Independent Double-gated Operation
In addition to optimize the fabrication and the structural parameters of the
devices, the use of circuit technique to suppress the VTH fluctuation represents another useful approach [40]. In this work various operation modes are investigated to address the feasibility. As we mentioned in previous chapter, DG-TFTs can be operated in three modes. This also applies to the NW-DGTFTs since the devices are configured with two independent gates. From the cross-sectional TEM images shown in Fig. 4-1, it could be seen that NW is surrounded by the top gate and inverse-T gate. In this case, when the top gate and inverse-T gate are both connected together to serve as the driving gate, it is denoted as the DG mode. When sweeping bias is applied to the top gate while the inverse-T gate is grounded, the mode is called the top-gate (TG) mode.
When the bias conditions used in the TG mode is interchanged, the mode is called the inverse-T gate (ITG) mode.
Figure 4-15 shows the VTH standard deviation in various operation modes. It is found that standard deviation of VTH is the largest in TG mode with the ITG mode falls in the middle among the three operation modes. Since the NW channel is positioned on the upper-step corners of the inverse-T gate (shown in Fig. 4-1), the effective conduction widths for ITG mode is larger than that of TG mode, resulting in a better gate controllability and a less VTH standard deviation. In DG mode, VTH
standard deviation is expected to be the least due to the best gate controllability among the three operation modes
In our device structure, VTH of the single-gate mode can be modulated by the voltage applied to the control gate (e.g., the inverse-T gate in TG mode, and the top gate in ITG mode). In this work we also address the VTH-fluctuation issues in these operation conditions. Figure 4-16 shows the distribution of transfer curves of twenty five devices characterized under TG and ITG modes, and Fig. 4-17 shows the VTH-fluctuation results as a function of VTH-control gate voltage in TG and ITG modes. It is found that the VTH standard deviation is reduced when VTH-control gate voltage is applied with -1V for the two single-gate modes. Moreover, such improvement in VTH-fluctuation is especially profound as the channel length is scaled down.
It is worth noting that the threshold voltage of the NW-TFTs device under DG mode of operation, VTH(DG), is about 0.42V. Here we consider the operation of the ITG mode when 1V is applied to the top-gate. The device now becomes a normally-on device since an inversion layer will be formed at the channel interface adjacent to the top gate surface. To turn off the device, the voltage applied to the inverse-T gate must be sufficiently negative to deplete the aforementioned conduction channel. Such situation will increase the effective gate dielectric thickness which is about the sum of the thickness of inverse-T gate oxide plus the fully depleted body, thus deteriorating the short-channel effect. As a result, a larger VTH-fluctuation is observed in
short-channel devices. This explains the results shown in Fig. 4-17 that VTH standard deviation in devices with channel length of 0.8μm is degraded when 1V is applied to the control gate. For long-channel devices, VTH standard deviation is almost independent of the voltage applied to the control gate since the short channel effect is not obvious. When top-gate voltage is less than VTH(DG), the conduction channel is formed solely at the channel interface adjacent to the inverse-T gate, and short channel effect can be suppressed. Thus, the VTH-fluctuation will be less sensitive to the channel length.
Besides, since the distance between inverse-T gate and inversion layer includes the fully depleted body at VTG > VTH(DG), the variation of NW body thickness is another factor that affects the VTH standard deviation. To further explain this phenomenon, we recall the model mentioned in pervious chapter:
Case 1: ( 1) 1
In the two cases, the main difference is the channel body thickness, Tsi. For Case 1 (corresponding to the case when -1V is applied to the top gate), the standard deviation of Tsi is part of the denominator, which will lead to a reduced standard deviation on VTH shift rate. For the Case 2 (corresponding to the case when 1V is applied to top gate), the standard deviation of Tsi is part the numerator, which will
amplify the VTH-fluctuation.
However, there exists an optimum top-gate (control gate) bias for minimizing the fluctuation. A high negative top-gate bias will result in a large vertical electric field in the Si channel. This will amplify the impact of small Tsi (i.e., channel thickness of the Si channel) variation on the modulation of the channel potential, thus leading to a larger VTH variation [45]. An example is shown in Fig. 4-18. We can see that the optimized voltage applied to the top-gate (i.e., the control gate) is in the range from 0 to -1V
Figure 4-19 shows the standard deviation of VTH under the TG mode, compared with the DG mode. It is seen that the DG mode has a better suppression in VTH
fluctuation, owing to the reduction in effective depletion width. Whereas TG mode with VITG=1V has the worst VTH variation. As mentioned above, this could be ascribed to the increase in effective gate dielectric thickness.