Sn-doped gallium oxide (Ga2O3) wrap-gate fin-array field-effect transistors (finFETs) were formed by top-down BCl3 plasma etching on a native semi-insulating Mg-doped (100) β-Ga2O3 substrate. The fin channels have a triangular cross-section and are approximately 300 nm wide and 200 nm tall. FinFETs, with 20 nm Al2O3 gate dielectric and ∼2 μm wrap-gate, demonstrate normally-off operation with a threshold voltage between 0 and +1 V during high-voltage operation. The ION/IOFF ratio is greater than 105 and is mainly limited by high on-resistance that can be significantly improved. At VG = 0, a finFET with 21 μm gate-drain spacing achieved a three-terminal breakdown voltage exceeding 600 V without a field-plate.

Gallium oxide (Ga2O3) is emerging as a potential disruptive electronic material for high-voltage electronics applications. The excitement of this material is due to its (1) ultra-wide bandgap of ∼4.8 eV with ∼8 MV/cm theoretical critical field strength,1 (2) up to four-inch native substrate availability and capability of melt-growth synthesis,2 and (3) a wide range of n-type doping achievable by halide vapor phase epitaxy (HVPE),3 molecular beam epitaxy (MBE),4 low-pressure chemical vapor deposition (LPCVD),5 metal-organic chemical vapor deposition (MOCVD),6 and metal-organic vapor phase epitaxy (MOVPE).7,8 The β-phase Ga2O3 unit crystal has a monoclinic structure and is reported as the most thermally stable and conducive for the single-crystal homoepitaxial growth.9,10 For heterogeneous integration, a notable cleavage plane is located along the (100) crystal plane, which has incited nanomembrane research for integration with arbitrary substrates and two-dimensional semiconductors.11–13 

The first transistor devices by homoepitaxial Ga2O3 were demonstrated with a Sn-doped Ga2O3 channel grown by MBE on (010) semi-insulating β-Ga2O3 substrates.1,14 Metal-oxide-semiconductor field-effect transistors (MOSFETs) followed later with a Si-doped channel and ohmic contacts by implantation with breakdown exceeding 750-V with a field-plate.15,16 Most recently, Sn-doped Ga2O3 MOSFETs homoepitaxially grown by MOVPE on (100) semi-insulating β-Ga2O3 achieved a record-high 3.8 MV/cm critical field strength surpassing GaN and SiC bulk theoretical field strengths.17 

For power electronics applications, a normally-off transistor is preferred for safe high-voltage operation and to mitigate off-state power dissipation. To achieve a high-current density, Ga2O3 MOSFETs require high doping concentration resulting in a negative threshold voltage (VTH). To shift toward positive VTH, non-planar fin-shaped channels offer enhanced electrostatic control of the channel by depleting it from the side walls without sacrificing doping. Achieving dense, parallel arrays of fin channels is most easily achieved by top-down plasma etching though reports of fin channels formed by metal-catalyzed wet-etching18 and self-assembly19 are promising to avoid plasma etch damage.

GaN-based fin-channel field-effect transistors (finFETs) have been reported with Si-doped GaN junctionless and high-electron mobility AlGaN/GaN heterostructure channels where the gate wraps around fins with enhanced electrostatics to nearly or fully deplete the channel.20–22 However, the main drawbacks for GaN are cost and the availability of native substrates for low-defect density homoepitaxial growth. In this letter, we present a finFET with arrays of parallel Sn-doped Ga2O3 fin channels formed by top-down plasma etching to achieve normally-off operation on a native (100) semi-insulating β-Ga2O3 substrate. The results show the feasibility of wrap-gate architecture to shift the VTH to positive values while maintaining volume current densities for consideration in future high-voltage device design.

Fin-array field-effect transistors (finFET) devices were fabricated from a 300-nm Sn-doped Ga2O3 channel grown homoepitaxially by MOVPE on a 100-mm2 Mg-doped semi-insulating (100) β-Ga2O3 substrate.7,8,23 First, arrays of ∼300-nm wide fin channels with a ∼900-nm pitch were formed by electron beam lithography followed by 150-nm Cr metal evaporation as the hard mask. A second 200-nm Cr hard mask was superimposed on the fin mask by projection lithography to create bulk mesa contacts for source and drain electrodes. Both Cr layers were etched by inductively coupled plasma (ICP) etching using BCl3 chemistry.24 The etch conditions were 120 W reactive ion etching (RIE) power and 300 W coil power with 20 sccm BCl3 and 16 mTorr chamber pressure. The etch selectivity of Ga2O3:Cr was approximately ∼2:1. To sufficiently remove the entire 300-nm channel between fins, an over-etch was required, which completely etched the fin Cr mask resulting in triangular-shaped fins. Residual Cr on the source and drain mesas was removed by commercially available Cr wet-etchant. Ohmic contacts consisted of Ti/Al/Ni/Au (20/100/50/50 nm) rapidly annealed for 1-min at 470 °C in nitrogen. A 20-nm Al2O3 gate dielectric was deposited by atomic layer deposition (ALD) at 250 °C and patterned by fluorine-based RIE to allow for Ni/Au (20/480 nm) interconnects and ∼2 μm long optical gate metal evaporation. Finally, a second 20-nm ALD Al2O3 layer was deposited and patterned on the sample to passivate the etched Ga2O3 surfaces between interconnects. The fabrication process is illustrated in Fig. 1(a).

FIG. 1.

(a) Fabrication process for Ga2O3 finFETs and (b) the tilted false-colored SEM image of a LSD = 4 μm finFET depicting the geometry of Ga2O3 fin channels and contacts.

FIG. 1.

(a) Fabrication process for Ga2O3 finFETs and (b) the tilted false-colored SEM image of a LSD = 4 μm finFET depicting the geometry of Ga2O3 fin channels and contacts.

Close modal

The finFET has a centered two-finger gate layout with each gate finger wrapping along 48 fins. The total source to drain distance (LSD) is ∼4m, and the fin-array spans approximately ∼3-μm of this source-drain distance. A tilted SEM image of the fin channels with wrap-gate and bulk-like ohmic contacts is shown in Fig. 1(b). The sidewall morphology appears relatively smooth as previously observed using high-power ICP plasma etching with BCl3.24 A representative cross-sectional SEM image of three fins is shown in Fig. 2(a). The darker contrast observed in the fin compared to the substrate is indicative of the Sn-doped channel and adequate electrical isolation between fins. Fig. 2(b) depicts that the fins are approximately ∼300 nm at the base with tapered sidewalls joining at ∼200 nm thickness. The 20-nm Al2O3 gate dielectric and Ni/Au gate metal conform to the fin on all sides.

FIG. 2.

(a) The cross-sectional SEM image of three Ga2O3 fins and (b) a high magnification SEM image of one Ga2O3 fin channel with associated dimensions.

FIG. 2.

(a) The cross-sectional SEM image of three Ga2O3 fins and (b) a high magnification SEM image of one Ga2O3 fin channel with associated dimensions.

Close modal

The mobility and doping concentration of the fins were measured from on-wafer Van der Pauw (VdP) test structures and device C-V measurements. It is widely reported that ionized donor concentration, ND, can vary significantly from the chemical Sn-doping concentration.8,17 The sheet resistance (RSH) and electron mobility (μ) were measured on a VdP structure near the reported device as ∼40 kΩ/sq and ∼24 cm2/Vs, respectively. We observed the larger geometry of the Cr mask used for the VdP mesa etched slower compared to the Cr fin-array mask; therefore, the VdP mesa was protected during the fin-array definition process. A forward and reverse C-V measurement of the finFET is shown in Fig. 3 indicating ∼0.8 V of hysteresis, which has been previously reported as mobile border traps in accumulation.25 In the inset, an ND ∼ 2.3 × 1017 cm−3 was extracted from the linear region of 1/C2 as a function of VGS. The area was estimated as LGWfinNfin where Wfin is ∼200 nm after considering a ∼70 nm backside depletion width and using a 3:2 width-to-height triangular fin cross-section. Finally, the flat-band capacitance, CFB, can be calculated by the measured oxide capacitance (Cox225 fF) in series with the semiconductor capacitance (CS).26 The corresponding forward and reverse sweep flat-band voltage, VFB, is 1.3 V and 2.1 V, respectively.

FIG. 3.

Forward and reverse C-VGS sweep of the finFET indicating the calculated flatband and oxide capacitance; and, (inset) C−2-VGS characteristics to extract carrier concentration.

FIG. 3.

Forward and reverse C-VGS sweep of the finFET indicating the calculated flatband and oxide capacitance; and, (inset) C−2-VGS characteristics to extract carrier concentration.

Close modal

In the absence of accurate models for Ga2O3, one-dimensional analytical expressions were used to estimate the depletion widths (Wd) on the two sides and bottom facet of the Sn-doped fins. The partial depletion width of the sides in the ungated region can be estimated by the built-in energy potential (Vbi) using the energy band lineup at the Al2O3/Ga2O3 interface20 

EGAl2O3=EVBMAl2O3+ΔEC+(ECGa203EF)+Vbi,
(1)

where EG, ΔEC, and EVBM are the bandgap, conduction band offset, and valence band maximum with respect to the Ga2O3 Fermi level energy (EF), respectively. Vbi(ug) represents band-bending in the ungated Ga2O3 due to the presence of interface traps. Kamimura et al. reported on the Al2O3/Ga2O3 interface with EG ∼ 6.8 eV, ΔEC ∼ 1.5–1.7 eV,25,27 and EVBM ∼ 3.8 eV25 using XPS measurements. For ND ∼ 2.3 × 1017 cm−3, the semiconductor EC−EF energy difference can be expressed as

Vtln(NDNC),
(2)

where Vt and NC are the thermal voltage (∼26 mV at 300 K) and effective conduction band density of states for Ga2O3.28 This analysis shows EC−EF is ∼73 meV and leaves a non-negligible Vbi(ug) ∼ 1.4 eV, which may be related to interface traps and/or pinning, which is neither well-understood nor reported.

A similar study of GaN finFETs deduced a Vbi ∼ 0.74 eV in the ungated region and was explained by Al2O3/GaN interfacial chemistry by XPS.20,29 Furthermore, in the gated region, the band-bending can increase an additional ∼1.15 eV due to the difference in metal work functions of Ni (Фm = 5.15 eV) and Ga2O3S = χs + EC−EF),20 where χs is the electron affinity of Ga2O3 (∼3.5 to 4.0 V).27,30 However, this does not consider trap-assisted tunneling for thin Al2O3 gate oxide,27 and it remains unclear how the Vbi(ug) compensates for the band-bending normally induced by a gate contact without thorough XPS characterization of our particular interface. For a simple case, however, where both energy barriers are combined in the gated region, the Vbi(g) is ∼2.5 V, which is reasonably close to the measured reverse sweep VFB. The maximum depletion width, Wd, for each region is calculated by the following equation:

Wd=2εoεSVbi(ug,g)qND,
(3)

where εo and εS are the permittivity of free space and Ga2O3 dielectric constant, respectively. This yields Wd ∼ 83 nm and ∼110 nm in the ungated and gated regions, respectively. A backside depletion width from the semi-insulating substrate is found to be ∼70 nm assuming a mid-gap interface trap density of ∼2 × 1017 cm−3.17,31 Therefore, we estimate an undepleted fin with approximately ∼26 nm (base) × ∼ 17 nm (height) in the ungated region contributes to the volume conduction mechanism in the finFET. In contrast, the depletion from the sides and substrate fully deplete the fin dimensions in the gated region to realize the normally-off operation. It should be noted that once VGS > VFB, the finFET is operating in accumulation similar to non-planar normally-off junctionless GaN finFETs.20,32

Fig. 4(a) shows the family of ID-VD curves from VGS = +4 V to 0 V. At VGS = +4 V, the on-current (ION) reaches ∼3.5 μA. An upper bound on expected current in the partially depleted fin-arrays can be approximated by the open channel current density (Jn) where VDS<|VGS-VTH| using the drift current equation

Jn=qμNDECH,
(4)

where ECH = VDS/LCH is the potential across the source-drain channel (LCH). Using the partially depleted fin-array cross-sectional area at VDS = 2 V and LCH = 3 μm, Jn = 5.9 kA/cm2 or ID ≈ 1.3 μA, which is close to the measured value in Fig. 4(a). For comparison, this simple analysis is also in agreement at VDS = 1 V (IDS = 0.55 mA) for the planar Sn-doped Ga2O3 MOSFET reported by Green et al. using the surface and substrate depletion widths with the reported ND and Ti/Au gate.17 The gate width, WG=WfinNfin, is ∼19 μm corresponding to an ION ∼ 0.18 μA/μm. The low ION is a main limitation of the fin-array topology reported here, but can be drastically improved in the future with higher-mobility materials and on-resistance optimization. For this device, the gate swing was limited by the conduction band offset of Al2O3/Ga2O3, which can be improved with, for example, ALD SiO2.33 The gate leakage characteristics are shown in Fig. 4(b) and indicate ultra-low gate leakage near 10−12 Amps before an onset of trap-assisted tunneling at forward bias appears.27 

FIG. 4.

(a) The ID-VD family of output curves from VGS = +4 to 0 V; (b) Two-terminal log(IG)-VG gate leakage performances from off-to-on state; and, (c) log(ID)-VG forward and reverse sweeps indicating a normally-off state at VGS = 0 V. An identical device with the entire epitaxial channel etched is included to illustrate the effect of parasitic substrate conduction.

FIG. 4.

(a) The ID-VD family of output curves from VGS = +4 to 0 V; (b) Two-terminal log(IG)-VG gate leakage performances from off-to-on state; and, (c) log(ID)-VG forward and reverse sweeps indicating a normally-off state at VGS = 0 V. An identical device with the entire epitaxial channel etched is included to illustrate the effect of parasitic substrate conduction.

Close modal

Fig. 4(c) shows the ID-VG characteristics at VDS = 10 V. Despite ION limitations, the finFET has >105ION/IOFF ratio. The device reaches an off-state approaching 10−12 Amps between 0 and +1 VGS indicating enhancement-mode operation. To rule out the parasitic conduction in the substrate between fins, an identical MOSFET with the epitaxial channel etched away was fabricated and shows minimal modulation of the remaining etched SI substrate. We attribute this parasitic modulation to uncompensated free carriers in the substrate being accumulated at the Al2O3/SI-Ga2O3 interface. The forward and reverse sweeps reveal trapping effects that may be a combination of the unoptimized Al2O3/Ga2O3 interface and density of interface traps (Dit) caused by the plasma etching of the fin side walls. Despite no surface treatment optimization, the subthreshold slope (SS) is 158 mV/dec, which is superior to previously reported Ga2O3 MOSFETs. The Dit can be estimated by the shift in forward and reverse VFB from Fig. 3 using the following expression:

Dit=CoxΔVFBqEGGa2O3,
(5)

which is approximately ∼3.9 × 1011 cm−2 eV−1 where Cox ∼ 3.8 fF/μm2. The area for Cox is calculated using the measured Cox, 20 nm thickness and a gate dielectric constant of 8.5; though, estimating the area using the sum of the fin side facets multiplied by LG gives nearly the same value. This Dit value is similar to previously reported Al2O3/Ga2O3 and SiO2/Ga2O3 MOS capacitors on (-201) n+ β-Ga2O3 substrates with Dit < 1.0 × 1012 cm−2 eV−1 after surface treatment optimization.33,34 The effect of surface plane orientation of the fin facets and the dielectric-Ga2O3 interface quality is unclear and requires further investigation.

FinFET high-voltage operation on wider devices with LGD = 16 μm and 21 μm was characterized with an Agilent B1505A on a Cascade Tesla probe station. At VGS = 0 V, the IDS is <10−7 Amps until a breakdown voltage (VBK) is reached. For each LGD, a VTH = +0.8 is measured at VDS = 10 V, which is shown by the inset of Fig. 5. It should be noted that the ION for large LGD devices have a very high on-resistance and do not saturate at VDS = 10 V. At VGS = 0 V, a VBK was measured at 567 and 612 V for LGD = 16 and 21 μm, respectively. As indicated in Fig. 5, VBK is destructive and limited by peak electric fields in the gate oxide.

FIG. 5.

The breakdown voltages of Ga2O3 finFETs with LG = 2 μm and LGD = 16, 21 μm while biased in the off-state at VGS = 0 V. The inset shows the transfer characteristics of the same device indicating a VTH = +0.8 V.

FIG. 5.

The breakdown voltages of Ga2O3 finFETs with LG = 2 μm and LGD = 16, 21 μm while biased in the off-state at VGS = 0 V. The inset shows the transfer characteristics of the same device indicating a VTH = +0.8 V.

Close modal

To conclude, we have fabricated an enhancement-mode Ga2O3 MOSFET enabled by arrays of Sn-doped fins on a semi-insulating (100) β-Ga2O3 substrate. A VBK exceeding 600-V at VGS = 0 V off-state was demonstrated and represents the highest breakdown voltage measured without field-plate for β-Ga2O3 transistors, and the highest breakdown for any transistor technology utilizing non-planar device channels.20–22,35–37 Future work includes understanding the role of traps at the dielectric-Ga2O3 interface and optimizing on-resistance by reducing the fin channel length and using highly doped ohmic cap layer.

The authors would like to acknowledge the support of bulk and epitaxial crystal growth by IKZ-Berlin and fabrication support by Mr. Jason Hickey and Mr. Andrew Browning at AFRL. X. Li acknowledges NSF DMR #1508140.

1.
M.
Higashiwaki
,
K.
Sasaki
,
A.
Kuramata
,
T.
Masui
, and
S.
Yamakoshi
,
Appl. Phys. Lett.
100
(
1
),
013504
(
2012
).
2.
M.
Higashiwaki
,
K.
Sasaki
,
A.
Kuramata
,
T.
Masui
, and
S.
Yamakoshi
,
Phys. Status Solidi A
211
(
1
),
21
(
2014
).
3.
H.
Murakami
,
K.
Nomura
,
K.
Goto
,
K.
Sasaki
,
K.
Kawara
,
Q. T.
Thieu
,
R.
Togashi
,
Y.
Kumagai
,
M.
Higashiwaki
, and
A.
Kuramata
,
Appl. Phys. Express
8
(
1
),
015503
(
2015
).
4.
K.
Sasaki
,
A.
Kuramata
,
T.
Masui
,
E. G.
Víllora
,
K.
Shimamura
, and
S.
Yamakoshi
,
Appl. Phys. Express
5
(
3
),
035502
(
2012
).
5.
S.
Rafique
,
L.
Han
,
M. J.
Tadjer
,
J. A.
Freitas
, Jr.
,
N. A.
Mahadik
, and
H.
Zhao
,
Appl. Phys. Lett.
108
(
18
),
182105
(
2016
).
6.
D.
Shinohara
and
S.
Fujita
,
Jpn. J. Appl. Phys.
47
(
9R
),
7311
(
2008
).
7.
G.
Wagner
,
M.
Baldini
,
D.
Gogova
,
M.
Schmidbauer
,
R.
Schewski
,
M.
Albrecht
,
Z.
Galazka
,
D.
Klimm
, and
R.
Fornari
,
Phys. Status Solidi A
211
(
1
),
27
(
2014
).
8.
M.
Baldini
,
M.
Albrecht
,
A.
Fiedler
,
K.
Irmscher
,
D.
Klimm
,
R.
Schewski
, and
G.
Wagner
,
J. Mater. Sci.
51
(
7
),
3650
(
2016
).
9.
S.
Geller
,
J. Chem. Phys.
33
(
3
),
676
(
1960
).
10.
L. M.
Foster
and
H. C.
Stumpf
,
J. Am. Chem. Soc.
73
(
4
),
1590
(
1951
).
11.
W. S.
Hwang
,
A.
Verma
,
H.
Peelaers
,
V.
Protasenko
,
S.
Rouvimov
,
H. G.
Xing
,
A.
Seabaugh
,
W.
Haensch
,
C.
Van de Walle
, and
Z.
Galazka
,
Appl. Phys. Lett.
104
(
20
),
203111
(
2014
).
12.
S.
Ahn
,
F.
Ren
,
J.
Kim
,
S.
Oh
,
J.
Kim
,
M. A.
Mastro
, and
S. J.
Pearton
,
Appl. Phys. Lett.
109
(
6
),
062102
(
2016
).
13.
J.
Kim
,
S.
Oh
,
M. A.
Mastro
, and
J.
Kim
,
Phys. Chem. Chem. Phys.
18
(
23
),
15760
(
2016
).
14.
M.
Higashiwaki
,
K.
Sasaki
,
T.
Kamimura
,
M. H.
Wong
,
D.
Krishnamurthy
,
A.
Kuramata
,
T.
Masui
, and
S.
Yamakoshi
,
Appl. Phys. Lett.
103
(
12
),
123511
(
2013
).
15.
M. H.
Wong
,
K.
Sasaki
,
A.
Kuramata
,
S.
Yamakoshi
, and
M.
Higashiwaki
,
IEEE Electron Device Lett.
37
(
2
),
212
(
2016
).
16.
K.
Sasaki
,
M.
Higashiwaki
,
A.
Kuramata
,
T.
Masui
, and
S.
Yamakoshi
,
Appl. Phys. Express
6
(
8
),
086502
(
2013
).
17.
A. J.
Green
,
K. D.
Chabak
,
E. R.
Heller
,
R. C.
Fitch
,
M.
Baldini
,
A.
Fiedler
,
K.
Irmscher
,
G.
Wagner
,
Z.
Galazka
,
S. E.
Tetlak
,
A.
Crespo
,
K.
Leedy
, and
G. H.
Jessen
,
IEEE Electron Device Lett.
37
(
7
),
902
(
2016
).
18.
Y.
Song
,
P. K.
Mohseni
,
S. H.
Kim
,
J. C.
Shin
,
T.
Ishihara
,
I.
Adesida
, and
X.
Li
,
IEEE Electron Device Lett.
37
(
8
),
970
(
2016
).
19.
X.
Miao
,
K.
Chabak
,
C.
Zhang
,
P. K.
Mohseni
,
D.
Walker
, Jr.
, and
X.
Li
,
Nano Lett.
15
(
5
),
2780
(
2015
).
20.
K.-S.
Im
,
C.-H.
Won
,
Y.-W.
Jo
,
J.-H.
Lee
,
M.
Bawedin
,
S.
Cristoloveanu
, and
J.-H.
Lee
,
IEEE Trans. Electron Devices
60
(
10
),
3012
(
2013
).
21.
B.
Lu
,
E.
Matioli
, and
T.
Palacios
,
IEEE Electron Device Lett.
33
(
3
),
360
(
2012
).
22.
K.-S.
Im
,
V.
Sindhuri
,
Y.-W.
Jo
,
D.-H.
Son
,
J.-H.
Lee
,
S.
Cristoloveanu
, and
J.-H.
Lee
,
Appl. Phys. Express
8
(
6
),
066501
(
2015
).
23.
Z.
Galazka
,
K.
Irmscher
,
R.
Uecker
,
R.
Bertram
,
M.
Pietsch
,
A.
Kwasniewski
,
M.
Naumann
,
T.
Schulz
,
R.
Schewski
, and
D.
Klimm
,
J. Cryst. Growth
404
,
184
(
2014
).
24.
J. E.
Hogan
,
S. W.
Kaun
,
E.
Ahmadi
,
Y.
Oshima
, and
J. S.
Speck
,
Semicond. Sci. Technol.
31
(
6
),
065006
(
2016
).
25.
T.
Kamimura
,
K.
Sasaki
,
M. H.
Wong
,
D.
Krishnamurthy
,
A.
Kuramata
,
T.
Masui
,
S.
Yamakoshi
, and
M.
Higashiwaki
,
Appl. Phys. Lett.
104
(
19
),
192104
(
2014
).
26.
K.
Piskorski
and
H. M.
Przewlocki
, in
MIPRO, 2010 Proceedings of the 33rd International Convention
(
IEEE
,
2010
), pp. 37–42.
27.
T.-H.
Hung
,
K.
Sasaki
,
A.
Kuramata
,
D. N.
Nath
,
P. S.
Park
,
C.
Polchinski
, and
S.
Rajan
,
Appl. Phys. Lett.
104
(
16
),
162106
(
2014
).
28.
ATLAS SILVACO
, The Simulation Standard (
2013
), p.
7
.
29.
T. L.
Duan
,
J. S.
Pan
, and
D. S.
Ang
,
Appl. Phys. Lett.
102
(
20
),
201604
(
2013
).
30.
I.
López
,
E.
Nogales
,
P.
Hidalgo
,
B.
Méndez
, and
J.
Piqueras
,
Phys. Status Solidi A
209
(
1
),
113
(
2012
).
31.
A.
Chandra
,
C. E. C.
Wood
,
D. W.
Woodard
, and
L. F.
Eastman
,
Solid-State Electron.
22
(
7
),
645
(
1979
).
32.
K.-S.
Im
,
Y.-W.
Jo
,
J.-H.
Lee
,
S.
Cristoloveanu
, and
J.-H.
Lee
,
IEEE Electron Device Lett.
34
(
3
),
381
(
2013
).
33.
K.
Zeng
,
Y.
Jia
, and
U.
Singisetti
,
IEEE Electron Device Lett.
37
(
7
),
906
(
2016
).
34.
H.
Zhou
,
S.
Alghmadi
,
M.
Si
,
G.
Qui
, and
P.
Ye
,
IEEE Electron Device Letters
37
(11),
1411
(
2016
).
35.
K. S.
Im
,
R. H.
Kim
,
K. W.
Kim
,
D. S.
Kim
,
C. S.
Lee
,
S.
Cristoloveanu
, and
J. H.
Lee
,
IEEE Electron Device Lett.
34
(
1
),
27
(
2013
).
36.
S.
Takashima
,
Z.
Li
, and
T. P.
Chow
,
IEEE Trans. Electron Devices
60
(
10
),
3025
(
2013
).
37.
F.
Yu
,
D.
Rümmler
,
J.
Hartmann
,
L.
Caccamo
,
T.
Schimpke
,
M.
Strassburg
,
A. E.
Gad
,
A.
Bakin
,
H.-H.
Wehmann
,
B.
Witzigmann
,
H. S.
Wasisto
, and
A.
Waag
,
Appl. Phys. Lett.
108
(
21
),
213503
(
2016
).