanowires. Si-OH and Ge-OH groups co-existe at the surface of SiGe nanowire while there is only Si-OH groups on the surface of Si nanowire. Afer silanization reaction, the number of APTMS molecules binding on surface is likely to be different between SiGe and Si nanowires.

Besides, we observed that p-type nanowires generated larger ΔG than n-type nanowires.

It’s attributed to the difference between depletion and accumulation. Nanowires operating in depletion mode exhibit larger ΔG than those in accumulation mode[66]. In this experiment, APTMS modification led to depletion for p-type nanowires and accumulation for n-type nanowires.

Fig. 3-12 shows the baseline conductance (G0) of SiGe and Si nanowires. Although the ion implantation dosages for p-type and n-type doping were the same, conductance of p-type

nanowires was much larger. This result may be duo to higher degree of dopant activation or higher mobility in p-type nanowires. Higher hole mobility is considered as the main reason that p-type SiGe nanowires exhibited much higher conductance than p-type Si[73-75]. The baseline conductance values of n-type SiGe and Si nanowires were about the same. Since the volume of Si nanowires was slightly larger, dopant activation and electron mobiliy were likely to be promoted in SiGe nanowires.

The conductance change corresponding to BS3 molecule binding of APTMS-modified nanowires depends on the number of APTMS molecules on the surface of nanowire. We estimated the percentage of BS3-linked APTMS molecules from the ration of ΔG. 30~60%

APTMS molecules on the surface of nanowires were bound to BS3 molecules in our experiment. It’s impossible to link every APTMS molecule to BS3 molecule because of electrostatic repulsive force. The optimum modification condition requires further investigation.

3.3 Sensitivity of Si0.8Ge0.2 and Si nanowire sensors to APTMS

Fig. 3-13 shows the sensitivity (S) of Si0.8Ge0.2 and Si nanowire sensors. Although p-type Si0.8Ge0.2 device generated the largest ΔG, the sensitivity of p-type Si0.8Ge0.2 device was the lowest because the baseline conductance was much larger than other devices. N-type devices exhibited higher sensitivity than p-type devices because of smaller baseline conductance.

N-type Si0.8Ge0.2 device had the highest sensitivity (20.5%) in our experiment. These results were similar to the experimental data reported by Li and co-workers[76]. In their experiment,

p-type Si nanowire sensor exhibited higher conductance and conductance change. However, n-type sensor provided larger sensitivity.

3.4 Conductance change and sensitivity of blank nanowires to DI water and BS3

We also treated as-fabricated Si0.8Ge0.2 and Si devices with deionized (DI) water and recorded the change of conductance, which is shown in Fig. 3-14. The conductance of nanowires tended to increase after DI water treatment. Possible causes of this result includes non-intentional binding, passivation effect[77], pH variation, contamination of container, and impurities in DI water. However, the magnitude of ΔG was much smaller than the response to APTMS binding.

The response of blank nanowires to BS3 solution is shown in Fig. 3-15. The conductance of p-type Si0.8Ge0.2 nanowires decreased while other nanowires generated both negative and positive ΔG. Conductance variation of blank nanowires after BS3 treatment was attributed to non-specific binding. It’s much smaller than the response of as-fabricated nanowires to APTMS molecules and the response of APTMS-modified nanowires to BS3 molecules.

Fig. 3-16 ~ 3-19 show the summary of sensitivity of p-type Si0.8Ge0.2, n-type Si0.8Ge0.2, p-type Si, and n-type Si nanowires respectively. The fact that sensitivity to APTMS molecules was apparently higher than to DI water and BS3 proved that the conductance changes shown in Fig. 3-11 were indeed resulted from specific binding of APTMS molecules, and APTMS modification was necessary for BS3 binding.

Chapter 4

Non-homogeneous SiGe Nanowire Sensors 4.1 Achieving non-homogeniety by dry oxidation at 900℃ for 4 minutes

Si0.93Ge0.07 nanowires were used in this experiment. The width of nanowires was about 75nm, and the height was about 210nm. In order to obtain Ge-rich layer at the outer region of nanowires, Si0.93Ge0.07 nanowires underwent dry oxidation at 900℃ for 4 minutes. We set the oxidation temperature at 900℃ to suppress inward diffusion of Ge atoms, aiming at confining Ge atoms at the outer region of nanowires. After oxidation, the width of nanowires was around 55nm, and the height of nanowires was about 190nm. Fig. 4-1 is the SEM image of one of the nanowires after oxidation. The undercut below the nanowire was due to the underlayer oxide etching during wafer cleaning. Fig. 4-2 is the Auger depth profile after oxidation. Although the real Ge concentration can’t be measured because of the lack of standard sample for calibration, it’s obvious that Ge concentration at the surface was greatly increased after oxidation, and SiGe was clearly divided into two layers according to concentration difference. Through this oxidation process, we successfully obtained non-homogeneous SiGe nanowire.

4.2 Electrical characteristics of Si_0.93Ge_0.07 nanowires after oxidation

Fig. 4-3 is the ID-VD characteristics of 8-μm-long nanowire. The curves were still linear after oxidation. Fig. 4-4 shows the conductance at different stages of surface modification.

After APTMS modification, we observed 10% conductance change.

4.3 Effect of annealing on conductance change

A lot of articles regarding Ge condensation technology reported that Ge pile-up and defects generation are concomitant events during SiGe oxidation[78]. Therefore, we carried out in-situ annealing after oxidation to investigate the effect of annealing after oxidation. SiGe nanowires were in-situ annealed in N2 ambient at 1000℃, 900℃, and 700℃. Fig. 4-5 is the schematic representation of oxidation and annealing process. After that, samples underwent phosphorus-ion implantation with the does of 3x10¹⁵cm^-2. The ion energy was 15keV. We utilized RTA process as dopant activation for this experiment to prevent diffusion of Ge atoms;

samples were annealed at 900℃ for 30 seconds.

First, we investigated the effect of 1000℃ annealing. ΔG of samples that were annealed for 5 minutes and 10 minutes, and without annealing are shown in Fig. 4-6. It was observed that ΔG of samples experienced 1000℃ annealing was smaller than that of the sample without annealing, and the performance of the 10-minute annealed sample was worse than that of the 5-minute annealed sample. We attribute this phenomenon to the diffusion behavior of Ge atoms. Fig. 4-7 is the Auger depth profile of Si, Ge, and O for the 10-minute annealed sample respectively. Ge diffusion became prominent at 1000℃, and Ge atoms diffused away from surface after 1000℃ annealing; therefore Ge-pileup was not observed in Fig. 4-7. Since oxide layer is considered as diffusion barrier for Ge atoms, out-diffusion could be ignored, and therefore the Ge diffusion is mainly from surface to the bulk of nanowires. After 10-minute

annealing at 1000℃, Ge distribution was quite uniform as shown in Fig. 4-7. Since Ge concentration near the surface was lowered after annealing at 1000℃, the conductance change caused by APTMS modification was lowered as well.

The effects of annealing at 900℃ and 700℃ were also investigated. ΔG of samples underwent 10-minute and 30-minute annealing at 900℃ and 60-minute annealing at 700℃

are plotted in Fig. 4-8. Since the Ge diffusion was very slow at 900℃, there was no sensitivity degradation after 900℃ annealing. On the contrary, ΔG was increased after annealing. This is attributed to the fact that defects that were generated during Ge condensation were healed out by annealing process, and the quality of the outer layer of nanowires was improved. As a result, the change of conductance was increased. Annealing for 30 minutes led to greater improvement than annealing for 10 minutes because more defects were healed. Fig. 4-9 is the Auger depth profile after annealing at 900℃ for 30 minutes. The Ge-rich region was slightly broader than that before annealing. Instead of spreading uniformly over the nanowire, which was the case of 1000℃ annealing, annealing at 900℃ for 30 minutes made the Ge distribution become locally uniform while maintaining Ge-rich layer next to the surface.

Annealing at 700℃ for 60 minutes also resulted in greater response because Ge atoms barely diffused inward, and some defects were recovered. However, 700℃ annealing was insufficient for healing all defects; namely, some defects could only be recovered at higher temperature when more energy is provided. Therefore, the response of nanowires annealed at 700℃ for 60 minutes was slightly greater than that of nanowires annealed at 900℃ for 10 minutes but still smaller than the conductance change of nanowires annealed at 900℃ for 30

minutes.

4.4 Effect of annealing on conductance of nanowires

Fig. 4-10 shows the conductance of samples underwent different annealing process. In our experiment, there is no direct relationship between conductance and annealing process.

Our explanation is that the size of nanowire is the dominant factor of conductance for wires in this scale. However, the oxidation process is difficult to control since nanowires we fabricated are polycrystalline, and oxidation of SiGe is very fast. Besides, the uniformity of the thickness of Si0.93Ge0.07 film deposited by UHVCVD also has great impact on conductance. We believe that the variation of conductance is the result of size variation from SiGe deposition and oxidation. Since conductance is related to the cross-section area that could be estimated by the product of width and height of nanowire in our experiment, it is greatly influenced by size variation.

4.5 Annealing at 900℃ for 30 minutes after ion implantation

In order to simplify thermal processes, we carried out 900℃ annealing as dopant activation instead of in situ annealing. Fig. 4-11 shows normalized ΔG of A and B sample. A sample was in situ annealed at 900℃ for 30 minutes after oxidation, and B sample was not in situ annealed but instead underwent 30 minutes 900℃ annealing after ion implantation. ΔG of B sample was about 25% larger than that of A sample. This improvement was attributed to the distribution of phosphorus atoms. After phosphorus ion implantation, both A and B sample

underwent 900℃ annealing. A sample and B sample were annealed for 30 seconds and 30 minutes respectively. Since the implantation energy in this experiment was quite low, the peak of phosphorus atom distribution in A sample was very close to the suface. As for B sample, diffusion of phosphorus atoms was more prominent, and the concentration of phosphorus at the surface was lowered. Consequently, electron mobility and ΔG increased.

Conductance of A and B sample was about the same, as shown in Fig. 4-12. Annealing at 900℃ for 30 minutes is likely to cause more dopant segregation at grain boundaries and more out-diffusion; hence, conductance did not increase for longer duration of annealing.

Sensitivity of A sample and B sample were 21%, and 25.5% respectively.

According to our observation, annealing at 900℃ for 30 minutes as dopant activation always led to higher ΔG than rapid thermal annealing. However, in some cases, nanowires that underwent 30 minutes annealing exhibited higher conductance and, hence, lower sensitivity. It depends on doping types, annealing temperature, and annealing duration.

4.6 Estimation of surface charge density

The capacitance between Al electrode and back gate was 2.175x10^-8F/cm², and the

The capacitance between Al electrode and back gate was 2.175x10^-8F/cm², and the