Performance Dependence on Width-to-Length Ratio of Si Cap/SiGe Channel MOSFETs

Abstract

This paper measures the n-and p-MOSFETs fabricated through 65-nm high-k/metal gate CMOSFET process flow. The [110] channels of the Si cap on SiGe with different width (W ) and length (L) ratios were compared with Si-only channels. The results show that a high W-L ratio in the [110] n-channel can alleviate the degradation of biaxial compressive stress. Meanwhile, a low W-L ratio in the p-channel can improve the performance; however, the ratio should at least be below two in this paper. The dominance of the longitudinal or transverse configurations successfully explains this phenomenon because of the reliance of the different levels of piezoresistance coefficient on the channel orientation. The threshold voltage shifts versus the W-L ratio in the p-channel is in agreement with the result. The result is verified by a quantitative current comparison at a high bias in the n-/p-channels between the strained SiGe and Si-only channels, which shows that an extreme W-L ratio in the original biaxially strained SiGe channel can result in longitudinal or transverse strain, thereby leading to different levels of performance degradation/improvement.

Third generation semiconductor device research: Optimizing CMOS and HEMT designs

“…To find out the relationship between the width and length of CMOS, we should first find out the relationship of the ratio between each N-channel and P-channel. After experiment, the numerical value of the dimension interference can be calculated, which suggests that the higher value of Width makes degradation in P-channel, and the lower value of Width leads to enhancement [8]. It also observed that a shift of Threshold Voltage exists while changing the dimensions.…”

“…Come back with how CMOS works properly, we can find two ways to improve the efficiency, one is about material, and the other one is about Width Length ratio (W/L). According to the research, the way to improve CMOS performance is about the W/L, which can also be changed with applying different material [8]. In the paper, it demonstrates the Si-only interface together with SiGe interface.…”

Threshold voltage and transconductance shifting reliance on strained-SiGe channel dimension

“…The present study provides a basis for channel width-to-length (W/L) ratio comparison, which indicates that the dominance of longitudinal or transverse configuration depends on this ratio. Moreover, this ratio may explain the reason that strained long SiGe p-channels are enhanced more than short p-channels are [6]. The mobility of long p-channels is enhanced because the piezoresistance coefficient of the longitudinal [1 1 0] p-channel is positive [7], and a relatively long channel has longitudinal dominance.…”

Threshold voltage and transconductance shifting reliance on strained-SiGe channel dimension

“…The long channels are not well defined, although they are valid from 10 m to tens of nanometer [1]- [3]. A recent report provided a basis for comparison using the channel width-to-length (W/L) ratio, which indicates that dominance of the longitudinal or transverse configurations depends on this ratio; moreover, this ratio may explain the underlying mechanism that causes the strained long SiGe p-channels to be enhanced more than the short p-channel [4]. Given that the piezoresistance coefficient of the longitudinal [110] p-channel is positive [5] and that the relatively long channel has longitudinal dominance, the mobility is enhanced.…”