RyuMinjeong1
ChoiWoo Young1
-
(Department of Electrical and Computer Engineering and Inter-university Semiconductor
Research Center (ISRC), Seoul National University, Seoul 08826, Korea)
Copyright © The Institute of Electronics and Information Engineers(IEIE)
Index Terms
Silicon-germanium tunnel field-effect transistor, ambipolar behavior, complementary-like logic
I. INTRODUCTION
Ambipolar transistors have emerged as one of the most promising devices to simplify
the whole process flow of complementary metal-oxide-semiconductor (CMOS) logic integrated
circuits. Depending on the bias conditions, ambipolar transistors can function in
both hole-dominated p-type and electron-dominated n-type modes. A single ambipolar
device incorporating both p- and n-type performances enables complementary-like operation
and can replace p- or n-channel unipolar MOSFETs with a single ambipolar device [1-4]. Especially, for memory applications, because each unipolar mode of ambipolar transistors
can be controlled independently, a large memory window can be achieved [1]. However, most of the previously-proposed ambipolar transistors adopt solution-process-based
materials such as organic semiconductors, carbon nanotubes, and conjugated polymers.
They suffer from the following practical problems: 1) difficulty in finding the electrodes
for efficient electron/hole injection, 2) incompatibility with standard CMOS process,
and 3) power dissipation increase and performance degradation due to leakage current
[1-4].
To overcome the abovementioned challenges, tunnel field-effect transistors (TFETs)
with ambipolar characteristics have been considered as an alternative [5-7]. In contrast to solution-process-based ambipolar FETs, TFETs feature the following
advantages: 1) well-balanced electron- and hole-transport properties, 2) good CMOS
process compatibility, and 3) extremely low leakage current and subthreshold swing
[5-9]. Fig. 1 illustrates the structure of a planar SiGe homojunction TFET. Fig. 2 shows the lateral surface energy band diagrams from the source to drain. As the source
and drain regions are doped with p-type and n-type dopants, respectively, TFETs can
conduct both holes and electrons through their surface channels. When both gate voltage
(V$_{\mathrm{G}}$) and drain voltage (V$_{\mathrm{D}}$) are positive, drain current
(I$_{\mathrm{D}}$) is dominated by the electrons injected through the source-to-channel
junction. On the contrary, when both V$_{\mathrm{G}}$ and V$_{\mathrm{D}}$ become
negative, I$_{\mathrm{D}}$ is dominated by the holes injected through the drain-to-channel
junction. This ambipolar conduction becomes prominent when the power supply voltage
(V$_{\mathrm{DD}}$) or V$_{\mathrm{D}}$ is higher than E$_{\mathrm{G}}$/q (E$_{\mathrm{G}}$:
bandgap, q: elementary charge) [7]. Also, in the off state, while a large tunneling barrier blocks any carrier injection,
low-level leakage current is generated by the Shockley-Read-Hall (SRH) generation
near the source and drain and by the direct or trap-assisted tunneling (TAT) from
the source to drain [7].
However, because band-to-band tunneling (BTBT) is a major carrier injection mechanism
whose probability is low, silicon TFETs have faced the problem with low current drivability
[6-9]. Previous studies improved this issue by adopting silicon-germanium (SiGe) channel,
source, and drain owing to the smaller band gap and consequent enhanced BTBT rate
than silicon [6-10].
Even if SiGe-TFET-based logic applications have been studied intensively in previous
works, most of them utilized TFETs as unipolar devices while suppressing the ambipolar
current. In this manuscript, we investigate the effect of Ge mole fraction variation
on the ambipolar characteristics of SiGe TFETs by using device and circuit simulation.
Fig. 1. Schematic of a planar homojunction SiGe TFET.
Fig. 2. Schematic of the ambipolar behavior of TFETs: (a) P-channel ambipolar state; (b) off state; (c) n-channel turn-on state.
II. SIMULATION ENVIRONMENTS
Device simulations were conducted using a commercial technology computer-aided design
(TCAD) simulator [11]. Table 1 summarizes the simulated device parameters regarding the dimension and material properties.
To obtain the symmetric I$_{\mathrm{D}}$-V$_{\mathrm{G}}$ curves, source and drain
doping concentrations are made equal. Table 2 summarizes the device simulation models such as Fermi statistics, non-local BTBT,
and bandgap modulation owing to the high doping concentration in the source and drain
regions. The BTBT parameters of Si$_{\mathrm{1-x}}$Ge$_{\mathrm{x}}$ as a function
of Ge mole faction (x) refer to the literature. For a particular electric field F,
the BTBT generation rate G is given as follows [12]:
For x ${\leq}$ 0.5, the indirect tunneling process is dominant and F$_{0}$ = 1 V/cm,
P = 2.5 [12]. The indirect BTBT parameters are A = 3.29${\times}$10$^{15}$, 2.61${\times}$10$^{15}$,
2.27${\times}$10$^{15}$ [cm$^{-3}$·s$^{-1}$] and B = 23.8, 18.1, 15.5 [MV·cm$^{-1}$]
for x = 0, 0.3, 0.5 of Si$_{\mathrm{1-x}}$Ge$_{\mathrm{x}}$ materials, respectively
[12].
For channel materials, unstrained Si$_{\mathrm{1-x}}$Ge$_{\mathrm{x}}$ is assumed
where strain is relieved due to interfacial dislocations as shown in Fig. 3(a). The conduction band offset (${\Delta}$E$_{\mathrm{c}}$) values and the valence band
offset (${\Delta}$E$_{\mathrm{v}}$) values between the Si and Si$_{\mathrm{1-x}}$Ge$_{\mathrm{x}}$
are indicated in the band diagrams [12,13]. In the other case, Si$_{\mathrm{1-x}}$Ge$_{\mathrm{x}}$ pseudomorphically grown
on Si can be compressively strained as shown in Fig. 3(b) [14,15]. While the higher x value of Si$_{\mathrm{1-x}}$Ge$_{\mathrm{x}}$ and strain reduce
bandgap energy, it is widely known that the most band offset between the Si to Si$_{\mathrm{1-x}}$Ge$_{\mathrm{x}}$
layer occurs in the valence band, which is called a type-Ⅱ staggered junction [14].
Although a SiGe-on-insulator (SGOI) layer is formed uniformly with the same Si$_{\mathrm{1-x}}$Ge$_{\mathrm{x}}$
material, only the surface region near the source- and drain-to-channel junctions
affects the TFET on-current [10]. In this work, the Si$_{\mathrm{1-x}}$Ge$_{\mathrm{x}}$ whose x is {\textless} 0.5
is simulated. In this range, the following two characteristics are observed: 1) stronger
indirect tunneling than direct tunneling [12] and 2) pure interfacial SiO$_{2}$ formed by thermal oxidation by rejecting Ge [16]. Also, regarding the interface trap density (D$_{\mathrm{it}}$), the high D$_{\mathrm{it}}$
is one of the major challenges on Si$_{\mathrm{1-x}}$Ge$_{\mathrm{x}}$-based CMOS
devices. Nevertheless, it was reported in the previous study that D$_{\mathrm{it}}$
can be reduced to 2 ${\times}$10$^{11}$ cm$^{-2}$ by applying defect elimination techniques
[17].
Fig. 3. (a) Band alignments of Si and unstrained Si1-xGexgrown on Si substrate; (b) A schematic of unstrained Si1-xGexfilm deposited on Si; (c) Band alignments of compressively strained Si1-xGexgrown on Si; (d) A schematic of compressively strained Si1-xGexas a result of pseudomorphic growth on Si.
Table 1. Simulation parameters of TFETs
Device Parameters
|
Values
|
Gate length (LG)
|
500 nm
|
Gate dielectric thickness (tSiO2)
|
2 nm
|
Si1-xGex-on-Insulator thickness (tSGOI)
|
30 nm
|
Channel doping concentration (Nch)
|
1016 cm-3
|
Source and drain doping
concentration (NS, ND)
|
1020 cm-3
|
Gate work function (ΦM)
|
4.66 eV
|
Interface trap density (Dit)
|
2×1011 cm-2
|
Table 2. Simulation models of TFETs
Device Physics
|
Models
|
Carrier Transport
|
Drift-diffusion
|
Bandgap
|
Bandgap narrowing
|
Recombination
|
Shockley-Read-Hall (SRH)
|
Tunneling
|
Dynamic nonlocal path BTBT
|
III. SIMULATION RESULTS AND DISCUSSION
1. Device Simulation Results
Fig. 4(a) shows the simulated I$_{\mathrm{D}}$-V$_{\mathrm{G}}$ curves of Si$_{\mathrm{1-x}}$Ge$_{\mathrm{x}}$
TFETs for three x values. I$_{\mathrm{D}}$ consists of turn-on current (I$_{\mathrm{turn-on}}$),
ambipolar current (I$_{\mathrm{amb}}$), and off-state leakage current (I$_{\mathrm{leak}}$).
Also, as shown in Fig. 4(b), two kinds of threshold voltages (V$_{\mathrm{th}}$’s) are extracted: one for the
turn-on state (V$_{\mathrm{turn-on}}$) and the other for the ambipolar state (V$_{\mathrm{amb}}$).
They are defined as the V$_{\mathrm{G}}$’s where I$_{\mathrm{turn-on}}$ and I$_{\mathrm{amb}}$
exceed I$_{\mathrm{leak}}$, respectively [5].
From Fig. 4(a) and (b), it is observed that the {\textbar}V$_{\mathrm{amb}}${\textbar} decreases
significantly with higher x value. On the contrary, V$_{\mathrm{turn-on}}$ remains
almost the same even if the x value varies. It can be explained by the energy band
diagrams presented in Fig. 5(a) and (b). V$_{\mathrm{turn-on}}$ is determined by the point where the lowest conduction
band edge (E$_{\mathrm{c}}$) of the channel exceeds the highest valence band edge
(E$_{\mathrm{v}}$) of the source [18]. However, as shown in Fig. 3, the ${\Delta}$E$_{\mathrm{c}}$ is almost zero regardless
of the x values of Si$_{\mathrm{1-x}}$Ge$_{\mathrm{x}}$. Thus, V$_{\mathrm{turn-on}}$
to induce an energy overlap window between the source valence band and the channel
conduction band remains almost unchanged. Meanwhile, I$_{\mathrm{turn-on}}$ is boosted
owing to the smaller bandgap and effective mass, and consequently enhanced electron
tunneling probability. In contrast, because the ${\Delta}$E$_{\mathrm{v}}$ increases
significantly with higher x value, it requires lower {\textbar}V$_{\mathrm{amb}}${\textbar}
to shift the E$_{\mathrm{v}}$ of the channel over the E$_{\mathrm{c}}$ of the drain,
as can be derived from Fig. 5(b). As a result, with increasing x value, {\textbar}V$_{\mathrm{amb}}${\textbar} decreases
and it increases I$_{\mathrm{amb}}$ more rapidly than the I$_{\mathrm{turn-on}}$ under
the same bias conditions.
Fig. 4. (a) ID-VG characteristics of simulated Si1-xGexTFETs with parameters corresponding to unstrained Si1-xGex; (b) The change of Vth and ID (extracted at |VG| = 1.5 V) in ambipolar and turn-on states depending on the Ge mole fraction.
Fig. 5. Simulated lateral energy band diagrams of unstrained Si1-xGexTFETs that represent: (a) source-to-channel junction which controls the n-type turn-on states; (b) drain-to-channel junction which controls the p-type ambipolar states, respectively. The y-axis value is set so that the quasi electron/hole Fermi level in the source region is zero.
2. Circuit Simulation Results
In this section, we investigate the influence of the x value on the transient switching
characteristics of ambipolar Si$_{\mathrm{1-x}}$Ge$_{\mathrm{x}}$ TFETs. The simple
inverter and NAND gate characteristics were simulated by HSPICE [19]. Our TFET compact model [18,20] was calibrated referring to the TCAD simulation data as shown in Fig. 6. The previous model, based on 1-D Zener tunneling only considered the I$_{\mathrm{turn-on}}$
of TFETs, neglecting the ambipolar behavior. Thus, we modified the model to operate
as a unified model that simultaneously describes the turn-on and ambipolar state characteristics.
Even if ambipolar FETs-based complementary-like logic gates may consist of the identical
devices [3], for V$_{\mathrm{DD}}$ downscaling, two identical devices with different gate work
functions are used. Fig. 7(a) and (b) represent the circuit schematics of ambipolar Si$_{\mathrm{1-x}}$Ge$_{\mathrm{x}}$
TFETs-based inverter and NAND gate, respectively. There are two signal propagation
paths determining the output. In the case of the pull-up part, connected from V$_{\mathrm{DD}}$
to the output node, the I$_{\mathrm{amb}}$ drives the transient response. On the contrary,
in the pull-down part, linked from the output node to the ground, the I$_{\mathrm{turn-on}}$
drives the transient response.
Fig. 7(c) shows the simulated transient response of the ambipolar Si$_{\mathrm{1-x}}$Ge$_{\mathrm{x}}$
TFET-based complementary-like inverters with increasing the x value. The tendency
for propagation delay can be predicted from the relatively small increase in I$_{\mathrm{turn-on}}$
and the significant increase in I$_{\mathrm{amb}}$. The rising time decreases more
abruptly than the falling time because I$_{\mathrm{amb}}$ dominates the rising time,
while I$_{\mathrm{turn-on}}$ dominates the falling time. Pull-up delay decreases from
650 ns to 8 ns while pull-down delay decreases from 185 ns to 20 ns. The dynamic power
consumption in the worst case increases from 0.94 nW to 2.21 nW when x increases from
0 to 0.5. The degree of decrease in delay and increase in power accompanied by higher
x value is consistent with the previous study which focused on the transient performance
of n-type pull-down TFETs based on Si and SiGe [21]. Also, in TFET inverters, voltage undershoot and overshoot occur at the output low-to-high
and high-to-low transitions, respectively [22]. As shown in Fig. 6(b), the gate-to-drain and gate-to-source capacitance (C$_{\mathrm{GD}}$ and C$_{\mathrm{GS}}$)
of TFETs in the linear region stays almost the same regardless of the x value. Therefore,
the extent of overshoot and undershoot is the same for different x values. Meanwhile,
C$_{\mathrm{GS}}$ in the ambipolar state increases more rapidly with a higher x value,
while the behavior of C$_{\mathrm{GD}}$ in the turn-on state changes little. The reason
is the same as explained in the previous section regarding the ${\Delta}$E$_{\mathrm{c}}$’s
and ${\Delta}$E$_{\mathrm{v}}$’s with different x values. In the transient behavior
of NAND gates as shown in Fig. 7(d), the propagation delay from low to high decreases more than the delay from high to
low with increasing the x value.
Thus, to design ambipolar Si$_{\mathrm{1-x}}$Ge$_{\mathrm{x}}$ TFET-based complementary-like
logic integrated circuits, the different tendency in the propagation delay related
to the I$_{\mathrm{turn-on}}$ and I$_{\mathrm{amb}}$ for x variations needs to be
considered for appropriate signal manipulation.
Fig. 6. (a) Transfer characteristics; (b) CGD/CGS-VG characteristics of Si1-xGexTFETs used for complementary-like logic operation. For TFET components with red lines, transient responses are driven by the Iamb. For blue lines, transient responses are driven by the Iturn-on.
Fig. 7. (a) Configuration of the ambipolar Si1-xGexTFET-based inverter; (b) NAND gates; (c), (d) their transient switching characteristics. The load capacitance is 1 fF. In the TFET symbols, the direction in which BTBT occurs mainly in the operating bias regime is indicated.
IV. CONCLUSIONS
The influence of the channel stoichiometry of SiGe TFETs on the ambipolar behavior
was investigated via simulations. At the device level, it is observed that a higher
Ge mole fraction enhances the carrier transports of both polarities. At the same time,
it has a more significant effect on the V$_{\mathrm{th}}$ shift and on-current boosting
for the hole-dominated regime than for the electron-dominated regime. Finally, when
applied to complementary-like logic operations, the pull-up network becomes stronger
than the pull-down network with increased x value in terms of current driving.
ACKNOWLEDGMENTS
This work was supported by Samsung Research Funding & Incubation Center of Samsung
Electronics under Project Number SRFC-TA2103-01.
References
Y. Ren, X. Yang, L. Zhou, J.-Y. Mao, S.-T. Han, and Y. Zhou, “Recent Advances in Ambipolar
Transistors for Functional Applications,” Adv. Funct. Mater., vol. 29, no. 40, Jul.
2019, Art. no. 1902105.
J. Zaumseil and H. Sirringhaus, “Electron and ambipolar transport in organic field-effect
transistors,” Chem. Rev., vol. 107, pp. 1296-1323, Mar. 2007.
S. Z. Bisri, C. Piliego , J. Gao, and M. A. Loi, “Outlook and emerging semiconducting
materials for ambipolar transistors,” Adv. Mater., vol. 26, pp. 1176-1199, Feb. 2014.
Y. Zhang, C.-H. Huang, and K. Nomura, “Voltage transfer characteristics of CMOS-like
inverters for ambipolar SnO thin-film transistors,” IEEE Electron Device Lett., vol.
43, no. 1, pp. 52-55, Jan. 2022.
J.-S Jang and W. Y. Choi, “Ambipolarity factor of tunneling field-effect transistors
(TFETs),” J. Semicond. Technol. Sci., vol. 11, no. 4, pp. 272-277, Dec. 2011.
F. Mayer et al., “Impact of SOI, Si1-xGexOI and GeOI substrates on CMOS compatible
Tunnel FET performance,” in IEEE Int. Electron Devices Meet., San Francisco, CA, USA,
2008, pp. 162-166.
A. C. Seabaugh and Q. Zhang, “Low-voltage tunnel transistors for beyond CMOS logic,”
Proc. IEEE, vol. 98, no. 12, pp. 2095-2110, Dec. 2010.
W. Y. Choi, B.-G. Park, J. D. Lee, and T.-J. King Liu, “Tunneling field effect transistors
(TFETs) with subthreshold swing (SS) less than 60 mv/dec,” IEEE Electron Device Lett.,
vol. 28, no. 8, pp. 743-745, Aug. 2007.
A. M. Ionescu and H. Riel, “Tunnel field-effect transistors as energy-efficient electronic
switches,” Nature, vol. 479, no. 7373, pp. 329-337, Nov. 2011.
N. Patel, A. Ramesha, and S. Mahapatra, “Drive current boosting of n-type tunnel FET
with strained SiGe layer at source,”, Microelectron. J., vol 39, no. 12, pp. 1671-1677,
Dec. 2008.
Sentaurus Device User Guide, Version T-2022.03, Synopsys, Inc., Mountain View, CA,
USA, 2022.
K.-H. Kao, A. S. Verhulst, W. G. Vandenberghe, B. Sorée, G. Groeseneken, and K. De
Meyer, “Direct and indirect band-to-band tunneling in germanium-based TFETs,” IEEE
Trans. Electron Devices, vol. 59, no. 2, pp. 292-301, Feb. 2012.
J.-H Han, M. Takenaka, and S. Takagi, “Analysis of interface trap density of plasma
post-nitrided Al2O3 /SiGe MOS interface with high Ge content using high-temperature
conductance method,” J. Appl. Phys., vol. 120, no. 12, Sep. 2016, Art. no. 125707.
D. J. Paul, “Si/SiGe heterostructures: from material and physics to devices and circuits,”
Semicond. Sci. Technol., vol. 19, no. 10, pp. R75-R108, Oct. 2004.
M. Arienzo, S. S. Iyer, B. S. Meyerson, G. L. Patton, and J. M.C. Stork, “Si-Ge alloys:
growth, properties and applications,” Appl. Surf. Sci., vol. 48/49, pp. 377-386, 1991.
P.-E. Hellberg, S.-L. Zhang, F. M. d’Heurle, and C. S. Petersson, “Oxidation of silicon-germanium
alloys. I. An experimental study,” J. Appl. Phys., vol. 82, no. 11, pp. 5773-5778,
Dec. 1997.
C. H. Lee et al., “Selective GeOx-scavenging from interfacial layer on Si1-xGex channel
for high mobility Si/Si1-xGex CMOS application,” in IEEE Symp. VLSI Technol., Honolulu,
HI, USA, 2016, pp. 1-2.
H. Lu, D. Esseni, and A. Seabaugh, “Universal analytic model for tunnel FET circuit
simulation,” Solid-State Electron., vol. 108, pp. 110-117, Jun. 2015.
HSPICE User Guide: Simulation and Analysis P-2019.06, Synopsys. Inc., Mountain View,
CA, USA, Jun. 2008.
H. Lu, T. Ytterdal, and A. Seabaugh, Notre Dame TFET Model (Version 2.1.0), nanoHUB,
2017.
A. Pal, A. B. Sachid, H. Gossner, and V. R. Rao, “Insights into the design and optimization
of tunnel-FET devices and circuits,” IEEE Trans. Electron Devices, vol. 58, no. 4,
pp. 1045-1053, Apr. 2011.
S. Mookerjea, R. Krishnan, S. Datta, and V. Narayanan, “On Enhanced Miller Capacitance
Effect in Interband Tunnel Transistors,” IEEE Electron Device Lett., vol. 30, no.
10, pp. 1102-1104, Oct. 2009.
Minjeong Ryu received the B.S. degree in the Department of Electrical and Computer
Engineering from Seoul National University, Korea, in 2022. She is currently working
toward the Integrated Ph.D. degree in the Department of Electrical and Computer Engineering
from Seoul National University, Korea. Her current research interests include tunneling
field-effect transistors, ferroelectric field-effect transistors, and in-memory computing
architectures.
Woo Young Choi received the B.S., M.S., and Ph.D. degrees in the School of Electrical
Engineering from Seoul National University, Korea, in 2000, 2002, and 2006, res-pectively.
From 2006 to 2008, he was with the Department of Electrical Engineering and Computer
Sciences, University of California (Berkeley, USA) as a post-doctor. From 2008 to
2021, he was with the Department of Electronic Engineering, Sogang University (Seoul,
Korea) as a professor. Since 2022, he has been a member of the faculty of Seoul National
University (Seoul, Korea), where he is currently an associate professor with the Department
of Electrical and Computer Engineering. He has authored or coauthored over 400 papers
in international journals and conference proceedings and holds over 60 Korean/US patents.
His current research interests include design, fabrication, characterization and theoretical
analysis of nanoscale MOSFETs and emerging extremely-low power logic/analog/memory
devices, three-dimensional integration and neuromorphic computing.