On the Bandstructure Velocity and Ballistic Current of Ultra Narrow
Silicon Nanowire Transistors as a Function of Cross Section Size,
Orientation and Bias
Neophytos Neophytou*, Sung Geun Kim**, Gerhard Klimeck**, and Hans Kosina*
*Technical University of Vienna, TU Wien, Vienna, 1040, Austria.
**Network for Computational Nanotechnology, Birk Nanotechnology Center,
Purdue University, West Lafayette, Indiana 47907-1285
*Corresponding author’s email: firstname.lastname@example.org
A 20 band sp3d5s* spin-orbit-coupled, semi-empirical, atomistic tight-binding (TB)
model is used with a semi-classical, ballistic, field-effect-transistor (FET) model, to theoretically
examine the bandstructure carrier velocity and ballistic current in silicon nanowire (NW)
transistors. Infinitely long, uniform, cylindrical and rectangular NWs, of cross sectional
diameters/sides ranging from 3nm to 12nm are considered. For a comprehensive analysis, n-type
and p-type metal-oxide-semiconductor (NMOS and PMOS) NWs in ,  and 
transport orientations are examined. In general, physical cross section reduction increases
velocities, either by lifting the heavy mass valleys, or significantly changing the curvature of the
bands. The carrier velocities of PMOS  and  NWs are a strong function of diameter,
with the narrower D=3nm wires having twice the velocities of the D=12nm NWs. The velocity
in the rest of the NW categories shows only minor diameter dependence. This behavior is
explained through features in the electronic structure of the silicon host material. The ballistic
current, on the other hand, shows the least sensitivity with cross section in the cases where the
velocity has large variations. Since the carrier velocity is a measure of the effective mass and
reflects on the channel mobility, these results can provide insight into the design of NW devices
with enhanced performance and performance tolerant to structure geometry variations. In the
case of ballistic transport in high performance devices, the  NWs are the ones with both
high NMOS and PMOS performance, as well as low on-current variations with cross section
Index terms – nanowire, velocity, atomistic, bandstructure, sp3d5s*, tight binding,
transistors, MOSFETs, variations, effective mass.
Motivation: The recent advancements in process and manufacturing of nanoelectronic
devices have allowed the manufacturability of nanowire (NW) devices, which are considered as
possible candidates for a variety of applications. For field-effect-transistor applications,
nanowires have recently attracted large attention because of the possibility of enhanced
electrostatic control, and the possibility of close to ballistic transport . Ultra scaled nanowire
transistors of diameters down to D=3nm and gate lengths down to LG=15nm, have already been
demonstrated by various experimental groups [2, 3, 4, 5, 6, 7]. Beyond the use in ultra-scaled
high performance logic and memory transistors, NWs have also attracted large attention as
biological sensors , optoelectronic [9, 10] and thermoelectric devices [11, 12]. Nanowire
properties can be engineered and optimized through size, crystal orientation and strain [13, 14].
The carrier mobility and electrical conductivity that determine to a large degree the on-current
and performance of devices, are quantities closely related to the bandstructure velocity of the
channel. A thorough understanding of the bandstructure velocity and the parameters that control
it, will aid the optimization and design of devices for a variety of electronic transport
applications ranging from the diffusive to the ballistic limits. The bandstructure velocity and
ballistic on-current in NWs as a function of the cross sectional size, transport orientation, carrier
type, and gate bias, is the focus of this work.
The properties of 1D silicon NWs [13, 14, 15, 16, 17, 18] and 2D thin-body devices [19,
20, 21] in the high symmetry orientations ,  and , have been addressed in a
variety of theoretical studies. The challenges in simulating ultra-thin-body (UTB) and NW
devices in which the atoms are countable in their cross section, call for sophisticated models
beyond the effective mass approximation, especially in describing the valence band. The tight-
binding [13, 14, 16, 17] and the k.p [20, 22] methods were used to calculate the electronic
properties of these nanostructures, both with unstrained and strained lattice and scattering
considerations [23, 24, 25, 26]. The performance in terms of on-current is associated with the
carrier velocities, which are linked to the effective mass and the carrier mobility. Strain
engineering, in that respect, is introduced in devices as a way to increase the carrier velocities
and improve performance [27, 28].
These theoretical studies, however, have been performed for specific NW cross sections
or thin-body widths. A comprehensive study that addresses the carrier velocities and ballistic on-
current in NWs as a function of: i) channel cross sectional size, ii) carrier type (n-type or p-type),
iii) transport orientation (, , ), iv) gate bias, and v) cross sectional shape
(cylindrical/rectangular), has not yet been reported. Such a comprehensive device exploration
will provide useful insight into optimization strategies of NW device performance for a variety
of applications. The analysis presented in this paper addresses all these design factors. The
nearest-neighbor atomistic tight binding (TB) model (sp3d5s*-SO) [29, 30, 31, 32] is used for the
NWs’ electronic structure calculation, coupled to a 2D Poisson solver for the electrostatic
potential. To evaluate transport characteristics, a simple semi-classical ballistic model [33, 34,
13, 14] is used. Cylindrical NWs of diameters from D=3nm to D=12nm, and rectangular NWs
from 3nm to 12nm wide/tall (all combinations of aspect ratios) in three different transport
orientations are considered. The design space could be further expanded by the use of strain as
partly shown in Ref. , but this is beyond the scope of this paper.
We find that cross section reduction introduces changes in the bandstructure features that
in general increase the carrier velocities. The increase can vary from ~0 to ~100% depending on
the NW category. High inversion conditions (large gate biases) also increase the carrier
velocities by ~50% as higher energy states are occupied. Device designs are identified for
optimized performance, as well as performance variation tolerance to cross section variations.
We note here that the ballistic model used in this study provides the upper limit for the
performance of the devices. In reality, even devices with ultra short channel lengths are not
100% ballistic. The carrier velocity trends, however, reflect on the effective mass and carrier
mobility and point to the direction of enhanced performance. In addition, imperfections and non-
idealities will also affect the performance. In the last section of the paper, therefore, the results
for one particular geometry are quantified in the presence of surface roughness scattering (SRS)
using a quantum ballistic, full band, atomistic tight-binding simulator [17, 35, 36], and the
magnitude of this additional variation is estimated.
This paper is organized as follows: In part II we describe the approach followed. In part
III we present the results for the velocity and on-current as function of diameter in cylindrical
NWs. In part IV we discuss and provide explanations for the results. In part V we extend our
analysis to rectangular NW devices. In part VI we provide design considerations for optimized
performance. Finally, part VII summarizes and concludes the work.
Atomistic modeling: To obtain the bandstructure of the NWs both for electrons and holes
for which spin-orbit coupling is important, a well calibrated atomistic model is used. The nearest
neighbor TB sp3d5s*-SO model captures all the necessary band features, and in addition, is
robust enough to computationally handle larger NW cross sections as compared to ab-initio
methods. As an indication, the unit cells of the NWs considered in this study contain from ~150
to ~6500 atoms, and the computation time needed varies from a few hours to a few days for each
bias point on a single CPU. The model itself and the parameterization used , have been
extensively calibrated to various experimental data of various natures with excellent agreement
[37, 38, 39, 40]. In particular, we highlight the match to experimental data considering the valley
splitting in slanted strained Si ultra-thin-body devices on disordered SiGe , and single
impurities in Si , without any material parameter adjustments. The model provides a simple
but effective way for treatment of the surface truncation by hydrogen passivation of the dangling
bonds on the surface of the nanowire . What is important for this work, the Hamiltonian is
built on the diamond lattice of silicon, and the effect of different orientations and cross sectional
shapes is automatically included, all of which impact the interaction and mixing of various bulk
The simulation approach: The devices simulated are cylindrical and rectangular NWs in
the ,  and  transport orientations, surrounded by SiO2. In the case of the
cylindrical devices, the oxide thickness is set to 8nm. These are typical NW device sizes that
have been reported in experimental studies [3, 4]. In the rectangular cases, the oxide used is
1.1nm, which is a more realistic thickness for ultimately desired high performance devices. The
simulation procedure consists of three steps as described in detail in Ref.  and is summarized
1. The bandstructure of the wire is calculated using the sp3d5s*-SO model. As an
example, Fig. 1a and 1b show the conduction and valence bands of the D=6nm cylindrical NW
in the  direction (positive-k parts of the dispersions). The different valleys, with different
effective masses, as well as band interactions, non-parabolicities and anisotropies of the Si bulk
bandstructure, especially for the valence band, are all captured by the model and appear in the
2. A semi-classical top-of-the-barrier ballistic model is used to fill the electronic
states and compute the transport characteristics [13, 14, 33, 34]. The range of validity of this
model is explored in detail in Ref.  with a much more computationally demanding 3D full
NEGF (Non-Equilibrium Green’s Function) model. There it was shown that the approach is valid
when the wire’s length/width ratio exceeds L/W>5 in which case source-drain tunneling is
3. The 2D Poisson equation is solved in the cross section of the wire to obtain the
electrostatic potential. It is added to the diagonal on-site elements of the atomistic Hamiltonian
as an effective potential for recalculating the bandstructure until self consistency is achieved.
Although the transport model used is a simple ballistic model, it allows for examining
how the bandstructure alone will affect the channel properties and transport characteristics, and
thus, it provides essential physical insight. It is the simplicity of the transport model, which
allows to shed light on the importance of the dispersion details and their effect on transport.
The average carrier velocity (or injection velocity vinj) of the wires, defined as ID/qn,
where ID is the ballistic drain on-current and n is the carrier density in the NW cross section,
depends strongly on carrier type, cross section diameter and orientation. Figure 2 presents these
results, while Fig. 3 extends the analysis by including the dependence on gate bias as well.
Figure 2a shows the carrier velocity for the cylindrical NMOS (dotted lines) and PMOS (circled
lines) NWs in the  (blue),  (red) and  (black) transport orientations at on-state
(|VG|=|VD|=1V) as a function of the wire’s diameter. Large differences in the carrier velocities of
the different NW orientation and carrier type are observed. Within the same NW category, large
variations are also observed as the diameter reduces. The changes are more prominent in the
PMOS , PMOS  and NMOS  cases. There, the velocity increases by up to a
factor of two as the diameter reduces. The other three cases show minor velocity variations. As it
will be explained in the “Discussion” section, this behavior originates purely from bandstructure
Figure 2b shows the charge density as a function of the diameter. The charge increases
almost linearly with cross section for devices at these diameter scales, following the increase in
the oxide capacitance
with the radius a. The charge, however, is of
similar magnitude in all cases, irrespective of orientation, NMOS or PMOS, because the gate
oxide dominates the overall capacitance in Si NWs. (The NMOS NWs have only slightly higher
charge than the PMOS ones). The total gate capacitance CG in our simulations is degraded from
COX by ~20-30% as also shown in Ref. . This degradation is the same irrespective of NW
type. NWs of different orientations or carrier type, have small differences in their quantum
capacitance CQ of the order of <10% in most cases, but these cause only small differences in the
total gate capacitance CG [13, 14, 15]. At VG=VD=1V, CQ varies from CQ~1.5 nF/m to CQ~9nF/m
as the diameter increases from D=3nm to D=12nm. (It increases with gate bias because CQ a
measure of the density of states at the Fermi level, and more and more subbands are occupied as
the bias increases). The values are very similar in magnitude for all NW typs. The oxide
capacitance, on the other hand, increases from COX = 0.124 nF/m to COX = 0.26 nF/m in the same
diameter range, values 12 X and 34 X smaller than CQ respectively. This indicates that COX still
dominates the electrostatics (more for the larger diameters than the smaller ones). Of course the
effect of CQ is more prominent for smaller oxide thicknesses. Still, the importance of CQ is
reduced, and at a specific diameter, the already small variations between the different NW types,
do not cause any variations in the charge density.
When considering ultrascaled transistors, transport can be close to the ballistic limit .
In this case the current through a device is simply given by the product of charge x velocity, ID =
qn x vinj. The on-current vs. diameter is shown in Fig. 2c. It increases with diameter, but its
magnitude, as well as its rate of increase is different for each NW case. The NWs with the larger
velocity variations have the smaller variations in ION as the diameter changes, as indicated by the
PMOS  ∆ID range vs. the NMOS  range in Fig. 2c. When the carrier velocity does not
vary significantly with diameter, the change in ID follows the change in the charge density with
diameter. This is the case for all NMOS NWs and the PMOS  NWs. In the PMOS  and
 cases where the carrier velocity increases as the diameter decreases, the counter-acting
effect of velocity increase and capacitance decrease makes the ballistic current more tolerant to
diameter variations. This behavior is better illustrated in Fig. 2d, which shows the on-current of
each NW category normalized to its higher value (the value of the D=12nm NW). The PMOS
 and PMOS  NWs have the least on-current reduction as the diameter decreases. As
the diameter is scaled by four times (from D=12nm to D=3nm), ID reduces by ~2 in the PMOS
 and PMOS  cases, whereas it reduces by ~3 times in the rest of the NW categories.
This behavior can potentially provide a mechanism for device designs with ballistic performance
more tolerant to structural variations, especially considering the difficulties in controlling the line
etch roughness in nanofabrication processes.
From Fig. 2a and 2c, it is evident that the designs are diameter dependent since at a
certain NW diameter different NWs perform differently. For the NMOS cases, at larger
diameters, the  NWs (dotted-blue) perform better, closely followed by the  NWs
(dotted-red). At smaller diameters this order is reversed. For the PMOS cases, at larger diameters
the performance of the  NW (circled-red) suffers compared to the  NWs (circled-
black). At smaller diameters, however, the  NWs suffer a smaller performance reduction,
and their performance becomes similar to the  NWs’ performance. The fact that the NMOS
, and especially the PMOS  NWs always lack on performance, leaves the 
direction the one with high performance for both carrier types in the entire diameter range. This
is enhanced by the fact that  NWs suffer less performance loss as the diameter reduces, for
both carrier types. In CMOS applications, for which both NMOS and PMOS need to be utilized,
 seems to be the optimal case.
The results of Fig. 2 are drawn at on-state at a fixed |VG|=|VD|=1V. Figure 3 generalizes
these results by showing the variation of the velocity, charge and current as a function of
diameter, orientation, and additionally gate bias. Row-wise, results for carrier velocity (first
row), charge (second row), and current (third row) vs. VG are shown. Column-wise, results for
the different transport orientations are shown:  – first column,  – second column, and
 – third column. The left/right panels of each figure (negative/positive VG), show results for
PMOS/NMOS NWs respectively. The NWs considered are cylindrical in cross section (as shown
in the inset of Fig. 3a) with diameters varying from D=12nm down to 3nm in decrements of
1nm. The arrows in each sub-figure indicate the direction of diameter reduction.
The carrier velocities of the NWs vs. VG are shown in Figures 3a, 3b, and 3c. In all cases,
the carrier velocities increase with increasing VG (positive for the NMOS and negative for the
PMOS cases). At higher inversion conditions, higher energy states are occupied with higher
carrier velocity (vinj~dE/dk). The increase in the carrier velocity with gate bias increase is as high
as ~50%, and appears in all NW cases.
The strong diameter dependence in the cases of PMOS  and  NWs (Fig. 3b, 3c,
left panels) is also observed through all biases. As the diameter scales from D=12nm to D=3nm,
the velocity doubles from vinj ~ 0.7 x 105 m/s to vinj ~ 1.4 x 105 m/s (values around VG=0V),
independent of gate bias. A similar variation trend, but at a smaller scale is observed for the
NMOS  NWs (Fig. 3b, right panel), for which the velocity increases from vinj ~ 0.9 x 105
m/s to vinj ~ 1.2 x 105 m/s as the diameter is reduced (values around VG=0V). On the other hand,
in the cases of PMOS  (Fig. 3a, left panel), NMOS  (Fig. 3a, right panel), and NMOS
 (Fig. 3c, right panel), the carrier velocity has only a small diameter dependence. In these
cases, the variation trends intermix at different gate biases, and cannot be identified as
monotonically increasing, or decreasing with diameter. Explanations for all these trends are
provided below in the “Discussion” section.
Comparing the magnitude of the velocities in each NW category, NMOS NWs in the
 and  directions have similar performance, with their velocities at low gate bias being
about vinj ~ 1 x 105 m/s. In the NMOS  case, the D=3nm NW has a slight advantage with the
velocity raising to vinj ~ 1.15 x 105 m/s. The NMOS  NW velocities are ~20% lower at vinj ~
0.8 x 105 m/s. In the PMOS NW cases, at a specific diameter size, the  NWs perform better,
followed by the  NW and finally by the  oriented NWs. The PMOS  NWs can
only deliver vinj ~ 0.5 x 105 m/s (low gate bias value), which makes them the NWs with the
lowest carrier velocities of all categories examined for all diameter sizes. This behavior holds for
all gate bias conditions.
In Fig. 3d-3f (second row) and Fig. 3g-3i (third row), we show the charge and ballistic
current respectively, of the NWs with diameters from D=8nm down to D=3nm for which a large
velocity variation is observed. As also shown in Fig. 2b, at the same gate overdrive the charge in
the NWs of the same diameter is of similar magnitude, changing linearly with the oxide
capacitance, irrespective of orientation, NMOS or PMOS.
The ballistic ID vs. VG characteristics in Fig. 3g, 3h and 3i are given by the product of the
charge times velocity ID = qn x vinj. In the NMOS NW cases (all right panels), where the carrier
velocity does not vary significantly with diameter, the change in ID follows the change in the
charge density with diameter. Same happens to the PMOS  NWs (Fig. 3g, left panel). At a
given VG value, as the diameter decreases from D=8nm to D=3nm, the ID is almost halved. In the
PMOS  and  cases Fig. 3h, 3i (left panels) where the carrier velocity increases as the
diameter decreases, the ballistic current is tolerant to diameter variations and reduces only by
~20% and 30% respectively. This is a general behavior at all gate biases.
The velocity variation trends with diameter and orientation are explained in Fig. 4 for
NMOS and Fig. 5 for PMOS NWs. These figures show the first occupied subband (subband
envelopes) of the NWs of each diameter at off-state conditions. Figures 4a, 4b and 4c show
results for NMOS NWs in , , and  orientations, respectively. For example, Fig.
4a shows the first occupied subband of the D=12nm (blue-square) down to the D=3nm (red-dot)
NMOS  NW. The subband edges of each NW are shifted to the same reference E=0eV for
comparison purposes. The arrows show the direction of diameter decrease. There are two
counteracting mechanisms that affect the carrier velocity in these NWs as the diameter
decreases: i) The Γ mass increases, a result of non-parabolicity in the dispersion of the Si bulk
bandstructure. From the bulk value of m*=0.19m0, it increases to 0.27m0 at D=3nm . ii) The
off-Γ valleys with heavier transport mass (m*=0.89m0), but light quantization mass, shift higher
in energy. The second mechanism is slightly stronger, and the combined effect is that the
velocities are slightly higher for NWs of smaller diameters. As the gate bias increases, however,
electrostatic confinement also increases the valley separation of the larger diameter NWs. No
clear trend in the velocities at all biases can, therefore, be identified as earlier described in Fig.
3a (right panel).
Figure 4b shows the first occupied subband of the  NWs of all diameters. As the
diameter reduces: i) The Γ mass slightly reduces (a result of the anisotropic dispersion of the Si
bulk bandstructure ), and ii) the heavier transport mass off-Γ valleys shift higher in energy.
Both effects tend to increase the carrier velocities. A clear trend in velocity reduction as the
diameter reduces is therefore observed, as shown in Fig. 3b (right panel).
In the case of the  NMOS NWs in Fig. 4c, the mass of the first conduction subband
slightly increases as the diameter reduces (the curvature reduces). This increase is only marginal,
and does not lead to any observable variations in the carrier velocities as shown in Fig. 3c - right
Figure 5 shows the same quantities as in Fig. 4, but for the PMOS NWs. Figure 5a shows
the first two occupied subbands for the  PMOS NWs as the diameter reduces from D=12nm
(blue-square) to D=3nm (red-dot). The subbands indicate no clear trend in their curvature as a
function of diameter. Instead, an oscillatory behavior is observed , with several band-
crossings between bands from wires of different diameters. (Showing the higher two bands in
this case indicates the oscillations more clearly). This reflects in the velocities of Fig. 3a (left
panel), for which no significant variation exists, and as the gate bias increases, the magnitude of
the velocities of wires with different diameters is also observed to interchange. The oscillatory
behavior of the subbands keeps the carrier velocities low. The low bias velocity of  PMOS
NWs is vinj ~ 0.5 x 105 m/s, whereas in all other NW categories the velocities are almost 2 X
The variation pattern in the subband envelopes of the  and  PMOS NWs in Fig.
5b, 5c is clearer. Here the subbands undergo a large transformation as the diameter decreases,
acquiring a larger curvature and lighter effective masses, and thus significantly higher carrier
velocities. This explains the velocity trend in Fig. 3b, 3c (left panels). The subband shape
behavior, and its large change under cross section reduction is described in detail in Ref. ,
and is a result of the anisotropy of the heavy-hole subband of the valence band shown in the inset
of Fig. 5b. There, the 45º lines drawn show the relevant energy lines that form the subbands for
NWs with the (1-10) surface quantized, as is the case for the  (and ) oriented NWs.
As the diameter reduces, subbands further away from the center of the Brillouin zone are
utilized, which have large curvatures and lower effective masses. The arrow along the 45º lines
shows the direction of decreased wire cross section, corresponding to larger k-value quantization.
The subband trend in Fig. 5b has its origin in this anisotropic energy surface. Of course, real NW
quantization involves many more interactions, but the basic trend of the heavy-hole band is
transferred to the NW subbands. A similar effect is responsible for the subband trend of the 
NWs shown in Fig. 5c.
V. Results for Rectangular NWs
After investigating the carrier velocity and current variations of cylindrical NWs, we
extend our analysis to rectangular NWs with widths/heights varying from 3nm to 12nm (all
aspect ratios), for the three orientations under consideration. Figures 6 and 7 show the results for
NMOS and PMOS NWs, respectively. Due to the large volume of the data for NWs with various
aspect ratios and gate biases, only the velocity (first row) and current (second row) results at on-
state (VG=VD=1V for NMOS, and VG=VD=-1V for PMOS) are presented. The lower left corners
of the sub-figures in Fig. 6 and 7 show the velocity/current of the 3nm x 3nm NWs, whereas the
upper right corners show the velocity/current of the 12nm x 12nm wires. Other than the
width/height of the NWs no other parameter is changed in the simulations.
Figures 6a, 6b and 6c (first row of Fig. 6) present the velocity results for the , 
and  oriented NMOS NWs, respectively. In all cases, cross section reduction results in
higher velocities (higher velocities in the lower/left than the upper/right part of the figures). In
the  NW case in Fig. 6a, following the cylindrical case arguments, the off-Γ valley is
pushed higher in energy and the overall velocity is higher. The velocity variation in the entire
figure is of the order of ~8%, ranging from 1.2 x 105 m/s to 1.3 x 105 m/s.
In the case of the  NMOS NW in Fig. 6b, the width and the height surfaces are in
the [1-10] and  directions, respectively, as shown in the third row of Fig. 6. The nature of
valley quantization is different for each surface. In the height in the  quantization direction,
the Γ valleys have light transport mass, but heavy quantization mass. The off-Γ valleys have the
reverse, heavy transport mass, but light quantization mass. Reducing the height lifts the lightly
quantized off-Γ valleys, just as in the  NW cases above, and increases the velocities (i.e the
velocities increase as one moves from top to bottom in Fig. 6b). On the other hand, in the [1-10]
width direction, the off-Γ valleys are more heavily quantized than the Γ valleys. Variations in the
width do not shift their energy minima strongly, and the carrier velocities are therefore almost
constant along that direction.
An extrapolation of our results, indicates that NMOS (001)/ channels (extension
beyond the lower-right corner of Fig. 6b), have higher velocities than NMOS (1-10)/
channels (extension beyond the upper-left corner of Fig. 6b). Experimental data on the channel
mobility vs. transport and quantization orientation in Si MOSFET channels [43, 44] show that
the mobility is also higher for NMOS (001)/ rather than (1-10)/ channels. Although
the mobility depends also on the scattering process and not only on bandstructure, the velocity
results point towards a possible explanation of the experimental behavior. Furthermore, for long
channel NWs devices with finite width, the fact that the velocity is not sensitive to variations in
the [1-10] direction, points towards utilizing this direction as the one for which the line etch
control is minimal in the fabrication process, so that the performance variation due to size
variations is reduced.
Comparing the magnitude of the carrier velocities, it is very similar in the  and 
oriented NWs in Fig. 6a and 6b since in both cases it is mostly determined by the Γ valleys. It
ranges from 1.1 to 1.3 x 105 m/s. On the other hand, the velocities of the  oriented NWs in
Fig. 6c, are determined by tilted Si conduction band ellipsoids of higher effective mass (m* ~
0.43m0 in bulk) . They are therefore ~30% lower, ranging only from 0.9 to 1.1 x 105 m/s.
Still, however, the same pattern is followed, where size reduction increases carrier velocities.
Figures 6d, 6e and 6f present the ballistic on-current results for the NWs in the three
orientations. Since the charge increases linearly with the cross section (by almost four times in
Fig. 2b), and the velocity decreases by only ~30%, the on-current increases monotonically
following the increase in the cross section. As in the cylindrical NW cases earlier, the on-current
is higher for the  NWs, closely followed by the  NWs, whereas the  NWs have
~25% lower on-current. The contour lines in the sub-figures, all tilted at ~45º, indicate the linear
increase in ID with cross section increase, as well as the symmetry between width/height
surfaces, even in the velocity asymmetric case of the  NWs.
Figures 7a, 7b and 7c present the carrier velocities for the rectangular PMOS NWs. The
velocities of the  and  NWs in Fig. 7b and 7c range from vinj ~ 1 x 105 m/s to 1.5 x 105
m/s, and vinj ~ 1.2 x 105 m/s to 1.8 x 105 m/s respectively, a variation of ~50%. For the 
NWs in Fig. 7a, the velocities range from vinj ~ 0.5 x 105 m/s to 0.8 x 105 m/s and are much
lower compared to all other NW cases of either carrier type. Similar to the NMOS case, in
general, cross section size reduction increases the carrier velocities. This is more evident in the
 and  NW orientation cases of Fig. 7a and Fig. 7c, respectively. Figure 7b, on the other
hand, shows a strong surface anisotropic behavior for the  PMOS NWs. Scaling of either
NW side (width in [1-10] or height in ), increases carrier velocities. In the case of scaling
the height, however, at ~6nm the velocity gets an upward jump before it starts to decrease again.
This is attributed to the anisotropic quantization mass in the two surfaces and detailed
explanations are provided in Ref. .
It is also worth mentioning here that in the case of the PMOS  channels,
experimental data [43, 44] show that long channel (1-10)/ MOSFET channels have higher
mobility than the (001)/ ones, the opposite of what is observed in the NMOS  channel.
The different carrier velocities in the two PMOS channels could point to the reasons that might
be responsible for this. Scaling of the [1-10] width quantizes the k-space along the light branch
of the anisotropic heavy-hole valley (inset of Fig. 5b). Scaling the  height does not provide
this advantage. Although this surface difference is only weakly reflected in the velocities of the
narrow NWs of Fig. 7b (left vs. lower parts), simulations using real 2D bandstructures could
potentially demonstrate the difference in velocities between the two surfaces. In general,
however, physical scaling of the channel in directions that utilize the larger curvature regions of
the bulk bandstructure is beneficial to the carrier velocity and device performance (as the scaling
of the [1-10] width in the PMOS (1-10)/ case).
Figures 7d, 7e and 7f show the PMOS NWs ballistic on-current results for the three
orientations. In the  case in Fig. 7d and  case in Fig. 7f, the on-current increases
monotonically as the cross section increases, similar to the NMOS cases. In the case of the 
NWs in Fig. 7e, the on-current has a more complex behavior, following the complex behavior of
the velocity in Fig. 7b. Here, the regions of 3nm-5nm and 7nm-12nm of height, and for any
width are design regions for low on-current variations to large cross section variations. Around a
height of ~6nm, however, large variations are observed, and device designs with such height
should be avoided. Comparing the magnitude of the on-current in the PMOS NW cases, it is in
general higher for the  NWs, closely followed by the  NWs, whereas the performance
of the  NWs is almost half compared to the other NWs.
VI. Design Considerations
Table 1 summarizes the performance comparisons between the NWs of the different
orientations for the small (D=3nm) and the larger (D=12nm) diameters. The indications “High”,
“Fair”, and “Low”, refer to the relative performance of the NWs of each row (orientation
comparison) and not necessarily on an absolute scale. The numbers within brackets correspond
to the performance order (both carrier velocity and current have the same order) of the different
orientated NWs within each row. Although this table is constructed according to the cylindrical
NW results, the same conclusions follow in the cases of the rectangular devices.
In the case of NMOS NWs, in Table 1a, both the  and  orientations have high
performance, with the  orientation having an advantage at larger diameters and the  at
smaller diameters. The  NWs have lower performance at smaller diameters and fair at
larger diameters compared to the two other orientations. The PMOS performance comparison is
shown in Table 1b. In this case, the  orientation is the most advantageous in all diameter
ranges, closely followed by the  orientation, whereas the  orientation performs purely
for all NW diameters. PMOS  NWs perform higher than all other NWs (PMOS or NMOS),
and are the optimal solution for applications that require high performance individual NWs.
Since the NMOS  and especially the PMOS  NWs perform purely, for CMOS
applications that both NMOS and PMOS high performance is required, the  orientation
seems to be the optimal solution.
Up to this point our analysis considered infinitely long, undistorted NWs with perfect
surfaces, assuming perfect manufacturability. In reality, structure imperfections exist in devices
and affect the performance [45, 46, 47, 48, 49]. Controlling the line edge roughness in
nanofabrication processes imposes challenges, and the lack of it leads to device-to-device
performance variations. The width is usually less well controlled since it is formed by etching,
whereas the height is controlled by growth and can be more precise. For the high performance
rectangular PMOS  NWs build on (001) substrates, the on-current is not significantly
sensitive to the width (Fig. 7e). It is also not significantly sensitive to the height, except at around
6nm of height, a design region that should be avoided. In the NMOS  case the (001)/
configuration is also beneficial since the velocity is almost constant in the [1-10] width direction.
Although the ballistic on-current in Fig. 6e is symmetric with respect to the width/height, devices
are not 100% ballistic, and therefore it is still beneficial to have the direction of least control
aligned with the direction of velocity invariance.
As a design strategy, therefore, our results demonstrate that out of all NWs examined, the
 oriented NWs are advantageous for CMOS technology applications in either design case: i)
if the design goal is driven by the highest performance assuming perfect manufacturing abilities,
or ii) if the design is driven by low device-to-device performance fluctuations. Design regions
with velocity and on-current insensitivity to geometry can provide low device-to-device
variations strategies for both, long channel devices, and short channel close-to-ballistic devices.
In the cases of the  channel orientations built on (001) substrates, design regions can be
identified for either case, while still keeping the performance high.
An important source of performance variation and degradation is surface roughness
scattering (SRS). To quantify the previous results, we examine the effect of SRS on the velocity
and current variations for the case of the NMOS  D=3nm cylindrical NW. For this purpose,
real-space 3D, full-band, quantum ballistic simulations are performed using the OMEN code [17,
35, 36], in which the roughness is treated in a realistic way by adding/subtracting atoms on the
surface of the NW . A ~15% degradation, as well as an additional ~30% variation in the
velocities is computed by using a sample size of 50 rough NWs. A 10% variation in the on-
current is also observed, in agreement with other works [46, 48, 49]. This is an indication of an
additional variation, on top of the variation expected due to bandstructure as a result of diameter
change. A proper and more elaborate investigation the effect of surface roughness scattering
which targets all orientations and a large range of diameter sizes will be presented elsewhere.
These numbers, however, further stress the need of device and circuit designs tolerant to
performance variations originating from structure variations and non-idealities. Bandstructure
features can provide a mechanism to partially help on this.
An atomistic tight-binding approach and a semi-classical ballistic model are used to
calculate the bandstructure velocity and ballistic current of nanowire devices, self-consistently
with the electrostatic potential. NMOS and PMOS NWs of diameters from D=12nm down to
D=3nm in , , and  orientations, of cylindrical and rectangular cross sectional
shapes are considered. The carrier velocities are strong function of orientation, band type,
diameter, and bias. Cross section scaling, in general increases the carrier velocities either by
raising the energy minima of the heavier transport mass valleys, or by significantly changing the
subbands’ curvature. PMOS  and  NWs have the largest velocity sensitivity to
diameter, in which cases the velocity doubles as the diameter scales from D=12nm to D=3nm.
The carrier velocity of NMOS  NWs is also sensitive to the diameter, but at a smaller
degree. On the other hand, the velocities of NMOS  and , and PMOS  NWs are
insensitive to the diameter. Gate bias also tends to increase carrier velocities by ~50%, as higher
energy states are occupied at inversion conditions. Trends in carrier velocity with diameter do
not completely reflect to the terminal current characteristics because the carrier density is also
subject to the change of the cross section geometry. The ballistic on-current shows less
sensitivity with cross section fluctuations in the cases of  PMOS and  PMOS NWs,
whereas in the rest of the cases it varies linearly with cross section. The PMOS  NWs are
the ones with the highest performance from all NW categories, whereas the PMOS  the
ones with the lowest performance. The  oriented NWs, on the other hand, are the ones with
both, high NMOS and PMOS carrier velocities and on-current, and therefore more suitable for
CMOS applications. In either case, if the design goal is driven by the highest performance
assuming perfect manufacturability, or if the design is driven by low device-to-device
performance fluctuations, this study suggests the (001)/ oriented NWs for both NMOS and
PMOS NW devices for either long channel, or short channel ballistic devices.
This work has been partially supported by funds from the Austrian Science Fund, FWF,
contract I79-N16. S.G. Kim was supported by Materials Structures and Devices Focus Center
MSD/MARCO. Dr. Mathieu Luisier is acknowledged for providing the quantum transport code
for the surface roughness calculations and for various discussions. Computational resources of
the Network for Computational Nanotechnology (NCN) operated by nanoHUB.org are
acknowledged. The simulations in this work can be duplicated with Bandstructure Lab on
 ITRS Public Home Page. http://www.itrs.net/reports.html
 N. Singh, F. Y. Lim, W. W. Fang, S. C. Rustagi, L. K. Bera, A. Agarwal,C. H. Tung,K. M.
Hoe, S. R. Omampuliyur, D. Tripathi, A. O. Adeyeye, G. Q. Lo, N. Balasubramanian, and D. L.
Kwong, “Ultra-narrow silicon nanowire gate-all-around CMOS devices: Impact of diameter,
channel-orientation and low temperature on device performance,” Int. Elec. Dev. Meeting, IEDM
 K. H. Cho, Y. C. Jung, B. H. Hong, S. W. Hwang, J.H. Oh, D. Ahn, S. D. Suk, K. H. Yeo,
D.-W Kim, D. Park, and W.-S. Lee, “Observation of single electron tunneling and ballistic
transport in twin silicon nanowire MOSFETs (TSNWFETs) fabricated by top-down CMOS
process,” Int. Elec. Dev. Meeting, IEDM 2006.
 K. H. Cho, K.H. Yeo, Y.Y Yeoh, S. D. Suk, M. Li, J. M. Lee, M.-S. Kim, D.-W. Kim, D.
Park, B. H. Hong, Y. C. Jung, and S. W. Hwang, “Experimental evidence of ballistic transport in
cylindrical gate-all-around twin silicon nanowire metal-oxide-semiconductor field-effect
transistors,” Appl. Phys. Lett., 92, 052102, 2008.
 M. Kobayashi and T. Hiramoto, “Experimental study on quantum confinement effects in
silicon nanowire metal-oxide-semiconductor field-effect transistors and single-electron
transistors,” J. Appl. Phys., 103, 053709, 2008.
 J. Xiang, W. Lu, Y. Hu, Y. Wu, H. Yan, and Charles M. Lieber, “Ge/Si nanowire
heterostructures as high-performance field-effect transistors,” Nature, vol. 441, no. 25, 2006.
 K. H. Yeo, S. D. Suk, M. Li, Y.-Y. Yeoh, K. H. Cho, K.-H. Hong, S. K. Yun, M. S. Lee, N.
Cho, K. Lee, D. Hwang, B. Park, D.-W. Kim, D. Park, and B.-I. Ryu, “Gate-all-around (GAA)
twin silicon nanowire MOSFET (TSNWFET) with 15nm length gate and 4nm radius nanowire,”
in Proc. of IEDM 2006, pp. 1-4, 2006.
 S. Huang, and Y. Chen, “Ultrasensitive fluorescence detection of single protein molecules
manipulated electrically on Au nanowire,” Nano Lett., vol. 8, no. 9, pp. 2829-2833, 2008.
 C. B. Winkelmann, I. Ionica, X. Chevalier, G. Royal, C. Bucher, and V. Bouchiat, “Optical
switching of porphyrin-coated silicon nanowire field effect transistors,” Nano Lett., 7, 6, pp.
 M. Law, L. E. Greene, J. C. Johnson, R. Saykally, and P. Yang, “Nanowire dye-sensitized
solar cells,” Nature Mat., vol. 4, pp. 455-459, 2005.
 A. I. Boukai, Y. Bunimovich, J. T.-Kheli, J.-K. Yu, W. A. G. III, and J. R. Heath, “Silicon
nanowires as effective thermoelectric materials,” Nature, vol. 451, pp. 168-171, 2008.
 A. I. Hochbaum, R. Chen, R. D. Delgado, W. Liang, E. C. Garnett, M. Najarian, A.
Majumdar, and P. Yang, “Enhanced thermoelectric performance of rough silicon nanowires,”
Nature, vol. 451, pp. 163-168, 2008.
 N. Neophytou, A. Paul, M. Lundstrom, and G. Klimeck, “Bandstructure effects in silicon
nanowire electron transport,” IEEE Trans. Elect. Dev., vol. 55, no. 6, pp. 1286-1297, 2008.
 N. Neophytou, A. Paul, and G. Klimeck, “Bandstructure effects in silicon nanowire hole
transport,” IEEE Trans. Nanotechnol., vol. 7, no. 6, pp. 710-719, 2008.
 N. Neophytou and G. Klimeck, “Design space for low sensitivity to size variations in 
PMOS nanowire devices: The implications of anisotropy in the quantization mass,” Nano
Lett.,vol. 9, no. 2, pp. 623-630, 2009.
 M. Luisier, A. Schenk, W. Fichtner, “Full-Band Atomistic Study of Source-To-Drain
Tunneling in Si Nanowire Transistors,” Proceedings of the 12th International Conference on
Simulation of Semiconductor Processes and Devices, SISPAD 2007, pp. 221-224.
 M. Luisier, and G. Klimeck, “Full-band and atomistic simulation of n- and p-doped double-
gate MOSFETs for the 22nm technology node,” Proceedings of the 13th International
Conference on Simulation of Semiconductor Processes and Devices, SISPAD 2008, pp. 17-20.
 G. C. Liang, J. Xiang, N. Kharche, G. Klimeck, C. M. Lieber, and M. Lundstrom,
"Performance analysis of a Ge/Si core/shell nanowire field-effect transistor," Nano Lett., vol. 7,
pp. 642-646, 2007.
 Y. Liu, N. Neophytou, T. Low, G. Klimeck, and M. S. Lundstrom, “A tight-binding study of
the ballistic injection velocity of ultra-thin body SOI MOSFETs,” IEEE Trans. Electr. Dev., vol.
55, no. 3, pp. 886-871, 2008.
 M. V. Fischetti, Z. Ren, P. M. Solomon, M. Yang, and K. Rim “Six-band k.p calculation of
the hole mobility in silicon inversion layers: Dependence on the surface orientation, strain, and
silicon thickness,” J. Appl. Phys., vol. 94, no. 2, 2003.
 M. Yang et al., “Hybrid-orientation technology (HOT): opportunities and challenges,” IEEE
Trans. Electr. Dev., vol. 53, no. 5, pp. 965-978, 2006.
 E. X. Wang, P. Matagne, L. Shifren, B. Obradovic, R. Kotlyar, S. Cea, M. Stettler, and M.
Giles, “Physics of hole transport in strained silicon MOSFET inversion layers,” IEEE Trans.
Electr. Dev., vol. 53, no. 8, 2006.
 E. Gnani, A. Gnudi, S. Regianni, and G. Baccarani, “Effective mobility in nanowire FETs
under quasi-ballistic conditions,” IEEE Trans. Electr. Dev., vol. 57, no. 1, pp. 336-344, 2010.
 S. Jin, M. V. Fischetti, and T.-W. Tang, “Modeling of electron mobility in gated silicon
nanowires at room temperature: Surface roughness scattering, dielectric screening, and band
parabolicity,” J. Appl. Phys., 102, 083715, 2007.
 R. Kotlyar, B. Obradovic, P. Matagne, M. Stettler, and M. D. Giles, “Assessement of room-
temperature phonon-limited mobility in gated silicon nanowires,” Appl. Phys. Lett., 84, no. 25,
 R. Kim, and M. Lundstrom, “Physics of carrier backscattering in one- and two-dimensional
nanotransistors,” IEEE Trans. Elect. Dev., vol. 56, no. 1, pp. 32-139, 2009.
 S.E. Thompson et al., “A 90-nm logic technology featuring strained-silicon,” IEEE Trans.
Electr. Dev., vol. 51, no. 11, pp. 1790-1797, 2004.
 M. Saitoh, S. Kobayashi, and K. Uchida, “Physical Understanding of Fundamental
Properties of Si (110) pMOSFETs Inversion-Layer Capacitance, Mobility Universality, and
Uniaxial Stress Effects,” Int. Elec. Dev. Meeting, IEDM 2007, pp. 711-714.
 T. B. Boykin, G. Klimeck, and F. Oyafuso, “Valence band effective-mass expressions in the
sp3d5s* empirical tight-binding model applied to a Si and Ge parametrization,” Phys. Rev. B,
vol. 69, pp. 115201-115210, 2004.
 G. Klimeck, F. Oyafuso, T. B. Boykin, R. C. Bowen, and P. von Allmen, Computer
Modeling in Engineering and Science (CMES), vol. 3, no. 5, pp. 601-642, 2002.
 G. Klimeck, S. Ahmed, H. Bae, N. Kharche, S. Clark, B. Haley, S. Lee, M. Naumov, H.
Ryu, F. Saied, M. Prada, M. Korkusinski, and T. B. Boykin, “Atomistic simulation of
realistically sized nanodevices using NEMO 3-D—Part I: Models and benchmarks,” IEEE
Trans. Electr. Dev., vol. 54, no. 9, pp. 2079-2089, 2007.
 J. C. Slater and G. F. Koster, “Simplified LCAO method for the periodic potential
problem,” Phys. Rev., vol 94, no. 6, 1954.
 M.S. Lundstrom, and J. Guo, “Nanoscale transistors: Device physics, modeling and
simulation,” Springer, 2006.
 A. Rahman, J. Guo, S. Datta, and M. Lundstrom, “Theory of ballistic nanotransistors,”
IEEE Trans. Electr. Dev., vol. 50, no. 9, pp. 1853-1864, 2003.
 M. Luisier, and G. Klimeck, “OMEN an Atomistic and Full-Band Quantum Transport
Simulator for post-CMOS Nanodevices,” Proceedings of the 8th IEEE Conference on
Nanotechnology, IEEE NANO 2008, pp. 18-21.
 M. Luisier, N. Neophytou, N. Kharche, and G. Klimeck, "Full-Band and Atomistic
Simulation of Realistic 40 nm InAs HEMT", Int. Elec. Dev. Meeting, IEDM 2008, pp. 887-891.
 R. C. Bowen, G. Klimeck, R. Lake, W. R. Frensley and T. Moise, “Quantitative Resonant
Tunneling Diode Simulation,” J. of Appl. Phys. 81, 3207 (1997)
 J. Wang, PhD thesis, Purdue University, 2005.
 N. Kharche, M. Prada, T. B. Boykin, and G. Klimeck, “Valley splitting in strained silicon
quantum wells modeled with 2º miscuts, step disorder, and alloy disorder,” Appl. Phys. Lett., vol.
90, no. 9, p. 092109, 2007.
 R. Rahman, C. J. Wellard, F. R. Bradbury, M. Prada, J. H. Cole, G. Klimeck, and L. C. L.
Hollenberg, “High precision quantum control of single donor spins in silicon,” Phys. Rev. Lett.,
vol 99, no. 3, p. 036403, 2007.
 S. Lee, F. Oyafuso, P. Von, Allmen, and G. Klimeck, “Boundary conditions for the
electronic structure of finite-extent embetted semiconductor nanostructures,” Phys. Rev. B, vol.
69, pp. 045316-045323, 2004.
 A. Paul, S. Mehrotra, G. Klimeck and M. Luisier, “On the validity of the top of the barrier
quantum tranport model for ballistic nanowire MOSFETs,” Proceedings of the 13th International
Conference on Computational Electronics, IWCE 2009, pp. 177-180.
 M. Yang, V. W. C. Chan, K. K. Chan, L. Shi, D. M. Fried, J. H. Stathis, A. I. Chou, E.
Gusev, J. A. Ott, L. E. Burns, M. V. Fischetti, and M. Ieong, “Hybrid-orientation technology
(HOT): Opportunities and challenges,” IEEE Trans. Electr. Dev., vol. 53, no. 5, pp. 965-978,
 L. Chang, M. Ieong, and M. Yang, “CMOS circuit performance enhancement by surface
orientation optimization,” IEEE Trans. Electr. Dev., vol. 51, no. 10, pp. 1621-1627, 2004.
 M. Luisier, A. Schenk, and Wolfgang Fichtner, “Atomistic treatment of interface roughness
in Si nanowire transistors with different channel orientations,” Appl. Phys. Lett., 90, 102103,
 J. Wang, E. Polizzi, A. Ghosh, S. Datta, and M. Lundstrom, “Theoretical investigation of
surface roughness scattering in silicon nanowire transistors,” Appl. Phys. Lett., 87, 043101, 2005.
 N. Seoane, A. Martinez, A. R. Brown, J. R. Barker, A. Asenov, “Current variability in Si
nanowire MOSFETs due to random dopants in the source/drain regions: A fully 3-D NEGF
simulation study,” IEEE Trans. Electr. Dev., vol. 56, no. 7, pp. 1388-1395, 2009.
 M. Lenzi, A. Gnudi, S. Reggiani, E. Gnani, M. Rudan, and G. Baccarani, “Semiclassical
transport in silicon nanowire FETs including surface roughness,” J. Comput. Electron. 7: 355-
 S. Poli, M. G. Pala, T. Poiroux, S. Deleonibus, and G. Baccarani, “Size dependence of
surface-roughness-limited mobility in silicon-nanowire FETs,” IEEE Trans. Electr. Dev., vol 55,
no. 11, 2008.
 [nanoHub] Bandstructure lab on nanoHUB.org
Figure 1: Dispersions of  NMOS and PMOS NWs of D=6nm
Figure 1 Caption:
Equilibrium dispersion relations for the D=6nm  NW (positive k-states). (a)
Electrons. (b) Holes. The edges of the first band are shifted to E=0eV. a0’ is the
unit cell length.
Figure 2: velocity/charge/ION variations in cylindrical NMOS/PMOS NWs at
|VG|=1V vs. diameter.
Figure 2 Caption:
Results for cylindrical NMOS and PMOS NWs at |VG|=1V and |VD|=1V vs.
diameter. (a) Carrier injection velocity. (b) Charge density. (c) On-current. (d) On-
current normalized to its maximum value (the D=12nm value).
Figure 3: velocity/charge/current variations in cylindrical NMOS/PMOS for
different diameters vs. bias
Figure 3 Caption:
Results for cylindrical NWs of diameters varying from D=12nm down to 3nm,
PMOS and NMOS vs. VG for |VD|=1V. PMOS NW results are shown for negative
VG, NMOS results for positive VG. The arrows indicate the direction of diameter
decrease in the cases for which a uni-directional trend is observed. (a-c) Carrier
velocities. (d-f) Charge density (for D=8nm down to D=3nm). (g-i) Ballistic
current (for D=8nm down to D=3nm). Left column -  oriented wires. Middle
column -  oriented wires. Right column -  oriented wires. Inset of (a):
The cylindrical NW cross section.
Figure 4: The first subband (subband envelope) of NMOS NWs as the diameter
Figure 4 Caption:
The first subband (subband envelope) of NMOS NWs as the diameter scales from
D=12nm to 3nm. The arrows indicate the direction of diameter decrease. (a) 
oriented wires. (b)  oriented wires. (c)  oriented wires. The minima of all
bands are shifted to E=0 eV.
Figure 5: The first subband (subband envelope) of PMOS NWs as the diameter
Figure 5 Caption:
The first subband (subband envelope) of PMOS NWs as the diameter scales from
D=12nm to 3nm. The arrows indicate the direction of diameter decrease. (a) 
oriented wires (the first two subbands are shown). (b)  oriented wires. (c)
 oriented wires. The minima of all bands are shifted to E=0 eV. Inset of (b):
The heavy-hole subband of the Si bandstructure with (1-10) surface quantization
Figure 6: The average velocity and current of rectangular NMOS NWs at ON-state
Figure 6 Caption:
2D surfaces of transport quantities for rectangular NMOS NWs as the width (W) /
height (H) directions vary from W=3nm to W=12nm, and H=3nm to H=12nm. (a-c)
The average carrier velocity in 105 m/s for ,  and  directed wires
respectively. (d-f) The current density in µA for ,  and  directed
wires respectively. The lower row indicates the NW directions and surface
Figure 7: The average velocity and current of rectangular PMOS NWs at ON-state
Figure 7 Caption:
2D surfaces of transport quantities for rectangular PMOS NWs as the width (W) /
height (H) directions vary from W=3nm to W=12nm, and H=3nm to H=12nm. (a-c)
The average carrier velocity in 105 m/s for ,  and  directed wires
respectively. (d-f) The current density in µA for ,  and  directed
wires respectively. The lower row indicates the NW directions and surface
31 Download full-text
Table 1: NMOS and PMOS comparisons
NMOS performance   
Small D (3nm) High (2) High (1) Low (3)
Large D (12nm) High (1) High (2) Fair (3)
PMOS performance   
Small D (3nm) Low (3) High (2) High (1)
Large D (12nm) Low (3) Fair (2) High (1)
Table 1 Caption:
Relative performance comparison for the NWs of the different orientations and
diameters: (a) NMOS. (b) PMOS.