Fig. 3-1 demonstrates the diagram of an n-MOSFET with pocket implantation. As can be imagined, pocket implantation would induce non-uniform threshold voltage distribution along the channel (shown in Fig. 3-1) and may produce additional oxide traps near the source/drain edge. According to Eq. (3-1), which is illustrated in Chapter 2.2,
x dx
the drain current flicker noise will be affected as the oxide traps distribution (Nt(Efn,x)) or channel carrier distribution (N(x)) is influenced by the pocket implantation process.
3.3.1 Noise Behavior for Different Pocket Dose
In order to verify the pocket implantation effect on drain current flicker noise, two sets of identical devices only with different pocket doses are used. The higher pocket dose is 2.2 times larger then the lower one. Fig. 3-2 shows the reverse short channel effect (RSCE) for both sets of devices, which is compared to make sure the sufficient of the pocket dose difference. As can be seen, the effective threshold voltage becomes higher when the channel length shrinks. This is caused from pocket implantation induced higher local threshold voltage at the source/drain edge. However, the short channel effect (SCE) becomes dominates as the channel length is smaller than 0.3µm. As can be seen, the raise of threshold voltage induced from RSCE for the higher pocket dose set is about 0.1V, while the lower one is only about 0.05V. Based on this considerably pocket-induced difference, the noise behavior for both sets of devices is compared.
The characteristics of normalized noise power spectrum density (SId/Id2
) versus gate overdrive voltage (Vg-Vt, Vt is the device threshold voltage) is used as the monitor of noise behavior, which is considered to be a fair index because of the normalization to the drain current and comparing at the same gate overdrive voltage. Fig. 3-3,4 shows the noise behavior for two different doses. Both long channel (L=10µm) and short channel (L=0.32µm) devices are compared. As mentioned before, the SHE becomes dominant as the channel length is smaller than 0.3µm. Since we focus only on the pocket implantation effect and it is not clear how the SCE affects the drain current flicker noise, so 0.32µm is chosen as the short channel device. The noise is measured at linear operation regime (Vd=0.2V), so that the inversion charge will not affected by the drain bias. All noise data are averaged from 3 to 10 devices until the flicker noise characteristic follows 1/f. In addition, since the noise follows 1/f, the noise data point shown in Fig. 3-3,4 is only at f=100Hz. As can be shown in Fig. 3-3, the noise in the two 10µm devices is almost the same without regard to a considerably different pocket dose. Fig. 3-4 shows the noise in two 0.32µm devices with the same pocket implant split. Unlike the result in the 10mm devices, the higher pocket dose device exhibits much worse noise behavior in the entire range of gate bias. Fig. 3-5 shows the channel length dependence of pocket implant effect on drain current flicker noise. The pocket implant induced noise degradation is larger in a shorter gate length device. As for these devices, the pocket-affected region could take over an important part in the entire cannel region.
3.3.2 Verified by Charge Pumping Method
As mentioned above, the worse noise behavior for higher pocket dose devices may cause from additional oxide traps creation or non-uniform threshold voltage distribution along the channel [3-9,10]. In order to verify whether the pocket implantation process would create additional oxide traps, a charge pumping technique [3-13,14] is used. Fig. 3-6 shows that the oxide (interface) traps density of the two pocket splits is about the same (Nit approximates to
1.56×1010 at Vghigh=0.5V). It is not unimaginable because that the additional oxide traps created from pocket implantation process would be annealed during the latter thermal process.
The distinct noise degradation in the higher pocket dose device in Fig. 3-4 therefore cannot be explained simply by implant caused oxide traps creation. Instead, pocket implant will result in a non-uniform threshold voltage distribution along the channel, which should be responsible for noise degradation. A simple analytic model will be illustrated latter to explain non-uniform threshold voltage enhanced noise degradation.
3.3.3 Three-Region Model of Pocket Implant Effect on Flicker Noise
In our model, the channel is divided into three regions, as illustrated in Fig. 3-1. Regions 1 and 3 represent a pocket implant region, where the local threshold voltage (Vt) is increased due to pocket implantation. Region 2 is the rest of the channel region and possesses a lower Vt. As mention above, the number fluctuation mechanism dominates the noise behavior of n-MOSFETs at a relatively low gate overdrive bias. So the mobility fluctuation mechanism can be neglected in Eq. (3-1). In addition, since the oxide (interface) trap density is not affected by the pocket implantation process (shown in Fig. 3-6), it is assumed to be uniform along the channel. Based on these assumptions, Eq. (3-1) can be simplified as follows:
dx
As the local Vt in region 1 and 3 becomes higher due to pocket implantation, the carrier density (N(x)) would be lower in these two regions. Therefore, some parts of the summation term in Eqs. (3-2) would also become higher, and then the normalized noise level would in terms increase. This is not surprised because that the noise (number fluctuation mechanism) would become more sensitive to the variation of carrier numbers as the amount of them is relatively small. In a word, the noise for high threshold region in the channel would be the main source of the drain current flicker noise as the number fluctuation mechanism
dominates.
Based on this concept, the summation term in the noise model can be divided into three regions with their own carrier number density. So Eq. (3-2) can be rewritten as follows:
where N1 and N3 represent conducting charge density in Region 1 and Region 3, which are modulated by pocket implant dosage. In the long channel devices (L=10µm), the noise component arising from the pocket implantation regions is relatively small. This argument is evident from Fig. 3-3 that the noise is nearly the same for different pocket splits in long channel devices. In other words, the second term in Eq. (3-3), i.e., L2 region, is dominant in a
“long” channel device. From the measured noise and threshold voltage in a long channel device, the oxide trap density, Nt(Efn), can be extracted. The result is shown in Table 3-1. The measurement and calculation results of noise level for long channel devices with two different pocket dose are show in Fig. 3-7,8. The noise shows good agreement with the model in the relatively lower gate overdrive voltage regime where the number fluctuation mechanism dominates.
For the noise calculation for short channel devices, the respective parameters in the three regions must be extracted first. The effective channel length is about Leff = Lmask-0.06µm, which is extracted based on the shift and ratio method [3-15]. To obtain the respective length and local Vt in the pocket implantation regions (i.e., Region 1 and Region 3), we use the method in [3-16,17] to extract them from the reverse short channel effect of the two types of devices (shown in Fig. 3-2). Table 3-1 shows the extraction results for both sets of short channel devices. Based on Eq. (3-3), the measurement and calculation results show good agreement with each other (shown in Fig. 3-9,10).
However, on the assumption of domination of number fluctuation mechanism, the noise can be well modeled for both long and short channel devices except for a higher gate overdrive bias regime. It is believed that the mobility fluctuation mechanism (αµ) should be considered as the gate overdrive bias is relatively high. This can also be deduced from the results shown in Fig. 3-4. As can be seen, the noise increase ratio becomes lower as the gate overdrive voltage is getting higher. This is due to that the level of 1/N(x) will reduce to compatible level to am as the gate overdrive voltage is high enough. That’s why the mobility fluctuation mechanism plays a more and more important role as the gate overdrive voltage gets higher. In addition, the mobility for higher pocket dose devices would be smaller than the lower one because of the degradation mechanism of impurity scattering. So the noise increase ratio would reduce in the relatively higher gate overdrive bias regime.