Mfmis VS Mfis

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 65, NO.

3, MARCH 2018 867

Physical Insights on Negative Capacitance


Transistors in Nonhysteresis and Hysteresis
Regimes: MFMIS Versus MFIS Structures
Girish Pahwa , Student Member, IEEE , Tapas Dutta, Member, IEEE , Amit Agarwal,
and Yogesh Singh Chauhan , Senior Member, IEEE

Abstract — We present a comprehensive comparison of


the two different types of ferroelectric negative capacitance
FET (NCFET) structures: metal-ferroelectric-metal-
insulator-semiconductor (MFMIS) and metal-ferroelectric-
insulator-semiconductor (MFIS). A new segmentation
approach is proposed to simulate MFIS NCFET, which
correctly takes care of the nonuniformity in potential
and horizontal electric field at the ferroelectric–oxide
interface. We show that MFMIS NCFET provides a higher
ON-current than MFIS NCFET except for the ferroelectrics
with very low remnant polarization (Pr ) in the high operating
voltage regime. We find that this behavior is caused by a
reduction or enhancement of the longitudinal electric field in
the channel of MFIS structure depending upon Pr of the fer-
roelectric and the operating voltage. Moreover, there exists
an optimum Pr which provides maximum ON-current for both
the devices. We also find that MFIS NCFET is more prone
to hysteresis and starts showing a hysteretic behavior at a Fig. 1. Different NCFET structures. (a) MFMIS. (b) MFIS. Circuit
lower ferroelectric thickness compared with MFMIS NCFET. equivalents of (c) MFMIS and (d) MFIS. Each unit in the MFIS equivalent
network represents an MFMIS structure. Vint denotes internal voltage
Index Terms — Ferroelectric, metal-ferroelectric- at the ferroelectric–oxide interface. For MFIS, Vint varies along the
insulator-semiconductor (MFIS), metal-ferroelectric-metal- longitudinal direction when a nonzero drain bias is applied and hence,
insulator-semiconductor (MFMIS), negative capacitance it has been modeled by the arrangement shown.
(NC), NCFET.
While in the other structure shown in Fig. 1(b), the ferro-
I. I NTRODUCTION electric is in a direct contact with the oxide layer (a metal-
ferroelectric-insulator-semiconductor (MFIS) FET [9]–[12]).
F ERROELECTRIC-BASED negative capacitance (NC)
transistors, which can enable sub-60-mV/dec switch-
ing [1], are being actively researched for ultralow power
Recently, Duarte et al. [13] and we [14] have reported that
MFIS provides a higher ON-state current compared with
applications [2], [3]. The experimental demonstrations of MFMIS, however, we find this not to be true for all the
NCFET have mainly employed two different types of NCFET ferroelectric materials. Furthermore, a detailed physical inter-
structures, as shown in Fig. 1. One of the structures involves pretation of the behavior of the two structures and a rel-
a metal layer (floating gate) between the ferroelectric and ative performance analysis is still missing in the literature.
the internal gate oxide (a metal-ferroelectric-metal-insulator- In this paper, we perform an extensive one-to-one comparison
semiconductor (MFMIS) FET [4]–[8]), as shown in Fig. 1(a). between MFMIS and MFIS structures of NCFET using a
compact modeling approach. We show that the relative value of
Manuscript received July 17, 2017; revised November 5, 2017; ON -current is a strong function of remnant polarization of the
accepted January 13, 2018. Date of publication January 31, 2018; date
of current version February 22, 2018. This work was supported in part ferroelectric when comparing MFMIS and MFIS structures.
by the Department of Science and Technology, in part by Science and We also investigate and explain the hysteresis behavior of the
Engineering Research Board, and in part by Semiconductor Research two structures highlighting the differences in their underlying
Corporation. The review of this paper was arranged by Editor D. Esseni.
(Corresponding author: Girish Pahwa.) hysteresis mechanisms for the first time.
G. Pahwa, T. Dutta, and Y. S. Chauhan are with the Nanolab,
Department of Electrical Engineering, IIT Kanpur, Kanpur 208016, India II. S IMULATION M ETHODOLOGY
(e-mail: girish@iitk.ac.in; chauhan@iitk.ac.in).
A. Agarwal is with the Department of Physics, IIT Kanpur, We use a compact modeling approach to simulate the two
Kanpur 208016, India. types of NCFET structures. For MFMIS, due to the presence
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org. of a metal layer, the ferroelectric capacitor and the baseline
Digital Object Identifier 10.1109/TED.2018.2794499 MOSFET can be considered as two different circuit elements

0018-9383 © 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

Authorized licensed use limited to: AJOU UNIVERSITY. Downloaded on March 23,2023 at 06:36:32 UTC from IEEE Xplore. Restrictions apply.
868 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 65, NO. 3, MARCH 2018

simply connected by a wire, as shown in Fig. 1(c). The MFMIS neglected second-order effects, such as mobility degradation,
structure can thus be simulated by a lumped model approach short channel effects, and so on, in order to compare the
by self-consistently solving the Landau–Devonshire (L–D) two structures only on the basis of the NC effect. Recently,
model of the ferroelectric with the MOSFET physics [13], it has been reported that the NC state in MFMIS NCFET
[15]–[19]. L–D equation relates the voltage drop across the may become unstable in practical devices in the presence
ferroelectric, Vfe and the gate charge density, Q G as Vfe = of leakage current [24] or domain formation [25], which
2α Q G + 4β Q G 3 , where α and β are the material-dependent need to be duly addressed [26]. We have assumed the
parameters. α and β can also be related to the ferroelectric devices to be leakage free in this paper. We have used the
material properties, the remnant polarization (Pr ), and the L–D equation parameters for a CMOS process compatible
coercive field (E c ) [20]. In this paper, we solve the L–D Gd:HfO2 ferroelectric in the simulations [27]. Pr and E c
equation with the Berkeley short-channel IGFET model values calculated from the L–D parameters are found to be
bulk (BSIM-BULK) (formerly BSIM6) MOSFET model [21], 12.13 μC/cm2 and ≈1 MV/cm, respectively [20]. We also
[22] in Verilog-A to simulate the MFMIS NCFET. discuss results with other possible material parameters of
However, for the MFIS structure, the above-mentioned HfO2 -based ferroelectrics later in Section III-C.
approach cannot be used, as the voltage at the ferroelectric–
oxide interface (internal voltage, Vint ) varies in the longitudinal III. R ESULTS AND D ISCUSSION
(source to drain) direction when a nonzero drain bias is A. Drain Current
applied. This structure has been modeled either by adopting a
Fig. 2(a)–(d) compares the I D –VG and I D –V D
full analytical approach [14], [23] or by a distributed modeling
characteristics and the corresponding derivatives for
approach [13] to calculate the channel current. In this paper,
MFMIS and MFIS NCFETs. It is evident that the MFMIS
we apply a segmentation approach to model MFIS NCFET
structure shows higher ON-current, transconductance (gm ),
as a series network of N identical MFMIS units, as shown
and output conductance (gds) compared with MFIS for
in Fig. 1(d), where each unit is modeled using L–D and BSIM-
Pr = 12.13 μC/cm2 . Note that both the structures show a
BULK as explained earlier. Ideally, N should be very large, but
lower threshold voltage and a larger OFF-current compared
our simulation results show that the electrical characteristics
with the reference MOSFET (tfe = 0) due to the internal
start to saturate when N becomes as large as 100 except for
voltage amplification caused by NC effect, which is present
the hysteretic operation regimes. Thus, for simplicity and to
even for VG = 0 as the flat band voltage (VFB ) = 0. Fig. 2(e)
avoid large simulation time, we have assumed N = 100 for the
shows the variation of the ON-current with the thickness of
nonhysteretic cases and N = 300 for the hysteretic cases. All
the ferroelectric layer (tfe ) at iso-IOFF of 1 nA/μm. Iso-IOFF
the simulations have been carried out using the commercial
is achieved using a VFB tuning. As shown in the figure,
Cadence Spectre SPICE simulator. For the MFIS device sim-
ON -current of both the structures increases with tfe with
ulation, SPICE satisfies conditions of channel potential con-
MFMIS excelling MFIS for all the tfe values. These results
tinuity in between the subtransistors and the channel current
are in quite contrast with previously reported results [13], [14]
continuity in all the subtransistors. Simulation using the above-
which use different ferroelectric materials, as explained later.
mentioned method provides node voltages and node currents
For the low values of drain voltage, both the structures show
associated with each node and other electrical quantities of
similar current values as evident in Fig. 2(b). This is expected,
interest used in this paper as follows. For any nth subtransistor
since, for low V D values the internal voltage (Vint in Fig. 1)
(an MFMIS NCFET), the local gate charge density, Q G n
does not vary much from source to drain making the MFIS
(=gate charge per unit area of the n th subtransistor) can
structure behavelike MFMIS.
be accessed from the BSIM-BULK Verilog-A code, which
calculates it using an analytical expression [21], [22]. The
local ferroelectric and internal gate capacitance densities can B. Hysteresis Behavior
then be calculated by the expressions Cfen = ∂ Q G n /∂ Vfen Fig. 3(a) compares the hysteresis characteristics of the two
and Cintn = ∂ Q G n /∂ Vintn , respectively, where Vfen and Vintn , structures observed for large tfe values. First, we note that
respectively, denote ferroelectric and internal node voltages. the MFIS device shows a smoother hysteresis compared with
The inversion charge density at each node (Q I n ) along the the sharp jumps at transition points in the MFMIS device,
channel can again be accessed from the BSIM-BULK code. consistent with [13]. In MFMIS, dipoles in the ferroelectric
BSIM-BULK evaluates inversion charge densities at the source capacitor behave in unison due to the presence of an equipo-
and drain of each subtransistor, which are further used to tential metal layer at the ferroelectric–oxide interface, causing
calculate the surface potential at these nodes [21], [22]. Once, a large change in the polarization and hence, the current.
surface potential is obtained as a function of position, the On the other hand, in MFIS, different dipoles require different
horizontal x-component of the electric field is calculated by gate voltages to reach hysteresis transition depending upon the
taking its negative gradient along the channel. value of the electron–hole quasifermi level difference (VC )
In this paper, we have assumed a long channel internal in the channel, as shown in Fig. 3(b). Hysteresis threshold
MOSFET for both the MFMIS and MFIS NCFETs with length voltages (VG values at which hysteretic jumps occur) increase
L = 1 μm, width W = 1 μm, body doping concentration with VC . Thus, as VG is swept forward, first of all, the dipole
Na = 5 × 1017 cm−3 , interfacial oxide thickness tox = at the source end (VC = 0) makes a hysteretic jump followed
1 nm, and electron mobility μ = 500 cm2 /V-s. We have by other dipoles one by one in continuation from the source

Authorized licensed use limited to: AJOU UNIVERSITY. Downloaded on March 23,2023 at 06:36:32 UTC from IEEE Xplore. Restrictions apply.
PAHWA et al.: PHYSICAL INSIGHTS ON NC TRANSISTORS IN NONHYSTERESIS AND HYSTERESIS REGIMES 869

Fig. 2. Comparison of (a) ID –VG , (b) ID –VD , (c) transconductance, and (d) output conductance of MFMIS and MFIS NCFETs for different remnant
polarization values. Solid lines correspond to Gd:HfO2 ferroelectric. MFMIS structure exhibits higher ION , gm , and gds compared with MFIS for the
high Pr . On the contrary, for the low Pr case, MFIS exhibits higher ION , gds , and gm (in the high VG range) than MFMIS. (e) ON-current variation
with ferroelectric thickness for MFMIS and MFIS NCFETs at iso-IOFF . VFB and tfe denote the flat band voltage and the thickness of the ferroelectric
layer, respectively. Ec = 1 MV/cm is used for all the plots.

Fig. 3. (a) Hysteresis characteristics of MFMIS and MFIS devices. MFMIS exhibits hysteresis at larger ferroelectric thickness compared with MFIS.
Hysteresis plots are shown for the equal maximum width of hysteresis (0.1 V) of the two structures. (b) Impact of electron–hole quasifermi level
difference on hysteresis characteristics of ferroelectric dipoles in MFIS NCFET. (c) Variation of ferroelectric (Cfe ) and internal MOS capacitances (Cint )
as the functions of the gate charge density. In this figure, for MFIS, Cfe , Cint , and QG signify local densities at the source end while for MFMIS they
signify the total value of the quantity/gate area.

toward the drain end. This results in a smooth hysteresis of hysteresis windows. Note that the MFMIS structure shows
behavior in I D –VG characteristics of MFIS NCFET. When VG the same hysteresis window at a much larger ferroelectric
is swept backward, dipoles again one by one return to lower thickness (tfe ) compared with the MFIS structure. Also,
polarization values but at different (lower) threshold voltages note that it does not exhibit hysteresis at tfe =18.5 nm
compared with those in the forward sweep, giving rise to a whereas the MFIS does, but still provides a comparable
hysteresis behavior. ON -current. Hysteresis in any ferroelectric–dielectric system
In Fig. 3(a), hysteresis characteristics of MFMIS and occurs when the absolute value of negative ferroelectric
MFIS structures are shown for the equal maximum widths capacitance becomes equal to the value of positive dielectric

Authorized licensed use limited to: AJOU UNIVERSITY. Downloaded on March 23,2023 at 06:36:32 UTC from IEEE Xplore. Restrictions apply.
870 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 65, NO. 3, MARCH 2018

Fig. 4. (a) Gate charge density as a function of ferroelectric voltage drop for MFIS NCFET along source to drain direction for different Pr values.
Gate charge density values for the corresponding MFMIS NCFETs are also shown by MFMIS bias points. (b) Inversion charge density profiles in
the channel along source to drain direction. (c) Internal gate voltage (Vint ) and surface potential (ψs ) profiles along source to drain direction. Internal
gate voltage levels for the corresponding MFMIS devices are also given (horizontal lines) for a comparison purpose. (d) Horizontal component of
electric field in the channel along source to drain direction.

capacitance [1], [28]. For MFIS, the onset of hysteresis is source end [29] where Q I denotes inversion charge density
decided by the hysteresis transition in the source end dipole, and E x denotes the horizontal component of the electric
i.e., by the capacitance divider formed there by ferroelectric field. Therefore, we investigate the Q I and E x profiles in the
and internal MOS capacitances. Note that, the capacitance channel for low and high values of Pr in the following to
matching at the source side remains unaffected by the drain explain the above-mentioned observations.
voltage (in a long channel MOSFET). On the other hand, Fig. 4(a) shows gate charge density versus ferroelectric
due to the lumped nature of MFMIS, hysteresis is governed voltage curves for MFIS NCFET for both the high and
by the capacitance divider formed between total ferroelectric low Pr values. The corresponding MFMIS gate charge den-
capacitance and total internal MOS capacitance, which is sities appear as single bias points in the figure (as all the
influenced by the drain voltage [29]. Fig. 3(c) shows these dipoles in MFMIS have the same polarization value). Note
capacitances as a function of gate charge density Q G . At tfe = that in MFIS, Q G always decreases in the source to drain
16.5 nm, ferroelectric capacitance |Cfe | matches exactly with direction due to drain to channel coupling regardless of the
internal MOS capacitance Cint for MFIS implying the onset value of Pr . For high Pr , all the dipoles in the ferroelectric of
of hysteresis. However, in the case of MFMIS, capacitance MFIS reside in the NC regime, as shown in Fig. 4(a). Fig. 4(b)
matching gets disturbed by reduction in Cint caused by a drain shows the inversion charge density profiles in the channel
to channel coupling resulting in a nonhysteretic behavior from source toward drain for MFMIS and MFIS structures.
at the same tfe . MFMIS shows hysteresis only when tfe is MFMIS shows a lower Q I in the channel compared with
increased to 22.5 nm at which |Cfe | reduces and matches with MFIS for the high Pr . This observation can be explained
the lower Cint . Thus, for logic applications, where hysteresis is by considering a simple case of Q I near the source which
unwanted, MFMIS can be considered as a better choice over is a direct function of internal voltage Vint . The Vint at the
a rather hysteresis susceptible MFIS structure for a given tfe . source side [Vint1 in Fig. 1(d)] in MFIS is decided only by
the local capacitor divider built of Cfe1 and Cint1 , which is
independent of V D . While, for MFMIS, it is decided by the
C. Impact of Ferroelectric Material Parameters total lumped capacitances, Cfe and Cint [or the total gate
In general, Pr value of HfO2 -based ferroelectrics can lie in charge, see Fig. 1(c)] which are the functions of V D too in
a range beginning from a small value of 1 to 30 μC/cm2 , while addition to the applied gate voltage. V D disturbs the matching
E c remains close to 1 MV/cm [6], [27], [30]–[33]. Therefore, of the capacitances, as discussed in Section III-B, giving rise
in this section, we explore the behavior of MFMIS and MFIS to a lower Vint , as shown in Fig. 4(c) and therefore, a lower
NCFETs with respect to different values of Pr . Q I in MFMIS compared with MFIS. However, Vint in MFIS
Fig. 2(a)–(d) shows electrical characteristics of MFMIS and structure decreases in the source to drain direction since,
MFIS NCFETs for two different Pr values. For a low Pr value Vint = VG −Vfe , and Vfe increases with a decrease in Q G in NC
of 2 μC/cm2 , MFIS shows a higher ON-current and output region. Note that, the surface potential (ψs ) shown in Fig. 4(c)
conductance (gds) compared with an MFMIS contrary to the still increases but rather gradually compared with MFMIS. The
trends at a higher Pr = 12.13 μC/cm2 . Moreover, MFIS shows decreasing Vint profile in MFIS thus, produces a counter effect
a slightly negative gds or a negative differential resistance for causing a reduction in the horizontal component of electric
high V D values [16]. Both the MFIS and MFMIS for low Pr field in the channel, as shown in Fig. 4(d) compared with
case show a sudden rise in transconductance (gm ) in low VG MFMIS for which Vint remains a constant throughout. This
region with MFMIS also exhibiting a peak in gm in accordance results in a lower ON-current for MFIS than MFMIS in the
with [13]. For MFMIS, gm starts to decrease in high VG region case of high Pr ferroelectric despite a higher inversion charge
while for MFIS, it maintains a monotonically increasing trend. density in the channel for MFIS [Fig. 4(b)].
The channel current in the strong inversion region can be In case of the low Pr ferroelectric, all the dipoles in MFMIS
approximated by μW E x Q I (drift current) calculated near the structure and most of the dipoles in MFIS structure are able to

Authorized licensed use limited to: AJOU UNIVERSITY. Downloaded on March 23,2023 at 06:36:32 UTC from IEEE Xplore. Restrictions apply.
PAHWA et al.: PHYSICAL INSIGHTS ON NC TRANSISTORS IN NONHYSTERESIS AND HYSTERESIS REGIMES 871

rapidly compared with MFMIS in most of the part of the


channel which significantly enhances the horizontal electric
field, as shown in Fig. 4(d). This leads to a larger ON-current
for MFIS than MFMIS [Fig. 2(b)] despite the former having
a lower inversion charge density compared with the latter.
In addition to this, the surface potential for MFIS in this
case can be seen to decrease near the drain due to a rapidly
falling Vint . This causes the electric field to reverse its direction
in the channel as also been reported in a recent TCAD
study [34]. In this region of the channel, the drift current also
reverses its direction, which is now compensated by a larger
diffusion current (=μW Vt d Q I /d x, where Vt denotes thermal
voltage [29]) in the opposite direction, to maintain a current
continuity in the channel. Note that there is a large Q I gradient
near the drain side for Pr = 2 μC/cm2 in Fig. 4(b).
Fig. 5(a) shows the variation of ON-current with supply
voltage VDD for different Pr values at iso-IOFF of 1 nA/μm.
For the higher Pr , MFMIS remains superior to MFIS for all
supply voltages. For lower Pr , a different trend is observed
with MFIS being better than MFMIS for higher VDD values
and vice versa for smaller VDD values. For low VDD , VG is
also small, and most of the dipoles now lie in NC region even
for low Pr ferroelectric, and thus, ION shows trends similar to
high Pr case with MFMIS excelling over MFIS.
The variation of ON-current with Pr , for a fixed VDD ,
is shown in Fig. 5(b). ION shows a similar behavior for
both the structures. In low Pr region, ION is low because
the device gets biased in a PC regime and its ferroelectric
dipoles saturate at low polarization values [see Fig. 4(a)].
As we increase Pr , a greater number of dipoles start to stay
in NC region, which increases the value of current. However,
in high Pr region, ION starts to fall with Pr due to an increase
in mismatch between the ferroelectric and internal MOSFET
Fig. 5. (a) Variation of ON-current with supply voltage for MFMIS capacitances [14], [18], [35]. The optimum remnant polariza-
and MFIS structures. For a higher Pr , MFMIS outperforms MFIS for
all the supply voltages. For lower Pr material, MFIS shows a better tion ≈ 6 μC/cm2 provides a maximum ON-current for both
performance compared with MFMIS at higher supply voltages. the structures. Furthermore, MFMIS in Fig. 5(b) can be seen
(b) ON-current variation with remnant polarization for MFMIS and MFIS to outperform MFIS for all the Pr except very low values
NCFETs. Both the figures are shown at iso-IOFF = 1 nA/μm.
(<4 μC/cm2 ), as explained earlier.

IV. C ONCLUSION
reach a positive capacitance (PC) state, as shown in Fig. 4(a).
Some of the dipoles close to the drain end in an MFIS structure We provide the new physical insights into the operation of
still reside in the NC region due to a strong drain to channel the MFMIS and the MFIS structures of NCFET. Our analysis
coupling there, which reduces their polarization. Interestingly, shows that for high remnant polarization ferroelectric-
MFIS exhibits a lower Q I compared with MFMIS for low Pr based NCFETs, MFMIS structure excels the MFIS in
ferroelectric devices which derives from a lower Vint for the terms of ON-current with respect to supply voltage and
same, as shown in Fig. 4(c) (we again focus near the source ferroelectric thickness variation. However, for very low
side for simplicity). This is due to the fact that for the dipoles remnant polarization materials, the MFIS shows a relatively
near the source, Vfe > 0 [Fig. 4(a)], which implies Vint < VG . better performance compared with the MFMIS at high supply
While for MFMIS, Q G bias point is lower in value (due voltages. We also show that the MFMIS incurs a hysteresis
to drain voltage effect) with Vfe < 0 in spite of lying in at a larger ferroelectric thickness compared with the MFIS
the PC branch, which implies Vint > VG . This explains a while maintaining comparable ON-current, which makes it a
lower Q I for the MFIS structure. Contrary to the high Pr more suitable choice for hysteresis-free logic applications.
case, Vint in the MFIS structure for low Pr increases from the
source toward the drain because Vfe decreases with a decrease R EFERENCES
in Q G in the PC region except very close to the drain end
where dipoles lie in the NC state and therefore, Vint starts to [1] S. Salahuddin and S. Datta, “Use of negative capacitance to provide
voltage amplification for low power nanoscale devices,” Nano Lett.,
decrease. It causes the surface potential for MFIS to increase vol. 8, no. 2, pp. 405–410, 2007, doi: 10.1021/nl071804g.

Authorized licensed use limited to: AJOU UNIVERSITY. Downloaded on March 23,2023 at 06:36:32 UTC from IEEE Xplore. Restrictions apply.
872 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 65, NO. 3, MARCH 2018

[2] X. Li et al., “Enabling energy-efficient nonvolatile computing with [20] D. Ricinschi, C. Harnagea, C. Papusoi, L. Mitoseriu, V. Tura, and
negative capacitance FET,” IEEE Trans. Electron Devices, vol. 64, no. 8, M. Okuyama, “Analysis of ferroelectric switching in finite media as a
pp. 3452–3458, Aug. 2017, doi: 10.1109/TED.2017.2716338. Landau-type phase transition,” J. Phys., Condens. Matter, vol. 10, no. 2,
[3] T. Dutta, G. Pahwa, A. R. Trivedi, S. Sinha, A. Agarwal, and pp. 477–492, Apr. 1998, doi: 10.1088/0953-8984/10/2/026.
Y. S. Chauhan, “Performance evaluation of 7-nm node negative capaci- [21] Y. S. Chauhan et al., “BSIM6: Analog and RF compact model for bulk
tance FinFET-based SRAM,” IEEE Electron Device Lett., vol. 38, no. 8, MOSFET,” IEEE Trans. Electron Devices, vol. 61, no. 2, pp. 234–244,
pp. 1161–1164, Aug. 2017, doi: 10.1109/LED.2017.2712365. Feb. 2014, doi: 10.1109/TED.2013.2283084.
[4] A. Rusu, G. Salvatore, D. Jimenez, and A. Ionescu, “Metal-ferroelectric- [22] BSIM6 Technical Manual. Accessed: Jun. 10, 2017. [Online]. Available:
metal-oxide-semiconductor field effect transistor with sub-60 mv/decade http://bsim.berkeley.edu/models/bsimbulk/
subthreshold swing and internal voltage amplification,” in IEDM [23] D. Jiménez, E. Miranda, and A. Godoy, “Analytic model for the surface
Tech. Dig., Dec. 2010, pp. 16.3.1–16.3.4, doi: 10.1109/IEDM.2010. potential and drain current in negative capacitance field-effect transis-
5703374. tors,” IEEE Trans. Electron Devices, vol. 57, no. 10, pp. 2405–2409,
[5] J. Jo, W. Y. Choi, J.-D. Park, J. W. Shim, H.-Y. Yu, and C. Shin, Oct. 2010, doi: 10.1109/TED.2010.2062188.
“Negative capacitance in organic/ferroelectric capacitor to imple- [24] A. I. Khan, U. Radhakrishna, K. Chatterjee, S. Salahuddin, and
ment steep switching MOS devices,” Nano Lett., vol. 15, no. 7, D. A. Antoniadis, “Negative capacitance behavior in a leaky ferroelec-
pp. 4553–4556, 2015, doi: 10.1021/acs.nanolett.5b01130. tric,” IEEE Trans. Electron Devices, vol. 63, no. 11, pp. 4416–4422,
[6] K.-S. Li et al., “Sub-60 mv swing negative-capacitance FinFET with- Oct. 2016, doi: 10.1109/TED.2016.2612656.
out hysteresis,” in IEDM Tech. Dig., Dec. 2015, pp. 22.6.1–22.6.4, [25] M. Hoffmann, M. Pešić, S. Slesazeck, U. Schroeder, T. Mikolajick, and
doi: 10.1109/IEDM.2015.7409760. T. Mikolajick, “Modeling and design considerations for negative capac-
[7] A. I. Khan, K. Chatterjee, J. P. Duarte, Z. Lu, A. Sachid, S. Khandelwal, itance field-effect transistors,” in Proc. Joint Int. EUROSOI Workshop
R. Ramesh, C. Hu, and S. Salahuddin, “Negative capacitance in short- Int. Conf. Ultimate Integr. Silicon (EUROSOI-ULIS), Apr. 2017, pp. 1–4,
channel FinFETs externally connected to an epitaxial ferroelectric capac- doi: 10.1109/ULIS.2017.7962577.
itor,” IEEE Electron Device Lett., vol. 37, no. 1, pp. 111–114, Jan. 2016, [26] A. I. Khan, U. Radhakrishna, S. Salahuddin, and D. Antoniadis,
doi: 10.1109/LED.2015.2501319. “Work function engineering for performance improvement in leaky
[8] J. Zhou et al., “Ferroelectric HfZrOx Ge and GeSn PMOSFETs negative capacitance FETs,” IEEE Electron Device Lett., vol. 38, no. 9,
with Sub-60 mV/decade subthreshold swing, negligible hysteresis, and pp. 1335–1338, Sep. 2017, doi: 10.1109/LED.2017.2733382.
improved Ids,” in IEDM Tech. Dig., Dec. 2016, pp. 12.2.1–12.2.4, [27] M. Hoffmann et al., “Direct observation of negative capacitance in
doi: 10.1109/IEDM.2016.7838401. polycrystalline ferroelectric HfO2 ,” Adv. Funct. Mater., vol. 26, no. 47,
[9] G. A. Salvatore, D. Bouvet, and A. M. Ionescu, “Demonstration pp. 8643–8649, 2016, doi: 10.1002/adfm.201602869.
of subthrehold swing smaller than 60 mV/decade in Fe-FET with [28] A. Khan, C. Yeung, C. Hu, and S. Salahuddin, “Ferroelectric neg-
P(VDF-TrFE)/SiO2 gate stack,” in IEDM Tech. Dig., Dec. 2008, pp. 1–4, ative capacitance MOSFET: Capacitance tuning and antiferroelec-
doi: 10.1109/IEDM.2008.4796642. tric operation,” in IEDM Tech. Dig., Dec. 2011, pp. 11.3.1–11.3.4,
[10] S. Dasgupta, A. Rajashekhar, K. Majumdar, N. Agrawal, A. Razavieh, doi: 10.1109/IEDM.2011.6131532.
S. Troiier-McKinstry, and S. Datta, “Sub-kT/q switching in strong [29] Y. Tsividis and C. McAndrew, Operation and Modeling of the MOS
inversion in PbZr0.52 Ti0.48 O3 gated negative capacitance FETs,” IEEE Transistor. Oxford, U.K.: Oxford Univ. Press, 2011.
J. Exploratory Solid-State Comput. Device Circuits, vol. 1, pp. 43–48, [30] S. Mueller et al., “Incipient ferroelectricity in Al-doped HfO2 thin
Dec. 2015, doi: 10.1109/JXCDC.2015.2448414. films,” Adv. Funct. Mater., vol. 22, no. 11, pp. 2412–2417, 2012,
[11] M. H. Lee et al., “Prospects for ferroelectric HfZrOx FETs with doi: 10.1002/adfm.201103119.
experimentally CET=0.98 nm, SSfor=42 mV/dec, SSrev=28 mV/dec, [31] J. Müller et al., “Ferroelectricity in simple binary ZrO2 and
switch-off <0.2V, and hysteresis-free strategies,” in IEDM Tech. Dig., HfO2 ,” Nano Lett., vol. 12, no. 8, pp. 4318–4323, 2012,
Dec. 2015, pp. 22.5.1–22.5.4, doi: 10.1109/IEDM.2015.7409759. doi: 10.1021/nl302049kPMID:22812909.
[12] M. H. Lee et al., “Physical thickness 1.x nm ferroelectric HfZrOx [32] H. J. Kim et al., “Grain size engineering for ferroelectric Hf0.5 Zr0.5 O2
negative capacitance FETs,” in IEDM Tech. Dig., Dec. 2016, films by an insertion of Al2 O3 interlayer,” Appl. Phys. Lett., vol. 105,
pp. 12.1.1–12.1.4, doi: 10.1109/IEDM.2016.7838400. no. 19, p. 192903, 2014, doi: 10.1063/1.4902072.
[13] J. Duarte et al., “Compact models of negative-capacitance FinFETs: [33] H. Ota, S. Migita, J. Hattori, K. Fukuda, and A. Toriumi, “Mater-
Lumped and distributed charge models,” in IEDM Tech. Dig., Dec. 2016, ial and device engineering in fully depleted silicon-on-insulator tran-
pp. 30.5.1–30.5.4, doi: 10.1109/IEDM.2016.7838514. sistors to realize a steep subthreshold swing using negative capac-
[14] G. Pahwa, T. Dutta, A. Agarwal, and Y. S. Chauhan, “Compact model itance,” Jpn. J. Appl. Phys., vol. 55, no. 8S2, p. 08PD01, 2016,
for ferroelectric negative capacitance transistor with MFIS structure,” doi: 10.7567/jjap.55.08pd01.
IEEE Trans. Electron Devices, vol. 64, no. 3, pp. 1366–1374, Mar. 2017, [34] H. Ota, T. Ikegami, J. Hattori, K. Fukuda, S. Migita, and
doi: 10.1109/TED.2017.2654066. A. Toriumi, “Fully coupled 3-D device simulation of negative capac-
[15] G. Pahwa et al., “Analysis and compact modeling of nega- itance FinFETs for sub 10 nm integration,” in IEDM Tech. Dig.,
tive capacitance transistor with high ON-current and negative out- Dec. 2016, pp. 12.4.1–12.4.4, doi: 10.1109/IEDM.2016.7838403.
put differential resistance—Part I: Model description,” IEEE Trans. [35] C. I. Lin, A. I. Khan, S. Salahuddin, and C. Hu, “Effects of the variation
Electron Devices, vol. 63, no. 12, pp. 4981–4985, Dec. 2016, of ferroelectric properties on negative capacitance FET characteristics,”
doi: 10.1109/TED.2016.2614432. IEEE Trans. Electron Devices, vol. 63, no. 5, pp. 2197–2199, May 2016,
[16] G. Pahwa et al., “Analysis and compact modeling of nega- doi: 10.1109/TED.2016.2514783.
tive capacitance transistor with high ON-current and negative out-
put differential resistance—Part II: Model validation,” IEEE Trans.
Electron Devices, vol. 63, no. 12, pp. 4986–4992, Dec. 2016,
doi: 10.1109/TED.2016.2614436.
[17] A. Aziz, S. Ghosh, S. Datta, and S. K. Gupta, “Physics-based
circuit-compatible SPICE model for ferroelectric transistors,” IEEE
Electron Device Lett., vol. 37, no. 6, pp. 805–808, Jun. 2016, Girish Pahwa (S’16) received the B.Tech.
doi: 10.1109/LED.2016.2558149. degree in electronics and communication engi-
[18] G. Pahwa, T. Dutta, A. Agarwal, and Y. S. Chauhan, “Designing energy neering from Delhi Technological University,
efficient and hysteresis free negative capacitance FinFET with negative New Delhi, India in 2014. He is currently pursuing
DIBL and 3.5X ION using compact modeling approach,” in Proc. IEEE the Ph.D. degree with IIT Kanpur, Kanpur, India.
Eur. Solid-State Device Res. Conf. (ESSDERC), Sep. 2016, pp. 41–46, His current research interests include the
doi: 10.1109/ESSDERC.2016.7599584. physics and modeling of negative capacitance
[19] Y. Li, Y. Lian, K. Yao, and G. S. Samudra, “Evaluation and optimization transistors.
of short channel ferroelectric MOSFET for low power circuit applica-
tion with BSIM4 and Landau theory,” Solid-State Electron., vol. 114,
pp. 17–22, Dec. 2015, doi: 10.1016/j.sse.2015.07.001.

Authorized licensed use limited to: AJOU UNIVERSITY. Downloaded on March 23,2023 at 06:36:32 UTC from IEEE Xplore. Restrictions apply.
PAHWA et al.: PHYSICAL INSIGHTS ON NC TRANSISTORS IN NONHYSTERESIS AND HYSTERESIS REGIMES 873

Tapas Dutta (M’16) received the Ph.D. degree Yogesh Singh Chauhan (SM’12) was with ST
in nanoelectronics and nanotechnology from Microelectronics, Noida, India, from 2003 to
the Institut National Polytechnique de Grenoble, 2004, IBM, Bengaluru, from 2007 to 2010,
Grenoble, France, in 2014. the Tokyo Institute of Technology, Tokyo, Japan,
He then joined IIT Kanpur, Kanpur, India, as a in 2010, and the University of California at Berke-
Post-Doctoral Researcher, where he involved in ley, Berkeley, CA, USA, from 2010 to 2012. He is
multiscale simulations of III–V and 2-D material- currently an Associate Professor at IIT Kanpur,
based FETs, and compact modeling of negative Kanpur, India. His current research interests
capacitance transistors and the BSIMSOI model. include the characterization, modeling, and
simulation of semiconductor devices.

Amit Agarwal received the M.Sc. and Ph.D.


degrees in theoretical physics from the Indian
Institute of Science, Bengaluru, India, in 2005
and 2009, respectively.
In 2012, he joined IIT Kanpur, Kanpur, India,
as an Assistant Professor. His current research
interests include theoretical condensed mat-
ter theory, particularly low-dimensional systems,
and nanoscale device modeling.

Authorized licensed use limited to: AJOU UNIVERSITY. Downloaded on March 23,2023 at 06:36:32 UTC from IEEE Xplore. Restrictions apply.

You might also like