NOLTA, IEICE

Invited Paper

Device simulation of negative-capacitance

field-effect transistors with a uniaxial
ferroelectric gate insulator

Junichi Hattori 1 a) , Tsutomu Ikegami 1 , Koichi Fukuda 1 ,

Hiroyuki Ota 1 , Shinji Migita 1 , and Hidehiro Asai 1
1
National Institute of Advanced Industrial Science and Technology (AIST),
1-1-1 Umezono, Tsukuba 305-8568, Japan

a)
[email protected]

Received August 7, 2019; Revised November 18, 2019; Published April 1, 2020

Abstract: We model the behavior of uniaxial ferroelectrics and simulate planar negative-
capacitance (NC) field-effect transistors (FETs) having a gate insulating film made of a uni-
axial ferroelectric. The behavior of such NC FETs strongly depends on the direction of the
ferroelectric polarization axis. When the direction is away from being parallel to the ferro-
electric film to some extent, the ferroelectric polarization becomes larger than the paraelectric
polarization and the ferroelectric film begins to act as a negative capacitor. The NC FETs can
then be switched on and off more steeply than conventional metal–oxide–semiconductor FETs.
This NC effect is maximized at that moment and becomes weaker as the direction approaches
perpendicular to the ferroelectric film.
Key Words: negative capacitance, ferroelectric, field-effect transistor, steep-slope transistor,
device simulation, TCAD

1. Introduction
The demand for integrated circuits to reduce the power consumption has been growing for applications
in portable, wearable [1, 2], and implantable electronic devices [1, 3] and wireless sensor networks [4,
5]. An effective way is to lower the operating voltage, which requires constituent transistors to be
switched on and off steeply within a narrow voltage range, or to have a low subthreshold swing S.
Since the best value of S for conventional metal–oxide–semiconductor (MOS) field-effect transistors
(FETs) is limited to about 60 mV/decade at room temperature, the introduction of new physical
phenomena such as quantum tunneling of carriers [6] has been considered so far to overcome this
limitation. Negative capacitance (NC) of ferroelectric materials is also one such phenomenon [7] and
FETs harnessing this are referred to as NC FETs. NC FETs have similar structures to MOS FETs
and in most cases a ferroelectric thin film is inserted into their gate stack as a gate insulator. The
ferroelectric film reduces the voltage applied to its underlying structure at a low gate voltage VG , while
enhances it at a high VG . Owing to this VG modulation, the underlying structure experiences a larger
variation in the applied voltage than a given variation in VG ; consequently, the drain current steeply

145
Nonlinear Theory and Its Applications, IEICE, vol. 11, no. 2, pp. 145–156 c
IEICE 2020 DOI: 10.1587/nolta.11.145
increases or decreases. NC FETs have recently been studied widely and intensively and their steep
switching with an S of less than 60 mV/decade has already been reported by a number of research
groups [8–10]. Simulation will play an important role in the development of such studies by assisting
the understanding of the behavior, prediction of the performance, and exploration of the suitable
structure and material of NC FETs. It is therefore urgently needed to establish the simulation
method for NC FETs, especially compatible with device simulators in technology computer-aided
design (TCAD) systems, because those simulators are universally used for the simulation of standard
electronic devices including MOS FETs.
A typical simulation method is intended mainly for an NC FET having a metal film beneath a
ferroelectric film in the gate stack. In this method, such an NC FET is divided into two components:
the ferroelectric capacitor sandwiched between the internal metal film and gate electrode and the
remaining FET structure. They are simulated separately and then the results obtained for the FET
structure are corrected on the basis of the voltage across the ferroelectric capacitor [11–16]. Although
this method assumes that the polarization is uniform in the ferroelectric film, some recently proposed
methods take into account the nonuniformity of the polarization [17, 18]. Owing to the difference in
algorithms or core calculation techniques, it seems to be difficult to fully incorporate these methods
into device simulators.
We have considered the simulation methods of NC FETs using our original homemade device
simulator named Impulse TCAD [19, 20]. Owing to its high customizability that all device properties
and all equations governing them can be defined by users, we successfully realized the simulation of NC
FETs and revealed their interesting features such as reverse drain-induced barrier lowering, negative
drain conductance, and fringing field effect [21–23]. Recently, we proposed a new simulation method
taking into account nonuniform polarization fields in ferroelectrics [24]. It has complete applicability
not only to Impulse TCAD but also to other standard device simulators, and high extensibility as well,
which enables device simulators, for example, to take into account a nature of ferroelectrics that the
polarization vectors tend to have the same direction and magnitude. In this method, the polarization
vectors are assumed to potentially point in any directions; however, some materials, including hafnia of
recent great interest, exhibit uniaxial ferroelectricity [25, 26], that is, the polarization vectors can only
point in one direction or the opposite direction in such materials. In this paper, we model uniaxial
ferroelectrics in a form that can be used in standard TCAD device simulators and simulate the
behavior of planar NC FETs having a gate insulating film made of a uniaxial ferroelectric on Impulse
TCAD. Then we investigate the relationship between the polarization axis and the performance
of such NC FETs and find that their switching becomes steep only when the angle between the
polarization axis and the ferroelectric film is within a certain range.

2. Simulation method
2.1 Governing equations for uniaxial ferroelectrics
In the simulation of NC FETs, the polarization field in the ferroelectric film is a key target. We
assumed that it can be written as
P FE = P d + P p . (1)

The first term on the right side is the component representing the uniaxial ferroelectricity, which
is parallel to the unit vector d and whose magnitude is |P |. Also, the second term represents the
paraelectric component and it was assumed to be simply proportional to the electric field E as

P p = χp ε0 E, (2)

where ε0 is the permittivity of vacuum. The polarization field P FE interacts with E and affects the
behavior of NC FETs. Introducing the electric potential ψ instead of directly dealing with E, which
is given by −∇ψ, we can write the energy functional of the ferroelectric film as [17, 20]

δ 1 ε0

2
U [ψ, P ] = αP 2 + βP 4 + γP 6 + |∇P |2 + |P p |2 + (ρ − ∇ · P FE )ψ − |−∇ψ| dV, (3)
2 2χp ε0 2

146
 Fig. 1. Schematic view of the CV of a node n in a computational mesh, each
of whose faces is shared with the CV of an adjacent node and intersects with
the edge between the two nodes at a right angle at the midpoint of the edge.

where α, β, γ, and δ are material parameters and ρ is the charge density. When the ferroelectric film
is in a stable state, this energy is expected not to change with infinitesimal variations in ψ and P .
The conditions for ψ and P to meet this expectation leads to the following two equations,

∇ · (εp E + P d) − ρ = 0 (4)

and
2αP + 4βP 3 + 6γP 5 − δ∇ · (∇P ) − d · E = 0, (5)
where εp is equal to (1 + χp )ε0 . The first equation is known as Poisson’s equation, and the second
equation is referred to as the Landau–Khalatnikov (LK) equation in this paper because it is also
derived from the LK model [7, 24, 27]. These two equations govern ψ and P in the ferroelectric film
in a steady state.

2.2 Solution of the governing equations in device simulation

The first step in numerical computation is generally to discretize the equation to be solved. To do so,
a finite volume method is often used in device simulation. In that method, a target device is divided
into polyhedrons including one node of a computational mesh, which polyhedrons are referred to as
control volumes (CVs). In this paper, we assumed that each face of a CV is shared with an adjacent
CV and intersects with the edge between the nodes of the two CVs at a right angle at the midpoint
unless the face is a part of the device boundary, as shown in Fig. 1. Integrating an equation over a
CV after describing the parameters and variables in the CV by their values at the associated node
and adjacent nodes, we can derive the discretized form of the equation.
The discretized form of Poisson’s equation of Eq. (4) for a node n can be written as

(εp Enm + Pnm )Snm − ρn Vn = 0, (6)
m

where the summation is taken over all adjacent nodes m, Vn is the volume of n’s CV, and Snm is
the area of the CV’s face shared with m’s CV. Also, Enm and Pnm are the electric and ferroelectric
polarization fields perpendicular to and on Snm and they were approximated as Enm = −(ψm −
ψn )/ℓnm and Pnm = (Pn + Pm )d · r nm /2, where r nm is the edge vector from n to m and ℓnm is its
length. On the other hand, the LK equation of Eq. (5) can be discretized as
 Pm − Pn
−δ Snm + (2αPn + 4βPn3 + 6γPn5 − d · E n )Vn = 0. (7)
m
ℓnm

Here, E n is the electric field at n and its component in each direction was approximated as

Enm (unm )i |(unm )i |Snm
(E n )i = m  , (8)
m |(unm )i |Snm

147
 Fig. 2. Schematic views of planar NC FETs (a) with and (b) without an
internal metal film. The regions labeled “G”, “S”, and “D” represent the
gate, source, and drain electrodes, respectively. The thicknesses of the film
components are shown in parentheses in their labels. The thickness of the gate
oxide film TOX is 1 nm in (a) and 5 nm in (b).

where i denotes the x, y, or z-direction and unm is the unit vector of r nm .

In device simulation, since the discretized equations are solved by the Newton–Raphson method,
the Jacobians of the left sides of Eqs. (6) and (7) must be derived. Impulse TCAD, however, does
this automatically and partially automates even the discretization of equations [20]. When solving
the following equation with Impulse TCAD,

∇ · f + s = 0, (9)

all a user has to do is to specify the discretized general forms of the inward flux term f and the source
term s together with variables in the run script. Then Impulse TCAD discretizes this equation as

fnm Snm + sn Vn = 0 (10)
m

and derives the Jacobian of this left side automatically. For example, if fnm is specified as −(εp Enm +
Pnm ) and sn as ρn , Impulse TCAD constructs Eq. (6).

3. Results and discussion

In this paper, we considered simple planar NC FETs as shown in Fig. 2 and assumed that the unit
vector in the positive ferroelectric polarization direction d is perpendicular to the gate width direction
and has the form
d = (cos θ, 0, sin θ), (11)
where θ is the angle between d and the ferroelectric film. We simulated NC FETs with various values
of this ferroelectric polarization angle θ and investigated the dependence of the behavior, especially
the subthreshold swing, on θ.

3.1 NC FETs with an MFMOS structure

First, we simulated the NC FET depicted in Fig. 2(a), which has a metal–ferroelectric–metal–oxide–
semiconductor (MFMOS) structure, that is, whose gate stack consists of four layers, namely, a gate
oxide film, an internal metal film, a ferroelectric film, and a gate electrode. This NC FET is of n-type
and the acceptor concentration in the body are 6 × 1017 /cm3 . The source/drain region has a depth
of 50 nm and is formed by donors with a concentration of 1 × 1020 /cm3 from one edge of the body to
a position just below the outer edge of the nearest gate sidewall. The donor concentration decreases
with increasing distance from the source/drain region according to a Gaussian function and becomes
equal to the acceptor concentration at a position just below the nearest gate edge. In the simulation,
the gate length LG was set to 100 nm, and the gate oxide film thickness TOX , internal metal film
thickness, and ferroelectric film thickness TFE were set to 1, 4, and 10 nm. Also, the charge density
ρ in the ferroelectric film was set to zero, and the material parameters listed in Table I were used.
The values of ferroelectric parameters α, β, and γ were chosen so that the remanent polarization Pr

148
 Table I. Parameters used in the simulation.
Component Material Parameter Value
Body Si Permittivity ε 11.7ε0
Gate oxide film SiO2 ε 3.9ε0 TOX /(1 nm)
Gate sidewall Si3 N4 ε 7.8ε0
Ferroelectric film Hafnia α −4.64 × 108 m/F
β −1.29 × 108 m5 /FC2
γ 9.88 × 1010 m9 /FC4
δ 1.56 × 10−8 m3 /F
εp 36ε0

Fig. 3. (a) Relationship of the ferroelectric polarization P d and the elec-

tric field E in the considered ferroelectric material plotted along the positive
ferroelectric polarization direction d. NC states lie on the line A–B. (b) Con-
tribution of P d to the permittivity divided by the permittivity of vacuum ε0
and calculated for the NC states.

is 20 µC/cm2 , the coercive electric field Ec is 1 MV/cm, and the ferroelectric polarization at Ec is
0.67Pr , as shown in Fig. 3(a). On the line A–B in the figure, such a ferroelectric is in an NC state and
the contribution of the ferroelectric polarization to the permittivity becomes negative, as shown in
Fig. 3(b). The ferroelectric polarization angle θ was set to be perpendicular to the ferroelectric film,
or 90◦ , the paraelectric permittivity εp was set to 36ε0 , and the value of the remaining ferroelectric
parameter δ was determined so that the domain wall thickness Tdw is 10 nm, which is given by [28]

6δ
Tdw = . (12)
−(2α + βPr2 )

Figure 4(a) shows the drain current ID vs gate-to-source voltage VGS characteristics of the NC
FET and the corresponding MOS FET at a drain-to-source voltage VDS of 0.05 V. That MOS FET
is obtained by merging the ferroelectric film and internal metal film into the gate electrode in the NC
FET. Comparing the two ID –VGS characteristics, we can see that the ID of the NC FET varies more
steeply than that of the MOS FET. This is because, in the NC FET, the gate voltage VG is modulated
by the NC of the ferroelectric film. This VG modulation is shown in Fig. 4(b) as the voltage across
the ferroelectric film ∆VG , to be precise, as the electric potential of the lower surface relative to that
of the upper surface. The MOS structure under the ferroelectric film experiences VG + ∆VG as an
effective gate voltage and this voltage determines the ID of the NC FET. The ferroelectric film forms
a series capacitor together with the gate oxide film and the depletion layer under the semiconductor
surface and the difference between VG and the flat band voltage is applied to this series capacitor.

149
 Fig. 4. (a) Drain current ID vs gate-to-source voltage VGS characteristics of
(solid line) the NC FET with an MFMOS structure shown in Fig. 2(a) and
(dotted line) the corresponding MOS FET at a drain-to-source voltage VDS of
0.05 V. (b) Voltage across the ferroelectric film ∆VG in the NC FET. In the
simulation, the ferroelectric polarization angle θ was 90◦ , that is, perpendicular
to the ferroelectric film.

Fig. 5. (a) ID and (b) ∆VG in NC FETs with an MFMOS structure calcu-
lated for θ’s of 0◦ , 30◦ , and 60◦ and a VDS of 0.05 V as a function of VGS .

Thus, as VG increases, ∆VG increases. Because the underlying MOS structure is subjected to this
∆VG variation in addition to the VG variation, ID varies with VG more steeply in the NC FET.
Next, we simulated NC FETs with the same structure as above for different ferroelectric polarization
angles θ’s. Figures 5(a) and 5(b) show ID and ∆VG in the NC FETs calculated for θ’s of 0◦ , 30◦ , and
60◦ and a VDS of 0.05 V as a function of VGS , and they indicate that the behavior of the NC FETs
is strongly dependent on θ. As can be seen from Fig. 5(b), the capacitance of the ferroelectric film is

150
 Fig. 6. (a) Subthreshold swing S of NC FETs with an MFMOS structure
at a VDS of 0.05 V as a function of θ. The dotted line represents the S of the
corresponding MOS FETs, whose value is 65.1 mV/decade. (b) Capacitance of
the ferroelectric film CFE multiplied by TFE /ε0 , where TFE is the thickness of
the ferroelectric film. The gray areas represent the region of values of θ where
the evaluation of S failed.

no longer negative when θ is 0◦ or 30◦ .

Figure 6(a) shows the subthreshold swing S of the NC FETs as a function of θ. We calculated S
by dividing the VGS width required to increase ID from Ith /103 to Ith by three, where Ith is the ID
when VGS is equal to the threshold voltage Vth , and is defined as Ith = (WG /LG )10−7 A with WG
being the gate width. In general, the S of planar MOS FETs is given by [29]
 
kB T Cdep
S= ln 10 1 + , (13)
q Cins
where kB is the Boltzmann constant, T is the operating temperature, q is the elementary charge, and
Cdep is the capacitance of the depletion layer. Also, Cins is the capacitance of the gate insulator and
is equal to that of the gate oxide film COX in the considered MOS FETs. On the other hand, in the
NC FETs, Cins is the series capacitance of the gate oxide film and ferroelectric film, and thus Eq. (13)
can be rewritten as  
kB T Cdep Cdep
SMFMOS = ln 10 1 + + . (14)
q COX CFE
This equation indicates that when the capacitance of the ferroelectric film CFE has a negative value,
the NC FETs have a better S than the MOS FETs. The capacitance CFE can be approximated as

sin2 θ 1 εp
CFE ≃ + , (15)
2α + 12βP 2 + 30γP 4 TFE TFE
where the first term represents the contribution of the ferroelectric polarization Cf and the second
term represents the contribution of the paraelectric polarization and electric field Cp . Although Cp
is always positive, Cf becomes negative when the ferroelectric film is in an NC state. Figure 6(b)
shows CFE calculated as a function of θ with an assumption that both the ferroelectric polarization
P d and the electric field E are zero. When VGS is equal to Vth , the voltage across the ferroelectric
film is considered to be lower than 0.4 V for most values of θ, as can be inferred from Figs. 4 and
5. This means that, in the subthreshold region, the vertical electric field in the ferroelectric film Ez
(≃ ∆VG /TFE ) rarely exceeds 0.4 MV/cm. The electric field parallel to the ferroelectric polarization

151
axis E (≃ Ez sin θ) will certainly be weaker. Figure 3(b) indicates that Cf hardly changes with such
E since the contribution of P d to the permittivity plotted as a function of E in that figure is equal
to Cf TFE / sin2 θ. Also, Cp is constant for any electric field. Therefore, it can safely be assumed that
CFE takes the value shown in Fig. 6(b) in the subthreshold region. Then it is immediately apparent
from Eq. (14) that when CFE varies with θ as shown in Fig. 6(b), S varies as shown in Fig. 6(a).
When the ferroelectric polarization axis is close to parallel to the ferroelectric film, that is, when
θ is close to 0◦ or 180◦ , Cf cannot cancel Cp and CFE has a positive value; consequently, the NC
FETs have a worse S than the MOS FETs. On the other hand, when the polarization axis is close to
perpendicular to the ferroelectric film, that is, when θ is close to 90◦ , Cf completely cancels Cp and
CFE becomes negative; consequently, the S of the NC FETs becomes better than that of the MOS
FETs. The further θ is from 90◦ , the closer the negative CFE is to zero, and as can be seen from
Eq. (14), the better S becomes in the NC FETs. Since CFE is a series component of the gate-to-source
capacitance CGS , if the negative CFE is very close to zero, CGS will become negative. A negative CGS
means that the number of carriers increases with decreasing VGS , that is, the lower VGS becomes, the
larger ID flows. Although this will never occur in reality, it is not predictable how real NC FETs
behave for a negative CGS . In Fig. 6(b), CFE is zero when θ is about 33◦ or 147◦ . When θ was close
to such an angle, the voltage across the ferroelectric film became very large. Then the simulation
became unstable and we could not obtain ID in the range of VGS required to calculate S. The range
of values of θ that caused such a situation are shown in gray in Fig. 6.

3.2 NC FETs with an MFOS structure

We also simulated the NC FETs shown in Fig. 2(b) with a metal–ferroelectric–oxide–semiconductor
(MFOS) structure, i.e., without the internal metal film of the MFMOS structure. In our previous
study [23], it was found that the performance of NC FETs is affected by the electrical coupling between
the gate stack and the source/drain region through the gate sidewall. To avoid the change of such
coupling, especially between a side surface of the ferroelectric film and the source/drain region, we
obtained the MFOS structure corresponding to the above-mentioned MFMOS structure by merging
the internal metal film into the gate oxide film in the MFMOS structure. Although the thickness of
the gate oxide film in the MFOS structure then became 5 nm, in the simulation, the permittivity was
increased by five times so that the capacitance did not change from before. Figure 7(a) shows the
subthreshold swing S of such NC FETs at VDS of 0.05 V as a function of the ferroelectric polarization
angle θ. Comparing this figure with Fig. 6(a) with attention paid to the region of values of θ where
the S of an NC FET is better than that of the corresponding MOS FET, we can see that such an
improvement in S is reduced by the absence of the internal metal film. This is due to an increase in
the magnitude of the ferroelectric film capacitance CFE . In the presence of the internal metal film, the
electric field applied to the ferroelectric film becomes almost uniform; however, in the absence of the
internal metal film, the electric field varies in the direction and magnitude depending on the position.
This non-uniformity of the electric field makes the simulation difficult to converge; consequently, the
range of values of θ where S could not be evaluated was extended. Since, as shown in Fig. 3(b), the
contribution εf of the ferroelectric polarization to the permittivity depends on the electric field, CFE
also differs depending on the position. Owing to a high electric potential in the source and drain
regions, a relatively high electric field is applied to both edges of the ferroelectric film. In the MOS
FET corresponding to the NC FET shown in Fig. 2(b), the electric field at both upper corners of
the gate oxide film is higher than 0.5 MV/cm at a VGS of 0 V. Such a relatively high electric field is
applied to both lower corners of the ferroelectric film in the NC FET, although it is modulated by
the ferroelectric film. Then, as seen in Fig. 3(b), the magnitude of εf increases and the magnitude
of CFE also increases. As a result, the VG modulation by the ferroelectric film becomes weaker near
its both edges and the improvement in S becomes smaller in the NC FET without the internal metal
film than in that with the internal metal film.
So far, VDS has been as low as 0.05 V and the electric field applied to the ferroelectric film has
been approximately symmetrical about the center plane of the gate. As VDS increases, however, this
symmetry gradually disappears, as shown in Fig. 8. Then the symmetry of the behavior of NC FETs

152
 Fig. 7. S of NC FETs with an MFOS structure calculated for VDS ’s of (a)
0.05 V and (b) 1 V as a function of θ. The values of S of the corresponding
MOS FETs, shown by the dotted lines, are 65.1 and 64.9 mV/decade.

Fig. 8. (a) Electric potential distribution in the MOS FETs corresponding

to NC FETs with an MFOS structure at a VGS of 0.4 V and a VDS of 1 V.
The black lines represent equipotential lines and the difference in the potential
between two neighboring lines is 0.05 V. (b and c) Polarization fields in the
ferroelectric films in the NC FETs with θ’s of 45◦ and 135◦ . The electric
potential distributions in the NC FETs are also shown by the equipotential
lines.

with respect to the ferroelectric polarization angle θ measured from the normal of the ferroelectric
film, 90◦ , such as the subthreshold swing S shown in Figs. 6(a) and 7(a) may also disappear. As can
be inferred from the electric field near both upper corner of the gate oxide film in MOS FETs shown

153
in Fig. 8(a), the electric field entering the ferroelectric film in NC FETs has an angle of about 45◦ at
the source-side corner and has an angle of 135◦ at the drain-side corner. For this electric field, when
θ is 45◦ , the last term on the left side of Eq. (5), d · E, becomes about |E| at the source-side corner
and becomes about zero at the drain-side corner. On the other hand, when θ is 135◦ , the opposite
occurs. Thus, when VDS is high to some extent and the electric field entering each lower corner of
the ferroelectric film has a different magnitude, an NC FET with a θ of 45◦ probably shows different
behavior than that with a θ of 135◦ . If one set θ to other than 90◦ to bring out better NC effects,
this kind of behavioral asymmetry may have negative effects on the performance of some circuits
where the source and drain of transistors are interchanged during operation. This is because, for the
ferroelectric film, the interchange of source and drain means that the ferroelectric polarization angle
is changed from θ to 180◦ − θ. Finally, we simulated the NC FETs without an internal metal film for
a high VDS of 1 V. Figure 7(b) shows the dependence of their S on θ and its symmetry still seems to
be preserved. Comparing the electric potential distributions in the NC FETs with θ’s of 45◦ and 135◦
respectively shown in Figs. 8(b) and 8(c), we can see that they are quite different near the drain-side
edge of the ferroelectric film. This difference causes a difference in the electric potential distribution
near the gate-drain edge of the semiconductor body, although it is smaller. In the other regions of
the semiconductor body except for the region very close to the gate-source edge, however, such a
difference is not clearly observed between the two NC FETs. They thus have a similar S. This holds
true for the other pairs of NC FETs.

4. Conclusions
We developed a method for standard device simulators to simulate the behavior of uniaxial ferro-
electrics and incorporated it into our original homemade device simulator named Impulse TCAD,
thereby realizing the simulation of planar NC FETs with a gate insulating film made of a uniaxial
ferroelectric. The behavior of such NC FETs varies greatly depending on the direction of the ferro-
electric polarization axis regardless of the presence or absence of an internal metal film. When the
angle between the direction and the ferroelectric film becomes large to some extent, the ferroelectric
polarization exceeds the paraelectric polarization and the ferroelectric film acts as a negative capac-
itor; consequently, the NC FETs can be switched on and off more steeply than the corresponding
MOS FETs. This NC effect is maximized at that threshold angle and becomes weaker as the angle
approaches perpendicular to the ferroelectric film. Choosing an angle close to the threshold angle as
the ferroelectric polarization angle, however, should be avoided, because the NC of the ferroelectric
film can potentially make the gate-to-source capacitance negative and then destabilize the behavior of
the NC FETs in some operating conditions. When the angle is not perpendicular to the ferroelectric
film, the behavior of NC FETs without an internal metal film may change by an interchange of the
source and drain. This kind of asymmetry, however, was not observed in the NC FETs with a gate
length of 100 nm and it might appear in the NC FETs having a shorter gate.

References
[1] C. Pang, C. Lee, and K.-Y. Suh, “Recent advances in flexible sensors for wearable and im-
plantable devices,” J. Appl. Polym. Sci., vol. 130, no. 3, pp. 1429–1441, June 2013.
[2] Y. Khan, A.E. Ostfeld, C.M. Lochner, A. Pierre, and A.C. Arias, “Monitoring of vital signs with
flexible and wearable medical devices,” Adv. Mater., vol. 28, no. 22, pp. 4373–4395, February
2016.
[3] K. Bazaka and M.V. Jacob, “Implantable devices: issues and challenges,” Electronics, vol. 2,
no. 1, pp. 1–34, March 2013.
[4] A. Darwish and A.E. Hassanien, “Wearable and implantable wireless sensor network solutions
for healthcare monitoring,” Sensors, vol. 11, no. 6, pp. 5561–5595, June 2011.
[5] P. Rawat, K.D. Singh, H. Chaouchi, and J.M. Bonnin, “Wireless sensor networks: recent devel-
opments and potential synergies,” J. Supercomput., vol. 68, no. 1, pp. 1–48, April 2014.
[6] W.Y. Choi, B.-G. Park, J.D. Lee, and T.-J.K. Liu, IEEE Electron Device Lett., vol. 28, no. 8,
pp. 743–745, August 2007.

154
 [7] S. Salahuddin and S. Datta, “Use of negative capacitance to provide voltage amplification for
low power nanoscale devices,” Nano Lett., vol. 8, no. 2, pp. 405–410, December 2008.
[8] Z. Krivokapic, U. Rana, R. Galatage, A. Razavieh, A. Aziz, J. Liu, J. Shi, H.J. Kim, R. Sporer,
C. Serrao, A. Busquet, P. Polakowski, J. Müller, W. Kleemeier, A. Jacob, D. Brown, A. Knorr,
R. Carter, and S. Banna, “14nm ferroelectric FinFET technology with steep subthreshold slope
for ultra low power applications,” Proc. IEEE Int. Electron Device Meeting (IEDM), 15.1,
pp. 357–360, December 2017.
[9] J. Li, J. Zhou, G. Han, Y. Liu, Y. Peng, J. Zhang, Q.-Q. Sun, D.W. Zhang, and Y. Hao, “Neg-
ative capacitance Ge PFETs for performance improvement: impact of thickness of HfZrOx ,”
IEEE Trans. Electron Devices, vol. 65, no. 3, pp. 1217–1222, March 2018.
[10] T. Srimani, G. Hills, M.D. Bishop, U. Radhakrishna, A. Zubair, R.S. Park, Y. Stein, T. Palacios,
D. Antoniadis, and M.M. Shulaker, “Negative capacitance carbon nanotube FETs,” IEEE Elec-
tron Device Lett., vol. 39, no. 2, pp. 304–307, February 2018.
[11] A.I. Khan, C.W. Yeung, C. Hu, and S. Salahuddin, “Ferroelectric negative capacitance MOS-
FET: capacitance tuning & antiferroelectric operation,” Proc. IEEE Int. Electron Device Meet-
ing (IEDM), 11.3, pp. 255–258, December 2011.
[12] H. Ota, S. Migita, J. Hattori, K. Fukuda, and A. Toriumi, “Material and device engineering in
fully depleted silicon-on-insulator transistors to realize a steep subthreshold swing using negative
capacitance,” Jpn. J. Appl. Phys., vol. 55, no. 8S2, 08PD01, August 2016.
[13] A. Saeidi, F. Jazaeri, I. Stolichnov, and A.M. Ionescu, “Double-gate negative-capacitance MOS-
FET with PZT gate-stack on ultra thin body SOI: an experimentally calibrated simulation study
of device performance,” IEEE Trans. Electron Devices, vol. 63, no. 12, pp. 4678–4684, December
2016.
[14] H. Ota, S. Migita, J. Hattori, K. Fukuda, and A. Toriumi, “Structural advantages of silicon-on-
insulator FETs over FinFETs in steep subthreshold-swing operation in ferroelectric-gate FETs,”
Jpn. J. Appl. Phys., vol. 56, no. 4S, 04CD10, April 2017.
[15] J. Seo, J. Lee, and M. Shin, “Analysis of drain-induced barrier rising in short-channel negative-
capacitance FETs and its applications,” IEEE Trans. Electron Devices, vol. 64, no. 4, pp. 1793–
1798, April 2017.
[16] H. Ota, J. Hattori, H. Asai, T. Ikegami, K. Fukuda, S. Migita, and A. Toriumi, “Design of
steep-slope negative-capacitance FinFETs for dense integration: importance of appropriate fer-
roelectric capacitance and short-channel effects,” Jpn. J. Appl. Phys., vol. 57, no. 4S, 04FD03,
April 2018.
[17] P. Lenarczyk and M. Luisier, “Physical modeling of ferroelectric field-effect transistors in the
negative capacitance regime,” Proc. Int. Conf. Simulation of Semiconductor Processes and De-
vices (SISPAD 2016), pp. 311–314, September 2016.
[18] A.K. Saha, P. Sharma, I. Dabo, S. Datta, and S.K. Gupta, “Ferroelectric transistor model based
on self-consistent solution of 2D Poisson’s, non-equilibrium Green’s function and multi-domain
Landau Khalatnikov equations,” Proc. IEEE Int. Electron Device Meeting (IEDM 2017), 13.5,
pp. 326–329, December 2017.
[19] National Institute of Advanced Industrial Science and Technology (AIST), “Impulse TCAD,”
https://unit.aist.go.jp/neri/en/ImpulseTCAD/.
[20] T. Ikegami, K. Fukuda, J. Hattori, H. Asai, and H. Ota, “TCAD Device simulator for exotic
materials and its application to negative capacitance FET,” J. Comput. Electron., vol. 18, no. 2,
pp. 534–542, June 2019.
[21] H. Ota, T. Ikegami, J. Hattori, K. Fukuda, S. Migita, and A. Toriumi, “Fully coupled 3-D
device simulation of negative capacitance FinFETs for sub 10 nm integration,” Proc. IEEE Int.
Electron Device Meeting (IEDM 2016), 12.4, pp. 318–321, December 2016.
[22] H. Ota, K. Fukuda, T. Ikegami, J. Hattori, H. Asai, S. Migita, and A. Toriumi, “Perspective
of negative capacitance FinFETs investigated by transient TCAD simulation,” Proc. IEEE Int.
Electron Device Meeting (IEDM 2017), 15.2, pp. 361–364, December 2017.
[23] J. Hattori, K. Fukuda, T. Ikegami, H. Ota, S. Migita, H. Asai, and A. Toriumi, “Fringing field

155
 effects in negative capacitance field-effect transistors with a ferroelectric gate insulator,” Jpn.
J. Appl. Phys., vol. 57, no. 4S, 04FD07, April 2018.
[24] J. Hattori, T. Ikegami, K. Fukuda, H. Ota, S. Migita, and H. Asai, “Device simulation of
negative-capacitance field-effect transistors with a ferroelectric gate insulator,” Proc. Int. Conf.
on Simulation of Semiconductor Processes and Devices (SISPAD 2018), pp. 214–219, September
2018.
[25] T.S. Böscke, J. Müller, D. Bräuhaus, U. Schröder, and U. Böttger, “Ferroelectricity in hafnium
oxide thin films,” Appl. Phys. Lett., vol. 99, no. 10, 102903, September 2011.
[26] T.D. Huan, V. Sharma, G.A. Rossetti, Jr., and R. Ramprasad, “Pathways towards ferroelec-
tricity in hafnia,” Phys. Rev. B, vol. 90, no. 6, 064111, August 2014.
[27] L.D. Landau and I.M. Khalatnikov, “On the anomalous absorption of sound near a second order
phase transition point,” Dokl. Akad. Nauk SSSR, vol. 96, pp. 469–472, 1954.
[28] P. Marton, I. Rychetsky, and J. Hlinka, “Domain walls of ferroelectric BaTiO3 within the
Ginzburg-Landau-Devonshire phenomenological model,“ Phys. Rev. B, vol. 81, no. 14, 144125,
April 2010.
[29] S.M. Sze and K.K. Ng, Physics of Semiconductor Devices, 3rd ed., Wiley, Hoboken, 2007, p. 315.

156