Article

Approaching the quantum limit in

two-dimensional semiconductor contacts

https://doi.org/10.1038/s41586-022-05431-4 Weisheng Li1,12, Xiaoshu Gong2,12, Zhihao Yu1,12, Liang Ma2,12, Wenjie Sun3, Si Gao3,4,

Çağıl Köroğlu5, Wenfeng Wang1, Lei Liu1, Taotao Li1, Hongkai Ning1, Dongxu Fan1, Yifei Xu1,
Received: 9 March 2022
Xuecou Tu1, Tao Xu6, Litao Sun6, Wenhui Wang2, Junpeng Lu2, Zhenhua Ni2, Jia Li7,
Accepted: 7 October 2022 Xidong Duan7, Peng Wang3, Yuefeng Nie3, Hao Qiu1, Yi Shi1 ✉, Eric Pop5,8,9, Jinlan Wang2 ✉ &
Xinran Wang1,10,11 ✉
Published online: 11 January 2023

Check for updates

The development of next-generation electronics requires scaling of channel
material thickness down to the two-dimensional limit while maintaining ultralow
contact resistance1,2. Transition-metal dichalcogenides can sustain transistor
scaling to the end of roadmap, but despite a myriad of efforts, the device
performance remains contact-limited3–12. In particular, the contact resistance has
not surpassed that of covalently bonded metal–semiconductor junctions owing to
the intrinsic van der Waals gap, and the best contact technologies are facing
stability issues3,7. Here we push the electrical contact of monolayer molybdenum
disulfide close to the quantum limit by hybridization of energy bands with semi-
metallic antimony (0112) through strong van der Waals interactions. The contacts
exhibit a low contact resistance of 42 ohm micrometres and excellent stability at
125 degrees Celsius. Owing to improved contacts, short-channel molybdenum
disulfide transistors show current saturation under one-volt drain bias with an
on-state current of 1.23 milliamperes per micrometre, an on/off ratio over 108 and
an intrinsic delay of 74 femtoseconds. These performances outperformed
equivalent silicon complementary metal–oxide–semiconductor technologies and
satisfied the 2028 roadmap target. We further fabricate large-area device arrays and
demonstrate low variability in contact resistance, threshold voltage, subthreshold
swing, on/off ratio, on-state current and transconductance13. The excellent
electrical performance, stability and variability make antimony (0112) a promising
contact technology for transition-metal-dichalcogenide-based electronics beyond
silicon.

Metal–semiconductor (M–S) contacts have essential roles in modern requirement of energy-level alignment (that is, a zero or negative
electronics owing to the continuous scaling of device dimensions14. Schottky barrier) and coherence of wavefunctions across the inter-
In an ideal M–S junction, there is a fundamental quantum limit in the face without backscattering15. The situation is more complicated for
junction contact resistance by assuming ballistic transport of the con- van der Waals (vdW) materials such as transition-metal dichalcoge-
ducting modes: nides (TMDs) owing to their dangling-bond-free surface. The existence
of a vdW gap introduces an additional tunnel barrier and reduces the
h π charge injection, making typical Rc values several orders of magni-
Rc,min = , (1)
2q 2 2n2D tude higher than the quantum limit15. Attempts to overcome such a
‘contact gap’ include edge contacts16,17, low-work-function metals18,
where h is Planck’s constant, q is the unit charge and n2D is the carrier ultrahigh vacuum evaporation8, low-energy metal integration4,6,19,
concentration in the semiconductor. Such ideal M–S contacts have, doping20,21, tunnelling contacts22,23 and, more recently, semi-metal
however, been rarely obtained experimentally owing to the stringent contacts7,9,10,24. These efforts have reduced the Rc to several hundred

National Laboratory of Solid State Microstructures, School of Electronic Science and Engineering and Collaborative Innovation Center of Advanced Microstructures, Nanjing University,
1

Nanjing, China. 2School of Physics, Southeast University, Nanjing, China. 3National Laboratory of Solid State Microstructures, Jiangsu Key Laboratory of Artificial Functional Materials, College
of Engineering and Applied Sciences and Collaborative Innovation Center of Advanced Microstructures, Nanjing University, Nanjing, China. 4College of Materials Science and Engineering,
Nanjing Tech University, Nanjing, China. 5Department of Electrical Engineering, Stanford University, Stanford, CA, USA. 6SEU-FEI Nano-Pico Center, Key Laboratory of MEMS of Ministry of
Education, Southeast University, Nanjing, China. 7Hunan Key Laboratory of Two-Dimensional Materials and State Key Laboratory for Chemo/Biosensing and Chemometrics, College of
Chemistry and Chemical Engineering, Hunan University, Changsha, China. 8Department of Materials Science and Engineering, Stanford University, Stanford, CA, USA. 9Precourt Institute for
Energy, Stanford University, CA, USA. 10 School of Integrated Circuits, Nanjing University, Suzhou, China. 11Suzhou Laboratory, Suzhou, China. 12These authors contributed equally: Weisheng Li,
Xiaoshu Gong, Zhihao Yu, Liang Ma. ✉e-mail: [email protected]; [email protected]; [email protected]

274 | Nature | Vol 613 | 12 January 2023

 a – c e f g h
MoS2 Sb (0112) – – – –
0.2 Sb (0112) Sb (0112) Sb (0112) Sb (0112)
Sb (0001) 6 Sb (0001) Sb (0001) Sb (0001)

5.41

6.04

3.33
9.72

3.30
0 EF 6

3.22
3.21
Energy (eV)

DOS of TMDs (10–2 state per nm2)

–0.2

Charge transfer (10–1 e nm–2)

4.05
PLDOS (10–2 state per nm2)
4 3

3.57

EvdW (eV nm–2)

4.28
–0.4

4.17
6

2.88
G X S Y G

3.77
4
b d

2.62
MoS2 Sb (0001)

2.57
3.36
0.2

3.35

2.93
2

2.68
0 EF 3
Energy (eV)

2.04

0.85

2.20
1.47
1.34
1.26

2.16
0.63
–0.2

0.74

1.81
2

0.22
0.13
0.18
0.17
0.07
0 0 2
–0.4

MoSe2
MoSe2
MoSe2
Mo dxy,x2–y2

Mo dz2

MoS2
Sb pz

Mo dyz,xz

WSe2
Sb px,y

MoS2

WSe2
MoS2

WSe2

WS2
WS2
WS2
G M K G

Fig. 1 | Electronic properties of Sb (0112)–MoS2 and Sb (0001)–MoS2 contacts Sb, Mo and S atoms, respectively. The differential charge density iscalculated
from DFT calculations. a,b, Atomic-projected electronic band structure of Sb by subtracting pre-contact three-layer Sb and MoS2 charge density from the
(0112)–MoS2 (a) and Sb (0001)–MoS2 (b) contacts. The red and blue symbols post-contact charge density. The isosurface level is 9 × 10 −5 e Bohr−3 and
represent orbitals from MoS2 and Sb, respectively. The size of the symbols 7 × 10 −4 e Bohr−3 for the left and right panels, respectively. e, Projected PLDOS of
represents the relative contribution. c,d, The charge density near EF (left) and different orbitals of Sb (0112)–MoS2 and Sb (0001)–MoS2 contacts near EF. f, The
the differential charge density (right) of Sb (0112) (c) and Sb (0001) (d) contacts integrated DOS of TMDs at EF. g, The charge transfer from Sb to TMDs by Bader
(red, positive; blue, negative). The blue, green and yellow spheres represent  charge analysis. h, The interfacial vdW interaction between TMDs and Sb.

ohm micrometres, which is comparable to but still higher than M–S hybridization of M–S energy bands needs to be realized at the Fermi
junctions composed of intimate chemical bonds (for example, doped energy (EF) and across the vdW gap. We note that such hybridized
and undoped silicon (Si) and titanium nitride (TiN)/gallium nitride states are within the conduction and valance bands and are spatially
(GaN)). Besides Schottky barrier height (SBH) and tunnel barrier width, delocalized, and are distinct from the localized metal-induced gap
M–S orbital hybridization also has a critical role in the charge injec- states26,27. The metal-induced gap states originate from defects and
tion efficiency25. A major limitation of previous studies is that, even can cause undesirable Fermi level pinning.
for the best semi-metal contacts, the electronic states of M–S remain Here we realize band hybridization between monolayer molybdenum
weakly coupled across the vdW gap7. To further reduce Rc, coherent disulfide (MoS2) and semi-metallic antimony (Sb) (0112). We first use

a b
Si Si (0002) Si (0004)
A1g
Sb (0112) 100 ºC
XRD indensity (a.u.)

Sb/MoS2/SiO2
Intensity (a.u.)

β-Sb Sb (0003) RT Sb/MoS2/SiO2

MoS2
Eg 1 100 ºC Sb/SiO2
E2g A1g

100 400 500 600 10 20 30 40 50 60 70 80

Raman shift (cm–1) 2T (º)
c d e
Sb SiO2
Sb (0112)
~0.315 nm

Sb ~0.285 nm
MoS2 area
S
Mo
S
Substrate

Fig. 2 | Characterization of the Sb (0112)–MoS2 contact. a, Raman spectrum 0.315 nm for Sb (0112) planes is marked. Scale bar, 2 nm. d, Zoom-in
of Sb film deposited on MoS2/Si substrate shows Sb Eg and A 1g peaks at 111 cm−1 atomic-resolution image from the red box in c. The Mo, S and Sb atoms are
and 149 cm−1, respectively. b, XRD θ–2θ diffractogram of Sb film with different overlayed on the image. The vdW gap of 0.285 nm is marked. Scale bar, 1 nm.
deposition conditions and substrate. RT, room temperature. c, Cross-section e, Optical microscope image of Sb films deposited on MoS2 and SiO2 substrate,
HAADF-STEM image of the Sb (0112)–MoS2 contact. The interplane distance of showing clear colour contrast. Scale bar, 20 μm.

Nature | Vol 613 | 12 January 2023 | 275

Article
a b –2
c
1.0 300
13
cm
90 10
200
×

Pint (cm2 V−1 s−1)

–2
20 0.5 1.
2
13 cm

Rtot (kΩ μm)

= 0
0 n 2D 6 × 1 –2 200 150

Rc (Ω μm)
Ids (μA μm–1)
. 13 c
m
60 –0.05 0 0.05 =1
Lc (μm) 10 –2
1.5 n 2D 2.1 × 13 cm
= 10 100
1.0 10 n 2D 3.0 ×
0.8 n 2D
= 100
30 0.6
0.4 50
0.2
0.1
0 0 0
–3 0 3 6 9 12 15 0 0.5 1.0 1.5 50 100 200 400
Vgs (V) Lc (μm) T (K)

d e f –
30 – –
Sb (0112) 30 Sb (0112) Sb (0112)
40 Sb (0001)
25 Sb (0001) Sb (0001)
25
20 30
20

Count
Count
Count

15 15 20
10 10
10
5 5
0 0 0
0 200 400 600 800 1,000 0 20 40 60 80 100 50 60 70 80 90
Rc (Ω μm) LT (nm) Pint (cm2 V−1 s−1)

Fig. 3 | Electrical properties and stability of the Sb (0112)–MoS2 contact. It is noted that the LT derived by the TLM method assumes the same MoS2 sheet
a, Transfer curves of a typical TLM structure with Lc ranging from 0.1 μm to resistivity underneath and outside the contact, which is probably violated
1.5 μm and Vds = 0.1 V. Inset: false-colour scanning electron microscope image considering the charge transfer between Sb (0112) and MoS2. This could lead to
of a TLM structure. Scale bar, 2 μm. b, Rc extraction using the TLM method from uncertainties in the obtained LT. c. Rc (red squares) and intrinsic mobility (black
the device in a. From top to bottom, n2D = 1.2 × 1013 cm−2, 1.6 × 1013 cm−2, rhombuses) as a function of temperature for another device. d–f, Rc (d), LT (e)
2.1 × 1013 cm−2 and 3 × 1013 cm−2, respectively. Inset: the magnified data plot near and μint (f) distribution of the Sb (0112) and Sb (0001) contacts, respectively.
Lc = 0, where the y intercept and x intercept represent 2Rc and 2LT, respectively. The solid lines are Gaussian fittings.

density-functional theory (DFT) calculations to validate our approach We further calculated the interfacial vdW interaction (Evdw) in the
(see Methods for calculation details). Figure 1a,b show the atomically heterobilayer. Previous studies have established that the degree of
projected band structure of Sb (0112 )–MoS2 and Sb (0001)–MoS2, band hybridization in TMD heterostructures can be increased by vdW
respectively. For Sb (0112)–MoS2, multiple hybridized bands composed interaction28. Our calculation indeed revealed a strong Evdw = 3.33 eV nm−2
of a molybdenum (Mo) d orbital and Sb p and s orbitals intersect EF. For for Sb (0112)–MoS2, which was much greater than that for Sb (0001)–
Sb (0001)–MoS2, however, the bands intersecting EF are mainly com- MoS2 (2.62 eV nm−2; Fig. 1h). This was attributed to the planar instead
posed of Sb p and s orbitals, and the MoS2 conduction band minimum of protruding distribution of Sb atoms in the (0112) plane. We further
is located above EF (Fig. 1b). To find the origin of band hybridization, extended our calculations to several other TMDs (Fig. 1f–h and Extended
we analysed the population of orbitals in the projected local density Data Fig. 2), as well as bismuth (Bi) (0112)–MoS2 (Extended Data Fig. 1e–h).
of states (PLDOS) near EF (Fig. 1e). It can be seen that, although the Sb In all the cases, the Sb (0112) and Bi (0112) contacts showed enhanced
px,y orbitals are comparable, the Sb pz orbital of Sb (0112) is 190% larger band hybridization and Bader charge transfer compared with their
than that of Sb (0001) owing to the closely packed atomic surface of (0001) counterparts, showing that our approach is general and robust.
Sb (0112) (surface Sb atom density, 10.43 nm−2 versus 5.34 nm−2). Fur- Besides band hybridization, another important aspect to realize
thermore, the population of Mo d orbitals in the Sb (0112)–MoS2 contact ohmic contact is the tunnel barrier width. To this end, we calculated
is also markedly enhanced (Fig. 1e). It is noted that the distribution of the electrostatic potential profile of the contacts (Extended Data
Sb pz and Mo d z 2 orbitals is mainly along the vertical direction. As a Fig. 1c,d). The tunnel barrier width of Sb (0112)–MoS2 was 1.35 Å, which
result, a considerable real-space overlap between the Sb pz orbital and was smaller than that of Sb (0001)–MoS2 (1.39 Å) as well as Bi (0001)–
the Mo d orbital emerges (Fig. 1c), which leads to strong band hybrid- MoS2 (1.66 Å) (ref. 7). This further ensures the low tunnelling resistivity
ization in the Sb (0112)–MoS2 contact. The hybridization is visualized of Sb (0112) compared with other semi-metallic contacts.
by plotting the spatial charge distribution at EF in Fig. 1c,d (left). We The evaporation of semi-metals on MoS2 under ambient temperature
can see that the charge is distributed in both Sb (0112) and MoS2 layers favoured (0001)-orientated films7,9,10. Owing to the enhanced thermo-
despite the vdW gap, in clear contrast to Sb (0001)–MoS2. We further dynamic stability (Fig. 1h), we used a mild substrate heating of about
investigated the charge transfer by calculating the differential charge 100 °C during evaporation to obtain (0112)-oriented Sb films (see Meth-
density as shown in Fig. 1c,d (right). The degree of hybridization and ods for details). Raman spectroscopy (Fig. 2a) revealed that the Sb was
charge transfer was quantified by the integrated DOS of MoS2 at EF semi-metallic β-phase29. Figure 2b compares the X-ray diffraction (XRD)
(Fig. 1f) and Bader charge transfer (Fig. 1g). In both cases, the Sb (0112)– θ–2θ (where θ is the X-ray incident angle) diffractograms of Sb depos-
MoS2 contact shows significant enhancement, which is favourable for ited on MoS2 under different temperatures. The Sb deposited under
M–S contacts. The charge transfer effectively pulls the MoS2 conduc- ambient temperature showed pronounced (0003) diffraction peak.
tion band minimum about 0.4 eV below EF (Fig. 1a). Therefore, the MoS2 Under high temperature, however, the (0003) peak almost disappeared
under Sb (0112) is degenerated doped and forms ohmic contact with and the (0112) peak became dominant. Such temperature-dependent
the channel, which is similar to the situation in Si14. In fact, the simulated orientation was consistent with molecular-beam epitaxy results30. The
local device density of states (LDDOS) suggest that the band bending Sb (0112) was further confirmed by cross-section high-angle annular
across the highly doped contact and the channel region is very steep dark-field scanning transmission electron microscopy (HAADF-STEM)
in real space (about 0.6 nm) and allows considerable carrier tunnelling (Fig. 2c,d). The Sb atomic planes parallel to MoS2 had an interplane
through this space-charge region in the on state (Extended Data distance of 0.315 nm, corresponding to the (0112) planes. In contrast,
Fig. 1a,b). the interplane distance of Sb (0001) was measured to be 0.379 nm

276 | Nature | Vol 613 | 12 January 2023

 a b 10–2 c
1.6 Pulse
10–3 d.c.
Au Au
10–4
– –
Sb (0112) Sb (0112) 10–5 1.2

Ids (mA μm–1)

Ids (A μm–1)
10–6
MoS2
10–7 Lc ≈ 20 nm
0.8
10–8
10–9
Dielectric 0.4
10–10
10–11
Gate
10–12 0
–8 –4 0 4 8 12 0 0.5 1.0 1.5
Vgs (V) Vds (V)
d 103 e 1.6 f 10
This work
Si
102 Ni (ref. 45) Co/h-BN FinFET
(ref. 22) Gra (ref. 40) 1.2
UHV Au Ni (ref. 17) Si planar HP 1

Delay (ps)
41
101 (ref. )
Ion (mA μm–1)
Ag/Au (ref. 42) Ni/Au
Rc (kΩ μm)

In/Au (ref. 4) 43
Au (ref. 44) (ref. )
0.8
100 Sn/Au (ref. 24)
Au (ref. 21) 0.1
Sn/Au (ref. 44) In/Au (ref. 44)
Gra
InGaAs Bi (ref. 7)
10–1 Si Sb–MoS2
0.4
GaN
(this work)
Quantum limit Si GAAFET 0.01
10–2
0 Si planar LP

0.1 1 10 101 102 103 10 100

Lc (nm) Lg (nm)
n2D (1013 cm–2)
Bi (0001) (ref. 7) Au (ref. 46) Ag/Au (ref. 42) Si planar HP MoS2 (Vdd = 1 V)
Sb (0001) (ref. 10) Au (ref. 31) Ag/Au (ref. 48) Si planar LP Rc = 240 Ω μm
Sb (0001) (ref. 9) Au (ref. 47) Sb (0112) d.c. Si FinFET Rc = 42 Ω μm
In/Au alloy (ref. 44) UHV Au (ref. 41) Si GAAFET
Au(AlOx doping) (ref. 21) Sn (ref. 24) Sb (0112) pulse

Fig. 4 | Short-channel MoS2 FET performance and benchmark. a, Cartoon respectively. It is noted that the data (red pentagram symbols) for Sb (0112)-
illustration of the FET structure. b, Transfer characteristics of a typical MoS2 contacted MoS2 FETs come from 587 devices with different Lc, including 1.5 μm
FET with Lc = 20 nm under Vds = 100 mV (green line), 550 mV (blue line) and (83 devices), 1 μm (83), 800 nm (83), 600 nm (82), 400 nm (82), 200 nm (80),
1 V (red line). Inset: false-colour SEM image of the device. Scale bar, 500 nm. 100 nm (80), 60 nm (5), 40 nm (5) and 20 nm (4). For better clarity, the average
c, The output characteristics of the same device in b. From bottom to top, Ion (red pentagrams) and standard deviation (error bars) for each Lc are plotted
Vgs = −4 V to Vgs = 10 V with 2-V steps. The solid and dotted lines are the results of to avoid too many data points. f, Benchmark of the delay time of Sb (0112)–MoS2
the d.c. and pulse current–voltage measurements, respectively. d, Rc as a function FETs with Si planar FET, FinFET and GAAFET. All data for MoS2 FETs are evaluated
of n2D for monolayer MoS2 and other semiconductors in the literature (Methods). at Vds = 1 V. The blue and red solid lines represent the fitting of the experimental
The black solid line represents the quantum limit for Rc. e, Benchmark of data points for Si planar HP FET and Sb (0112)–MoS2 FETs, respectively. 
monolayer MoS2 FETs Ion at Vds = 1 V in the literature (Methods). The yellow, blue Gra, graphene; h-BN, hexagonal boron nitride; Vdd, operating voltage; Lg, gate
and pink ellipses represent the performance of high-performance (HP, solid length; UHF, ultra-high vacuum. The pink dotted line is the projected result
symbols) and low-power (LP, open symbols) Si planar FET, FinFET and GAAFET, assuming Rc = 42 Ω μm.

(Extended Data Fig. 3c,d). The STEM revealed a pristine and intimate 5.1 nm at n2D = 3 × 1013 cm−2, as shown in Fig. 3b. To our knowledge,
contact between the crystalline Sb and MoS2 without metal-induced both values are the lowest reported for TMD materials, and the Rc is
defects. Low-temperature photoluminescence spectroscopy also about three-times lower than the previous record using Bi (0001)7.
revealed no defect-related emission after Sb deposition (Extended It is noteworthy that the small LT ensures an ultra-compact device foot-
Data Fig. 3a). The vdW distance between the Sb and the sulfur (S) atoms print, which is critical for very-large-scale integration32. Moreover, the
was measured to be about 0.285 nm (Fig. 2d), which was smaller than small LT was further corroborated by the negligible on-state current
that in MoS2–indium4 and MoS2–Bi (0001)7 contacts. The high-resolution (Ion) drop for contact lengths down to about 60 nm (Extended Data
STEM image clearly resolved the atomic positions, in good agreement Fig. 4g–i). Furthermore, the variable-temperature measurements in
with the theoretical model in Fig. 1. The absence of Sb–S chemical bond Fig. 3c show the temperature-dependent Rc and intrinsic mobility μint
formation was confirmed by X-ray photoemission spectroscopy (after eliminating Rc) from 50 K to 400 K. The mobility showed a typical
(Extended Data Fig. 3b). We note that the Sb (0112) was formed only on phonon-limited behaviour (μ ≈ T−γ, where γ is the characteristic expo-
top of MoS2. Figure 2e presents an optical micrograph of patterned nent of phonon-limited transport and T is the temperature) at high
MoS2 on silicon dioxide (SiO2) substrate, covered by 20-nm Sb. The Sb temperature with γ = 1.8, and defect-limited saturation behaviour at
film showed a clear colour contrast on MoS2 and SiO2 owing to the low temperature33. Remarkably, the Rc was temperature independent
amorphous nature of SiO2 (Fig. 2b). in the entire range, suggesting that the cross-junction transport was
Next, we evaluated the electrical properties of the contact by measur- through tunnelling instead of thermionic emission14. This was strong
ing field-effect transistors (FETs) using MoS2 single crystals as the chan- evidence for ohmic contact. The ohmic nature was supported by two
nel31 (see Methods for fabrication and measurement details). Figure 3a additional evidences: (1) the extracted SBH was about −10 meV using
shows the transfer (drain-to-source current, Ids, versus gate-to-source the flat-band model7,18 (Extended Data Fig. 4a–c) and (2) the output
voltage, Vgs) characteristics of a transfer-length-method (TLM) device. (Ids versus drain-to-source voltage, Vds) curves maintained excellent
Using TLM, Rc and transfer length (LT) were derived to be 42 Ω μm and linearity at 50 K (Extended Data Fig. 4c).

Nature | Vol 613 | 12 January 2023 | 277

Article
a 10–4 b c i 0

Vth (V)
10–6 –2

Ids (A μm–1)
10–8 –4

10–10

SS (mV dec–1)
300
Lc = 0.1 μm Lc = 0.2 μm
10–12

10–4 d e 100

10–6 10

Ids (A μm–1)

log(Ion/Ioff)
10–8 8

10–10 6
1
Lc = 0.4 μm Lc = 0.6 μm

Ion (mA μm–1)

10–12

10–4 f g h
0.1
10–6
Ids (A μm–1)

Gm (μS μm–1)
10–8 100

10–10
10
Lc = 0.8 μm Lc = 1.0 μm Lc = 1.5 μm
10–12
–5 0 5 10 –5 0 5 10 –5 0 5 10 0.1 0.2 0.4 0.6 0.8 1.0 1.5
Vgs (V) Vgs (V) Vgs (V) Lc (μm)

Fig. 5 | Variability of Sb (0112)–MoS2 FETs. a, Top: optical microscope image of ratio, Gm and Ion. The colour of the symbols corresponds to the channel length
TLM arrays fabricated on HfO2/Si substrate. Scale bar, 400 μm. Bottom: in b–h. For each group of data, the boxplot statistics and Gaussian fitting were
magnified view of a TLM structure. Scale bar, 70 μm. b–h, Transfer performed and meaningful statistical data were obtained, including mean
characteristics of MoS2 FETs with channel lengths of 0.1 μm (b), 0.2 μm (c), (square symbols), standard deviation, lower quartile (25%), median (50%),
0.4 μm (d), 0.6 μm (e), 0.8 μm (f), 1.0 μm (g) and 1.5 μm (h) under Vds = 1 V. upper quartile (75%), interquartile range (25–75%), and maximum and minimum
i, Boxplots with Gaussian fitting of key device metrics, including Vth, SS, on/off (cross symbols).

To demonstrate the advantage of the Sb (0112) contact, we fabricated barrier lowering. More importantly, the device readily reached current
control TLM devices using an Sb (0001) contact (Fig. 2b). Figure 3d,e saturation at Vds = 1 V as evident from both transfer and output char-
shows the statistics of Rc and LT derived from 156 TLM devices (115 high acteristics. The Ion (1.23 mA μm−1) was 44.7% and 104% higher than
temperature and 41 ambient temperature). The average Rc of the Sb short-channel devices with Bi (0001) and Sb (0001) contacts, respec-
(0112) contact was 209 ± 100 Ω μm, which was 3.47-times lower than tively, under the same drain bias7,9. Both the saturation voltage and Ion
the Sb (0001) counterpart. A similar trend was observed for LT. The were unprecedented in TMDs owing to the improved contacts.
high confidence of Rc was supported by the standard deviation σ and TMD FETs are difficult to remove heat from, because the thin TMD
correlation coefficient R2 from the fitting process as well as the TLM and the vdW gaps around it are highly thermally resistive. As such,
structure extending to sub-100-nm channel length, Lc (Extended Data self-heating can lead to very high temperatures in the channel, degrad-
Fig. 5). We note that the improvement using the Sb (0112) contact was ing mobility and current, or leading to device failure. The high input
not due to the MoS2 quality, as supported by nearly the same average power (>1 mW μm−1) in direct-current (d.c.) measurements cause sig-
mobility for the two cases (Fig. 3f). In addition, variable-temperature nificant self-heating, as indicated by the negative differential resistance
measurements revealed a small but positive SBH of about 20 meV for at Vds > 1 V (Fig. 4c). To estimate the steady-state temperature distribu-
the Sb (0001) contact (Extended Data Fig. 4d–f), consistent with the tion of this device at its maximum power, we used finite-element-method
recent results from the Taiwan Semiconductor Manufacturing Com- simulations35. The temperature distribution along the drain–source
pany (TSMC)9. Owing to the improved contacts of Sb (0112), the FETs direction is shown in Extended Data Fig. 9. For plausible ranges of mate-
delivered an average 38%-higher Ion at 100-nm Lc (Extended Data Fig. 6). rial and interface thermal properties, we found peak temperature ris-
Besides electrical performance, the stability of the contact is also ing above ambient of 630 ± 150 K and 38 ± 20 K in the MoS2 channel
critical to applications. The stability of Bi is poor owing to its low melt- and the contacts, respectively. It is noted that the contacts remain
ing point below 300 °C, which posts severe limits in back-end processes. significantly cooler than MoS2, as they are relatively thick and thermally
In contrast, the melting point of Sb is above 600 °C. To this end, we conductive. The large uncertainties in the temperature are mainly
studied the device stability at 125 °C (which is the highest standard for owing to the unknown thermal boundary resistance between MoS2 and
electronic products34) in a nitrogen environment as shown in Extended Sb (0112), which is a subject of future research. To probe the intrinsic
Data Fig.  7. No obvious degradation of Ion, threshold voltage limits of device Ion, we used pulse measurements with 160-ns pulses
(Vth) and subthreshold swing (SS) was observed during 24 h for the Sb (Methods). At Vds = 1 V (1.5 V), the pulsed Ion was increased to 1.44 mA μm−1
contact. In contrast, the Ion of Bi-contact devices decreased by 41% (1.54 mA μm−1), accompanied by higher saturation voltage. The boost
under the same conditions. The excellent stability of the Sb (0112) con- of Ion under pulse measurements was attributed to the combined effects
tact also ensures that the devices can work reliably in the high-current of reduced self-heating and hot-electron charge trapping36.
regime, as shown in the following. We next benchmarked our device performance with existing TMDs
To fully determine the potential of the Sb (0112) contact, we fabri- and Si complementary metal–oxide–semiconductor (CMOS) tech-
cated scaled MoS2 FETs with sub-50-nm Lc (Fig. 4a). Figure 4b,c shows nologies. We first summarized the Rc of various materials in Fig. 4d.
the Ids–Vgs and Ids–Vds characteristics of a representative device with It was apparent that our Rc was not only one of the lowest among TMDs,
Lc ≈ 20 nm (more device data are shown in Extended Data Fig. 8). The but also lower than covalently bonded contacts in Si and compound
device showed excellent short-channel immunity, with an on/off ratio semiconductors3. At high n2D, the Rc was within a factor of two from the
over 108, an SS of about 180 mV dec−1 and negligible drain-induced quantum limit3. The reduced Rc led to a significant boost of Ion under

278 | Nature | Vol 613 | 12 January 2023

aggressively scaled Vds = 1 V (Fig. 4e). The Ion under d.c. and pulse con- 6. Liu, Y. et al. Approaching the Schottky–Mott limit in van der Waals metal-semiconductor
junctions. Nature 557, 696–700 (2018).
ditions was 44.7% and 68.8% higher than the previous record in mon- 7. Shen, P. C. et al. Ultralow contact resistance between semimetal and monolayer
olayer TMDs7. More importantly, our devices delivered similar Ion with semiconductors. Nature 593, 211–217 (2021).
high-performance planar Si CMOS and superior Ion than fin-shaped 8. English, C. D., Shine, G., Dorgan, V. E., Saraswat, K. C. & Pop, E. Improved contacts to MoS2
transistors by ultra-high vacuum metal deposition. Nano Lett. 16, 3824–3830 (2016).
FET (FinFET) and gate-all-round FET (GAAFET) technologies37,38, and 9. Chou., A.-S. et al. Antimony semimetal contact with enhanced thermal stability for
simultaneously showed immunity to short-channel effects owing to high performance 2D electronics. In 2021 IEEE International Electron Devices Meeting
the sub-1-nm channel thickness. We further benchmark with Si tech- 7.2.1–7.2.4 (IEEE, 2021); https://doi.org/10.1109/IEDM19574.2021.9720608.
10. O’Brien, K. P., Penumatcha, C. J. D. A. & Maxey, K. Advancing 2D monolayer CMOS
nologies by comparing the intrinsic gate delay (τ = CVds/Ion, where C is through contact, channel and interface engineering. In 2021 IEEE International Electron
the gate capacitance)39. As shown in Fig. 4f, the intrinsic gate delay is Devices Meeting 7.1.1–7.1.4 (IEEE, 2021); https://doi.org/10.1109/IEDM19574.2021.9720651.
74 fs for 20-nm MoS2 FETs, which is faster than Si planar devices and 11. Schulman, D. S., Arnold, A. J. & Das, S. Contact engineering for 2D materials and devices.
Chem. Soc. Rev. 47, 3037–3058 (2018).
FinFETs by 6.9 times and 15.7 times, respectively37–39. The solid red line 12. Das, S. et al. Transistors based on two-dimensional materials for future integrated circuits.
shows the fitting of experimental data using Rc = 240 Ω μm, projecting Nat. Electron. 4, 786–799 (2021).
an intrinsic gate delay of 20 fs at 5-nm Lc. This is an order of magnitude 13. Lanza, M., Smets, Q., Huyghebaert, C. & Li, L. J. Yield, variability, reliability, and stability of
two-dimensional materials based solid-state electronic devices. Nat. Commun. 11, 5689
lower than the International Roadmap for Devices and Systems (IRDS) (2020).
target for 202837. By using the lowest Rc = 42 Ω μm obtained in this work, 14. Sze, S. M., Li, Y. & Ng, K. K. Physics of Semiconductor Devices (Wiley, 2021).
we project an intrinsic delay of 3.6 fs at 5-nm Lc. 15. Landauer, R. Spatial variation of currents and fields due to localized scatterers in metallic
conduction. IBM J. Res. Dev. 1, 223–231 (1957).
To assess the suitability of Sb (0112) contacts for large-scale integra- 16. Jain, A. et al. One-dimensional edge contacts to a monolayer semiconductor. Nano Lett.
tion, we fabricated TLM arrays and examined the variability of key 19, 6914–6923 (2019).
device performances (Fig. 5). Figure 5b–h shows the Ids–Vgs character- 17. Cheng, Z. et al. Immunity to contact scaling in MoS2 transistors using in situ edge
contacts. Nano Lett. 19, 5077–5085 (2019).
istics of 573 FETs from 96 TLM structures with Lc ranging from 0.1 μm 18. Das, S., Chen, H. Y., Penumatcha, A. V. & Appenzeller, J. High performance multilayer
to 1.5 μm. The variability of Vth, SS, on/off ratio, Ion and transconductance MoS2 transistors with scandium contacts. Nano Lett. 13, 100–105 (2013).
(Gm) as a function of Lc are summarized in Fig. 5i and Extended Data 19. Jung, Y. et al. Transferred via contacts as a platform for ideal two-dimensional transistors.
Nat. Electron. 2, 187–194 (2019).
Table 1 (see Methods for details). On the basis of these data, several 20. Fang, H. et al. Degenerate n-doping of few-layer transition metal dichalcogenides by
important conclusions can be drawn. (1) We did not observe a signifi- potassium. Nano Lett. 13, 1991–1995 (2013).
21. McClellan, C. J., Yalon, E., Smithe, K. K. H., Suryavanshi, S. V. & Pop, E. High current
cant change of variability as a function of Lc. (2) Upon scaling to 100 nm,
density in monolayer MoS2 doped by AlOx. ACS Nano 15, 1587–1596 (2021).
the SS and on/off ratio did not degrade, but a slight negative shift of 22. Cui, X. et al. Low-temperature ohmic contact to monolayer MoS2 by van der Waals
Vth was observed owing to the short-channel effect (this was not obvi- bonded Co/h-BN electrodes. Nano Lett. 17, 4781–4786 (2017).
23. Wang, J. et al. High mobility MoS2 transistor with low Schottky barrier contact by using
ous for individual devices). However, the Vth shift could be reduced by
atomic thick h-BN as a tunneling layer. Adv. Mater. 28, 8302–8308 (2016).
scaling the effective oxide thickness together with Lc. (3) Both Ion and 24. Chou, A. S. et al. High on-state current in chemical vapor deposited monolayer MoS2
Gm were increased by nearly an order of magnitude upon scaling. The nFETs with Sn ohmic contacts. IEEE Electron Device Lett. 42, 272–275 (2021).
25. Kang, J. H., Liu, W., Sarkar, D., Jena, D. & Banerjee, K. Computational study of metal
boost of on-state performance from 200-nm Lc to 100-nm Lc suggested
contacts to monolayer transition-metal dichalcogenide semiconductors. Phys Rev. X 4,
that the devices were still not contact-limited, further substantiating 031005 (2014).
the low Rc. Extended Data Table 1 summarizes the variability data and 26. Louie, S. G. & Cohen, M. L. Electronic-structure of a metal–semiconductor interface.
Phys. Rev. B 13, 2461–2469 (1976).
compares with the existing literature on TMDs.
27. Qiu, H. et al. Hopping transport through defect-induced localized states in molybdenum
In conclusion, we overcame the inherent vdW gap and realized disulphide. Nat. Commun. 4, 2642 (2013).
near-quantum-limit electrical contact between MoS2 and semi-metallic 28. Jones, A. J. H. et al. Visualizing band structure hybridization and superlattice effects in
twisted MoS2/WS2 heterobilayers. 2D Mater. https://doi.org/10.1088/2053-1583/ac3feb
Sb (0112) by hybridization of M–S energy bands at the Fermi energy.
(2021).
The Rc surpassed that of covalently bonded M–S contacts and reached 29. Ji, J. et al. Two-dimensional antimonene single crystals grown by van der Waals epitaxy.
as low as 42 Ω μm. Short-channel MoS2 FETs delivered a high Ion of Nat. Commun. 7, 13352 (2016).
30. Chen, H. A. et al. Single-crystal antimonene films prepared by molecular beam epitaxy:
1.23 mA μm−1 (1.54 mA μm−1) under d.c. (pulse) measurements, an on/ selective growth and contact resistance reduction of the 2D material heterostructure.
off ratio over 108 and an intrinsic delay of 74 fs. These performances ACS Appl. Mater. Interfaces 10, 15058–15064 (2018).
outperformed equivalent Si CMOS technologies and satisfied the 2028 31. Li, T. et al. Epitaxial growth of wafer-scale molybdenum disulfide semiconductor single
crystals on sapphire. Nat. Nanotechnol. 16, 1201–1207 (2021).
roadmap target. Besides n-type MoS2, we also demonstrated a low Rc 32. Smets, Q. et al. Ultra-scaled MOCVD MoS2 MOSFETs with 42 nm contact pitch and
in ambipolar tungsten diselenide (WSe2) devices (Extended Data 250 µA/µm drain current. In 2019 IEEE International Electron Devices Meeting 23.2.1–23.2.4
Fig. 10), suggesting that Sb (0112) may be a universal contact technol- (IEEE, 2019); https://doi.org/10.1109/IEDM19573.2019.8993650.
33. Cui, X. et al. Multi-terminal transport measurements of MoS2 using a van der Waals
ogy for TMD-based electronics beyond Si. heterostructure device platform. Nat. Nanotechnol. 10, 534–540 (2015).
34. Commercial and Industrial-Grade Products White Paper CTWP011 (Cactus Technology,
2019).
Online content 35. Wu, R. et al. Filling the gap: thermal properties and device applications of graphene.
Sci. China Inf. Sci. 64, 140401 (2021).
Any methods, additional references, Nature Portfolio reporting summa- 36. Nathawat, J. et al. Transient hot-carrier dynamics and intrinsic velocity saturation in
ries, source data, extended data, supplementary information, acknowl- monolayer MoS2. Phys. Rev. Mater. 4, 014002 (2020).
37. International Roadmap for Devices and Systems (IRDS, 2020); https://irds.ieee.org/
edgements, peer review information; details of author contributions 38. International Technology Roadmap for Semiconductors (ITRS, 2015); https://www.
and competing interests; and statements of data and code availability semiconductors.org/resources/2015-international-technology-roadmap-for-
are available at https://doi.org/10.1038/s41586-022-05431-4. semiconductors-itrs/
39. Chau, R. et al. Benchmarking nanotechnology for high-performance and low-power logic
transistor applications. IEEE Trans. Nanotechnol 4, 153–158 (2005).
1. Akinwande, D. et al. Graphene and two-dimensional materials for silicon technology.
Nature 573, 507–518 (2019).
2. Ahmed, Z. et al. Introducing 2D-FETs in device scaling roadmap using DTCO. In 2020 IEEE Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
International Electron Devices Meeting 22.5.1–22.5.4 (IEEE, 2020); https://doi.org/10.1109/ published maps and institutional affiliations.
Iedm13553.2020.9371906.
3. Wang, Y. & Chhowalla, M. Making clean electrical contacts on 2D transition metal Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this
dichalcogenides. Nat. Rev. Phys. 4, 101–112 (2022). article under a publishing agreement with the author(s) or other rightsholder(s); author
4. Wang, Y. et al. Van der Waals contacts between three-dimensional metals and self-archiving of the accepted manuscript version of this article is solely governed by the
two-dimensional semiconductors. Nature 568, 70–74 (2019). terms of such publishing agreement and applicable law.
5. Kappera, R. et al. Phase-engineered low-resistance contacts for ultrathin MoS2 transistors.
Nat. Mater. 13, 1128–1134 (2014). © The Author(s), under exclusive licence to Springer Nature Limited 2023

Nature | Vol 613 | 12 January 2023 | 279

Article
Methods an FEI Helios 600i dual-beam focused ion beam (FIB) system. A tung-
sten protection layer was in situ deposited before FIB bombardment,
Chemical-vapour-deposition growth and transfer of monolayer followed by etching the surrounding area to form the lamella using a
MoS2 30-kV acceleration voltage and a 21-nA gallium ion beam. The target
We grew single-crystal monolayer MoS2 on A-axis c-plane (C/A) sapphire lamella was exfoliated from the original substrate and transferred to
(0001) substrates with miscut angle αA ranging from 0.2° to 1° that had TEM half grid inside the FIB chamber. HAADF-STEM characterization
undergone 1,000 °C annealing for 4 h as described in our previous was acquired on a FEI Titan G2 60–300 aberration-corrected S/TEM
work31. The transfer of MoS2 was done by a polymer-based transfer microscope with an acceleration voltage of 300 kV.
method. Poly(methyl methacrylate) (PMMA; Allresist A4, 950K) was
spin-coated on the MoS2/sapphire substrate at 2,000 r.p.m. After baking DFT calculations
at 150 °C to evaporate solvent completely, a thermal release tape (TRT) First-principles DFT calculations were performed by using the Vienna
was laminated and the PMMA/MoS2 was gently peeled off the sapphire Ab initio Simulation Package (VASP 5.4)49 with the exchange-correlation
substrate in deionized water. The TRT/PMMA/MoS2 stack was trans- functional based on Perdew–Burke–Ernzerhof 50 generalized gradient
ferred onto the target substrate and baked at 120 °C to release the TRT. approximation. The projected augmented wave potentials was adopted
to describe the interactions between valence electrons and ion cores51.
Chemical-vapour-deposition growth of monolayer WSe2 The Grimme’s DFT-D3 scheme was employed to include the vdW inter-
We grew monolayer WSe2 on 100-nm SiNx/Si substrate using a actions52. A slab model stacked with monolayer TMDs and three-layer
chemical-vapour-deposition tube furnace with a reverse-flow reac- semi-metal (Sb or Bi) (0001)/(0112 ) plane was used to simulate the
tor. An alumina boat loaded with WSe2 (Alfa Aesar, 99.98%) powder MoS2–Sb contact. The supercells of the semi-metal (0001)/(0112) sur-
was placed into the heating zone of a quartz tube, and the substrate faces and TMDs were carefully built to ensure a lattice mismatch smaller
was placed on the alumina boat and placed in the variable-temperature than 3%, where a vacuum space of >15 Å was added to separate images
zone. The chemical-vapour-deposition growth process was carried along c direction of the supercells. All structures were fully relaxed until
out at 1,180 °C under ambient pressure with a reverse argon gas flow the maximum force component acting on each atom was less than
of 80 standard cubic centimetres per minute. Once the desired growth 0.02 eV Å−1 with an energy convergence criterion of 10−4 eV and an kinetic
temperature was reached, the chemical vapour source was carried energy cut-off of 400 eV. Considering the very large systems simulated
downstream under a forward argon gas flow of 80 standard cubic cen- in this work, a k-spacing of <0.04 Å−1 was used for the structural relaxa-
timetres for a growth period of 4 min. After the growth, the furnace tion of metal (Sb (0001), Sb (0112), Bi(0001) and Bi (0112))–TMD (MoS2,
was cooled naturally. MoSe2, WS2 and WSe2) contacts. For the static electronic structure cal-
culations, the Brillouin zone integrations were sampled on a k-spacing
Transistor fabrication and measurement of <0.01 Å−1 for semi-metal–TMD contacts, respectively. The dipole
Monolayer MoS2 film was transferred onto hafnium oxide (HfO2)/highly correction along the c axis of supercell was considered in all DFT cal-
doped Si substrate using the PMMA-assisted transfer followed by chan- culations as well. The electrostatic potential profiles of Sb (0112)–MoS2
nel isolation through electron beam lithography and dry etching by and Sb (0001)–MoS2 contacts along the vertical direction were calcu-
O2. After removing PMMA by acetone and N-methyl-2-pyrrolidone lated using Poisson’s equation. The real-space band diagrams and the
(NMP), electron beam lithography was used to define the source and LDDOSs of the device composed of monolayer MoS2 and double-ended
drain contacts followed by electron beam evaporation of 20-nm Sb (0112) electrodes without doping (zero bias) and with electron dop-
Sb/30-nm Au and lift-off. We controlled the substrate temperature ing (similar to high gate bias) were simulated by using the DFT method.
during electron beam evaporation through infrared heating and adjust-
ing the cooling water flow rate. The substrate temperature was meas-
ured using the surface temperature indicating strips. Notably, the Sb Data availability
(0112) contact did not require any thermal annealing after device fab- Source data are provided with this paper.
rication, which made this technology back-end-of-line compatible with
CMOS. WSe2 FETs were fabricated on in-situ SiNx/Si substrate by the 40. Guimaraes, M. H. et al. Atomically thin ohmic edge contacts between two-dimensional
same methods without transfer process.  materials. ACS Nano 10, 6392–6399 (2016).
Electrical measurements were carried out by Keithley 4200 semicon- 41. Smithe, K. K. H., English, C. D., Suryavanshi, S. V. & Pop, E. Intrinsic electrical transport and
performance projections of synthetic monolayer MoS2 devices. 2D Mater. https://doi.org/
ductor parameter analyser in a closed-cycle cryogenic probe station 10.1088/2053-1583/4/1/011009 (2016).
with a base pressure of about 10−6 torr. Pulse testing was performed 42. Smithe, K. K. H., Suryavanshi, S. V., Munoz Rojo, M., Tedjarati, A. D. & Pop, E. Low variability
using a Keithley 4200 pulse measurement unit in conjunction with in synthetic monolayer MoS2 devices. ACS Nano 11, 8456–8463 (2017).
43. Zhu, Y. et al. Monolayer molybdenum disulfide transistors with single-atom-thick gates.
model 4225-RPM remote amplifier. During the pulse measurement, Nano Lett. 18, 3807–3813 (2018).
a synchronous pulse voltage sequence was applied to the drain and 44. Kumar, A. et al. Sub-200 Ω·μm alloyed contacts to synthetic monolayer MoS2. In 2021
gate electrodes, the minimum pulse width can be as low as 160 ns IEEE International Electron Devices Meeting 7.3.1–7.3.4 (IEEE, 2021); https://doi.org/10.1109/
IEDM19574.2021.9720609.
and the voltage duty cycle was 10 times. To reduce parasitic effects 45. Somvanshi, D. et al. Nature of carrier injection in metal/2D-semiconductor interface and
during pulse measurement, all the cables were shortened as much its implications for the limits of contact resistance. Phys. Rev. B 96, 205423 (2017).
as possible. The data on the MoS2 FETs shown in Fig. 4d,e are from 46. Daus, A. et al. High-performance flexible nanoscale transistors based on transition metal
dichalcogenides. Nat. Electron. 4, 495–501 (2021).
refs. 3,4,7,9,10,17,21,22,24,31,37,38,40–48. 47. Lembke, D. & Kis, A. Breakdown of high-performance monolayer MoS2 transistors. ACS
Nano 6, 10070–10075 (2012).
Characterization 48. Sanne, A. et al. Radio frequency transistors and circuits based on CVD MoS2. Nano Lett.
15, 5039–5045 (2015).
XRD was performed on a Bruker D8 Discover system. AFM was per- 49. Kresse, G. & Furthmuller, J. Efficient iterative schemes for ab initio total-energy
formed by an Asylum Cypher S system. Raman and photoluminescence calculations using a plane-wave basis set. Phys. Rev. B 54, 11169–11186 (1996).
50. Perdew, J. P., Burke, K. & Ernzerhof, M. Generalized gradient approximation made simple.
spectroscopy were carried out by a home-built system equipped with
Phys. Rev. Lett. 77, 3865–3868 (1996).
488-nm laser excitation and Princeton Instruments SP-2500 spectrome- 51. Blochl, P. E. Projector augmented-wave method. Phys. Rev. B 50, 17953–17979 (1994).
ter. X-ray photoemission spectroscopy (PHI 5000 VersaProbe Scanning 52. Jonchiere, R., Seitsonen, A. P., Ferlat, G., Saitta, A. M. & Vuilleumier, R. Van der Waals
effects in ab initio water at ambient and supercritical conditions. J. Chem. Phys. 135,
XPS Microprobe with a monochromatized Al Kα excitation source) was
154503 (2011).
performed with calibration to the chemical state of C–C of adventitious 53. Sebastian, A., Pendurthi, R., Choudhury, T. H., Redwing, J. M. & Das, S. Benchmarking
carbon at 284.8 eV. Cross-section TEM samples were fabricated using monolayer MoS2 and WS2 field-effect transistors. Nat. Commun. 12, 693 (2021).
Acknowledgements This work was supported by the National Key R&D Program of China analysis. Ç.K. and E.P. performed the thermal analysis. S.G., P.W., T.X. and L.S. performed
(grant numbers 2021YFA0715600, 2022YFB4400100, 2021YFA1500700, the TEM and data analysis. Wenfeng Wang, L.L. and T.L. performed the chemical-vapour-
2017YFA0204800 and 2021YFA1202903); the Leading-edge Technology Program of Jiangsu deposition growth of MoS2. J. Li and X.D. performed the chemical-vapour-deposition growth
Natural Science Foundation (grant numbers BK20202005 and BK20222007); the National of WSe2. Wenhui Wang, J. Lu and Z.N. performed the Raman characterization and data analysis.
Natural Science Foundation of China (grant numbers T2221003, 61927808, 61734003, W.L., Z.Y., L.M., J.W. and X.W. co-wrote the manuscript with input from other authors. All authors
61851401, 91964202, 61861166001, 51861145202, 22033002, 62204113, 62204124, 22222302, contributed to discussions.
11774153 and 11874199); the China Postdoctoral Science Foundation (grant numbers
2022M711549 and 2022T15036); Jiangsu Funding Program for Excellent Postdoctoral Talent Competing interests The authors declare no competing interests.
(grant number 20220ZB63); the Natural Science Foundation of Jiangsu Province (grant
number BK20220773); the Strategic Priority Research Program of Chinese Academy of Additional information
Sciences (grant number XDB30000000); the Key Laboratory of Advanced Photonic and Supplementary information The online version contains supplementary material available at
Electronic Materials, Collaborative Innovation Center of Solid-State Lighting and Energy-Saving https://doi.org/10.1038/s41586-022-05431-4.
Electronics, and the Fundamental Research Funds for the Central Universities, China. Correspondence and requests for materials should be addressed to Yi Shi, Jinlan Wang or
Xinran Wang.
Author contributions X.W. and Y.S. conceived and supervised the project. W.L., Z.Y., H.N., D.F., Peer review information Nature thanks Abhinandan Borah, Hyeon-Jin Shin and the other,
Y.X., X.T. and H.Q. contributed to the transistor fabrication, measurements and data analysis. anonymous, reviewer(s) for their contribution to the peer review of this work.
X.G., L.M. and J.W. performed the DFT calculations. W.S. and Y.N. performed the XRD and data Reprints and permissions information is available at http://www.nature.com/reprints.
Article

Extended Data Fig. 1 | Electronic properties of Sb/Bi-MoS2 contacts. (0112)-MoS2 (c) and Sb (0001)-MoS2 (d) contacts along the vertical direction,
a, b, The simulated band diagram and local device density of states (LDDOS) of respectively. e, f, Atomic-projected electronic band structures of Bi (0112)-
the monolayer MoS2 with double-end Sb (0112) electrodes: without electron MoS2 (e) and Bi (0001)-MoS2 (f) contacts. g, The interfacial vdW interaction
doping (zero gate bias) (a) and with 1.5 electron doping (similar to high gate between MoS2 and Bi (0112)/Bi (0001). h, The charge transfer from Bi (0112)/Bi
bias) (b). The Fermi level is set to zero.c, d, Electrostatic potential profiles of Sb (0001) to MoS2 by Bader charge analysis.
Extended Data Fig. 2 | The atomic-projected electronic band structures (0112) (a), WS2-Sb (0112) (c), WSe2- Sb (0112) (e), MoSe2-Sb (0001) (b), WS2-Sb
and the charge density near EF of TMDs-Sb contact. a–f, Left panels, (0001) (d) and WSe2-Sb (0001) (f) contacts. The corresponding isosurface level
atomic-projected electronic band structures (left panel) of MoSe2-Sb (0112) (a), of the charge density near EF (right panel) are 1.3×10 −4 e/Bohr3 for MoSe2-Sb
WS2-Sb (0112) (c), WSe2- Sb (0112) (e), MoSe2-Sb (0001) (b), WS2-Sb (0001) (d) and contact (a, b), 8×10 −5 e/Bohr3 for WS2-Sb (c, d) and WSe2-Sb (e, f) contact,
WSe2-Sb (0001) (f) contacts. Right panels, charge density near EF of MoSe2-Sb respectively.
Article

Extended Data Fig. 3 | Optical, chemical and structural characterization of substrate. b, High-resolution XPS spectra of 2 nm Sb deposited on monolayer
Sb-MoS2 interface. a, The Low-temperature (6 k) PL spectra of. After MoS2/sapphire, where the absence of Sb–S bond signal from Sb2S3 indicates no
depositing Sb film, the main exciton peak did not show widening, and no chemical bond formation between Sb and MoS2. c, Cross-section HAADF-STEM
obvious defect-related emissions were observed. This proved that the image of Sb (0001)-MoS2 contact. Scale bar, 1 nm. d, Zoom-in atomic-resolution
deposition of Sb was friendly to 2D materials without creating defects. image from c. The interplane distance was 0.379 nm. Scale bar, 1 nm.
The sharp peak at 1.79 eV was the photoluminescence signal from the sapphire
Extended Data Fig. 4 | Schottky barrier extraction and small footprint MoS2 various gate bias of the same device in d. f, Gate voltage dependence of the
FETs. a, The temperature-dependent Ids-Vgs transfer curves of Sb (0112)-MoS2 barrier height. The deviation from the linear trend (red solid line) defines the
FET. Vds = 0.1V. b, The Arrhenius plot at various gate bias of the same device in a. flat band voltage and shows positive Schottky barrier height. g, Transfer curves
c, Gate voltage dependence of the barrier height. The deviation from the linear of two MoS2 FETs with the same Lc = 60 nm but with contact length (Lcontact) of
trend (red solid line) defines the flat band voltage and shows negative Schottky 60 nm and 1 μm. Vds = 0.1 V. Inset is the false-colour SEM image of the device.
barrier height. Inset shows the linear output curves of the same FET at a low Scale bar, 200 nm. h–i, Output curves of the MoS2 FET with contact length of
temperature of 50 K with excellent linearity. d, The temperature-dependent 1 μm (h) and 60 nm (i), respectively.
Ids-Vgs transfer curves of Sb (0001)-MoS2 FET. Vds = 0.1V. e, The Arrhenius plot at
Article

Extended Data Fig. 5 | Extraction of low Rc from TLM devices with Sb (0112) Lc equal to 60 nm, 80 nm, 200 nm, 400 nm, 600 nm, 800 nm, 1 μm and 1.5 μm.
-MoS2 . a–e, The left is the transfer curves of TLM devices. Vds = 0.1V, and the Ultralow Rc extraction from the TLM devices with sub-100-nm Lc devices
right is the plot of Rtot versus Lc from the left devices, from which the 2Rc can be further enhances confidence in ohmic contact of Sb (0112) -MoS2 contact.
extracted from the y-axis intercepts. Symbols are experimental data and lines f, g, Histogram of σ (a) and R 2 (b) from the linear fitting process of TLM method
are linear fits at the right figure. It is noted that TLM in e contain 9 devices with corresponding to the results of Sb (0112)-MoS2 contact (115 TLMs) in Fig. 3d and e.
Extended Data Fig. 6 | Comparison of MoS2 FETs with Sb (0001) and Sb (0112) The mean (square symbols), lower quartile (Q1, 25%), median (Q2, 50%), upper
contacts. a, Transfer characteristics of 145 MoS2 FETs (65 with Sb (0001) quartile (Q3, 75%), interquartile range (25%-75%) and maximum/minimum
contact (black lines) and 80 with Sb (0112) contact (red lines)). Lc = 100 nm, (cross symbols) are presented. The Ion of Sb (0112) contact is significantly
Vds = 1V. b, The boxplot with Gaussian fitting of Ion at the same carrier density. improved compared with Sb (0001) contact due to the improvement of Rc.
Article

Extended Data Fig. 7 | Stability of Sb (0112)-MoS2 contact. a, Thermal stability 10 V with 2 V step. e, Transfer characteristics of a typical Bi-contact MoS2 FET
of Ion for Sb and Bi contact measured at different time in 125 °C nitrogen measured at different time in 125 °C nitrogen environments. Lc = 100 nm,
environment. b, Transfer characteristics of a typical Sb-contact MoS2 FET Vds = 1V. f, g, The output characteristics of the same device at 125 °C in the
measured Lc = 100 nm, Vds = 1V. c, d, The output characteristics of the same initial state (f) and after 24 h (g). From bottom to up, Vgs = −2 V to 10 V with
device in the initial state (c) and after 24 h (d). From bottom to up, Vgs = −2 V to 2 V step.
Extended Data Fig. 8 | Short-channel MoS2 FETs with Sb (0112)-contact. measurements, respectively. c, Transfer characteristics of a MoS2 FET with
a, Transfer characteristics of a MoS2 FET with Lc = 40 nm under Vds = 0.2 V and 1 V. Lc = 20 nm under Vds = 0.2 V and 1 V. Inset shows the corresponding SEM image.
Inset shows the corresponding SEM image. Scale bar, 500 nm. b, The output Scale bar, 500 nm. d, The output characteristics of the same devices in c. From
characteristics of the same devices in a. From bottom to up, Vgs = −2 V to 10 V bottom to up, Vgs = −2 V to 10 V with 2 V step. The solid and dotted lines are the
with 2 V step. The solid and dotted lines are the results of the DC and pulse I-V results of the DC and pulse I-V measurements, respectively.
Article

Extended Data Fig. 9 | The steady-state temperature distribution of the contacts) serving as the symmetry plane. A contact resistance of 100 Ω·μm is
short-channel MoS2 FET in Fig. 4b, c by finite-element method simulation. assumed in order to determine the fraction of the power dissipated at the
a, Steady-state distribution of temperature rises across the centre cross- contacts. b, The temperature rise above ambient in MoS2 (black) and the
section of the device along the current direction, for Vds = 1 V and Ids = 1.23 mA/μm. contact just above MoS2 (red) plotted as a function of position along the
Only half of the geometry is shown, x = 0 (halfway between the drain and source direction of current flow.
Extended Data Fig. 10 | Ambipolar monolayer WSe2 FETs with Sb (0112) with 4 V step with the same devices at c. e, Transfer curves of p-type TLM
contact. a, Transfer curves of n-type TLM devices with Sb (0112) contact. devices with Sb (0112) contact. Vds = −0.1 V. f, Plot of Rtot versus Lc from p-type
Vds = 0.1 V. b, Plot of Rtot versus Lc from n-type TLM devices with Sb (0112) contact TLM devices with Sb (0112) contact (red symbols) and Ti contact (blue symbols),
(red symbols) and Ti contact (blue symbols), from which the 2Rc can be from which the 2Rc can be extracted from the y-axis intercepts. Symbols are
extracted from the y-axis intercepts. Symbols are experimental data and lines experimental data and lines are linear fits at f. g, h, Performance comparison of
are linear fits at b. c, d, Performance comparison of n-type monolayer WSe2 p-type monolayer WSe2 FETs with Sb (0112) contact (red) and Ti contact (blue).
FETs with Sb (0112) contact (red) and Ti contact (blue). Lc = 100 nm. c shows the Lc = 100 nm. g shows the transfer curves. Vds = −0.1 V. h shows the output curves
transfer curves at Vds = 0.1 V. d shows the output curves at Vgs from 0 V to 40 V at Vgs from 0 V to −50 V with −5 V step with the same devices at g.
Article
Extended Data Table 1 | Variability data analysis and benchmark42,53

Vth is the threshold voltage of the device, which is extracted by using the linear extrapolation method. SS is the subthreshold swing extracted from the device in the subthreshold region. Ion
is the drain current when the device is in the ‘on’ state. log(Ion/Ioff) is the logarithm value of the ratio of the drain currents when the device is in the ‘on’ and ‘off’ states, respectively. Gm is the
transconductance value of the device at Vds = 1 V. Missing values are represented by spaces. The results for Vth, SS, log(Ion/Ioff), Ion, Gm and Rc are all obtained by statistical analysis of many devices
(#devices is the statistical number of the devices) and presented as “mean ± standard deviation”.