Chapter 2 Experiment
2.2 Measurement Method and Experimentally Assessed Effective Inversion-Layer
In the measurement method, the conventional inversion-layer mobility is usually extracted according to
where G
d
is drain conductance. When Ig is small sufficiently, G d
can be presented byc h
d
I V
, and the channel current (I
ch
) should be the same as drain current (Id
) and source current (Is
). However, while gate oxides are thin enough to encounter direct tunneling current in long channel device, high gate leakage current would affect the accurate where IGS
and IGD
are the current from the source to the gate and the current from the drain to gate, besides Is<0 and Id
>0. Figure 2.6 illustrated that Is
is larger than the current from the source into the channel (Ich
) due to the current flows from source to5
gate(I
GS
), but Id
is smaller than the current from channel to gate (Ich
) because the current tunnels from drain into gate (IGD
). When Vd
is sufficiently small, the IGS
must be the same as the IGD
, hence the channel current can be defined as2
Therefore, inversion layer (channel) mobility is measured by
Considering the difference between L
mask
and the metallurgical channel length Lm
, as well as the issue about the parasitic source/drain resistance (Rsd
), Eq.(2.5) could be defined more accurately by
6
Chapter 3 Inverse Modeling
In this section, we make use of a TCAD tool named Sentaurus to reproduce I
ch -V g
, especially for the subthreshold conduction. This procedure is so-called inverse modeling. First, we use the parameters which is got from C-V fitting: n+
polysilicon doping concentration Npoly
(= 11020
cm-3
), gate oxide (SiO2
) physical thickness tox
(=1.27 nm), and p-type substrate doping concentration P
sub
(= 41017
cm-3
). Second, we add two extra peak doping impurities: the source/drain extension Nsde
and the halo implant Phalo
. Remarkably, a fairly good fitting was achieved for different gate lengths, different temperatures, and different drain voltages, all with the same source/drain extension doping concentration Nsde
(= 4.95 1020
cm-3
) and halo implant Phalo
(=2.5 10
19
cm-3
). Figure 3.1 to Figure 3.4 show Ich -V g
fitting results. Ich -V g
calibration leads to device doping profiles for Lm
=1μm, 48nm and 33nm as shown in Figure 3.5 to Figure 3.7. Corresponding extension overlap and hence the metallurgical channel length Lm
can be drawn.TCAD further delivers the inversion-layer charge density qN
inv
as well as the source/drain series resistanceR
sd . The method we extracted those parameters will be illustrated clearly as follows. Also, we will introduce the physical model (drift-diffusion) we used in above calibration.3.1 Drift-diffusion Model
The drift-diffusion model is widely used for the simulation of carrier transport in semiconductors and is defined by the Poisson equation and continuity equations. The
7
three governing equations for charge transport in semiconductor devices are Poisson equation and the electron and hole continuity equations. Poisson equation is:
trap
is the concentration of ionized acceptors, and
trap is the charge density contributed by traps and fixed charges.The electron and hole continuity equations are
t
where
R
net is the net electron–hole recombination rate,j
n
is the electron current density, and
j
p
is the hole current density.
Combining equation 3.1~ equation 3.3, we can derive current densities for electrons and holes as given by: electron and hole quasi-Fermi potentials, respectively.
8
3.2 Calculation of Inversion Layer Charge Density
In this section, we will introduce the method we calculated inversion layer density from TCAD calibration model. However, we want to explain why we don’t use the same method as we used in extraction for additional mobility from gate plasmons in long-channel device [8]. From Figure 3.5 to Figure 3.7, we can observe that when channel length is shrunk, the substrate doping concentration under the SiO
2
/Si interface is obviously larger. This is due to the halo implant Phalo
in short-channel device. The altered substrate doping concentration may influence the inversion layer density directly and further bring about mobility degradation [3]. Consequently, 1D simulator [10] is not enough in this work. Instead, we use 2D simulator TCAD [12] to calculate inversion charge layer density involving the halo implant effect.First, we use the calibration model to see the free electron density in vertical direction as shown in Figure 3.8. Then we integrate the electron density vertically under the oxide/substrate interface to 30nm deep, which is the quantum confinement region. So, the area under the e-density curve in Figure 3.8 is the N
inv
at this x position.Then we cut L
m
=1μm, 48nm and 33nm devices into many pieces to do the integration, and in order not to contact the source/drain extension in the calculation, we only counted the channel between them. The region we made the calculation is between two dashed lines which is depicted in Figure 3.9 to Figure 3.11. After that, we did the same calculation for different gate voltages from 0V to 1.8V and temperatures from 292K to 380K. Corresponding inversion layer charge density results are shown in Figure 3.12 for Lm
=1μm, Figure 3.13 for Lm
=48nm and Figure 3.14.for Lm
=33nm.Since the N
inv
for short-channel device is not a constant value, unlike the long-channel device. We further did an average for the inversion layer charge density, i.e. we9
integrated the N
inv
along the channel and then divided the channel length. Finally, we get the Ninv
for Lm
=1μm ,48nm and 33nm versus gate voltage as shown in Figure 3.15, which is the Ninv
mentioned in the mobility calculation in Euation 2.6.3.3 Extraction of Parasitic Source/Drain Resistance
Since the parasitic resistance cannot be neglected in short channel device, we will explain the method we use in this work. First, we used the calibration model from TCAD simulation as shown in Figure 3.5 to Figure 3.7. Then, we only retained the drain region and constructed a metal contact beneath the oxide/substrate interface to 1 and 3nm deep. For this metal contact, one side connects the device and the other side is grounded as shown in Figure 3.16. After that, device is operated at drain voltage of 0.05V. Hence, we can get the current flow into the metal contact when we applied gate voltage. Therefore, we can derive
R
sd as below:I V R
sd
2
. (3.6) Figure 3.17 shows theR
sd under various bias conditions from 0.8V to 1.8V with the values of around 100~110Ω -μ m.Besides, we also provide another calculation method for
R
sd extraction. The method of extracting parasitic source/drain resistance (R
sd ) is well described elsewhere [13]. TheR
sd can be derived as follows. For the intrinsic MOSFET operated in linear region, drain current can be expressed as10
While minimizing the error between μ
ana
and μuniversal
as described in [13]. Herein,
ana is approximated through iteration to minimize its error. Figure 3.18 shows the I-V curves for short-channel devices whereV
bs1 =0V andV
bs2 = -0.4V. The result for extractingR
sd value is derived in Figure 3.19, and its value is about 123Ω -μ m for11
ana = 0.33, which is reasonable and consistent with the value (
ana 1/2~1/3) used in [13]. Also, theR
sd value obtained by using this method is reasonably consistentwith the extracted one (
R
sd =120Ω-μm) in [14]. Thus, we useR
sd =100~120Ω-μm to calculate the mobility for short-channel device.According to inverse modeling, the extension overlap and hence the metallurgical channel length L
m
are easy to obtain. Combining the inversion layer charge densityN inv
and source/drain resistanceR
sd , the measured effective mobility (μeff
) are obtained as demonstrated in Figure 3.20 versus inversion layer density Ninv
for differentL
m withR
sd of 100 to 120Ω-μm as parameters.12
Chapter 4
Analysis and Discussion
With the inverse modeling in Chapter 3, we extracted the mobility with
R
sd=100~120Ω -μ m for different metallurgical channel length L
m
of 1μ m, 48nm and 33nm. In Figure 3.20 we found the mobility degradation in short-channel device. Due to the controversial mechanism for the mobility degradation in short-channel device, we want to use the temperature-dependent experimental method as in [8] to determine the main source of degradation. First, we extracted additional mobility
add( 48 nm )
and
add( 33 nm )
of short-channel device with respect to long-channel one (Lm=
1μ m), and further obtained their temperature dependencies at high Ninv
(= 1x1013
cm-2
).Besides, we also extracted the extra additional mobility
add ,extra and decided the main source of this scattering mechanism. Moreover, we further provided an important evidence to verify our hypothesis about this additional scattering mechanism.4.1 Additional Mobilities
While comparing measured effective electron mobility (
eff ) between the long-channel and short-channel device shown in Figure 3.20, additional scattering in short-channel device can be dealt with using Matthiessen’s rule which essentially is valid under high Ninv
as follows:13
shows the extracted temperature-dependent additional mobility versus Ninv
for Lm
=48 and 33nm withR
sd = 100 to 120 Ω-μm. Note that the devices under study are all characterized by the same oxide thickness, so it is expected that scattering coming from above channel should be the same. It can be seen that additional mobility for Lm
= 48 nm
add( 48 nm )
increases with Ninv
, whereas for Lm
= 33nm
add( 33 nm )
exhibits a saturating trend which dominates in both cases ofR
sd . We furtherextracted additional mobility at N
inv
=1x1013
cm-2
versus temperature withR
sd as a parameter. In addition, we also provided a temperature-coefficient γ to clarify the temperature-oriented trend, there is a relationship between the additional mobility and temperature as follows: add T
(4.2)Apparently, corresponding temperature dependencies are also opposite to each other:
γ = 0.37~0.55 for L
m = 48nm and γ = -0.13~-0.17 for L m = 33nm as shown in Figure
4.2 for varyingR
sd .4.2 Main Source of Mobility Degradation in Short-Channel Device
Several mechanisms are considered to be involved in the process of mobility
14
degradation for short-channel device. Among them, S/D plasmons, bandgap narrowing, short-range Coulomb scattering due to halo implant or pockets and defects near S/D are believed to be the most probable culprits to degrade the performance [1]-[5]. Additional mobility of L
m
= 48 nm increases with temperature, consistent with the result of [3] about mobility degradation in short-channel device, indicating that the responsible mechanisms are those of short-range Coulomb centers (due to halo implant or pockets [3] and/or defects near S/D [4],[5]).As to L
m
= 33 nm, it exhibits a slight decrease with temperature (Figure 4.2), suggesting other mechanisms. The corresponding mobility component can be written as:Surprisingly, temperature-dependent
add ,extra in Figure 4.4 is satisfactorily close to that of gate-plasmon-limited ones [8]. The corresponding γ has a value of –0.69 to -0.83. Here, we have to rule out the mechanism of 2D charge sharing from S/D [3],[6].The reasons are twofold. First, although the width W
D
of simulated substrate depletion region increases with decreasing Lm
(Figure 3.5 to Figure 3.7), which may act as a signature of 2D charge sharing, the subband separation does not appear to decrease but increase (see Table 2), greatly contrary to [3],[6]. Here, we consider two lowest subbands as an estimation of population distribution versus energy. Meanwhile, this calculation is done for population for 1nm below surface. Figure 4.5 to Figure 4.7 show the subband energy below surface 1nm for gate voltage from 1.0V~1.5V. This obvious difference is simply because in our work, the halo implant is used and its15
effect is enhanced with decreasing L
m
. Second, we conducted bulk-phonon-limited mobility simulation using the simulation program [7],[10]. The resulting γ lies at a value from -1.5 to -1.6, which is much more negative than that of
add ,extra (Figure 4.8). It is therefore argued that only for Lm
less than about 40 nm can long-range Coulomb effects become noticeable.4.3 Evidence of Long-Range Coulomb Interactions
Finally, in Figure 4.9 we quote simulated transconductance at V
d
= 1.0 V and Vg
of 0.75 and 1.0 V above threshold [1] versus Lm
, for comparison with measured transconductance at Vd
= 0.8 and 1.0 V in this work. Evidently, the dimension of our devices, which is carefully chosen to meet the criterion, lies across the activation point of long-range Coulomb effects. With above evidence, we further argue that the long-range Coulomb interactions would be the main factor for performance degradation in ultra-short devices.16
Chapter 5 Conclusion
In this work, TCAD-based inverse modeling has been carried out with aim reconstruct the process parameters. Consequently, interesting and useful results have been created. First, the overlap region for short-channel device can be accurately determined. Also the halo implant P
halo
, which has a significant impact on the substrate doping concentration and further affect the inversion layer charge density, is also solved by calculating the free electron density in the channel region from 2D simulation. Moreover, the parasitic source/drain resistance (R
sd ) is also extracted from the calibration model. In addition, we also provide experimental method to estimate theR
sd for the comprehensive analysis.Second, the resulting temperature power-law exponent (γ) as extracted from our experimentally-determined additional mobility data points out that the long-range Coulomb interactions exist in the metallurgical channel length less than about 40 nm.
Thus, experimental evidence of long-range Coulomb interactions has been drawn.
Furthermore, underlying physical origins have all been distinguished for short channel device. Therefore, long-range Coulomb effect, which is not to be ignored in ultra-short devices, has been for the first time experimentally corroborated in the device samples under study.
17
References
[1] M. V. Fischetti , S. Jin , T.-W. Tang , P. Asbeck , Y. Taur , S. E. Laux , M. Rodwell and N. Sano, “Scaling MOSFETs to 10 nm: Coulomb effects, source starvation, and virtual source model,” J. Comput. Electron., vol. 8, p.60 , 2009.
[2] M. V. Fischetti and S.E. Laux, “Long-range Coulomb interactions in small Si devices. Part Ⅰ: Performance and reliability,” J. Appl. Phys., vol. 89, no. 2, pp.
1232-1248, January 2001.
[3] K. Rim, S. Naeasimha, M. Longstreet, A. Mocuta, and J. Cai, “Low field Mobility characteristics of sub-100 nm unstrained and strained si MOSFETs,” in IEDM
Tech. Dig. , pp. 43-46, 2002.
[4] Antoine Cros, Krunoslav Romanjek, Dominique Fleury, Samuel Harrison, Robin Cerutti, Philippe Coronel, Benjamin Dumont, Arnaud Pouydebasque, Romain Wacquez, Blandine Duriez, Romain Gwoziecki, Frederic Boeuf, Hugues Brut, Gerard Ghibaudo and Thomas Skotnicki, “Unexpected mobility degradation for very short devices: A new challenge for CMOS scaling,” in IEDM Tech. Dig., pp.
663-666, 2006.
[5] Vincent Barral, Thierry Poiroux, Daniela Munteanu, Jean-Luc Autran, and Simon Deleonibus, ‘Experimental investigation on the quasi-ballistic transport: part II—backscattering coefficient extraction and link with the mobility,” IEEE Trans.
Electron Devices, vol. 56, no. 3, pp. 420-430, March. 2009.
[6] P.Packan, S.Cea, H.Deshpande, T.Ghani, M.Giles, O.Golonzka, M.Hattendorf, R.Kotlyar, K.Kuhn, A.Murthy, P.Ranade, L.Shifren, C.Weber and K.Zawadzki,
“High performance Hi-K + metal gate strain enhanced transistors on (110) Silicon,” in IEDM Tech. Dig., pp. 63-66, 2008.
[7] M. V. Fischetti, “Long-range Coulomb interactions in small Si devices Part II.
18
Effective electron mobility in thin-oxide structures,” J. Appl.Phys., vol. 89, no. 2, pp. 1232–1250, Jan. 2001.
[8] Ming-Jer Chen, Li-Ming Chang, Shin-Jiun Kuang, Chih-Wei Lee, Shang-Hsun Hsieh, Chi-An Wang, Sou-Chi Chang, and Chien-Chih Lee,
“Temperature-oriented mobility measurement and simulation to assess surface roughness in ultrathin-gate-oxide ( ~1 nm) nMOSFETs and Its TEM evidence,”
IEEE Trans. Electron Devices, vol. 59, no. 4, pp. 949-955, April. 2012 .
[9] Schred, http://nanohub.org/resources/schred.
[10] M. J. Chen, C. C. Lee, and K. H. Cheng, “Hole effective masses as a booster of self-consistent six-band k‧p simulation in inversion layers of pMOSFETs,”
IEEE Trans. Electron Devices, vol. 58, pp. 931-937, April 2011.
[11] S. Takagi and M. Takayanagi, “Experimental evidence of inversion-layer mobility lowering in ultrathin gate oxide metal-oxide-semiconductor field-effect-transistors with direct tunneling current,” Jpn. J. Appl. Phys., vol.
41, pt. 1, no. 4B, pp. 2348-2352, Apr. 2002.
[12] TCAD. http://www.synopsys.com/Tools/TCAD/Pages/default.aspx.
[13] D.W. Lin, M. L. Cheng, S.W.Wang, C. C.Wu, and M. J. Chen, “A novel method of MOSFET series resistance extraction featuring constant mobility criteria and mobility universality,” IEEE Trans. Electron Devices, vol. 57, no. 4, pp.
890–897, Apr. 2010.
[14] K. Romanjek, F. Andrieu, T. Ernst, and G. Ghibaudo, "Improved split C-V method for effective mobility extraction in sub-0.1-μm Si MOSFETs," IEEE
Electron Devices Letters, vol. 25, no. 8, pp. 583-585, Aug. 2004.
19
Figure 1.1 Schematic of electron transport under long-range Coulomb interactions with S/D plasmons
.
oxide
N + N +
S/D plasmons
P + halo P + halo
20
Figure 1.2 Flowchart of inverse modeling in this work.
CV & IV fitting
Evidence for Long-range Coulomb Interactions
from S/D
Inverse Modeling
Doping Profile, Ninv, Rsd
Effective Mobility
Additional Mobility and Temperature Dependence
Power-Law
To Determine Main Source of Mobility Degradation for Short
Channel Device
21
0.0 0.5 1.0 1.5 2.0
-5.0x10
-60.0 5.0x10
-61.0x10
-51.5x10
-52.0x10
-52.5x10
-53.0x10
-53.5x10
-54.0x10
-5I b I g
I d
C u rr e n t (A )
Gate Voltage (V) T=292K
T=330K T=360K T=380K
I s
V
d=0.05V L g =1 m
Figure 2.1 Temperature-dependent terminal currents at V
d
=0.05V versus Vg
for Lg
=1μm.22
0.0 0.5 1.0 1.5 2.0
0.0 5.0x10
-51.0x10
-41.5x10
-42.0x10
-42.5x10
-4V
d=0.05V L g =65nm
I g &I b I s &I d
C u rr e n t (A )
Gate Voltage (V) T=292K
T=330K T=360K T=380K
Figure 2.2 Temperature-dependent terminal currents at V
d
=0.05V versus Vg
for Lg
=65nm.23
0.0 0.5 1.0 1.5 2.0
0.0 5.0x10
-51.0x10
-41.5x10
-42.0x10
-42.5x10
-4V
d=0.05V L
g
=50nm
Cu rr en t ( A)
Gate Voltage (V) T=292K
T=330K T=360K T=380K
I s &I
d
I g &I
b
Figure 2.3 Temperature-dependent terminal currents at V
d
=0.05V versus Vg
for Lg
=50nm.24
-3 -2 -1 0 1
0.0 0.4 0.8 1.2 1.6 2.0
C ap ac ita nc e ( F/ cm
2)
Gate Voltage (V) Experiment
Schred [9]
Simulation [10]
W/L=10/1 m N
poly=1e20cm
-3P
sub=4e17cm
-3t
ox=1.27nm
Figure 2.4 Comparison of the measured and simulated gate capacitance versus gate voltage.
25
Figure 2. 5 TEM image of the sample.
<001>
<110>
26
Figure 2.6 The schematic diagram for current flow of nMOSFETs with large gate tunneling current. Besides, I
S
<0 and Id
>0.27
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 10
-1010
-910
-810
-710
-610
-510
-410
-310
-2
L
m(nm) 1000 48 33 Exp
TCAD
C ha nn el C ur re nt (A )
T=292K
Gate Voltage (V)
W=1 m t
ox=1.27nm
N
sde=4.95e20cm
-3P
halo=2.5e19cm
-3P
sub=4e17cm
-3V
d=1.0V
V
d=0.05V
Figure 3.1 Measured and simulated I
ch
versus Vg
at T=292K and Vd
=0.05 and 1V.28
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 10
-1010
-910
-810
-710
-610
-510
-410
-310
-2
L
m(nm) 1000 48 33 Exp
TCAD
C ha nn el C ur re nt (A )
T=330K
Gate Voltage (V)
W=1 m t
ox=1.27nm
N
sde=4.95e20cm
-3P
halo=2.5e19cm
-3P
sub=4e17cm
-3V
d=1.0V
V
d=0.05V
Figure 3.2 Measured and simulated I
ch
versus Vg
at T=330K and Vd=
0.05 and 1V.29
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 10
-1010
-910
-810
-710
-610
-510
-410
-310
-2
L
g(nm) 1000 65 50 Exp
TCAD
C ha nn el C ur re nt (A )
T=360K
Gate Voltage (V)
W=1m t
ox=1.27nm
N
sde=4.95e20cm
-3P
halo=2.5e19cm
-3P
sub=4e17cm
-3V
d=1.0V
V
d=0.05V
Figure 3.3 Measured and simulated I
ch
versus Vg
at T=360K and Vd
=0.05 and 1V.30
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 10
-1010
-910
-810
-710
-610
-510
-410
-310
-2
L
m(nm) 1000 48 33 Exp
TCAD
C ha nn el C ur re nt (A )
T=380K
Gate Voltage (V)
W=1m t
ox=1.27nm
N
sde=4.95e20cm
-3P
halo=2.5e19cm
-3P
sub=4e17cm
-3V
d=1.0V
V
d=0.05V
Figure 3.4 Measured and simulated I
ch
versus Vg
at T=380K and Vd
=0.05 and 1V.31
Figure 3.5 Calibrated 2D simulation structure for L
g
=1μm and hence Lm
=1μm. WD
is the width of the mid-channel depletion region.W D
=45.53nm
P-type substrate
Depletion edge
Spac er Spac er
Drain
Silicide Silicide
Halo Halo
Con t act Con t act N + -Poly Gate
Source
Doping Conc. (cm -3 ) L g =1µm
L m ≈1µm
32
Figure 3.6 Calibrated 2D simulation structure for L
g
=65nm and hence Lm
=48nm. WD
is the width of the mid-channel depletion region.
N + -Poly Gate
W D = 57.27nm Depletion edge P-type substrate
Spacer Spacer
Con tact Con tact
Silicide Silicide
Source Drain
Halo
L g =65nm
Doping Conc. (cm -3 )
L m =48nm
33
Figure 3.7 Calibrated 2D simulation structure for L
g
=50nm and hence Lm
=33nm. WD
is the width of the mid-channel depletion region.
N + -Poly Gate
W D = 65.75nm Depletion edge P-type substrate
Spacer Space r
Con tact Con tact
Silicide Silicide
Source Drain
Halo
L g =50nm
Doping
Conc. (cm -3 )
L m =33nm
34
-5.0x10
-70.0 5.0x10
-71.0x10
-61.5x10
-62.0x10
-62.5x10
-60.0
5.0x10
191.0x10
201.5x10
202.0x10
202.5x10
20L=65nm
e D e n s it y ( cm
-3)
Y (cm) area=N
invFigure 3.8 The free electron density under the surface 30nm at one position of the channel region.
35
Figure 3.9 The schematic diagram for inversion layer charge density of metallurgical length L
m
=1μ m.36
Figure 3.10 The schematic diagram for inversion layer charge density of metallurgical length L
m
=48nm.37
Figure 3.11 The schematic diagram for inversion layer charge density of metallurgical length L
m
=33nm.38
Figure 3.12 The calculated N
inv
along the channel direction under interface 30nm for Lm
= 1μm at T= (a) 292K, (b) 330K, (c) 360K, and (d) 380K.39
Figure 3.12 The calculated N
inv
along the channel direction under interface 30nm for Lm
= 1μm at T= (a) 292K, (b) 330K, (c) 360K, and (d) 380K.40
-0.04 -0.03 -0.02 -0.01 0.00 0.01 0.02 0.03 0.04 0.05 0.0
-0.04 -0.03 -0.02 -0.01 0.00 0.01 0.02 0.03 0.04 0.05 0.0
Figure 3.13 The calculated N
inv
along the channel direction under interface 30nm for Lm
= 48nm atT= (a) 292K, (b) 330K, (c) 360K, and (d) 380K.41
-0.04 -0.03 -0.02 -0.01 0.00 0.01 0.02 0.03 0.04 0.05 0.0
-0.04 -0.03 -0.02 -0.01 0.00 0.01 0.02 0.03 0.04 0.05 0.0
Figure 3.13 The calculated N
inv
along the channel direction under interface 30nm for Lm
= 48nm at T= (a) 292K, (b) 330K, (c) 360K, and (d) 380K.42
-0.02 -0.01 0.00 0.01 0.02 0.03
0.0
-0.02 -0.01 0.00 0.01 0.02 0.03
0.0
Figure 3.14 The calculated N
inv
along the channel direction under interface 30nm for Lm
= 33nm at T= (a) 292K, (b) 330K, (c) 360K, and (d) 380K.43
-0.02 -0.01 0.00 0.01 0.02 0.03
0.0
-0.02 -0.01 0.00 0.01 0.02 0.03
0.0
Figure 3.14 The calculated N
inv
along the channel direction under interface 30nm for Lm
= 33nm at T= (a) 292K, (b) 330K, (c) 360K, and (d) 380K.44
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 0.0
5.0x10
121.0x10
131.5x10
132.0x10
132.5x10
13N in v ( cm -2 )
Gate Voltage (V)
T=292K T=330K T=360K T=380K L
m=1 m L
m=48nm L
m=33nm
Figure 3.15 Simulated N
inv
versus gate voltage.45
Figure 3.16 Simulation structure for R
sd
assessment.46
0.8 1.0 1.2 1.4 1.6 1.8
80 90 100 110 120 130
R sd ( m )
Gate Voltage (V) V
d=0.05V
R
sdcalculated by TCAD simulation
Figure 3.17Simulated R
sd
versus Vg
at T=292K.47
48
0.0 0.2 0.4 0.6 0.8 1.0
0 20000 40000 60000 80000 100000
Er r ([ cm
2/(V *s ) ]
2)
R
sd=123.1 - m
=0.33
0=700cm
2/V*s
Figure 3.19 The extracted R
sd
by using the experimental method [13].49
(a)
(b)
Figure 3.20 Extracted temperature-dependent effective mobility versus N
inv
for different Lm
with Rsd
= (a) 100Ω-μm and (b) 120Ω-μm.T=292K T=330K T=360K T=380K
T=292K T=330K T=360K T=380K
50
T=292K T=330K T=360K T=380K
T=292K T=330K T=360K T=380K
(b)
Figure 4.1 Extracted temperature-dependent additional mobility versus N
inv
for Lm
=48 and 33nm with Rsd
= (a) 100Ω-μm and (b) 120Ω-μm.51
240 280 320 360 400 440 480
200 300 400 500 600 700 800 1000 900 1100
33nm device 48nm device
R
sd=120 - m
R
sd=120 - m R
sd=100 - m R
sd=100 - m
N
inv=1x10
13cm
-2
add T
A dd iti on al M ob ili ty
add( cm 2 /V s )
Temperature (K)
Figure 4.2 Extracted additional mobility at N
inv
=1x1013
cm-2
versus temperature with Rsd
as a parameter. The power-law coefficient γ is obtained by data fitting.52
5 10 15 20 25
300 400 500 600 700 800 900 1000
add, extra=1/
add
(33nm)-1/
add
(48nm) R
sd=120 - m
R
sd=100 - m
ad d, e xt ra ( cm
2/V s )
N
inv( x10
12cm
-2)
T=292K T=330K T=360K T=380K
Figure 4.3 Extracted temperature-dependent μ
add
,extra
versus Ninv
with Rsd
as a parameter.53
240 280 320 360 400 440 480 450
500 550 600 650 700 750 800 850
add T
N
inv=1x10
13cm
-2
add, extra=1/
add
(33nm)-1/
add
(48nm)
R
sd=120 - m
R
sd=100 - m
ad d , e xt ra ( cm
2/V s )
Temperature (K)
Figure 4.4 Extracted μ
add, extra
at Ninv
=1x1013
cm-2
versus temperature, along with corresponding power-law coefficient γ.54
Figure 4.5 Simulated conduction-band edge, subband levels, and Fermi level versus position for L
m
= 1μm at Vg
= (a) 1V, (b) 1.2V, and (c) 1.5V.55
-0.06 -0.04 -0.02 0.00 0.02 0.04 0.06 -0.2
-0.06 -0.04 -0.02 0.00 0.02 0.04 0.06 -0.2
-0.06 -0.04 -0.02 0.00 0.02 0.04 0.06 -0.2
Figure 4.6 Simulated conduction-band edge, subband levels, and Fermi level versus position for L
m
= 48nm at Vg
= (a) 1V, (b) 1.2V, and (c) 1.5V.56
-0.06 -0.04 -0.02 0.00 0.02 0.04 0.06 -0.2
-0.06 -0.04 -0.02 0.00 0.02 0.04 0.06 -0.2
-0.06 -0.04 -0.02 0.00 0.02 0.04 0.06 -0.2
Figure 4.7 Simulated conduction-band edge, subband levels, and Fermi level versus position for L
m
= 33nm at Vg
= (a) 1V, (b) 1.2V, and (c) 1.5V.57
5 6 7 8 9 10 11 12
-1.6 -1.4 -1.2 -1.0 -0.8 -0.6 -0.4
R
sd=120 - m R
sd=100 - m
Po w er -L aw E xp on en t
N
inv( x10
12cm
-2)
from simulated bulk-phonon-limited mobility (P
sub=1x10
18cm
-3)
from extracted
add, extra
Figure 4.8 Comparison of temperature power-law exponent of extracted μ
add, extra
and simulated bulk phonon-limited mobility, plotted versus Ninv
.58
10 100 1000
0 500 1000 1500 2000 2500 3000 3500
Tr ab sc on du ct an ce (S /m )
Metallurgical Channel Length (nm)
2D-MC: full-Coulomb effects 2D-MC: no Coulomb effects Vd=0.8V (this work)
Vd=1.0V (this work)
Figure 4.9 Comparison of measured transconductance in this work and simulated ones with and without Coulomb effects [1].
59
Table 1 The length of the overlap region for different L
gate
, and the metallurgical channel length (Lm
=Lgate
-ΔL).60
(a)
(b)
(c)
Table 2 The subband population for different metallurgical channel length at gate voltages of (a) 1V, (b) 1.2V, and (c) 1.5V.