Книги+1 / 2013 [Chandan_Kumar_Sarkar]_Technology_CAD
.pdf
Study of Deep Sub-Micron VLSI MOSFETs through TCAD |
243 |
loaded. The device is sliced using an x-cut tool [4]. It is to be noted that the electrostatic potential values obtained through TCAD are all computed from an arbitrarily defined reference potential. In particular, for silicon, the standard approach is to set the reference potential equal to the Fermi potential of an intrinsic semiconductor. The figure illustrates that with a certain gate bias, the electrostatic potential in the semiconductor is maximum at the surface (y = 0) and gradually decreases to zero beyond the depletion depth where the bulk material is neutral.
The field patterns (i.e., the electrostatic potential contours in the depletion region of a long channel and a short channel bulk MOSFET) are shown in Figures 6.6(a) and 6.6(b), respectively. These figures are obtained after the respective structures are simulated using the tool ‘Sentaurus device’ of TCAD and then visualized using the tool ‘Tecplot SV’. The long device (Figure 6.6a) has an effective channel length, Leff of 1 µm, and the short device (Figure 6.6b) has an effective channel length of 65 nm. As seen from Figure 6.6(a), the potential contours are almost parallel to the oxide-silicon interface. The electric field is thus one-dimensional, being along the vertical direction only for almost the entire length of the channel. However, in Figure 6.6(b) the field is two-dimensional (i.e., the components of the electric field along both directions are appreciable). It is also seen that for a given gate bias, the electrostatic potential at a particular depth from the oxidesilicon interface is higher for the shorter device. In other words, the surface potential (electrostatic potential at the surface) of the shorter device is more, with a greater band bending at the oxide-semiconductor interface. This is the key difference between a short channel and a long channel MOS transistor. The depletion width is thus more for the shorter device, as seen from Equation (6.1) [5]:
Wd = |
2εSiψs |
(6.1) |
|
qNa |
|||
|
|
where Wd is the depletion width, and ψs is the surface potential.
The white colored contour in Figure 6.6(b) is the depletion width contour. The greater depletion depth in the shorter device means that the depletion charge is effectively reduced. This leads to a reduction in the threshold voltage of the shorter device. Equation (6.2) shows the dependence of the threshold voltage on the depletion charge density [5]:
Vth = VFB + 2ψB + |
Qd |
(6.2) |
|
WLeff Cox |
|||
|
|||
|
|
where Vth is the threshold voltage, VFB is the flat-band voltage, ψB is the difference between Fermi level and intrinsic level, Qd is the depletion charge
244 |
Technology Computer Aided Design: Simulation for VLSI MOSFET |
0
Y[um]
0.5 |
|
|
Electrostatic Potential [V] |
|
|
|
|
|
1.6E+00 |
|
|
|
|
1.2E+00 |
|
|
|
|
7.8E–01 |
|
|
|
|
3.7E–01 |
1 |
|
|
|
–3.6E–02 |
|
|
|
–4.5E–01 |
|
–1 |
–0.5 |
0 |
0.5 |
1 |
|
|
X[um] |
|
|
(a)
Y[um]
–0.05
0
0.05
Electrostatic Potential [V] 1.6E+00
1.2E+00
7.8E–01
0.1 |
3.7E–01 |
–3.6E–02
–4.5E–02
–0.1 –0.05 0 0.05 0.1 X[um]
(b)
FIGURE 6.6 (See color insert)
Electrostatic potential contours in (a) long channel Leff = 1 m and (b) short channel Leff = 65 nm (Vgs = 1.0V,Vds = 0.05V).
density, and Cox is the gate oxide capacitance per unit area. Thus, as Qd decreases, Vth also decreases.
The two-dimensional field pattern in a short channel device is due to the close proximity of the source and drain regions. In a short channel device,
Study of Deep Sub-Micron VLSI MOSFETs through TCAD |
245 |
the source-drain distance is comparable to the depletion width in the vertical direction [6]. The source-drain potential exerts a strong influence on the gate potential over a significant portion of the channel length. According to the charge sharing model [5,6], when the drain bias is low, all of the depletion charges beneath the gate are not imaged on the gate charges. Rather, some of them are the terminating centers of the field lines originating near the source and drain junctions. Thus there is an effective reduction in the depletion charge density that ultimately leads to the threshold voltage roll-off. This is in contrast to the long channel devices where almost all of the depletion charges are imaged on all the gate charges.
The threshold voltage roll-off as obtained from the TCAD simulator is shown in Figure 6.7. Different devices of varying effective lengths are simulated. The threshold voltages of the different devices are then determined using the tool ‘Inspect’ by the constant current (CC) technique [7]. The reference current is taken to be 10–6 A/µm. It is seen that for long channel lengths the threshold voltage remains almost constant. However, as the channel length is decreased the threshold voltage falls from its long-channel value. This is due to the gradual reduction of the gate control over the channel in the shorter devices. The drain field starts to exert its influence over the gate field in the shorter devices.
The simple assumptions of the charge sharing model, however, render it invalid for high drain and substrate biases. It is therefore unable to explain the drain-induced barrier lowering (DIBL) that is discussed next.
reshold Voltage, Vth (V)
0.6 |
|
|
|
|
|
|
|
0.5 |
|
|
|
|
|
|
|
0.4 |
|
|
|
|
|
|
|
0.3 |
|
|
|
|
|
|
|
0.2 |
|
|
|
|
|
|
|
0.1 |
|
|
|
|
|
|
|
|
|
|
|
|
Vsb = 0.1 V |
|
|
|
|
|
|
|
|
Vsb = 0.0 V |
|
0 |
0 |
200 |
400 |
600 |
|
800 |
1000 |
|
|
|
E ective Gate Length, Le (nm) |
|
|
||
FIGURE 6.7
Threshold voltage roll-off for two different substrate biases (Vgs = 1.0 V,Vds = 0.05 V).
246 |
Technology Computer Aided Design: Simulation for VLSI MOSFET |
6.4.2 Drain-Induced Barrier Lowering (DIBL)
For long channels, as seen in Figure 6.6(a), the surface potential is flat over most of the channel region of the device. The surface potential is mainly controlled by the gate voltage and acts as a barrier to the electrons (for n-channel MOSFET). The electrons are not able to surmount this barrier below the threshold condition. However, in case of short devices, the source and the drain fields penetrate deeper into the middle of the channel. This lowers the barrier between the source and the drain. The electrons are then able to overcome the reduced barrier and move toward the drain end. This increases the subthreshold current. The threshold voltage of the short channel device is thus lower than that of a long channel device. As the drain bias is increased the barrier is lowered further, resulting in a further decrease in the threshold voltage. This phenomenon is referred to as the DIBL effect. The surface potential plots as obtained from TCAD simulation for the long and the short channel devices are shown in Figure 6.8 to illustrate the DIBL effect. The shorter device has been studied with two drain biases.
In the figure, the variation of surface potential along the length of the device is obtained by using the y-cut tool of ‘Tecplot SV’ for each of the devices. The surface potential is then plotted against normalized channel length where 0 is the center of the channel, –1 is the source end, and 1 is the drain end.
It is shown in Figure 6.8 that with a very small drain bias (Vds = 0.05V), the surface potential at the source end is nearly equal to the built-in potential
1.00
0.95Le = 1 µm, Vds = 0.05 V
0.90Le = 65 nm, Vds = 0.05 V
|
0.85 |
Le = 65 nm, Vds = 0.5 V |
|
(V) |
0.80 |
|
|
0.75 |
|
|
|
Potential |
|
|
|
0.70 |
|
|
|
|
|
|
|
Surface |
0.65 |
|
|
0.60 |
|
|
|
|
|
|
|
|
0.55 |
|
|
|
0.50 |
|
|
|
0.45 |
|
|
|
0.40 |
|
|
|
0.35 |
0 |
1 |
|
–1 |
Normalized Channel Length
FIGURE 6.8
Surface potential plots as obtained through device simulation.
Study of Deep Sub-Micron VLSI MOSFETs through TCAD |
247 |
barrier, Vbi of the source-substrate junction. As the drain bias is increased (Vds = 0.5V), the surface potential at the drain end rises to (Vbi + Vds ). Thus the surface potential at the drain end rises by 0.45 V (0.50–0.05 V). Keep in mind that the surface potential values in TCAD are measured relative to a reference. The absolute values are to be calculated accordingly. The DIBL effect for the 65 nm MOSFET as obtained from the simulation results is shown in Figure 6.9. The DIBL coefficient obtained is –0.18.
The threshold voltage roll-off and DIBL can be minimized in three ways:
(1) reduction of the oxide thickness to achieve better gate control over the channel, (2) reduction of the depletion width, and (3) reduction of the sourcedrain junction depth. The depletion width is reduced by increasing the doping concentration in the channel. The source-drain junction depth is effectively reduced by introducing the source-drain extension structure.
Figure 6.10 shows the DIBL effect as obtained from the simulation results
of a MOSFET with a lower SDE depth. Leff = 65 nm, xj = 25 nm,tox = 2.2 nm, and power supply VDD = 1.2V. The threshold voltages are calculated using
the constant current technique. The DIBL coefficient in this case is calculated to be –0.11.
Figure 6.11 shows the subthreshold characteristics of the 65 nm MOSFET as obtained from TCAD simulation results. It shows plots of the drain current (in log scale) against the gate voltage for two different drain voltages. It is seen that with the increase in the drain bias, the subthreshold current increases. For very short devices, the subthreshold slope degrades because the surface potential is controlled more by the drain than the gate. Eventually, the gate loses all its control, and punch-through takes place when a high
reshold Voltage, Vth (V)
0.24
0.22
0.2
0.18
0.16
0.14 0
L = 65 nm
0.1 |
0.2 |
0.3 |
0.4 |
0.5 |
|
Drain-to-source Voltage, Vds (V) |
|
||
FIGURE 6.9
The DIBL effect, Leff = 65nm, xj = 40nm, tox = 3 nm.
248 |
Technology Computer Aided Design: Simulation for VLSI MOSFET |
reshold Voltage, Vth (V)
0.2 |
|
|
|
|
|
|
|
0.19 |
|
|
|
|
|
|
|
0.18 |
|
|
|
|
|
L = 65 nm |
|
0.17 |
|
|
|
|
|
|
|
0.16 |
|
|
|
|
|
|
|
0.15 |
|
|
|
|
|
|
|
0.14 |
|
|
|
|
|
|
|
0.13 |
|
|
|
|
|
|
|
0.12 |
|
|
|
|
|
|
|
0.11 |
|
|
|
|
|
|
|
0.1 |
0 |
0.1 |
0.2 |
0.3 |
0.4 |
0.5 |
0.6 |
|
|
|
Drain-to-source Voltage, Vds (V) |
|
|
||
FIGURE 6.10
The DIBL effect, Leff = 65 nm, xj = 25 nm, tox = 2.2 nm.
|
10–3 |
|
|
|
Vds = 0.5 V |
|
|
|
|
|
|
|
|
(A/um) |
10–4 |
|
|
|
Vds = 0.05 V |
|
10–5 |
|
|
|
|
|
|
ds |
|
|
|
|
|
|
I |
|
|
|
|
|
|
Current, |
10–6 |
|
|
|
|
|
|
|
|
|
|
|
|
Drain |
10–7 |
|
|
|
|
|
|
|
|
|
|
|
|
|
10–8 |
|
|
|
|
|
|
10–9 |
|
|
|
|
|
|
0 |
0.2 |
0.4 |
0.6 |
0.8 |
1 |
|
|
|
|
Vgs (V) |
|
|
FIGURE 6.11
Subthreshold characteristics for two different drain biases (Leff = 65 nm).
Study of Deep Sub-Micron VLSI MOSFETs through TCAD |
249 |
drain current persists irrespective of the gate voltage. The subthreshold swings (SSs) for the low and high drain biases as obtained from Figure 6.11 are 90 mV/decade and 95 mV/decade, respectively.
6.5 Mobility Degradation
The mobility of carriers in the MOSFET channel is significantly lower than that in the bulk silicon [6]. At the surface boundary, lattice (or phonon) scattering is increased due to the presence of crystalline discontinuity. At high vertical fields, surface roughness scattering severely degrades the carrier mobility. Channel mobility is also affected by the oxide and interface traps at the Si-SiO2 interface [6].
Figure 6.12 shows the plot of transconductance, gm against Vgs for the 65-nm device for two different substrate biases under low and high drain biases. Figure 6.12(a) is for a small drain bias, with Vds = 0.05 V, and Figure 6.12(b) is for a comparatively higher drain bias with Vds = 1.0 V. The transconductance plots are obtained by differentiating the curves of the Ids −Vgs characteristics. This is done by using the ‘diff’ command in the ‘Inspect’ tool. In Figure 6.12(a) it is observed that when the drain bias is low, the transconductance starts to fall off from a high value beyond a certain gate voltage. This is explained as follows. When the vertical electric field is very high (in this case it is greater than 1.6 × 106 V/cm for an oxide thickness of 3 nm and gate bias greater than 0.5 V), the mobility decreases very rapidly due to reasons discussed earlier. Thus, at low drain bias, the drain current and the transconductance are degraded significantly at high gate voltages. Also note that at low Vds, the parasitic source-drain resistance plays a significant role in determining the drain current. The peak drain current and consequently the transconductance decrease in the linear region (low Vds ) due to this resistance.
However, if the drain bias is high as in Figure 6.12(b), the transconductance does not fall off. This is due to velocity saturation. The mobility degradation of the carriers is somewhat counterbalanced by the high drift velocity of the carriers under high drain bias. For very short channel transistors, the drain current Ids becomes limited by the velocity saturation effect. This saturated drain current is given as [6]
Idsat = CoxWvsat (Vgs − Vth ) |
(6.3) |
Consequently, the transconductance is also fixed at a constant value.
It is interesting to note from Figure 6.12 that the maximum value of the transconductance at low Vds is much lower than that at high Vds. This is explained as follows.
250 |
Technology Computer Aided Design: Simulation for VLSI MOSFET |
gm (S/um)
gm (S/um)
0.00016
0.00014
0.00012
0.0001
8E–05
6E–05
4E–05
2E–05
0 0
0.001
0.0008
0.0006
0.0004
0.0002
0
|
|
|
|
Vsb = 0.5 V |
|
|
|
|
|
Vsb = 0 V |
|
0.2 |
0.4 |
0.6 |
0.8 |
1 |
|
|
Gate-to-source Voltage, Vgs (V) |
|
|
||
Vbs = 0.5 V
Vbs = 0 V
0.2 |
0.4 |
0.6 |
0.8 |
Gate-to-source Voltage, Vgs (V)
FIGURE 6.12
Transconductance versus Gate-to-source Voltage; (top)Vds = 0.05V and (bottom) Vds = 1.0V.
Study of Deep Sub-Micron VLSI MOSFETs through TCAD |
251 |
The drain current, Ids may be written as
Ids = |
Vds |
Rds + RCh |
where RCh is the channel resistance, and Rds due to the lightly doped SDE regions, or,
RCh = Vds
Ids Rds =0
(6.4)
is the source-drain resistance
or Equation (6.4) may be written as
|
|
RCh |
|
|
Ids Rds =0 |
(6.5) |
|||||
Ids = |
|
Vds |
|
= |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||
|
|
|
|
|
|
|
|
|
|
||
|
|
R |
|
|
Rds Ids |
|
|
|
|||
|
|
|
ds |
|
|
|
Rds =0 |
|
|
||
|
1 + RCh |
|
|
1 + |
|
|
|||||
|
|
|
|
Vds |
|
|
|||||
Therefore, when Vds is low, it is due to Rds , that Ids drops from its value, Ids Rds =0. Thus the maximum value of the transconductance is lowered at low drain bias.
6.6 Drain Characteristics
A significant effect that critically affects the I-V characteristics of a deep submicron MOS transistor is velocity saturation.
6.6.1 Velocity Saturation
In a short channel device, the saturation of the drain current may take place at a much lower Vds value than the Vds value of the longer device. This limits the saturation current of the device. This is illustrated in Figure 6.13 where two curves are drawn: the upper one is for the long channel, and the lower one corresponds to the shorter device. It is seen that the shorter device experiences an early saturation (i.e., the current saturates at a lower drain voltage). This occurs due to velocity saturation and is explained as follows. In the presence of velocity saturation,
1 |
= |
m |
+ |
1 |
(6.6) |
|
Vgs − Vth |
ξsatLeff |
|||
Vdssat |
|
|
|||
where ξsat is the critical electric field beyond which the velocity saturates, and m is the bulk charge factor [8]. Equation (6.6) shows that the short chan-
nel Vdssat is an average of ξsatLeff and long channel Vdssat(= VGS−Vth ). Thus short
m
channel Vdssat is smaller than long channel Vdssat. Hence the drain current for the shorter MOS transistor saturates earlier compared to the long channel value.
252 |
Technology Computer Aided Design: Simulation for VLSI MOSFET |
Drain Current, Ids (A/um)
0.0007
|
Le = 1000 nm |
0.0006 |
Le = 65 nm |
|
0.0005 |
|
|
|
|
|
|
|
|
0.0004 |
|
|
|
|
|
|
|
|
0.0003 |
|
|
|
|
|
|
|
|
0.0002 |
|
|
|
|
|
|
|
|
0.0001 |
|
|
|
|
|
|
|
|
0 |
0 |
0.2 |
0.4 |
0.6 |
0.8 |
1 |
1.2 |
1.4 |
|
|
|
Drain-to-Source Voltage, Vds (V) |
|
|
|||
FIGURE 6.13
Drain current versus drain-to-source voltage (Vgs = 1.1 V).
6.6.2 Output Resistance
Figure 6.14 shows the simulated drain characteristics of a 65 nm MOSFET. As seen in Figure 6.14, for the short channel MOSFET, the drain current increases beyond saturation. There are two reasons for this. First, due to the increase of drain voltage, threshold voltage falls and hence the current increases. This is the DIBL effect. Second, as Vds is increased beyond the saturation voltage, the saturation point (the point along the channel length where carriers attain the saturation velocity) where the surface channel collapses moves slightly toward the source [6]. That is, the conducting channel length deceases. As a result, the current increases beyond the saturation point. This is the channel length modulation effect.
The increase in the current beyond the saturation point implies that the output conductance is finite. A plot of the output resistance against the drain-to-source voltage is shown in Figure 6.15. It is seen from the curves that in the subthreshold region, the drain current is low so that the output resistance is high. In the strong inversion region, the drain current is high so that output resistance is low. In weak inversion, with the increase of drain bias, the drain current increases due to various second-order effects such as channel length modulation, DIBL effect, etc. Thus the output resistance value falls. However, in strong inversion, the increase of drain current with the increase of drain bias due to the above effects is somewhat counterbalanced by the carrier mobility degradation effect due to the applied gate bias. Therefore the overall increase of drain current is small. Hence the output resistance remains nearly constant.
