04408768
TRANSCRIPT
-
7/27/2019 04408768
1/12
48 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 55, NO. 1, JANUARY 2008
Analytical Modeling of the Suspended-Gate FETand Design Insights for Low-Power Logic
Kerem Akarvardar, Christoph Eggimann, Dimitrios Tsamados, Yogesh Singh Chauhan,Gordon C. Wan, Adrian Mihai Ionescu, Senior Member, IEEE,
Roger T. Howe, Fellow, IEEE, and H.-S. Philip Wong, Fellow, IEEE
AbstractAn analytical model for the suspended-gate field-effect transistor (SGFET), dedicated to the dc analysis of SGFETlogic circuits, is developed. The model is based on the depletionapproximation and expresses the pull-in voltage, the pull-out volt-age, and the stable travel range as a function of the structuralparameters. Gate position is explicitly expressed as a functionof the gate voltage, thus enabling the convenient integration ofthe analytical SGFET relationships into the standard MOSFETmodels. Starting from the new SGFET model, the influence ofthe mechanical hysteresis on the circuit steady-state behavior is
discussed, the potential of using the SGFET as an ultra-low powerswitch is demonstrated, and the operation of the complementarySGFET inverter is analyzed.
Index TermsElectrostatic actuators, inverter, microelectro-mechanical system (MEMS), MOSFET, nanoelectromechanicalfield-effect transistor (NEMFET), nanoelectromechanical systems(NEMS), resonant-gate FET, subthreshold swing, suspended-gateFET (SGFET).
I. INTRODUCTION
IN ADVANCED CMOS technology nodes with supply volt-
ages 1 V, one of the major issues is the limited scalabilityof the threshold voltage. Due to the fundamental minimum
value of the subthreshold swing, lowering the threshold voltagebelow about 150 mV results in intolerably high off-current
[1]. In this paper, we analyze the suspended-gate field-effect
transistor (SGFET) as a candidate to circumvent this limitation.
Due to their extremely low standby power consumption and
ideal switching characteristics, SGFETs can be used as sleep
transistors for efficient power management and partitioning
in highly scaled CMOS ICs. SGFETs can also be used to
implement low-power (full-SGFET or SGFET/MOSFET) logic
circuits.
Manuscript received May 17, 2007; revised August 9, 2007. The works of
K. Akarvardar, G. C. Wan, and H.-S. P. Wong are partially supported by theFocus Center for Circuit & System Solutions (C2S2), one of five researchcenters funded under the Focus Center Research Program, a SemiconductorResearch Corporation Program; C. Eggimann, D. Tsamados, Y. S. Chauhan,and A. M. Ionescu were supported in part by the Integrated Project MINAMI;and R. T. Howe were supported in part by a grant from the Charles PowellFoundation. The review of this paper was arranged by Editor T-J. K. Liu.
K. Akarvardar, G. C. Wan, R. T. Howe, and H.-S. P. Wong are with theDepartment of Electrical Engineering and the Center for Integrated Systems,Stanford University, Stanford, CA 94305 USA (e-mail: [email protected]).
C. Eggimann, D. Tsamados, and A. M. Ionescu are with the Swiss FederalInstitute of Technology Lausanne, 1015 Lausanne, Switzerland.
Y. S. Chauhan is with the Semiconductor Research and Development CenterIBM, Bangalore 560045, India.
Color versions of one or more of the figures in this paper are available onlineat http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TED.2007.911070
The idea of combining an electrostatically actuated mechan-
ical switch with a MOSFET was first introduced 40 years ago
[2]. The motivation behind this earliest resonant-gate FET
was to simultaneously obtain a stable and high-Q frequency se-lection provided by the electromechanical beam and an efficient
readout provided by the FET. After [2], hybrid MEMS/FET
structures were mainly used in gas [3][5] and pressure sensing
[6]. The use of the resonant-gate FETs in radiation-resistant
applications was also proposed [7].First in 2002, the SGFET (a variant of the resonant-gate FET)
was conceived as an abrupt current switch [8]: Dynamic thresh-
old operation and the possibility of a subthreshold swing below
60 mV/dec limit were reported. Subsequently, several other
papers described the use of the SGFET as a switch, resonator,
and memory [9][14]. The CMOS-compatible fabrication of
the SGFETs and the experimental characteristics, confirming
the theoretical predictions on the steep, < 60 mV/dec, sub-threshold swing and dynamic threshold operation were also
reported [9], [15]. Recently, the accumulation-mode SGFET
was introduced as a promising device for high-performance
logic circuits [16]. It is shown that such a structure can meet,
in theory, International Technology Roadmap for Semiconduc-tors performance specifications for low-power applications at
25-nm gate length.
Previous work on SGFET modeling is based on the Enz
KrummenacherVittoz (EKV)-based [17] expression of the
gate-to-channel capacitance [8]. This model is valid in all
regimes of operation. However, it requires time-consuming
iterative calculations for the simultaneous resolution of the
force-balance and capacitance equations. Additionally, it does
not provide simple guidelines for the transistor operation since
the variation of the transistor parameters as a function of the
structural parameters (dimensions and material properties) is
not explicitly shown.This is why, in this paper, by considering the operation
region of interest for the logic circuits, we derived a fully
explicit static model for the SGFET. Our new analytical model
provides physical insights related to the SGFET operation and
basic design rules suitable for hand calculations. We develop
simple relations for the pull-in and pull-out voltages, the travel
range, and the SG position with respect to the gate voltage.
In addition, by combining our model with a conventional
MOSFET currentvoltage model, we estimate the behavior and
performance of the SGFET logic circuits.
This paper is organized as follows. In Section II, we in-
troduce our analytical model following a brief description of
0018-9383/$25.00 2008 IEEE
-
7/27/2019 04408768
2/12
AKARVARDAR et al.: ANALYTICAL MODELING OF THE SGFET AND DESIGN INSIGHTS FOR LOW-POWER LOGIC 49
Fig. 1. N-channel SGFET. (a) Three-dimensional structure: The channel width is equal to the beam length (W = WFET = Lbeam), and the channel length isequal to the beam width (L = LFET = Wbeam). (b) Cross section parallel to device length. (c) Equivalent capacitor circuit. (d) Symbol.
the SGFET operation. In Section III, we discuss the optimal
choice of the structural parameters for logic applications. Thecurrentvoltage characteristics of the SGFET are presented in
Section IV. In Section V, we illustrate the use of our model
and the operation of complementary SGFET logic circuits by
introducing the SGFET inverter.
II. OPERATION AND ANALYTICAL MODELING
The 3-D structure, 2-D cross-section, equivalent capacitor
circuit, and symbol of the n-channel SGFET are shown in
Fig. 1. The dimensional parameters are defined as follows: L isthe SGFET channel length, h is the thickness of the suspended-
gate (SG), tox is the gate-oxide thickness, tgap0 is the initialgap thickness, and W is the SGFET channel width that canbe assumed equal to the beam (suspended bridge) length if
W tgap0; the beam width is equal to channel length L.An SGFET combines an electrostatically actuated NEMS
switch and an inversion-mode MOSFET (Fig. 1). It is distin-
guished from a regular MOSFET by the presence of an air
gap between the doubly clamped gate electrode and the gate
oxide. The SG structure in Fig. 1(a) is usually realized by
the sacrificial etching of a material (such as cured polyimide,
polycrystalline, or amorphous silicon) that is deposited on the
gate insulator before the gate formation [9], [15], [18]. The SG
material is typically polysilicon [18] or aluminum (AlSi) [15].
The bottom range of the values reported so far for the SGFETair gap is around a few hundred nanometers [9], [11], [15]
(130 nm minimum [9]). An important challenge regarding the
fabrication of SGFETs featuring CMOS-compatible actuationvoltages is to realize both the gap and the gate electrode in the
10-nm range. Atomic-layer deposition seems a very promising
technique for both structural and sacrificial layers, due to its
monolayer-level thickness control, and can be used for fabri-
cating SGFETs with vertical or lateral dimensions controlled
at the atomic scale [19]. It is important to mention that gap
values below 10 nm were already demonstrated in biosensors
[20]; however, the layer over the gap was much thicker and
not movable. On the other hand, fully released and functional
h = 20 nm-thick nanocomposite AlMo resonators were alsoreported [21].
The operation of the SGFET is explained as follows: At
flatband condition, (VG = VFB), the charge density at thegate electrode and inside the semiconductor is zero, yielding
x = tgap0, where x shows the actual distance between the gateoxide and the gate electrode [Fig. 1(b)]. As VG increases, apositive charge (and also an equal amount of negative charge
inside the semiconductor) is built up in the gate electrode,
giving rise to an electrostatic force pulling down the SG and
resulting in x < tgap0 [Fig. 2(a)]. Until the gate voltage reachesthe pull-in voltage Vpi, the electrostatic force can be balancedby the counteracting elastic force. However, for VG Vpi(corresponding to a critical surface potential s = pi and acritical gap thickness xpi), the electrostatic force overcomes the
elastic component, and the gate collapses (is pulled in) on thegate oxide [Fig. 2(b)]. When the gate is pulled in, the abrupt
-
7/27/2019 04408768
3/12
50 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 55, NO. 1, JANUARY 2008
Fig. 2. Cross-section of the n-channel SGFET parallel to device width.(a) Gate up (VFB < VG < Vpi, tgap0 > x > xpi). (b) Gate down (VG
Vpi, x = 0).
increase in gate capacitance leads to an abrupt reduction of
the threshold voltage and, consequently, a sharp increase in the
drain current [8]. As will be detailed later, the SGFET features
mechanical hysteresis, which means that the pull-out of the SG
requires a pull-out voltage VG = Vpo that is lower than Vpi.In the following, the SGFET pull-in and pull-out voltages
will be separately modeled, and an explicit relationship between
the gate position and the gate voltage will be developed.
A. Pull-In ModelingWe start with the force-balance equation related to the SG
[8], [16]
W LgapV2gap
2x2= k(tgap0 x). (1)
The left-hand side of (1) designates the electrostatic attraction
force applied to the SG, whereas the right-hand side designates
the counteracting elastic force [Fig. 1(b)]. gap is the gap per-mittivity, and Vgap is the voltage drop across the gap. The elasticforce is represented by a linear spring constant k. This is a sim-
plified assumption since the nonlinear stretching component ofthe spring constant, which can lead to a nonnegligible restoring
force (and can alter the pull-out behavior [22]), is neglected.
In (1), the van der Waals attraction [23], [24] between the
SG and the substrate is not taken into account. However, it is
worth mentioning that the impact of the van der Waals forces
on the SGFET characteristics becomes nonnegligible if the air
gap is extremely scaled: As an example, for tgap0 2 nm andVgap = 1 V, the van der Waals forces are theoretically evenhigher than the electrostatic force.
Vgap is expressed as a function of the actual gap thicknessand the VG-dependent semiconductor charge density Qsc as
Vgap = Qsc
gap/x. (2)
The denominator of (2) shows the actual gap capacitance per
unit area, Cgap [Fig. 1(c)]. The substitution of (2) into (1) yields
x = tgap0 W L
2gapkQ2sc. (3)
Note that the second term on the right-hand side of (3) corre-
sponds to the gate displacement x [Fig. 1(b)]. Equation (3) isvalid for x > xpi. Beyond this limit, the system is no longer inequilibrium, and the gate snaps down to the gate oxide, leading
to x = 0.For a uniformly distributed electrostatic force along the beam
and neglecting the residual stress, the spring constant k is givenin terms of the structural parameters by [25], [26]
k =32ELh3
W3. (4)
In a simple MEMS switch, consisting of two parallel metallic
plates separated by an air gap, the stability analysis yields
xpi = 2tgap0/3. When a second capacitance Cf is connectedin series with Cgap, xpi is reduced to
xpi =2Cgap0/Cf
3tgap0 (5)
where Cgap0 = gap/tgap0 is the minimum gap capacitance[27], [28]. Cf induces a negative feedback on Vgap and is nor-mally used to increase the travel range of the moving electrode
in MEMS switches. Note from (5) that, for Cgap0/Cf 2, theinstability (i.e., pull-in) is completely suppressed [27], [28]. In
the case of the SGFET, Cf is equal to the series equivalent ofCox with Csc [Fig. 1(c)].
Simple relationships for the pull-in voltage Vpi, the SGposition at pull-in, xpi, and the surface potential at pull-in(pi) are obtained starting from the depletion approximation.Since our ultimate goal is to use the SGFET in logic circuits
by taking advantage of the sharp onoff transition, we are
naturally interested in the case where the pull-in (and hence
pull-out) occurs before the formation of the inversion channel
(this implies that pi < 2F, where F is the substrate Fermipotential). Therefore, in terms of our objective, the depletion
approximation does not lead to a limitation.
Although SGFETs can be designed in such a way that the
gate is pulled-in in the strong inversion region (see, for instance,
currentvoltage characteristics in [8]), this case exacerbates theshort-channel effects due to the weak gate-to-channel coupling
in the OFF-state [16] and, therefore, will not be considered here.
By contrast, the occurrence of the pull-in in weak inversion
enables to suppress short-channel effects since the threshold
voltage and the subthreshold swing are determined by the
mechanical pull-in of the gate. As will be shown in the next
section, the weak inversion switching allows the SGFET to
eliminate the usual subthreshold region, where the slope of
the currentvoltage characteristic is finite, and to reduce the
threshold voltage without increasing the off-current.
Using the depletion approximation, the depletion charge is
given as a function of the surface potential by
Qd(s) =
2SiqNAs (6)
-
7/27/2019 04408768
4/12
AKARVARDAR et al.: ANALYTICAL MODELING OF THE SGFET AND DESIGN INSIGHTS FOR LOW-POWER LOGIC 51
where Si is the silicon permittivity, qis the elementary charge,and NA is the substrate doping. Substituting (6) into (3) forQsc, the gate position is expressed as a function of the surfacepotential
x(s) = tgap0 W LSiqNA
gapks. (7)
According to (7), as long as the substrate is in depletion, the
gate position is a linear function of the surface potential.
The limit gate position at pull-in (xpi) is written in termsof the limit surface potential at pull-in (pi) by using (5)and by expressing Cf as a series combination of the oxidecapacitance Cox and the semiconductor capacitance Csc (indepletion, Csc = Si/xdi, where xdi is the depletion depth ats = pi)
xpi =2 ( + pi)
3
tgap0 (8)
where
Cgap0
Cox(9a)
Cgap0
2
SiqNA. (9b)
The surface potential at pull-in is expressed by replacing s bypi and x by the expression of xpi [given by (8)] in (7), andsolving the resulting quadratic equation for pi
pi =+
2 2
22(10)
where
3W LSiqNA
tgap0gapk(11a)
2(1 + ) + 2. (11b)
Equation (10) is the largest of the two distinct roots of
the quadratic equation, which provides an accurate xpi.Equations (8) and (10) are the relationships expressing xpi and
pi, respectively, in terms of the structural parameters, and theyare valid for 0 pi < 2F.
The pull-in voltage is defined as the gate voltage leading
to s = pi. From the equivalent capacitor divider circuit inFig. 1(c), the effective gate voltage can be expressed as the sum
of the voltage drops across the gap (Vgap), across the gate oxide(Vox), and on the semiconductor (s)
VG(s) = VFB Qsc(s)x(s)
gap
Qsc(s)
Cox+ s (12)
where VFB is the flatband voltage related to the work function
difference and the oxide charge density. Substituting x(s) byxpi, Qsc(s) by Qd(pi), and s by pi in (12), the pull-in
voltage of the SGFET, provided that the switching occurs in the
weak inversion, is expressed as
Vpi = VFB +
pi + pi (13)
where
1 + Coxgap/xpi
(14)
and = (2SiqNA)0.5/Cox is the usual MOSFET body effect
coefficient. The term inside the parenthesis in (14) corresponds
to the increase in capacitance once the gate is pulled in.
Equation (13) is a general relationship for Vpi. It reducesto the well-known pull-in voltage of the simple NEMS switch
[29], Vpi(sw), for NA (metallic bottom electrode case,leading to pi 0 and xpi (2 )tgap0/3) and VFB 0(same material for both electrodes):
lim Vpi(SGFET)NA
VFB0
= Vpi(sw) =8k(tgap0 + tox/r)3
27gapW L . (15)
In (15), r is the dielectric constant of the gate oxide material.
B. Pull-Out Modeling
In this section, we present a simple relationship for the
SGFET pull-out voltage, starting from the forces acting on
the gate while the gate is pulled in. We take into account
the restoring elastic force and the opposing electrostatic and
adhesion forces.
In the SGFET, once the gate is pulled in, the gate capacitance
increases abruptly, and so do the surface potential and the
charge density. The abrupt increase in charge density can also
be explained by the abrupt reduction of the threshold voltage.
For VG > Vpi, the SGFET behaves as a conventional MOSFET.When VG is swept back from a larger value than the pull-involtage, pull-out does not occur at VG = Vpi because the sur-face potential is higher than pi. This leads to a higher chargedensity and, to a higher electrostatic force than those at the
onset of pull-in (while x = xpi). The release of the gate is alsoretarded (if not completely prevented) by the surface adhesion
forces [30]. Therefore, VG should be reduced to Vpo < Vpiin order for pull-out to happen.
The force-balance equation in the gate-down state, just be-fore the pull-out, can be approximated in the first order as
W LoxV2ox2t2ox
+ Fa = ktgap0 (16)
where the first term on the left-hand side represents the elec-
trostatic force applied to the gate, whereas the term on the
right-hand side shows the elastic restoring force of the doubly
clamped beam. Fa is the surface adhesion force. In (16), we as-sumed that the spring constant is given by (4) even after the gate
is pulled in [Fig. 2(b)]. The restoring elastic force can be more
accurately calculated by taking into account the influence of
the nonlinear stretching component on the spring constant [22].Furthermore, we neglected the peeling of the gate [31][33]
-
7/27/2019 04408768
5/12
52 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 55, NO. 1, JANUARY 2008
when VG is swept back toward Vpo, and we assumed that thegate stays in contact with the whole gate oxide area until the
occurrence of the pull-out.
In the absence of capillary forces and at small roughness
values, the adhesive interactions are dominated by the attractive
van der Waals forces between noncontacting surfaces rather
than by the areas that are actually in contact [34], [35]. Dueto the surface roughness, which prevents the intimate contact
of dry MEMS surfaces, the adhesion energies are very low,
typically in microjoules per square meter range [34], [35].
When the gate is in the down state, Fa can be expressed as
Fa = 2W L
D0(17)
where is the interfacial adhesion energy per unit area [34]. D0is an offset corresponding to the closest approach of the two sur-
faces and is determined by the average surface roughness [34].
For VG = Vpo, the depletion approximation leads to
Vox|VG=Vpo =
po, where po is the surface potential atpull-out. From (16) and (17), po is given by
po =ox
SiqNA
ktgap0
W L
2
D0
. (18)
Since po > 0 for an n-channel SGFET, the condition
Fa < ktgap0 (19)
needs to be fulfilled in order for the beam not to stick to the
substrate. In other words, in the absence of the electrostatic
force (at flatband condition), the restoring force should be large
enough to overcome the surface adhesion force.The pull-out voltage is given as the sum of the flatband
voltage, the voltage drop on the gate-oxide, and the surface
potential at pull-out
Vpo = VFB +
po + po. (20)
Equation (20) stands as a general pull-out expression: It re-
duces to the pull-out equation of the simple MEMS switch
featuring a dielectric layer [29] for NA , VFB 0,and 0
lim Vpo(SGFET)NA
VFB0
0
= Vpo(sw) =
2ktgap0t2
oxoxW L
. (21)
It is worth mentioning that, besides the surface adhesion
forces, the hysteresis window can also be enlarged by the oxide
surface charge, whose density depends on the gate position
(pulled in or pulled out). Indeed, this property is exploited in
[12] to build a capacitorless 1T memory cell. However, it is
experimentally shown that this effect is mostly significant when
the gate oxide is degraded, for instance, by an oxygen plasma
process that induces traps on the oxide surface [9], [12]. More
promising future SGFET memory architectures are likely to use
controlled thin storage layers in the gate dielectric instead ofthe oxide traps; thin nanocrystal or ferroelectric layers can be
engineered to achieve information storage in SGFET devices
with relatively low operation voltages (< 510 V) [36].
C. SG Position as a Function of Gate Voltage
To obtain a relationship between the gate position and the
gate voltage, (3) and (12) need to be solved together. However,the relationship resulting from these equations involves a third-
degree polynomial and does not provide a simple solution for
x(VG) and s(VG) even when the depletion approximationis used.
To obtain a simple, yet reasonably accurate expression
for x(VG) yielding x = tgap0 for VG = VFB and x = xpi atVG = Vpi, we first impose x(s) = xpi in (12) and solve theresulting quadratic equation for s while Qsc = Qd:
s,up =
2
VG VFB +
2
4
2. (22)
Equation (22) is valid when the gate is in the up state and
VG Vpi. When the gate is pulled down and VG Vpo, sis given by the usual MOSFET relationship [37] obtained by
imposing Qsc = Qd and x = 0 in (12) or by replacing in(22) by :
s,MOS =
2
VG VFB +
2
4
2. (23)
Again, the quadratic equations providing (22) and (23) enable
also a second root as solution, which is discarded since it does
not correspond to the physical situation. Equations (22) and
(23) are compared to the iterative solution of (3) and (12) inFig. 3(a). In the iterative solution, the exact charge equation
[37] including acceptors, holes, and electrons is used. Naturally,
s,MOS is in very good agreement with the numerical solutionfor po < s < 2F above which the depletion approxima-tion loses its validity. s,up is also in good agreement withthe iterative solution, particularly for VG values close to Vpiand VFB (since, for VG VFB, Qsc 0, and the sensitivityto x is very weak). The slight discrepancy between the exactsolution and (22) for the gate-voltage range VFB < VG < Vpiis of minor importance since, in the currentvoltage character-
istic, this gate-voltage range corresponds to the bottom of the
subthreshold region with very low drain-current values.The expression for x(VG) is obtained by substituting (22)into (3) for s while Qsc = Qd:
x(VG) = tgap0 W LSiqNA
gapk
2
VG VFB +
2
4
2.
(24)
Equation (24), which is valid for xpi x(VG) tgap0, iscompared to the iterative solution in Fig. 3(b). Both models
reproduce the same trend; however, (24) underestimates the gap
height for the intermediate x values due to the approximationmade in (22). On the other hand, the difference between the
approximate and exact solutions for Vpi and Vpo in Fig. 3(a)and (b) originates from the depletion approximation and is
-
7/27/2019 04408768
6/12
AKARVARDAR et al.: ANALYTICAL MODELING OF THE SGFET AND DESIGN INSIGHTS FOR LOW-POWER LOGIC 53
Fig. 3. Exact (numerical) model versus analytical model. (a) Variation ofthe surface potential as a function of the gate voltage. (b) Variation of thenormalized gap thickness as a function of the gate voltage. W = 650 nm,L = 100 nm, h = 10 nm, E = 170 GPa (Si) (k = 2 N/m), tgap0 = 10 nm,
tox = 2 nm (SiO2), NA = 3.1017
cm3
, VFB = 0, = 25 J/m2
, andD0 = 0.2 nm. Numerical model refers to the iterative solutions of (3) and (12)and uses the exact charge equation featuring acceptors, holes, and electrons.
negligible (kT/q). Note that VFB is assumed equal to zeroin this particular example, leading to s = 0 in Fig. 3(a) andx(VG)/tgap0 = 1 in Fig. 3(b) for VG = 0.
III. OPTIMAL DESIGN WINDOW FOR
SGFET LOGIC SWITCHES
The use of SGFETs in logic circuits imposes the following
conditions for the pull-in and pull-out voltages:
Vpo > 0 (25a)
Vpi < VDD (25b)
pi < 2F. (25c)
In addition to these, the threshold voltage when the gate
is pulled in (the conventional MOSFET threshold voltage),
VT,MOS, should naturally satisfy
VT,MOS = VFB +
2F + 2F < VDD. (25d)
The scaling of the supply voltage VDD is normally imposed by
(25b), rather than (25d), despite the high body doping that tendsto increase VT,MOS.
The impact of the structural parameters on the constraints
mentioned earlier is shown in Fig. 4(a)(d) where the varia-
tion of the pull-in and pull-out voltages as a function of the
(a) device width, (b) gap height, (c) SG thickness, and
(d) gate materials Youngs modulus is shown. Note that the
dimensions used in Fig. 4 are in the nanometer scale, and
they are about three orders of magnitude smaller than thetypical dimensions of MEMS switches (several micrometers
for the vertical dimensions and hundreds of micrometers for
the beam length) in order to meet the requirements of a small
device footprint and a low-voltage actuation (the range of the
parameters in Fig. 4 is selected such that Vpi 2 V). As ageneral trend, both Vpi and Vpo increase as the beam (SG) getsstiffer, i.e., as the spring constant increases by lowering W orincreasing h or E. Vpi and Vpo increase also for a larger tgap0due to the lowered electrostatic force. For tgap0 = h = 10 nmand for a polysilicon SG (E = 170 GPa), sub-1 V operation(Vpi VFB < 1 V) requires roughly W 700 nm. Furtherlateral scaling requires a corresponding vertical scaling (the
reduction of tgap0 and/or h below 10 nm) or the use of a gatematerial with a smaller Youngs modulus (Al or Ti). However,
these solutions may not be viable due to the pronounced impact
of the van der Waals forces (as tgap0 is reduced) and also dueto the pull-out requirements. Fig. 4 reveals that an excessive
increase of W or an excessive lowering of the tgap0, h, or Emay reduce the elastic restoring force (ktgap0) to an intolerablylow value, which could result in sticking according to (19).
Therefore, the scaling of VDD depends (indirectly) on the pull-out characteristics as well. The maximum value of W andthe minimum value of tgap0, h, and E that would not lead tosticking (while the remaining parameters are fixed) correspond
to po = 0 [and hence to Vpo = VFB according to (20)], andthey are indicated with arrows in Fig. 4.
The strong sensitivity of Vpi and Vpo with respect to thehorizontal and vertical dimensions imposes a particularly tight
process control to obtain uniform SGFET characteristics. The
dependence of the pull-out (and hence the hysteresis window
width) on the surface adhesion forces is the main technological
challenge related to the fabrication and design of SGFETs.
A reliable fabrication of SGFETs with a well-controlled pull-
out requires in-depth understanding and control of the surface
adhesion forces.
Another important parameter in SGFETs is the stable travel
range (maximum ofx) of the SG before it hits the gate oxidedue to instability. Fig. 5 shows that the travel range is a strongfunction of the substrate doping concentration, and it can even
be extended to the whole gap height (x = tgap0 or xpi = 0) inlowly doped substrates (solid curve) [27]. However, the use of
the SGFET in logic circuits requires instability (in contrast to
inertial sensor applications of microstructures) to improve the
on/off capacitance or current ratio, and therefore, a large travel
range is not desired. The travel range can be minimized (i.e., xpican be maximized) by increasing the substrate doping, as shown
in Fig. 5: A highly doped substrate emulates a metallic bottom
electrode, and the xpi/tgap0 ratio converges to the maximumvalue given by (5) [27].
A second reason which makes a low substrate dopingundesirable is the possible insufficiency of the electrostatic
-
7/27/2019 04408768
7/12
54 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 55, NO. 1, JANUARY 2008
Fig. 4. Variation of the pull-in and pull-out voltages as a function of the (a) device width, (b) gap height, (c) SG thickness, and (d) Youngs modulus of the gatematerial. L = 100 nm, tox = 2 nm (SiO2), NA = 10
18 cm3, VFB = 0.13 V, = 20 J/m2, D0 = 0.2 nm, and VT,MOS = 1.4 V. The variation ofVpo for
Vpo > VT,MOS is not shown since (20) is only valid in depletion.
Fig. 5. Normalized stable travel range as a function of the substrate dopingconcentration. tgap0 = 10 nm, h = 10 nm, L = 100 nm, tox = 2 nm (SiO2),
and VFB = 0.13 V.
force induced by the depletion charge to create instability.
Such structures with lowly doped substrates and relatively
high spring constants present no interest for logic appli-
cations since they do not enable weak inversion switching
(the channel would be already in inversion when the pull-
in occurs). In Fig. 5, for W = 800 nm and E = 170 GPa,the pull-in cannot be induced by the depletion charge alone
if NA < 6.1016 cm3. For W = 800 nm and E = 1 TPa
(Youngs modulus of carbon-nanotube), this limit shifts to
NA < 3.1017 cm3. In summary, as the beam gets stiffer, the
minimum substrate doping required to induce weak inversionswitching increases.
IV. CURRENTVOLTAGE CHARACTERISTICS
SGFET currentvoltage characteristics can be obtained start-
ing from any MOSFET model just by replacing the oxide
capacitance Cox in the original model equations with a seriesequivalent of Cox and Cgap = gap/x(VG), where x(VG) isgiven by (24).
For a long-channel device, as long as the pull-in occurs
in weak inversion (pi < 2F), the drain voltage VD hasnegligible influence on the surface potential [37] and therefore
does not modulate Vpi. This means that our model equationsthat are independent of VD can be safely integrated into theconventional MOSFET models. In a short-channel device, the
influence ofVD on Vpi and Vpo can be minimized by increasingthe substrate doping. As the channel length is reduced, the
fringing fields and the electrostatic force applied to the SG by
the source and drain regions (that are neglected in this paper)
become also important.
A transfer characteristic example for the SGFET, obtained by
integrating our analytical relationships into the MOSFET EKV
model [17], is shown in Fig. 6 (solid curve) along with the con-
ventional (non-SG) MOSFET characteristic (dotted curve). On
the same figure, the SGFET characteristic, obtained by using
the EKV model and the numerical solution of x(VG), Vpi, andVpo, is also shown (dashed curve). Note that the discrepancybetween the exact and the numerical model originated from
the approximation that is made while deriving (24) is trivial
since the corresponding drain-current values are well below the
junction leakage floor (assumed here as 1 pA/m).
According to Fig. 6, the SGFET operates as an idealswitch with an infinite subthreshold slope. However, the
-
7/27/2019 04408768
8/12
AKARVARDAR et al.: ANALYTICAL MODELING OF THE SGFET AND DESIGN INSIGHTS FOR LOW-POWER LOGIC 55
Fig. 6. Analytically and numerically calculated SGFET transfer characteris-tics (MOSFET characteristic is also shown for comparison). W = 500 nm,L = 100 nm, h = 10 nm, E = 170 GPa (Si), tgap0 = 10 nm (k =
4.35 N/m), tox = 2 nm (SiO2), VFB = 1.1 V, NA = 1018 cm3, VD =
1.5 V, = 250 cm2/(V s), = 45 J/m2, and D0 = 0.2 nm. Leakage flooris assumed equal to 1 pA/m.
Fig. 7. Schematic of the complementary SGFET inverter.
currentvoltage characteristic has hysteresis. This example
illustrates the interest of the SGFET for logic circuits:
By selecting Vpo just above 0 V (Vpo = 90 mV in thisexample) and the conventional MOSFET threshold just above
Vpo (VT,MOS = 280 mV in this example), SGFET provides
many orders of magnitude reduction in the off-current withoutdeteriorating the on-current, which means a tremendous im-
provement in Ion/Ioff ratio. Fig. 6 shows also the importanceof the tight Vpo control in order to take the full advantage of theSGFET low-leakage feature.
V. SGFET INVERTER
In this section, we present the operation and characteristics
of the complementary SGFET inverter in order to show the
efficiency of our simple model to design SGFET logic circuits
and to illustrate the advantages of the SGFETs mentioned in the
previous section.
The schematic of the complementary SGFET inverter isshown in Fig. 7. The architecture is the same as that of a CMOS
inverter except that the complementary MOSFETs are replaced
by complementary SGFETs.
The steady-state behavior of the proposed circuit is explained
as follows.
1) For Vin = VDD, Vout = 0. This yields VGS,n = VDD >Vpi,n, which means that the gate of the n-channel SGFET
is pulled in. On the other hand, |VGS,p| = 0 < |Vpo,p|,implying that the gate of the p-channel device is pulled
out (the subscripts n and p show the voltages or
parameters related to the n- and p-channel SGFETs,
respectively).
2) For Vin = 0, Vout = VDD. This leads VGS,n = 0 < Vpo,n,implying that the NFET is pulled out. On the other hand,
|VGS,p| = VDD > |Vpi,p|, which means that the PFET ispulled in.
In summary, in each static state, one of the transistors is
pulled in, whereas the other is pulled out. The pulled-out
transistor has a much smaller off-current than that of a regular
MOSFET, as shown in Fig. 6; therefore, the overall static power
dissipation in the SGFET inverter is reduced as compared to its
CMOS counterpart.
In SGFETs, once the gate is pulled-in in the subthreshold
region, the channel region either stays in depletion or it is
driven to inversion (even to strong inversion), depending on
the relative position of Vpi with respect to VT,MOS. We in-vestigate the SGFET inverter characteristics by distinguishing
two different situations based on the relationship between Vpiand VT,MOS. We assume a fully symmetrical case (i.e., samechannel mobility (), |Vpo|, |Vpi|, and |VT,MOS| for PFET andNFET) to focus the attention on the intrinsic behavior of the
SGFET.1) |Vpi| < |VT,MOS|: The transfer characteristics for both
types of SGFETs, satisfying |Vpi| < |VT,MOS|, are shown inFig. 8(a) in semilogarithmic and linear scales. Since, in this
case, the semiconductor surface stays in depletion after the pull-
in, the threshold voltage of the SGFET is equal to VT,MOS. Con-sequently, in linear scale, the SGFET transfer characteristics are
the same as those of the conventional MOSFETs.
The static characteristics of the SGFET inverter, designed
with the SGFETs featuring the characteristics in Fig. 8(a),
are shown in Fig. 8(b). The availability of an analytical
currentvoltage model that does not require iterations is crucial
for obtaining such characteristics in a short computing duration.Fig. 8(b) shows exactly the same voltage transfer characteristic
(and the supply current characteristic in linear scale) as that of
a CMOS inverter since the threshold voltages of the SGFETs
are not modified by the mechanical pull-in and pull-out events
that both occur in the subthreshold region. However, the SGFET
inverter consumes less static power than the CMOS inverter
because, in each static state, one of the SGs is pulled up. For this
configuration, the switching threshold of the SGFET inverter
Vth is equal to VDD/2 as in a CMOS inverter.2) |Vpi| > |VT,MOS|: The SGFET transfer curves in linear
scale corresponding to this second case are shown in Fig. 9(a)
(the n-channel SGFET characteristic in semilogarithmic scale
is similar with that in Fig. 6). The |Vpi| > |VT,MOS| conditionis satisfied just by reducing the beam length from its value in
-
7/27/2019 04408768
9/12
56 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 55, NO. 1, JANUARY 2008
Fig. 8. |Vpi| < |VT,MOS| case. (a) N- and p-channel SGFET transfer charac-teristics in linear and logarithmic scales (|VD| = 1.5 V). (b) ComplementarySGFET inverter static characteristics. W = 600 nm, L = 100 nm, h = 10 nm,E = 170 GPa (k = 2.5 N/m), tgap0 = 10 nm, tox = 2 nm (SiO2), NA =
1018 cm3, |VFB| = 1.05 V, = 250 cm2/(V s), VDD = 1.5 V, =
2 J/m2, and D0 = 0.2 nm.
Fig. 8, W = 600 nm, to W = 510 nm in Fig. 9 and hencemaking the beam stiffer (the spring constant is increased). The
other parameters remained as in Fig. 8 except that a higher is assumed in Fig. 9 to locate |Vpo| between 0 and |VT,MOS|.Notice in Fig. 9(a) that the threshold voltage of the SGFETs is
no longer given by VT,MOS while sweeping VG from0 to |VDD|,but it is now equal to Vpi (i.e., to Vpi,n or Vpi,p).
The modulation of the threshold voltages by the mechanical
pull-in results in a fundamental change in the SGFET-inverter
transfer characteristics [Fig. 9(b)]: The voltage transfer curve
now exhibits hysteresis with two different switching thresholds
(Schmitt-trigger behavior). This is because, in this configura-
tion, the logic state change is not possible unless one of thebeams, which was previously pulled out, is pulled in. As a
result, the two switching threshold voltages are dispersed on
either side of VDD/2: Vth = Vpi,n while sweeping VG from 0to VDD, and Vth = VDD |Vpi,p|while sweeping VG from VDDto 0. Even though the switching event is delayed in Schmitt-
trigger-like operation with respect to that in a CMOS inverter,
the SGFET inverter features all the advantages of a Schmitt-
trigger operation, such as extremely sharp switching and the
ability to operate in noisy environments.
The variation of the supply current as a function of the input
voltage, corresponding to the voltage transfer characteristic in
Fig. 9(b), is shown in Fig. 9(c). It is noticed that, in this config-
uration, the short-circuit current drawn from the power supplyaround the switching threshold can be considerably reduced
Fig. 9. |Vpi| > |VT,MOS| case. (a) N- and p-channel SGFET transfer char-acteristics in linear scale (|VD| = 1.5 V). (b) Voltage transfer characteristic ofthe complementary SGFET inverter (comparison with the conventional CMOSinverter). (c) Variation of the short-circuit supply current as a function of theinput voltage (comparison with the conventional CMOS inverter). Same param-
eters as in Fig. 8 except W = 510 nm (k = 4.1 N/m) and = 40 J/m2
.
as compared with that of a conventional inverter. This addi-
tional feature reinforces the low-power aspect of the SGFET
inverter.
The SGFET inverter with hysteresis stands as a very promis-
ing static memory cell (SRAM) with considerably reduced
transistor count and a very simple read scheme: The data are
stored at the output node, and they are written by applying logic
0 or 1 to the input. In standby or during read operation,
Vin = VDD/2 is applied, allowing the output to retain its state.In SGFET SRAM, the hysteresis window width is given by
2Vpi VDD and can be adjusted by the circuit designer (by
modifying the beam length W on the layout). In Fig. 10, a volt-age transfer-characteristic example with an enlarged hysteresis
-
7/27/2019 04408768
10/12
AKARVARDAR et al.: ANALYTICAL MODELING OF THE SGFET AND DESIGN INSIGHTS FOR LOW-POWER LOGIC 57
Fig. 10. Voltage transfer characteristic of the complementary SGFETinverter for |Vpi| |VT,MOS|. Same parameters as in Fig. 8 except
W = 480 nm (k = 4.9 N/m) and = 60 J/m2.
window is shown. This optimized characteristic, enabling a hys-
teresis window width of VDD/2 and centered around VDD/2,is obtained simply by reducing W from 510 nm (Fig. 9) to480 nm (Fig. 10). A large hysteresis window reduces the short-
circuit supply current to a negligible level.
VI. CONCLUSION
SGFET static characteristics and design criteria were ana-
lyzed. Basic formulas for the pull-in and pull-out voltages were
provided, and the SG position is explicitly expressed in terms
of the gate voltage. By using our model, key device parameters
were highlighted, a considerable Ion/Ioff ratio improvementin SGFETs is demonstrated, and the conditions for a low-
power and low-voltage operation are determined. Challenges
related to surface adhesion forces and to strong sensitivity of the
transistor parameters to the beam geometry are also indicated.
The operation and performance assessment of the complemen-
tary SGFET inverter based on the developed analytical model
suggests a significantly reduced OFF-state power dissipation
and an improved functionality as compared to classical CMOS.
Promising applications of the SGFETs include the header and
footer switches for power management and SRAM configura-
tion switches for field-programmable-gate-arrays.
REFERENCES
[1] International Technology Roadmap for Semiconductors: Process Inte-gration, Devices, and Structures, 2006. [Online]. Available: http://www.itrs.net/Links/2006Update/2006UpdateFinal.htm
[2] H. C. Nathanson, W. E. Newell, R. A. Wickstrom, and J. R. Davis, Theresonant gate transistor, IEEE Trans. Electron Devices, vol. ED-14, no. 3,
pp. 117133, Mar. 1967.[3] G. F. Blackburn, M. Levy, and J. Janata, Field-effect transistor sensitive
to dipolar molecules, Appl. Phys. Lett., vol. 43, no. 7, pp. 700701,Oct. 1983.
[4] T. Sato, M. Shimizu, H. Uchida, and T. Katsube, Light-addressablesuspended-gate gas sensor, Sens. Actuators B, Chem., vol. 20, no. 2/3,pp. 213216, Jun. 1994.
[5] Z. Gergintschew, P. Kometzky, and D. Schipanski, The capacitively con-trolled field effect transistor (CCFET) as a new low power gas sensor,Sens. Actuators B, Chem., vol. 35/36, no. 13, pp. 285289, Oct. 1996.
[6] J. A. Voorthuyzen and P. Bergveld, The PressFET: An integrated electret-MOSFET based pressure sensor, Sens. Actuators, vol. 14, no. 4, pp. 349360, Aug. 1988.
[7] J. Huang, R. T. Howe, and H.-S. Lee, Vacuum-insulated-gate field-effecttransistor, Electron. Lett., vol. 25, no. 23, pp. 15711573, Nov. 1989.
[8] A. M. Ionescu, V. Pott, R. Fritschi, K. Banerjee, M. J. Declerq, P. Renaud,
C. Hibert, P. Fluckiger, and G. A. Racine, Modeling and design of alow-voltage SOI suspended-gate MOSFET (SG-MOSFET) with a metal-over-gate architecture, in Proc. ISQED, 2002, pp. 496501.
[9] N. Abele, R. Fritschi, K. Boucart, F. Casset, P. Ancey, andA. M. Ionescu, Suspended-gate MOSFET: Bringing new MEMS func-tionality into solid-state MOS transistor, in IEDM Tech. Dig., 2005,pp. 479481.
[10] N. Abele, V. Pott, K. Boucart, F. Casset, K. Segueni, P. Ancey, andA. M. Ionescu, Comparison of RSG-MOSFET and capacitive MEMSresonator detection, Electron. Lett., vol. 41, no. 5, pp. 242244,Mar. 2005.
[11] N. Abele, K. Segueni, K. Boucart, F. Casset, B. Legrand, L. Buchaillot,P. Ancey, and A. M. Ionescu, Ultra-low voltage MEMS resonatorbased on RSG-MOSFET, in Proc. IEEE Int. Conf. MEMS, 2006,pp. 882885.
[12] N. Abele, A. Villaret, A. Gangadharaiah, C. Gabioud, P. Ancey, andA. M. Ionescu, 1T MEMS memory based on suspended gate MOSFET,in IEDM Tech. Dig., 2006, pp. 509512.
[13] B. Pruvost, H. Mizuta, and S. Oda, 3-D design and analysis of functionalNEMS-gate MOSFETs and SETs, IEEE Trans. Nanotechnol., vol. 6,no. 2, pp. 218224, Mar. 2007.
[14] T. Nagami, H. Mizuta, N. Momo, Y. Tsuchiya, S. Saito, T. Arai,T. Shimada, and S. Oda, Three-dimensional numerical analysis ofswitching properties of high-speed and nonvolatile nanoelectromechan-ical memory, IEEE Trans. Electron Devices, vol. 54, no. 5, pp. 11321139, May 2007.
[15] S. Frederico, C. Hibert, R. Fritschi, P. Fluckiger, P. Renaud, andA. M. Ionescu, Silicon sacrificial layer dry etching (SSLDE) for free-
standing RF MEMS architectures, in Proc. IEEE Int. Conf. MEMS, 2003,pp. 570573.
[16] H. Kam, D. T. Lee, R. T. Howe, and T.-J. King, A new nano-electro-mechanical field effect transistor (NEMFET) design for low-power elec-tronics, in IEDM Tech. Dig., 2005, pp. 463466.
[17] C. C. Enz, F. Krummenacher, and E. A. Vittoz, An analytical MOS tran-sistor model valid in all regions of operation and dedicated to low-voltageand low-current applications, Analog Integr. Circuits Signal Process.,vol. 8, no. 1, pp. 83114, Jul. 1995.
[18] H. M. Kotb, A. C. Salaun, T. M. Brahim, and O. Bonnaud, Air-gap polycrystalline silicon thin-film transistors for fully integratedsensors, IEEE Electron Device Lett., vol. 24, no. 3, pp. 165167,Mar. 2003.
[19] M. K. Tripp, C. Stampfer, D. C. Miller, T. Helbling, C. F. Hermann,C. Hierold, K. Gall, S. M. George, and V. M. Bright, The mechanicalproperties of atomic layer deposited alumina for use in micro- and nano-
electromechanical systems, Sens. Actuators A, Phys., vol. 130, pp. 419429, Aug. 2006.[20] D.-Y. Jang, Y.-P. Kim, H.-S. Kim, S.-H. K. Park, S.-Y. Choi, and
Y.-K. Choi, Sublithographic vertical gold nanogap for label-free elec-trical detection of protein-ligand binding, J. Vac. Sci. Technol., vol. 25,no. 2, pp. 443447, Mar. 2007.
[21] Z. Lee, C. Ophus, L. M. Fischer, N. N. Fitzpatrick, K. L. Westra, S. Evoy,V. Radmilovic, U. Dahmen, and D. Mitlin, Metallic NEMS componentsfabricated from nanocomposite AlMo films, Nanotechnology, vol. 17,no. 12, pp. 30633070, Jun. 2006.
[22] G. M. Rebeiz, RF MEMS: Theory, Design, and Technology. Hoboken,NJ: Wiley, 2003.
[23] R. Maboudian and R. T. Howe, Critical review: Adhesion in surface mi-cromechanical structures, J. Vac. Sci. Technol. B, Microelectron. Process.Phenom., vol. 15, no. 1, pp. 120, Jan. 1997.
[24] A. Granaldi and P. Decuzzi, The dynamic response of resistive mi-croswitches: Switching time and bouncing, J. Micromech. Microeng.,
vol. 16, no. 7, pp. 11081115, Jul. 2006.[25] P. Osterberg, H. Yie, X. Cai, J. White, and S. Senturia, Self-consistent
simulation and modeling of electrostatically deformed diaphragms, inProc. IEEE MEMS Conf., 1994, pp. 2832.
[26] J. Muldavin and G. M. Rebeiz, High isolation CPW MEMS shuntswitchesPart I: Modeling, IEEE Trans. Microw. Theory Tech., vol. 48,no. 6, pp. 10451052, Jun. 2000.
[27] J. I. Seeger and S. B. Crary, Stabilization of electrostatically actuatedmechanical devices, in Proc. Int. Conf. Solid-State Sens. Actuators, 1997,pp. 11331136.
[28] E. K. Chan and R. W. Dutton, Electrostatic micromechanical actuatorwith extended range of travel, J. Microelectromech. Syst., vol. 9, no. 3,pp. 321328, Sep. 2000.
[29] MEMS: A Practical Guide to Design, Analysis and Applications,J. G. Korvink and O. Paul, Eds. Norwich, NY: William Andrew, 2006,ch. 14.
[30] L. L. Mercado, S.-M. Kuo, T.-Y. T. Lee, and L. Liu, Mechanics-basedsolutions to RF MEMS switch stiction problem, IEEE Trans. Compon.Packag. Technol., vol. 27, no. 3, pp. 560567, Sep. 2004.
-
7/27/2019 04408768
11/12
58 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 55, NO. 1, JANUARY 2008
[31] C. H. Mastrangelo and C. H. Hsu, Mechanical stability and adhesion ofmicrostructures under capillary forcesPart II: Experiments, J. Micro-electromech. Syst., vol. 2, no. 1, pp. 4455, Mar. 1993.
[32] C. H. Mastrangelo, Adhesion-related failure mechanisms in microme-chanical devices, Tribol. Lett., vol. 3, no. 3, pp. 223238, 1997.
[33] N. Tas, T. Sonnenberg, H. Jansen, R. Legtenberg, and M. Elwenspoek,Stiction in surface micromachining, J. Micromech. Microeng., vol. 6,no. 4, pp. 385397, 1996.
[34] J. A. Knapp and M. P. de Boer, Mechanics of microcantilever beams sub-ject to combined electrostatic and adhesive forces, J. Microelectromech.Syst., vol. 11, no. 6, pp. 754764, Dec. 2002.
[35] F. W. Delrio, M. P. De Boer, J. A. Knapp, E. D. Reedy, Jr.,P. J. Clews, and M. L. Dunn, The role of van der Waals forces in adhesionof micromachined surfaces, Nat. Mater., vol. 4, no. 8, pp. 629634,2005.
[36] G. Salvatore, D. Bouvet, andA. M. Ionescu, Advanced 1T MEMmemorycell architectures, Internal Research Report, 2007. Integrated ProjectMINAMI.
[37] Y. Tsividis, Operation and Modeling of the MOS Transistor, 2nd ed.New York: McGraw-Hill, 1999.
Kerem Akarvardar received the B.S. and M.S.degrees in electrical engineering from Istanbul Tech-nical University, Istanbul, Turkey, in 1996 and2000, respectively, the DEA degree in microelec-tronics from Joseph Fourier University, Grenoble,France, in 2003, and the Ph.D. degree in micro/nanoelectronics from Institut National Polytechniquede Grenoble, Grenoble, in 2006.
He was a Process Engineer with the NationalResearch Institute of Electronics and Cryptology(UEKAE-YITAL), Kocaeli, Turkey, from 1996 to
2000. During his Ph.D. studies, he was with the Institute of Microelectronics,Electromagnetism, and Photonics, Grenoble. He is currently a Postdoctoral Re-searcher with the Center for Integrated Systems, Stanford University, Stanford,CA. His research interests include emerging logic and memory devices, inparticular advanced SOI FETs and NEMS-based devices.
Christoph Eggimann received the B.S. and M.S.degrees in microengineering from the Swiss FederalInstitute of Technology, Lausanne, Switzerland, in2005 and 2007, respectively. He also studied at theUniversity of Waterloo, Waterloo, ON, Canada, andmost recently at Stanford University, Stanford, CA.
He is currently with the Swiss Federal Institute ofTechnology Lausanne, Lausanne, Switzerland. Hisscientific interests include micro- and nanosystemtechnology and analog microelectronics.
Dimitrios Tsamados was born in Athens, Greece,in 1976. He received the B.Sc. degree in physicsfrom the Aristoteles University of Thessaloniki,Thessaloniki, Greece, in 1999, the M.Sc. degree inoptics, optoelectronics, and microwaves from theInstitut National Polytechnique de Grenoble (INPG),Grenoble, France, in 2000, the M.Sc. degree inmicroelectronics from the Universit Joseph Fourier,Grenoble, in 2001, and the Ph.D. degree in micro-and nanoelectronics from the INPG for his workon RF-microelectromechanical-system reliabilityin 2005.
Since September 2005, he has been with the Swiss Federal Institute
of Technology Lausanne, Lausanne, Switzerland, working on suspended-gate field-effect-transistor (FET) numerical simulation and modeling, carbonnanotubenanoelectromechanical systems, and organic FETs.
Yogesh Singh Chauhan received the B.E. degree inelectronics and telecommunication engineering fromthe Shri Govindram Seksaria Institute of Technol-ogy and Science, Indore, India, in 2001, and theM.Tech. degree in microelectronics, very large scaleintegration (VLSI), and display technologies fromthe Indian Institute of Technology Kanpur, Kanpur,India, in 2003, and the Ph.D. degree in compact
modeling of high-voltage MOSFETs from the SwissFederal Institute of Technology Lausanne, Lausanne,Switzerland, in 2007.
During his M.Tech. studies, he was with Samtel India, Ltd., working onthe design of current-programmed active matrix displays. He was with STMi-croelectronics, Noida, India, as an Associate Design Engineer from 2003 to2004, focusing on VLSI I/O library design and validation on MAT10 quality.He was with the European Commission research project ROBUSPIC (ROBUstmixed signal design methodologies for Smart Power ICs) from March 2004 toJuly 2007. He is currently with the Semiconductor Research and DevelopmentCenter IBM, Bangalore, India, as an Advisory Research Engineer. His respon-sibilities include compact modeling of both active and passive semiconductordevices. He is the author or a coauthor of 19 papers in international refereed
journals and conferences. He is also the reviewer ofSolid State Electronics. Hiscurrent research interests include simulation, modeling, and characterization ofsemiconductor devices.
Dr. Chauhan is a member of the IEEE Electron Devices Society. He is
an active Reviewer of IEEE TRANSACTIONS ON ELECTRON DEVICES. Hewas also the Reviewer of the IEEE International Conference of VLSI Designin 2003.
Gordon C. Wan received the B.S degree (with high-est honors) in electrical engineering and mathematicsfrom the University of Texas at Austin, Austin, TX,in 2005, and the M.S.E.E degree from Stanford Uni-versity, Stanford, CA, in 2007, where he is currentlyworking toward the Ph.D. degree in electrical engi-
neering at the Center for Integrated Systems.His research interests include nanoscale physics
and device modeling, including carbon nanotubesand nanoelectromechanical systems.
Adrian Mihai Ionescu (S91M93SM00)received the Ph.D. degree in microelectronics fromthe University Politehnica of Bucharest, Bucharest,Romania, in 1994, and the Ph.D. degree in physicsof semiconductors from the Institut National Poly-
technique de Grenoble, Grenoble, France, in 1997.He was with LETI-CEA, Grenoble, and CNRS,
France, and he was a Visiting Researcher with theCenter for Integrated Systems, Stanford University,Stanford, CA. He is currently an Associate Profes-sor with the Swiss Federal Institute of Technology
Lausanne, Lausanne, Switzerland, where he is the Director of the Laboratoryof Micro/Nanoelectronic Devices and was the Director of the Institute ofMicroelectronics and Microsystems of EPFL from 2002 to 2006. He is theauthor or a coauthor of around 100 research papers. His current researchinterests include the design, modeling, and characterization of submicrometerMOS devices, single-electron devices and few-electron circuit architectures,SOI novel applications, and RF microelectromechanical system.
Dr. Ionescu was appointed as the National Representative of Switzerland forthe European Nanoelectronics Initiative Advisory Council. He received threeBest Paper Awards in international conferences and the Annual Award of the
Technical Section of the Romanian Academy of Sciences in 1994. He served inthe ISQED and IEDM conference technical committees in 2003 and 2004 andas the Technical Program Committee Chair of ESSDERC in 2006.
-
7/27/2019 04408768
12/12
AKARVARDAR et al.: ANALYTICAL MODELING OF THE SGFET AND DESIGN INSIGHTS FOR LOW-POWER LOGIC 59
Roger T. Howe (F96) received the B.S. degree inphysics from Harvey Mudd College, Claremont, CA,and the M.S. and Ph.D. degrees in electrical engi-neering from the University of CaliforniaBerkeley,Berkeley, in 1981 and 1984, respectively.
After his academic stint with Carnegie MellonUniversity, Pittsburgh, PA, and the MassachussettsInstitute of Technology, Cambridge, from 1984 to
1987, where he held faculty positions, he returnedto the University of CaliforniaBerkeley where hewas a Professor until 2005. He is a currently a
Professor with the Department of Electrical Engineering, Stanford University,Stanford, CA, where he is currently with the Center for Integrated Systems.His research interests include microelectromechanical-system (MEMS) design,micro-/nanomachining processes, and parallel-assembly processes. The focusof his research has been on processes to fabricate integrated microsystems,which incorporate both silicon integrated circuits and MEMS. He has madecontributions to the design of MEMS accelerometers, gyroscopes, electrostaticactuators, and microresonators.
Dr. Howe was a corecipient of the 1998 IEEE Cledo Brunetti Award. He waselected to the National Academy of Engineering in 2005 for his contributions toMEMS processes, devices, and systems. He is the Cofounder of Silicon Clocks,Inc., a start-up company producing timing products.
H.-S. Philip Wong (F00) received the B.Sc. degree(Hons.) in electrical engineering from the Universityof Hong Kong, Kowloon, Hong Kong, in 1982, theM.S. degree in electrical engineering from the StateUniversity of New York, Stony Brook, in 1983,and the Ph.D. degree in electrical engineering fromLehigh University, Bethlehem, PA, in 1988.
He was with the IBM T. J. Watson Research
Center, Yorktown Heights, NY, in 1988. While hewas with IBM, he worked on charge-coupled deviceand CMOS image sensors, double-gate/multigate
MOSFET, device simulations for advanced/novel MOSFET, strained silicon,wafer bonding, ultrathin-body SOI, extremely short gate FET, germaniumMOSFET, carbon-nanotube FET, and phase-change memory. He held variouspositions from Research Staff Member to Manager and Senior Manager. Whilehe was a Senior Manager, he had the responsibility of shaping and executingIBMs strategy on nanoscale science and technology as well as exploratory sili-con devices and semiconductor technology. Since September 2004, he has beenwith Stanford University as Professor of Electrical Engineering. His researchinterests are in nanoscale science and technology, semiconductor technology,solid-state devices, and electronic imaging. He is interested in exploring newmaterials, novel fabrication techniques, and novel device concepts for futurenanoelectronic systems. Novel devices often require new concepts in circuitand system designs. His research also includes explorations into circuits andsystems that are device-driven. His current research covers a broad range of
topics including carbon nanotubes, semiconductor nanowires, self-assembly,exploratory logic devices, and novel memory devices.
Dr. Wong is a member of the Emerging Research Devices Working Group,International Technology Roadmap for Semiconductors. He served on the IEEEElectron Devices Society (EDS) as an elected AdCom member from 2001to 2006. He serves on the International Electron Devices Meeting (IEDM)Committee from 1998 to 2007, was the Technical Program Chair in 2006,and is the General Chair in 2007. He served on the International Solid-StateCircuits Conference (ISSCC) Program Committee from 1998 to 2004 and wasthe Chair of the Image Sensors, Displays, and MEMS Subcommittee from2003 to 2004. He was the Editor-in-Chief of the IEEE TRANSACTIONS ONNANOTECHNOLOGY in 20052006. He is a Distinguished Lecturer of the IEEEEDS and Solid-State Circuits Society. He has taught several short courses at theIEDM, ISSCC, Symposium on VLSI Technology, IEEE International Silicon-on-Insulator Conference, European Solid-State Device Research Conference,and The International Society for Optical Engineering conferences.