Results and discussion

2.4.1 Characteristics of Poly-Si TFTs on FSG layer

Figure 2-3 shows the transfer characteristics (IDS-VGS) of the conventional and the proposed poly-Si TFTs with different FSG buffer layers. The measurements was performed at drain voltage of VDS=5V. The measured and extracted parameters from the devices are listed in Table 2-2. The threshold voltage, subthreshold swing, on-current (VGS=25V), and off-current (VGS=-10V) were measured at VDS=5V.

In the Fig. 2-3, we can see that the poly-Si TFTs fabricated on the FSG buffer

layers exhibit better on-state and off-state characteristics than those of the control sample. Notably, under a large negative gate bias, the leakage currents of the TFTs on FSG layers (3.08×10^-9A, 3.32×10^-10A, and 9.95×10^-10A for FSG1, FSG2, and FSG3, respectively) are over one order of magnitude lower than that on the conventional oxide buffer layer (1.14×10^-7A). This is attributed to the facts that the reduced the traps by the incorporation of fluorine in the poly-Si films during ELA [13] and the released tensile stress at the poly-Si/buffer-oxide-layer interface [16]. Moreover, the threshold voltage and subthreshold swing of the poly-Si TFTs on FSG layers (4.77V

& 1.422V/dec., 4.82V & 1.44V/dec., and 4.95V & 1.45V/dec. for FSG1, FSG2, and FSG3, respectively) were found to be superior to those of the control sample (5.07V

& 1.55V/dec.).

It has been revealed that the threshold voltage and subthreshold swing are more sensitive to the density of deep trap states near midgap associated with the dangling bonds [4]. For this reason, it is inferred that the dangling bonds within the poly-Si channel and SiO2/poly-Si interface were effectively passivated by fluorine atoms.

Figure 2-4 shows the field-effect mobility of the conventional and the proposed poly-Si TFTs. They were found to be 63.9cm²/V.s, 69.1cm²/V.s, 65.4cm²/V.s, and 43cm²/V.s for control, FSG1, FSG2, and FSG3, respectively. As can be seen, except FSG3, the field-effect mobility was improved by the incorporation of fluorine atoms within the poly-Si channel. It was demonstrated that the field-effect mobility is significantly influenced by the tail states near the band edge caused by the strain bonds in poly-Si and SiO2/Si interface [4]. Therefore, we believe that fluorine atoms introduced into poly-Si channel can not only passivate the dangling bonds, but also release the strain bonds within the poly-Si channel and SiO2/poly-Si interface.

However, for the degradation phenomenon in FSG3, we will explain later.

The evidence of the fluorine incorporation can be firmly demonstrated with the

SIMS profiles of fluorine, as shown in Fig. 2-5. It was clearly observed that considerable fluorine atoms were introduced and confined in the poly-Si for the FSG samples and, in particular, two fluorine peaks were located at the top and bottom interfaces. Therefore, we believe that the grain boundary trap states, both the top and bottom interface states were terminated by fluorine atoms. Moreover, in order to verify the effect of fluorine passivation, the effective trap state density (Nt) was calculated. Figure 2-6 shows the plot of ln(IDS/VGS) versus 1/VGS at low drain voltage and high gate voltage for all samples. The effective trap state density caculated form the slopes for the control, FSG1, FSG2, FSG3 were 5.64×10¹²cm^-2, 3.91×10¹²cm^-2, 3.97×10¹²cm^-2, and 4.01×10¹²cm^-2, respectively. These figures strongly hint that the fluorine can effectively passivate the present trap states in poly-Si channel region.

However, the FSG3 shows a detrimental effect on the performance of the resulting TFT. This is attributed to the moisture absorption. According to previous report, the moisture absorption increased with increasing fluorine content in the FSG layers [21]. The absorbed moisture would easily form OH or react with fluorine to form HF, which in turn deteriorate the devices and result in the degraded performance and reliability [22].

2.4.2 Uniformity and Reliability of Poly-Si TFTs on FSG layer

Figure 2-7 displays the statistical distributions of the field-effect mobility (µeff), the leakage current (Ioff), and the threshold voltage (Vth) of the conventional and the proposed poly-Si TFTs with different FSG buffer layer deposition conditions. The vertical bars in the figure indicate the minimum and maximum values of the devices characteristics and the squares present the average values. The average values of the µeff for the control, FSG1, FSG2 and FSG3 samples were 57.7cm²/V.s, 66.7cm²/V.s,

63.9cm²/V.s, and 45.2cm²/V.s with standard deviations of 4.05, 2.98, 3.09 and 4.15, respectively. This tendency indicates that with moderate fluorine content in FSG layers the average values and the deviations of µeff can be greatly improved. Also, the average values of the Ioff for the control, FSG1, FSG2, and FSG3 samples were 6.8×10^-8A, 7.8×10^-9A, 1.3×10^-9A, and 1.8×10^-9A with standard deviations of 8.14×10^-8, 2.46× 10^-9, 6.55×10^-10 and 1.93×10^-9, respectively. The average of the Vth

for the control, FSG1, FSG2, and FSG3 samples were 5.50V, 5.2V, 4.89V, and 4.88V with standard deviations of 0.21, 0.037, 0.22 and 0.13, respectively. These results imply that fabricating poly-Si TFTs on FSG buffer layers with moderate fluorine concentration can improve the device performance as well as the uniformity of its characteristics. The uniformity of the poly-Si TFTs is strongly affected by the random distribution of grain boundaries. Therefore, using fluorine to terminate those trap states within poly-Si channel can effectively alleviate the influence of grain boundaries.

Then, we would discuss the reliability issue of the conventional and the proposed poly-Si TFTs with different FSG buffer layer deposition conditions. The hot-carrier stress was performed at VD,stress=20V, VG,stress=10V, and source electrode grounded for 1000sec to investigate the device reliability. Figure 2-8 plots the variations of on-state current (Ion), threshold voltage (Vth), and field-effect mobility (µeff) over stress time. The variations of Ion, Vth,and µeff were defined as (Ion,stressed－ Ion,initial)/Ion,initial×100%, (Vth,stressed－Vth,initial)/Vth,initial×100%, and (µeff,stressed－µeff,initial) /µeff,initial×100%, respectively, where Ion,stressed, Vth,stressed, µeff,stressed, Ion,initial, Vth,initial, and µeff,initial represent the measured values before and after electrical stress.

Notably, the control shows relatively large variations in Ion, Vth, and µeff after 1000sec stress, whereas the FSG2 has slight change in Ion and Vth and stays almost unchanged in µeff. These results imply that poly-Si TFTs fabricated on the FSG layer

greatly reduced the device degradation under hot carrier stress, which is due to the formation of the rather strong Si-F bonds. Since the calculated percentages of fluorine content in the FSG layers are 2%, 4% and 7% for FSG1, FSG2, and FSG3, respectively, we deduce based on the above experimental results that the trap states can be effectively passivated when the fluorine content in the FSG is above 2%; while the absorbed moisture in the FSG as the content of fluorine is above 4% starts to induce visible corrosion of the poly-Si structures after competing with the trap states termination. Definitely, the corrosion becomes more severe as the content of fluorine reached 7%. As a result, the optimized condition of fluorine content of FSG is probably within 2% to 4%.