Articles

https://doi.org/10.1038/s41563-020-0631-x

Atomically thin half-van der Waals metals enabled

by confinement heteroepitaxy
Natalie Briggs   1,2,3,14, Brian Bersch1,2,14, Yuanxi Wang   2,3,14, Jue Jiang4, Roland J. Koch   5,6, Nadire Nayir3,7,
Ke Wang8, Marek Kolmer   9, Wonhee Ko9, Ana De La Fuente Duran1, Shruti Subramanian1,2,
Chengye Dong1,2, Jeffrey Shallenberger8, Mingming Fu9, Qiang Zou9, Ya-Wen Chuang4, Zheng Gai   9,
An-Ping Li   9, Aaron Bostwick5, Chris Jozwiak   5, Cui-Zu Chang   4, Eli Rotenberg   5, Jun Zhu2,4,
Adri C. T. van Duin1,3,7,8,10,11,12, Vincent Crespi2,3,4,8 and Joshua A. Robinson   1,2,3,8,13 ✉

Atomically thin two-dimensional (2D) metals may be key ingredients in next-generation quantum and optoelectronic devices.
However, 2D metals must be stabilized against environmental degradation and integrated into heterostructure devices at
the wafer scale. The high-energy interface between silicon carbide and epitaxial graphene provides an intriguing framework
for stabilizing a diverse range of 2D metals. Here we demonstrate large-area, environmentally stable, single-crystal 2D gal-
lium, indium and tin that are stabilized at the interface of epitaxial graphene and silicon carbide. The 2D metals are covalently
bonded to SiC below but present a non-bonded interface to the graphene overlayer; that is, they are ‘half van der Waals’ metals
with strong internal gradients in bonding character. These non-centrosymmetric 2D metals offer compelling opportunities for
superconducting devices, topological phenomena and advanced optoelectronic properties. For example, the reported 2D Ga is a
superconductor that combines six strongly coupled Ga-derived electron pockets with a large nearly free-electron Fermi surface
that closely approaches the Dirac points of the graphene overlayer.

M
ajor advances in fundamental science have resulted from services: it only partially passivates the SiC surface underneath,
the exfoliation, stacking and encapsulation of atomically thereby sustaining the high-energy interface that drives intercala-
thin 2D layers1. The next step towards a technological tion; it lowers the energy of the (otherwise exposed) upper surface
impact of 2D layers and heterostructures is to transition sophisti- of the metal, thus facilitating 2D morphologies; and it protects the
cated ‘pick and place’ devices to a wafer-scale platform. However, newly formed 2D metal from environmental degradation after
the sensitivity of 2D systems to interfacial reactions and environ- intercalation through in  situ healing of the graphene defects. The
mental influences—especially for 2D metals or small-gap semi- stability of these 2D metals in air over a period of months greatly
conductors—poses challenges for large-scale integration. Very few facilitates ex situ characterization and enables facile processing and
metals resist degradation of their top few atomic layers upon envi- device fabrication. Competing methods to obtain thin, single-crys-
ronmental exposure, and for a 2D metal, these layers constitute the tal metal films are specific to certain metallic species (for example
entire system. A general platform for producing environmentally Au (ref. 3) or TiN (ref. 4)) and generally produce thicker layers of a
stable and wafer-scale 2D metals that are not prone to interfacial few nanometres without in situ encapsulation, whereas CHet pro-
interactions would represent a significant advance. Inspired by the duces stable, atomically thin crystalline 2D forms of diverse metals
success of wide-bandgap 2D gallium nitride2, we focus on the metal that do not otherwise exist in nature.
alone and demonstrate a platform dubbed confinement heteroepi- Unlike traditional EG intercalation methods5–8, CHet (Fig. 1a)
taxy (CHet), where the interface between epitaxial graphene (EG) uses plasma-treated EG and high-pressure (300 Torr) thermal evapo-
and silicon carbide (SiC) stabilizes crystalline 2D forms of group ration to create continuous films of crystalline 2D metals. Nominally
III (Ga, In) and group IV (Sn) elements. Defect engineering of monolayer EG is grown by silicon sublimation from 6H-SiC (0001)
the graphene overlayer enables uniform, large-area intercalation (ref. 9), after which exposure to an oxygen plasma generates defects
at the high-energy SiC/EG interface; this interface then templates in the EG layers. Metallic precursors such as Ga, Sn and In situated
intercalant crystallization at a thermodynamically defined num- in a crucible directly beneath the EG are then heated with the EG/
ber of atomic layers. The unreactive nature of as-grown EG on SiC SiC to 700–800 °C. The vaporized metal diffuses through EG defects
(graphene plus buffer layer) results in the EG performing multiple to reach the EG/SiC interface. X-ray photoelectron spectroscopy

Department of Materials Science and Engineering, The Pennsylvania State University, University Park, PA, USA. 2Center for 2-Dimensional and Layered
1

Materials, The Pennsylvania State University, University Park, PA, USA. 32-Dimensional Crystal Consortium, The Pennsylvania State University, University
Park, PA, USA. 4Department of Physics, The Pennsylvania State University, University Park, PA, USA. 5Advanced Light Source, Lawrence Berkeley National
Laboratory, Berkeley, California, USA. 6The Molecular Foundry, Lawrence Berkeley National Laboratory, Berkeley, California, USA. 7Department of
Mechanical Engineering, The Pennsylvania State University, University Park, PA, USA. 8Materials Research Institute, The Pennsylvania State University,
University Park, PA, USA. 9Center for Nanophase Materials Sciences, Oak Ridge National Laboratory, Oak Ridge, TN, USA. 10Department of Chemistry,
The Pennsylvania State University, University Park, PA, USA. 11Department of Engineering Science and Mechanics, The Pennsylvania State University,
University Park, PA, USA. 12Department of Chemical Engineering, The Pennsylvania State University, University Park, PA, USA. 13Center for Atomically Thin
Multifunctional Coatings, The Pennsylvania State University, University Park, PA, USA. 14These authors contributed equally: Natalie Briggs, Brian Bersch,
Yuanxi Wang. ✉e-mail: [email protected]

Nature Materials | www.nature.com/naturematerials

Articles Nature Materials

a Time
metal intercalation (Ga, In or Sn) releases the buffer layer5–7 and
removes carbon–oxygen bonding. XPS signatures remain stable
Plasma
even after more than eight months in air (Supplementary Fig. 2).
Intercalation also yields a pronounced (~×5) decrease in ID relative
Elemental
metal

to plasma-treated EG, and a ×3–4 and ×1–2 increase in IG and I2D,

Si
sublimation
respectively, relative to as-grown EG (Fig. 1c). The decreased ID,
lack of D′ and D + G modes, loss of carbon–oxygen bonding and
air stability of the resulting 2D metals all suggest that the EG heals
Graphene formation Defect generation Metal intercalation during intercalation at elevated temperatures13–15 and subsequently
b 5 c 12
acts as a barrier to oxidation post-synthesis. The strengthening of
Ga-intercalated
Gr G IG and I2D after intercalation may be attributed to metal/EG charge
4 SiC
10
2D
transfer and/or plasmon resonance16–18. A comparison of Auger
D In-intercalated electron spectroscopy (AES; Fig. 1d and Supplementary Fig. 3) and
Counts per second (×104)

8
O2/He plasma-treated EG Raman mapping (Supplementary Fig. 4) of EG/Ga samples pre-
Counts (×104)

3
Buffer Gr SiC
6
Ga-intercalated
pared using pristine versus plasma-treated EG reveals that plasma
treatments greatly increase the lateral coverage of intercalated Ga.
C–OH
C–O–C
2 C=O Sn-intercalated

As-grown EG Gr
4
O2/He plasma- Beyond a simple increase in intercalation sites, first-principles cal-
1
Buffer SiC
2
treated EG
culations indicate that Ga/EG bonding is strengthened with increas-
As-grown EG
ing vacancy size, where large, unpassivated defects bond covalently
0
290 288 286 284 282 280
0
1,400 1,800 2,400 2,800
with Ga (Fig. 1e and Supplementary Fig. 5). First-principles calcula-
Binding energy (eV) Wave number (cm–1) tions suggest that the termination of multivacancy graphene defects
d As-grown EG Plasma-treated EG
with oxygen moieties facilitates metal adhesion and transport to the
EG EG
underlying EG/SiC interface through intermediate metal binding
/SiC /S
iC
EG/Ga/ EG/Ga/
energies that are weaker than those of unpassivated vacancies but
stronger than those for hydroxyl terminations or the basal plane
SiC SiC

EG
/Ga EG
/Ga
(Fig. 1e and Supplementary Fig. 5).
/SiC
Two-dimensional Ga, In and Sn are one to three atomic layers
10 µm /SiC 10 µm

e Bare defect C–O–C C=O C–OH

thick and registered to the SiC substrate (Fig. 2a,c,e). The dominant
Gr–Ga
bonding: –7.4 eV –2.1 eV –3.4 eV –1.4 eV
thickness observed can be explained by first-principles equilibrium-
phase stability calculations (Fig. 2b,d,f and Supplementary Fig. 6)
that predict the layer number as a function of the metal chemical
H
O
potential. Based on first-principles calculations, the strong metal–
C SiC bonding stabilizes a 2D metal compared to the bulk 3D form,
Ga
but this advantage is diluted by successive thickening of the inter-
calant so that the bulk phase is preferred beyond a certain thickness.
Fig. 1 | CHet with defect-engineered epitaxial graphene. a, Schematic
These calculations predict a stability range of one to three layers for
of CHet showing EG growth, O2/He plasma treatment and intercalation
Ga, one to two layers for In and one layer for Sn, all in encourag-
steps. b,c, XPS showing C 1s for as-grown EG (bottom), O2/He plasma-
ingly close agreement with scanning transmission electron micros-
treated EG (middle), and Ga-intercalated, O2/He plasma-treated EG (top)
copy (STEM) images (Fig. 2a,c,e), which display three layers of Ga,
demonstrating the creation and annihilation of C–O bonds during CHet; this
two of In and one to two layers of Sn (in the case of Sn, the blurring
is confirmed by c, Raman spectroscopy of as-grown EG, O2/He plasma-
of the second Sn layer indicates a lower structural stability for this
treated EG and metal-intercalated, O2/He plasma-treated EG, where the
layer). This close correspondence suggests near-equilibrium growth
defect peak (D peak) intensity is dramatically reduced as a result of the
conditions during CHet, that is, a relatively high metal atom mobil-
intercalation process. SiC Raman spectra are subtracted from the Raman
ity, which is supported by the uniform metal coverage achieved. The
data shown in c. The Gr label in b corresponds to the blue fitted peak, which
high energy of the initial EG/SiC interface, which follows from the
denotes graphene layers in the C 1s spectra. d, SEM and accompanying
ineffectiveness of EG in terminating SiC dangling bonds, provides a
Ga AES maps of Ga-intercalated, as-grown EG and Ga-intercalated,
strong thermodynamic driving force for 2D metal formation, sug-
O2/He plasma-treated EG. Defects formed due to plasma treatments lead
gesting a broader applicability of CHet to support additional species
to significantly improved intercalation uniformity. The AES colour scales
as 2D metals. The metal thicknesses shown in Fig. 2a,c,e are uni-
show low (dark) to high (light) Ga signal across the mapped regions.
form across the terraces of SiC, although some SiC terraces show
e, DFT modelling of Ga atoms on graphene sheets containing unpassivated
uniform two-layer Ga.
and passivated defects. DFT models of Ga on graphene sheets containing
An epitaxial relationship of the 2D metals to the SiC substrate
defects passivated with C–O–C, C=O, and C–OH suggest that oxygen
is further supported by the observation in low-energy electron dif-
termination provides favourable energies for metal attraction and
fraction (LEED) of EG and SiC patterns, without distinct spots cor-
intercalation through the graphene sheet. The Ga binding energy to each
responding to a structurally unique intercalant layer (Fig. 2g and
defect is shown in each model.
Supplementary Fig. 8). The SiC diffraction pattern is shared by
the metal. The 2.72 Å lateral Ga spacing observed via STEM also
closely matches that of SiC (Fig. 2a and Supplementary Fig. 9).
(XPS; Fig. 1b) and Raman spectroscopy (Fig. 1c) reveal how the EG First-principles structural optimization provides further insights
defects evolve during CHet. O2/He plasma treatments create defects into the epitaxial relationship of Ga to SiC (0001). Structural relax-
in the EG that contain C=O and C–OH or C–O–C bonds visible ations were performed for one to three layers of Ga initialized at
in the C 1s spectra (Fig. 1b)10,11; these defects correlate with a ×15 sites projecting onto the silicon, carbon and/or hollow sites of SiC
increase in the Raman D/G intensity ratio (ID/IG; Fig. 1c)12. Upon (Fig. 2h, Supplementary Figs. 10,11 and Supplementary Table 1).
metal intercalation, signatures of carbon–oxygen bonding and the Adding a top bilayer graphene affects only the band fillings without
EG buffer layer peak disappear from the C 1s region (Fig. 1b), and changing the relative stability of the Ga structures (Supplementary
a metallic Ga 3d peak appears (Supplementary Fig. 1). Ultimately, Table 1), therefore bilayer graphene is not considered in these

Nature Materials | www.nature.com/naturematerials

Nature Materials Articles
a b g j
1
EG/Ga/SiC

Bulk Ga
SiO2
3
Two-layer EG 0
1 EG
2 Bulk Ga
SiC ep
Ga
2 stable st
–1 C
1 3 Si
2D
SiC stable
1 nm –2 10 nm
–4.0 –3.5 –3.0 –2.5 165 eV

c d h k
1 SiC top view

Bulk In
Silicon

Energy (eV)
Three-layer EG 0 Carbon
1 Hollow
In In
–1 2
SiC step
SiC
1 nm –2 Ga C Si 5 nm
–3.5 –3.0 –2.5 –2.0

e f 1
i l
Ga/SiC side view

Bulk Sn
SiO2
0
Two-layer EG
1
Sn –1 Ga

SiC –2
1 nm 10 nm SiC step
–5.0 –4.5 –4.0 –3.5
µGa, µIn or µSn (eV)

Fig. 2 | Atomic structure of CHet-grown 2D metals. a, Cross-sectional STEM showing three layers of Ga between EG and SiC. The inset shows the
different interlayer spacings of the Ga layers. b, Corresponding energy minimization calculations showing preferred Ga layer numbers. c–f, Cross-sectional
STEM and layer-number calculations for In and Sn. EDS (Supplementary Fig. 7) confirms that intercalant layers match the respective precursor elements
and are not oxidized. The grey, vertical lines in b, d and f indicate the metal chemical potential of the bulk metal, and the red, green and blue lines indicate
the energy as a function of chemical potential for one, two and three layers, respectively. The lowest line at a given potential indicates the layer number
of the ground state. g, Acquired LEED pattern for EG/Ga/SiC indicating the presence of EG and SiC with no additional spots for a unique Ga structure
(or In or Sn structure; see Supplementary Fig. 8), indicating the Ga is lattice matched to EG or SiC. h,i, Top-down schematic (h) of hexagonal SiC with
silicon, carbon and hollow sites labelled; intercalated Ga layers exhibit an ABC stacking over the SiC substrate (i). j–l, Step bunching in the SiC surface
can lead to discontinuities in the 2D metals, necessitating that SiC steps be only a few atoms in height to maintain the structural and electrical continuity
of the 2D metal films.

structural stability calculations. The ground-state locations for Ga (assuming a free-electron model)20 or In on Si(111) (ref. 21) and
the first, second and third layers of Ga are above the silicon (GaSi), substantially larger than the band velocity of graphitic π states22,23.
carbon (GaC) and hollow (Gahollow) sites, respectively (Fig. 2i). This The calculated band structure (Fig. 3b) shows colour-coded pro-
ABC stacking resembles a face-centred cubic (fcc) lattice cleaved jections of the total wavefunction onto the plane-wave components
along the (111) plane, matching the hexagonal arrangement of of the graphene (black) and Ga/SiC (blue) primitive cell, where
SiC (0001), and is possibly related to the metastable distorted fcc bands are unfolded from the supercell Brillouin zone24. The most
phase of Ga(iii) (ref. 19). The Ga registry weakens for increased prominent features are three avoided crossings between s and p Ga
metal thickness: the GaSi stacking site for single-layer Ga is 0.14 eV bands, one along Γ-MGa and two along Γ-KGa (comparing with a
per unit cell (that is, per 8.30 Å2), more stable than that of GaC or freestanding frozen bilayer of Ga (Supplementary Fig. 13), the Ga
Gahollow, but the ground states of Ga multilayers are preferred by only s-band starts ~9 eV below the Fermi level). A large Ga-derived near-
0.05 eV against competing phases. Cross-sectional STEM (Fig. 2a, free-electron-like Fermi surface (Fig. 3c,d) closely approaches the
inset) supports the calculated evolution with increasing thickness, graphene Dirac points at Kgr and K′gr with further electron pockets
where the interlayer spacing between the first and second Ga lay- around the corners of the Ga Brillouin zone. The position of the gra-
ers (2.19 Å) is significantly smaller than that between the second phene Dirac point 0.2–0.3 eV below the Fermi level (Supplementary
and third (2.36 Å). Density functional theory (DFT) results show Fig. 14) indicates that Ga (or In, Supplementary Fig. 12) electron-
a pronounced shift from largely covalent bonding in the bottom dopes graphene by 8–10 × 1012 cm–2. This is supported by the Fermi
Ga layer (templated against SiC) to comparably strong but more surface of bilayer Ga/SiC calculated without graphene (Fig. 3e)
metallic bonding between the upper Ga layers and much weaker where the Fermi level is upshifted by 0.4 eV to be consistent with
interaction between the upper Ga layer and the graphene over- the measured ARPES band alignment (see Supplementary Fig. 11);
layer (Supplementary Table 1). Experimentally, epitaxy and metal this artificial upshift is not needed when using hybrid functionals,
continuity are interrupted by large steps in the SiC surface which generally yield more accurate band alignments (see
(Fig. 2j), while smaller steps (Fig. 2k) are less disruptive and single- Supplementary Information for details on the supercell strain
atom steps (Fig. 2l) are metal-continuous and appear to maintain imposed in the calculations). Finally, the measured Fermi surface
epitaxy across the step. clearly shows that the graphene Brillouin zone is rotated 30° from
The structural quality of 2D Ga is reflected in the excellent the underlying Ga/SiC zone, providing direct evidence of the orien-
agreement of angle-resolved photoemission spectroscopy (ARPES) tational epitaxy between Ga and SiC.
measurements with calculations of the structurally optimized epi- The air stability of 2D metals formed by CHet enables ex  situ
taxial system (Fig. 3a, also Supplementary Fig. 12 for 2D In). The transport measurements without complex capping methodolo-
high valence electron count of p-block metals and the small lat- gies. A co-linear four-point-probe configuration with In dot con-
eral unit cell of epitaxial 2D Ga yield an exceptionally large s-band tacts reveals a largely metallic behaviour for 2D Ga on small-step
Fermi velocity of 2 × 106 m s–1, comparable to that of bulk Al and SiC from room temperature down to 4 K, below which a sharp,

Nature Materials | www.nature.com/naturematerials

Articles Nature Materials

a 0 c d
Ga
Ga
Energy (eV)

–1 4 /
Hybridized s and p BZSiC
Gra

Ga orbitals
phe

–2
Г
ne

3
–3
Kgr Г Г KGa MGa

Ga
b Mgr Kgr Mgr
Г Г 2 Ga Dirac
4
Gr BZ
3 ky
e
Ga/SiC BZ
Kgr MGa
2 1
KGa

1
Energy (eV)

Ga/SiC BZ
0 0

–1

–2 –1

–3

–4 –2 –1 0 1
MGa Г Г KGa MGa
kx

Fig. 3 | Electronic structure of CHet-grown 2D Ga. a,b, Measured ARPES spectra for EG/Ga/SiC showing EG and Ga bands near MGa and KGa
(a) and calculated band structure in black and blue overlaid with measured ARPES (purple) (b). The effective unfolded band structures of 2 × 2
graphene + √3 × √3 R30º bilayer Ga/SiC along the Γ-MGa (Γ-Kgr) and the Γ-KGa (Γ-Mgr) directions are projected onto graphene (black) and Ga/SiC (blue),
as indicated by the Brillouin zone (BZ) paths in the inset (Mgr in the left panel and MGa on the right are from repeated zones). ARPES measurements along
the same paths are superimposed. c,d, ARPES-measured Fermi surface showing nearly free-electron-like circular contours, where kx and ky are the electron
crystal momenta in the in-plane directions. Superimposed black, white and purple circles in d correspond to nearly free-electron-like circular contours from
Ga and are drawn to aid in the comparison of the experimental data with calculated data. e, DFT-calculated Fermi surface of bilayer Ga/SiC with the Fermi
level shifted (to the blue dashed line in panel b) to match the measured band filling.

four-order-of-magnitude drop in resistance occurs (Fig. 4a). As transition temperature30 of TBKT = 3.1 K (2.9 K) for transport paral-
defined in the Methods section, Tconset = 3.95 K and Tczero = 3.2 K, val- lel (perpendicular) to the SiC steps (Supplementary Fig. 17), indi-
ues that are higher than those in bulk orthorhombic α-Ga (1.08 K)25, cating a largely isotropic 2D superconducting transition in these
similar to those of metastable monoclinic β-Ga (5.9–6.5 K) and samples. Note that I–V measurements up to the critical current (Ic)
below those of metastable amorphous Ga (8.4 K)26–28. As expected, are limited by the experimental setup, and TBKT is likely closer to
the transition drops and broadens with increasing perpendicu- 3.88 K (Supplementary Fig. 17) if extracted from curves closer to
lar magnetic field B⊥ (Fig. 4b). Resistance (R) versus B⊥ indicates Ic. Full R(T) and R(B) curves for ‘large-step’ 2D Ga are provided in
Bc = 130 mT at 2 K with a corresponding coherence length ξ0 ~50 nm Supplementary Fig. 18. Epitaxial graphene synthesis must limit step
(Fig. 4c and Methods). A linear extrapolation of Bc(T) from R(B) bunching to ensure uniform superconducting films with isotropic
suggests a zero-Kelvin critical field Bc0 ≈ 260 mT and ξ0 ~36 nm, transport. SiC step height and bunching can be mitigated by reduc-
higher than that of α-Ga (Bc0 ≈ 6 mT)25 and β-Ga (Bc0 ≈ 54 mT ing the temperature of the pre-EG-growth H2 etch step31, or through
(ref. 28); Supplementary Fig. 15a). Considering the very high Fermi the delivery of carbon to the SiC surface during EG growth to pro-
velocity and moderate Tc, the system is likely not in the clean limit. mote buffer layer nucleation and suppress mass transport across SiC
Scanning tunnelling microscopy/spectroscopy show quite uniform that leads to step formation32,33.
differential tunnelling conductance (dI/dV) spectra at multiple loca- When considering the electronic origin of the superconductivity
tions on the sample (Fig. 4d,f and Supplementary Fig. 15). A super- in 2D Ga/SiC, each of the constituent materials appear to be poor
conducting gap with well-defined coherence peaks at ±0.6 meV is superconductors: SiC is semiconducting, EG is not as heavily doped
seen at 2.2 K (Fig. 4h) and disappears at ~3 K, below the transport Tc (Supplementary Fig. 14) as in graphite intercalation compound
(Fig. 4g), perhaps due to a moderately reduced proximity-induced superconductors, and 2D Ga exhibits nearly free-electron charac-
superconducting order parameter in the two- to three-layer EG teristics that would not lead to a substantial Tc, since electron–lat-
overlayer29. The observation of a significant superconducting order tice interactions are weak in nearly free-electron metals. We show,
parameter at the upper EG surface bodes well for the formation of using first-principles calculations, that another Fermi surface mani-
superconducting heterostructures via the proximity effect. fold—small electron pockets around K—supports superconductiv-
When crystallographic steps on SiC (0001) are less than 1 nm ity. Using DFT, we calculate the electronic densities of states (DOSs)
tall (Fig. 2k,l), transport measurements that orient the current par- near the Fermi level for α-Ga, β-Ga and one- to three-layer 2D Ga
allel or perpendicular to the steps have similar Tconset (~4 K) and (Fig. 5a,b) and also the Eliashberg spectral function α2F(ω) for
Tczero (~3.2 K) values (Fig. 4i). Although steps more than 5 m tall three-layer GaSiGaCGaC epitaxial to SiC without the graphene cap
(Figs. 2j,4j) yield a similar Tconset of ~3.8 K for both directions, only (Methods and Supplementary Fig. 19). α-Ga contains Ga dimers
parallel transport displays a fully developed superconducting tran- whose incipient covalency suppresses the DOS near the Fermi level,
sition with a Tczero of ~2.5 K, suggesting that transport perpendicu- while β-Ga recovers a more nearly free-electron-like behaviour
lar to large steps encounters a finite series resistance at the steps. (Fig. 5a). Two-layer and three-layer Ga/SiC (Fig. 5b) exhibit a DOS
Small-step 2D Ga exhibits a Berezinskii–Kosterlitz–Thouless (BKT) at EF similar to that of β-Ga. Figure 5c compares α2F(ω) and the

Nature Materials | www.nature.com/naturematerials

Nature Materials Articles
a 80 b c

60 60

60
100

10
40 40
40
R (Ω)

R (Ω)
R (Ω)
1

0.1 0T
20 0.02 T 20 2K
20 0.01 0.04 T 2.5 K
0.06 T 2.9 K
∆Tc 0.08 T 3.3 K
0.001
2 3 4 5 0.10 T 3.7 K
T (K) 0.12 T 4.1 K
0 0 0
0.14 T 4.5 K

0 50 100 150 200 250 300 3 4 5 –0.2 –0.1 0.0 0.1 0.2
T (K) T (K) B (T)

d f g h i 120 j
Defect T = 2.2 K 3.0 K T = 2.2 K Parallel 320 Parallel
Perpendicular Perpendicular
280
0.10 T 100 240
2.9 K
Normalized dI /dV (a.u.)

200
2.8 K 0.05 T 80 1 nm 10 nm
80
dI /dV (a.u.)
dI /dV (a.u.)

10 nm 2.7 K 0.03 T

R (Ω)
R (Ω)
60 1 µm 1 µm
e 2.6 K
–10 nm
0.02 T –1 nm
2.5 K 40 40
2.4 K 0.01 T
20
2.3 K 0T

2 nm 0 0
–2 –1 0 1 2 –2 –1 0 1 2 –2 –1 0 1 2 2 3 4 2 3 4
Sample bias Sample bias Sample bias T (K)
T (K)
(mV) (mV) (mV)

Fig. 4 | Superconductivity in 2D Ga grown via CHet. a, Zero-field R(T) curve for an optimized 2D Ga film (O2/He plasma-treated EG, small-step, parallel
configuration) from 300 K to 2 K. Inset of a, log-scale plot from 5 K to 2 K of the same curve. b, R(T) curves showing a suppression in Tc and increase in
residual resistance at 2 K with increasing out-of-plane magnetic field. c, Resistance versus out-of-plane magnetic field (R(B)) curves showing a similar
suppression in the superconducting state with increasing temperature. The kink may be attributed to two superconducting phases, possibly from discrete
two-layer Ga and three-layer Ga regions, step-edge interactions or distinct superconducting transitions within a three-layer Ga region. Data shown in a–c
is from the same sample and measurement configuration. d,e, Large-area and atomic-resolution scanning tunnelling microscopy images of the EG surface
topography, respectively. f, Differential conductance (dI/dV) spectra (normalized to 3.0 K spectra) taken at 2.2 K for the three different regions shown in d.
g,h, Temperature-dependent (zero-field) and perpendicular magnetic-field-dependent (2.2 K) dI/dV spectra, respectively, of a region on the right terrace
in d, marked by the red dot. i,j, R(T) plots comparing perpendicular and parallel current directions performed on small-step and large-step samples (both
using O2/He plasma-treated EG), respectively. Black arrows indicate approximate Tczero values, and dashed lines are meant to aid the eye (T = 4 K line).
Insets, corresponding EG/SiC step-edge morphology and height measured by atomic force microscopy.

cumulative electron–phonon coupling strength λ(ω) to the pro- character. A set of six pockets is indeed clearly seen in the ARPES
jected phonon DOS for the three individual Ga layers and the inter- measurement in Fig. 3c, 30º rotated from the EG Dirac points; these
facial Si atom. The dominant contributions to the overall λ = 1.62 correspond to pockets near K in Fig. 3b,e, and are also similar in size
come from Ga vibrations below 120 cm–1, with modest evidence to the pockets in the more accurate hybrid-functional-calculated
near 180 cm–1 for distinct coupling channels specific to the (more- bands of Supplementary Fig. 11 (see discussions on Supplementary
covalent) lower layers of Ga and little sign of involvement from the Fig. 21 for details on the exact character of the pocket). The domi-
interfacial Si. This result contrasts with previous reports attributing nant contribution from these pockets is reassuring, since such
2D superconductivity to interfacial bonding34. Figure 5d shows the strong coupling would not be expected from the more free-electron-
momentum-resolved electron–phonon coupling35 (λk) across the like sheets of the Fermi surface. Using the McMillan–Allen–Dynes
Ga/SiC Brillouin zone for electronic states within ±0.5 eV of the formula36,37 with λ = 1.62 and μ* from 0.1 to 0.15 yields a Tc of 3.5 to
Fermi energy (Supplementary Fig. 19 provides a similar plot for a 4.1 K, in good agreement with experiment. Literature reports of the
narrower energy range). Large contributions to λk come from elec- electron–phonon coupling strength in β-Ga suggest that it is also a
tron pockets near the two symmetry-inequivalent K points. Further reasonably strongly coupled superconductor38, unlike more weakly
decomposition of λk=K by phonon wavevectors shows that the coupled α-Ga (ref. 36).
strongest contributions arise from intervalley scattering between The large λ = 1.62 derived from the Ga states alone suggests that
the two K valleys (see discussion on Supplementary Figs. 20,21). the EG layers are unlikely to be the driver of superconductivity in
Interestingly, this reciprocal-space structure to the coupling is simi- CHet-derived 2D Ga. ARPES (Fig. 3) provides further evidence
lar to that of MoS2, but here we have a p-block metal (with a large towards this conclusion: the Fermi level is only 0.2–0.3 eV above
companion near-circular nearly free-electron Fermi surface) with a the EG Dirac point, corresponding to n ≈ 8–10 × 1012 cm–2, which is
small but crucial Fermi surface manifold of doped semiconductor ×10–100 lower than in superconducting Li-doped or Ca-intercalated

Nature Materials | www.nature.com/naturematerials

Articles Nature Materials

a 1.0 b Online content

Any methods, additional references, Nature Research reporting
DOS (states per eV per atom)

0.8 summaries, source data, extended data, supplementary informa-

tion, acknowledgements, peer review information; details of author
0.6 contributions and competing interests; and statements of data and
code availability are available at https://doi.org/10.1038/s41563-
0.4 020-0631-x.

0.2
Received: 26 May 2019; Accepted: 3 February 2020;
β-Ga
One-layer Ga/SiC
Two-layer Ga/SiC
Published: xx xx xxxx
α-Ga Three-layer Ga/SiC
0
–2 –1 0 1 2 –1.0 –0.50 0 0.5 1.0 References
Energy (eV) Energy (eV) 1. Rhodes, D., Chae, S. H., Ribeiro-Palau, R. & Hone, J. Disorder in van der
Waals heterostructures of 2D materials. Nat. Mater. 18, 541–549 (2019).
c d 2. Al Balushi, Z. Y. et al. Two-dimensional gallium nitride realized via graphene
0.5
← → 2 encapsulation. Nat. Mater. 15, 1166–1171 (2016).
α 2F 1 λ 3. Maniyara, R. A. et al. Tunable plasmons in ultrathin metal films. Nat. Photon.
λk
0 0 13, 328–333 (2019).
4 4. Shah, D., Reddy, H., Kinsey, N., Shalaev, V. M. & Boltasseva, A. Optical
Projected phonon DOS

Gatop
properties of plasmonic ultrathin TiN films. Adv. Opt. Mater. 5,
3
Γ 1700065 (2017).
Gamid
2 5. Riedl, C., Coletti, C. & Starke, U. Structural and electronic properties of
epitaxial graphene on SiC (0001): A review of growth, characterization,
Gainter 1
K’ transfer doping and hydrogen intercalation. J. Phys. D Appl. Phys. 43,
374009 (2010).
Siinter M 6. Emtsev, K. V., Zakharov, A. A., Coletti, C., Forti, S. & Starke, U. Ambipolar
0 100 200 300 400 500 K doping in quasifree epitaxial graphene on SiC (0001) controlled by Ge
intercalation. Phys. Rev. B 84, 125423 (2011).
Phonon frequency (cm–1)
7. Gierz, I. et al. Electronic decoupling of an epitaxial graphene monolayer by
gold intercalation. Phys. Rev. B 81, 235408 (2010).
Fig. 5 | Theoretical calculations on heterostructures of graphene and 8. Virojanadara, C., Watcharinyanon, S., Zakharov, A. A. & Johansson, L. I.
2D Ga. a, Electronic DOS versus energy DFT calculations for the two bulk Epitaxial graphene on 6H-SiC and Li intercalation. Phys. Rev. 82,
phases of Ga including the stable α-Ga (low Tc) and the metastable β-Ga 205402 (2010).
9. Subramanian, S. et al. Properties of synthetic epitaxial graphene/molybdenum
(high Tc) phases. b, DOS versus energy calculations for one- to three-layer
disulfide lateral heterostructures. Carbon 125, 551–556 (2017).
Ga/SiC. c, The Eliashberg spectral function α2F(ω) (red) compared with 10. Moulder, J. F. & Chastain, J. Handbook of X-ray Photoelectron Spectroscopy: a
the projected phonon DOS of the three types of Ga atoms in three-layer Reference Book of Standard Spectra for Identification and Interpretation of XPS
Ga (shades of green) and the top Si atoms at the interface (blue). The Data (Perkin-Elmer, 1992).
cumulative electron–phonon coupling strength λ(ω) is superimposed in 11. Beamson, G. & Briggs, D. High Resolution XPS of Organic Polymers: the
Scienta ESCA300 Database (Wiley, 1992).
black at the top of c. d, Momentum-resolved electron–phonon coupling λk
12. Eckmann, A. et al. Probing the nature of defects in graphene by Raman
shows that the dominant contribution to coupling strength λ comes from spectroscopy. Nano Lett. 12, 3925–3930 (2012).
the electron pockets near the K and Kʹ points in the Ga Brillouin zone. The 13. Vishwakarma, R. et al. Transfer free graphene growth on SiO2 substrate at
colour bar shows the range of λk values depicted in the image. Graphene is 250 °C. Sci. Rep. 7, 43756 (2017).
not included in calculations. 14. Araby, M. I. et al. Graphene formation at 150 °C using indium as catalyst.
RSC Adv. 7, 47353–47356 (2017).
15. Fujita, J. et al. Near room temperature chemical vapor deposition of graphene
with diluted methane and molten gallium catalyst. Sci. Rep. 7, 12371 (2017).
EG39,40. In those cases, superconductivity is attributed to a partially 16. Yi, C. et al. Evidence of plasmonic coupling in gallium nanoparticles/
graphene/SiC. Small 8, 2721–2730 (2012).
filled band near the Γ point at a much higher level of charge transfer 17. Losurdo, M. et al. Demonstrating the capability of the high-performance
into EG than is observed in Ga-intercalated EG. Superconductivity plasmonic gallium-graphene couple. ACS Nano 8, 3031–3041 (2014).
in low-angle twisted bilayer graphene41 likely exhibits a different 18. Khorasaninejad, M. et al. Highly enhanced Raman scattering of graphene
origin from that in EG/Ga, wherein the required interlayer twist is using plasmonic nano-structure. Sci. Rep. 3, 2936 (2013).
not present within the EG itself, although the environmental stabil- 19. Voloshina, E., Rosciszewski, K. & Paulus, B. First-principles study of the
connection between structure and electronic properties of gallium. Phys. Rev.
ity of 2D Ga coupled with the controllable thickness of EG (one- to B 79, 045113 (2009).
three-layer post-intercalation) suggests prospects in mechanically 20. Ashcroft, N. W. & Mermin, N. D. Solid State Physics (Holt, Rinehart and
stacking an additional (twisted) graphene monolayer onto EG and Winston, 1976).
2D Ga to create one of many hybrid superconducting systems possi- 21. Yoshizawa, S., Kim, H., Hasegawa, Y. & Uchihashi, T. Disorder-induced
ble based on the platform of highly stable CHet-derived 2D metals. suppression of superconductivity in the Si(111)-(7 × 3)-In surface: scanning
tunneling microscopy study. Phys. Rev. B 92, 041410 (2015).
CHet stabilizes 2D forms of 3D metals. The overlying graphene 22. Yang, L., Deslippe, J., Park, C.-H., Cohen, M. L. & Louie, S. G. Excitonic
layers utilized in CHet not only help confine the 2D metals, but also effects on the optical response of graphene and bilayer graphene. Phys. Rev.
serve as a seal to prevent the oxidation of ultrathin non-noble met- Lett. 103, 186802 (2009).
als, as revealed by the ability to perform extensive ex situ charac- 23. Zhang, Y., Tan, Y.-W., Stormer, H. L. & Kim, P. Experimental observation
terization of these materials. Whereas previously Au was arguably of the quantum Hall effect and Berry’s phase in graphene. Nature 438,
201–204 (2005).
the only environmentally stable elemental metal, the realization of !
24. Popescu, V. & Zunger, A. Extracting E versus k effective band structure
air-stable 2D heterostructures containing single-crystal, elemental I
from supercell calculations on alloys and impurities. Phys. Rev. B 85,
metals and superconductors at the interface of EG and SiC opens 085201 (2012).
the door to stabilizing diverse 2D allotropes of 3D metals and their 25. Gregory, W. D., Sheahen, T. P. & Cochran, J. F. Superconducting
alloys across the periodic table with potentially novel properties, transition and critical field of pure gallium single crystals. Phys. Rev. 150,
315–321 (1966).
all of which are candidates for incorporation into advanced multi- 26. Chen, T. T., Chen, J. T., Leslie, J. D. & Smith, H. J. T. Phonon spectrum of
component heterostructures for next-generation quantum42, pho- superconducting amorphous bismuth and gallium by electron tunneling.
tonic43 and electronic applications. Phys. Rev. Lett. 22, 526–530 (1969).

Nature Materials | www.nature.com/naturematerials

Nature Materials Articles
27. Wühl, H., Jackson, J. E. & Briscoe, C. V. Superconducting tunneling in the 37. Allen, P. B. & Dynes, R. C. Transition temperature of strong-coupled
low-temperature phases of gallium. Phys. Rev. Lett. 20, 1496–1499 (1968). superconductors reanalyzed. Phys. Rev. B 12, 905–922 (1975).
28. Parr, H. & Feder, J. Superconductivity in β-phase gallium. Phys. Rev. B 7, 38. Garno, J. P. Simple high vacuum evaporation system with low-temperature
166–181 (1973). substrate. Rev. Sci. Instrum. 49, 1218–1220 (1978).
29. Werthamer, N. R. Theory of the superconducting transition temperature 39. Ludbrook, B. M. et al. Evidence for superconductivity in Li-decorated
and energy gap function of superposed metal films. Phys. Rev. 132, monolayer graphene. Proc. Natl Acad. Sci. USA 112,
2440–2445 (1963). 11795–11799 (2015).
30. Reyren, N. et al. Superconducting interfaces between insulating oxides. 40. Ichinokura, S., Sugawara, K., Takayama, A., Takahashi, T. & Hasegawa, S.
Science 317, 1196–1199 (2007). Superconducting calcium-intercalated bilayer graphene. ACS Nano 10,
31. Oliveira, M. H. Jr, Schumann, T., Ramsteiner, M., Lopes, J. M. J. & Riechert, 2761–2765 (2016).
H. Influence of the silicon carbide surface morphology on the epitaxial 41. Cao, Y. et al. Unconventional superconductivity in magic-angle graphene
graphene formation. Appl. Phys. Lett. 99, 111901 (2011). superlattices. Nature 556, 43–50 (2018).
32. Kruskopf, M. et al. A morphology study on the epitaxial growth of graphene 42. Fu, L. & Kane, C. L. Superconducting proximity effect and Majorana
and its buffer layer. Thin Solid Films 659, 7–15 (2018). fermions at the surface of a topological insulator. Phys. Rev. Lett. 100,
33. Kruskopf, M. et al. Comeback of epitaxial graphene for electronics: large-area 096407 (2008).
growth of bilayer-free graphene on SiC. 2D Mater. 3, 041002 (2016). 43. Boltasseva, A. & Shalaev, V. M. Transdimensional photonics. ACS Photon. 6,
34. Zhang, T. et al. Superconductivity in one-atomic-layer metal films grown on 1–3 (2019).
Si(111). Nat. Phys. 6, 104–108 (2010).
35. Margine, E. R. & Giustino, F. Anisotropic Migdal-Eliashberg theory using
Wannier functions. Phys. Rev. B 87, 024505 (2013). Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
36. McMillan, W. L. Transition temperature of strong-coupled superconductors. published maps and institutional affiliations.
Phys. Rev. 167, 331–344 (1968). © The Author(s), under exclusive licence to Springer Nature Limited 2020

Nature Materials | www.nature.com/naturematerials

Articles Nature Materials

Methods width as seen in the log plot inset in Fig. 4a. Tc(0.5RN) is defined as the temperature
EG synthesis and plasma treatment. EG is synthesized via silicon sublimation at which the sample reaches half of its normal resistance. Tczero is defined as the
from the (0001) plane of semi-insulating 6H-SiC (II-VI Inc.) at 1,800 °C, 700 Torr temperature at which resistance effectively reaches a zero-resistance state (the
Ar, for 15 min (ref. 9). EG layers were plasma treated using a Tepla M4L plasma noise floor of the PPMS system, ~0.01 Ω). The transition width ∆Tc is defined as
etch tool, using 150 sccm O2 and 50 sccm He under a pressure of 500 mTorr and the change in temperature between the Tconset and Tczero values. Various Tc values are
power of 50 W for 60 s. provided to help in comparisons with other works in the literature, which may use
different values. Critical field Bc2(0.9RN) is defined as the magnetic field at which
Two-dimensional metal intercalation. Metal intercalation was performed using the sample reaches 90% of its normal resistance. Coherence length is estimated
from Bc2 ðT Þ ¼ 2πξ 2 where Φ is the flux quantum.
Φ
an STF-1200 horizontal tube furnace fitted with a 1 inch outer diameter quartz 0
tube. A custom-made alumina crucible from Robocasting Enterprises was used to I
hold 1 × 1 cm EG/SiC substrates, which were placed with EG layers on the Si face Scanning tunnelling spectroscopy/microscopy. Ga-intercalated EG/SiC was
of SiC facing downwards, towards the inside of the crucible. Then, 30–60 mg of studied using an ultrahigh-vacuum low-temperature scanning tunnelling
metallic Ga (Sigma Aldrich, 99.999%), In powder (Alfa Aesar, −325 mesh, 99.99%) microscope with an in situ out-of-plane magnetic field at the Center for Nanophase
or Sn granules (Alfa Aesar, 99.5%) were placed in the crucible directly beneath the Materials Sciences at Oak Ridge National Laboratory. The sample was preheated
EG/SiC substrate. The crucible with EG/SiC and the respective metal precursor to 200 °C to remove surface adsorbates at ultrahigh vacuum with a base pressure
was then loaded into the tube furnace and evacuated to ~5 mTorr. The tube was of 2×10–10 Torr before transferring in situ to the scanning tunnelling microscope
then pressurized to 300–700 Torr with Ar. At this time, the furnace was heated to stage. Scanning tunnelling microscopy or spectroscopy was conducted using a
700–800 °C under a ramp rate of 20° min–1 and Ar flow of 50 sccm. The furnace was mechanically cut Pt–Ir tip. All Pt–Ir tips were conditioned and checked using
held at the growth temperature for 30 min, then cooled to room temperature. a clean Au (111) surface before each measurement. Topographic images were
acquired in constant-current mode with the bias voltage applied to the samples.
XPS. XPS measurements were carried out with a Physical Electronics Versa Probe All the spectroscopies were obtained using the lock-in technique with the bias
II equipped with a monochromatic Al Kα X-ray source (hν = 1,486.7 eV) and a modulation at 973 Hz. The scanning tunnelling microscopy image in Fig. 4d
concentric hemispherical analyser. High-resolution spectra were obtained over an was taken at Vb = 10 mV and It = 400 pA. The image in Fig. 4e was taken at
analysis area of 200 μm at a pass energy of 29.35 eV for C 1s, Si 2p, Ga 3d and Ga 2p Vb = −100 mV, and It = 100 pA. The dI/dV spectra in Fig. 4f–h were measured at
regions. O 1s regions were collected with a pass energy of 46.95 eV. The acquired Vb = 5 mV, It = 400 pA and ΔV = 0.1 mV.
spectra were fitted Lorentzian line shapes, and the asymmetric graphene peak fit
was derived from exfoliated highly oriented pyrolytic graphite and H-intercalated Theory.
EG reference samples. Spectra were charge referenced to this graphene peak i. Graphene defect generation/passivation and Ga adsorption: All DFT calcula-
in C 1s corresponding to 284.5 eV. A U 2 Tougaard background was used to fit tions investigating the role of plasma treatment on graphene defects and Ga
XPS spectra. intercalation were performed in Quantum Espresso44, using projected aug-
mented wave (PAW) pseudopotentials45,46 and the Perdew–Burke–Ernzerhof
Raman spectroscopy. Raman Spectroscopy was performed with a Horiba LabRam parametrization of the generalized gradient approximation exchange-correla-
Raman system using a wavelength of 488 nm and a power of 4.6 mW. Spectra were tion functional (GGA-PBE, ref. 47,48). A 5 × 5 × 1 Γ-centred k-point mesh was
acquired with an integration time of 30 s, using a grating with 600 grooves mm–1. applied for Brillouin zone integration. Plane-wave expansions were truncated
at an energy cut-off of 408 eV for wavefunctions and at 4,080 eV for charge
Cross-sectional STEM. Cross-sectional samples for STEM imaging were prepared densities. The Marzari–Vanderbilt cold smearing scheme49 was applied with
by in situ lift-out via milling in a FEI Helios NanoLab DualBeam 660 focused a broadening of 0.1 eV. Structural relaxations used the Broyden–Fletcher–
ion beam (FIB). Prior to the FIB, ~40/5/10 nm of SiO2/Ti/Au was deposited via Goldfarb–Shanno algorithm with a force threshold of 0.025 eV Å–1. A vacuum
electron-beam evaporation in a Kurt J. Lesker Lab18 evaporator, to improve layer of 20 Å was inserted in the direction normal to the graphene sheets to
the contrast during STEM imaging at low magnifications. Cross sections were minimize the spurious interactions across the periodic boundary. The models
prepared using a Ga+ ion beam at 30 kV, then stepped down to 1 kV to avoid ion- in the figures were visualized using OVITO50 and VESTA51 software.
beam damage to the sample surface. ii. Two-dimensional Ga phase stability and electronic structure calculations: All
High-resolution STEM of sample cross sections was performed in a FEI dual DFT calculations on phase stabilities and electronic structure were performed
aberration-corrected Titan3 G2 60–300 S/TEM at 200 kV using a high-angle using the GGA-PBE exchange-correlation functional47,48 and the PAW pseu-
annular dark-field (HAADF) detector. The HAADF detector (Fischione) had dopotentials45,46. Plane-wave expansions were truncated at an energy cut-off
a collection angle of 51–300 mrad for Z-contrast imaging. A beam current of of 500 eV. All structural relaxations were performed using dipole corrections
70 pA, beam convergence of 30 mrad (C2 aperture of 70 μm) and camera length to the total energy52 and to the electrostatic potential53 in the out-of-plane
of 115 mm were used for STEM image acquisition. The STEM energy-dispersive direction, until the remaining forces were within 0.01 eV Å–1. Brillouin zone
X-ray spectroscopy (EDS) maps were collected using the superX EDS system, samplings were performed on grids with k-point densities equivalent to
which has four EDS detectors surrounding the sample. that of a 20 × 20 × 1 grid for a 1 × 1 Ga/SiC unit cell. All Ga/SiC calculations
were performed using seven repeating units of SiC along the z direction as
LEED. LEED measurements of EG/Ga/SiC, EG/In/SiC and EG/Sn/SiC samples substrate, capped by Ga from above and by H from below. Graphene/Ga/SiC
were performed using LEED spectrometer BDL800IR-MCP manufactured by OCI calculations were performed using five repeating units along the z direction
Vacuum Microengineering. Samples were first degassed at 200 °C for 30 min under for a 2 × 2 graphene + √3 × √3 R30º Ga/SiC supercell, and three repeat-
ultrahigh vacuum to desorb surface moisture and contaminants. LEED patterns ing units for a 5 × 5 graphene + 4 × 4 R0º Ga/SiC supercell to alleviate the
were then acquired at room temperature using constant primary beam currents of computational demand of accommodating more atoms in the large supercells.
10 nA and beam energies of 50 eV–250 eV, in 1 eV steps. Band unfolding was performed using the GPAW package54; all other calcula-
tions were performed by the Vienna Ab Initio Package (VASP; ref. 55). Fermi
ARPES. ARPES measurements were performed at the Microscopic and Electronic surfaces of Ga/SiC were calculated on a 40 × 40 × 1 grid and interpolated onto
STRucture Observatory (MAESTRO) beamline at the Advanced Light Source at a 200 × 200 × 1 grid for plotting. Band structures at the hybrid functional
Lawrence Berkeley National Lab. The sample was annealed at 550 K for 30 min level were calculated using the range-separated form of Heyd, Scuseria and
in the end-station before measurements to remove adsorbates from the transfer Ernzerhof (ref. 56; HSE06, with a range-separation parameter of 0.2 Å–1) and
of the sample through air. Measurements of EG/Ga/SiC and Gr/In/SiC structures using structures relaxed at the PBE level. Self-consistent HSE06 calculations
were performed using a photon energy of 140 eV and 110 eV, respectively. were performed on a 12 × 12 × 1 k-point grid. For the DOS of bilayer Ga in
Photoemission spectra were collected by moving the sample around one angle Fig. 5b, we artificially shifted EF by 0.5 eV to account for the additional (un-
while using the angle-resolved mode of a Scienta R4000 electron analyser for the determined) electron doping so that the band alignment agrees with ARPES
collection of the other angular axis. measurements. As for the DOS calculations carried out on hexagonal two-
layer and three-layer Ga/SiC, the GaSiGaC and GaSiGaCGaC stacking sequences
Transport measurements and Tc, Bc and TBKT extraction. Transport were used, respectively. The GaSiGaCGaC stacking sequence occupies one of
measurements were carried out in a Quantum Design physical property the lower-energy configurations out of all the possible stacking sequences
measurement system (PPMS). Contacts were made to the EG/Ga heterostructure for three-layer Ga and most closely matches the band structure as directly
by lightly scratching the film surface with a diamond scribe or tweezers, and then measured in ARPES (Supplementary Figs. 10,11 and Supplementary Table 1).
lightly pressing or soldering In dots onto the scratched region. This was done Thus, GaSiGaC and GaSiGaCGaC stackings were used to calculate the DOS, and
in a casual attempt to make side contact to the 2D Ga. In dots were arrayed in the GaSiGaCGaC stacking was used to calculate Tc (Supplementary Fig. 19a).
a standard co-linear four-point-probe configuration with contact pitch on the iii. Electron–phonon interactions: All calculations related to electron–phonon
order of hundreds of micrometres. All resistance measurements were made with interactions were performed in a cell with only two SiC units, due to the
an excitation current of 1 µA. Tconset is defined as the intersection of the linear heavy computational demand of these routines; SiC slabs were passivated
extrapolations of the transition and normal regions. The linear fit for the transition from below by H atoms that are each artificially given the same mass as a Si
region is the region of maximum slope, which is consistent for the entire transition atom (so that phonon frequencies more closely resemble that of a semi-

Nature Materials | www.nature.com/naturematerials

Nature Materials Articles
infinite SiC substrate). The starting-point electronic charge density was calcu- 54. Enkovaara, J. et al. Electronic structure calculations with GPAW: a real-space
lated on a 12 × 12 × 1 Γ-centred k-point grid. Electronic wavefunctions were implementation of the projector augmented-wave method. J. Phys. Condens.
then computed for a 6 × 6 × 1 grid. The phonon dispersion was calculated Matter 22, 253202 (2010).
using density functional perturbation theory based on the same 6 × 6 × 1 grid. 55. Kresse, G. & Furthmüller, J. Efficient iterative schemes for total-energy
All computations were performed by the Quantum ESPRESSO package using calculations using a plane-wave basis set. Phys. Rev. B 54,
the local density approximation exchange-correlation functional, Hartwig- 11169–11186 (1996).
sen–Goedeker–Hutter norm-conserving pseudopotentials and a plane-wave 56. Heyd, J., Scuseria, G. E. & Ernzerhof, M. Hybrid functionals based on a
expansion cut-off of 1,090 eV57. To achieve a dense sampling of electron– screened Coulomb potential. J. Chem. Phys. 118, 8207–8215 (2003).
phonon coupling matrix elements across the Fermi surface, we constructed 57. Hartwigsen, C., Goedecker, S. & Hutter, J. Relativistic separable dual-space
electronic and phonon Wannier (EPW) functions based on wavefunctions Gaussian pseudopotentials from H to Rn. Phys. Rev. B 58, 3641–3662 (1998).
and phonon modes sampled on the coarse 6 × 6 × 1 grid and generated 58. Noffsinger, J. et al. EPW: a program for calculating the electron–phonon
interpolations onto a 96 × 96 × 1 grid, as implemented by the EPW code58,59. coupling using maximally localized Wannier functions. Comp. Phys.
Wannier functions were initialized by projecting the following orbitals onto Commun. 181, 2140–2148 (2010).
Bloch wavefunctions: two s and one pz for each Ga, one sp3 for each Si, and 59. Giustino, F., Cohen, M. L. & Louie, S. G. Electron-phonon interaction using
one sp3 for each C. An outer disentanglement window (one that captures all Wannier functions. Phys. Rev. B 76, 165108 (2007).
targeted bands with the chosen orbital characters) coincided with the entire 60. Souza, I., Marzari, N. & Vanderbilt, D. Maximally localized Wannier
energy range (Supplementary Fig. 19b)60. An inner window (where all Bloch functions for entangled energy bands. Phys. Rev. B 65, 035109 (2001).
states were included within the projection manifold) spanned the energy
range from the lower bound of Supplementary Fig. 19b up to 1 eV above the Acknowledgements
Fermi level. See the Supplementary Information for more details on the equa- Funding for this work was provided by the Northrop Grumman Mission Systems’
tions and references used. University Research Program, Semiconductor Research Corporation Intel/Global
Research Collaboration Fellowship Program, task 2741.001, National Science
Foundation (NSF) CAREER Awards 1453924 and 1847811, the Chinese Scholarship
Data availability Council, an Alfred P. Sloan Research Fellowship, NSF DMR-1708972 and 1808900,
The data that support the findings of this study are available at 10.6084/
and the 2D Crystal Consortium NSF Materials Innovation Platform under cooperative
m9.figshare.c.4830711 or from the authors on reasonable request. See author
agreement DMR-1539916. A portion of this research was conducted at the Center for
contributions for specific data sets.
Nanophase Materials Sciences, which is a DOE Office of Science User Facility, and at the
Pennsylvania State University Materials Research Institute’s Material Characterization
Code availability Laboratory. This research used resources of the Advanced Light Source, which is a
Code used for computational investigations presented in this manuscript is DOE Office of Science User Facility under contract no. DE-AC02-05CH11231. We
available at gitlab.com/QEF/q-e/tree/qe-6.3 (EPW v5.0.0, Quantum Espresso v6.3) acknowledge Haiying Wang for help with STEM sample cross-section preparation
and www.vasp.at (VASP). via FIB; Vince Bojan, Nabil Bassim and Heshem Elsherif for help with AES; and Max
Wetherington for Raman spectroscopy support.
References
44. Giannozzi, P. et al. QUANTUM ESPRESSO: a modular and open-source Author contributions
software project for quantum simulations of materials. J. Phys. Condens. N.B, B.B., Y.W., V.C. and J.A.R. wrote the paper with input from the co-authors. N.B.
Matter 21, 395502 (2009). performed CHet and XPS characterization and assisted in the Raman spectroscopy and
45. Kresse, G. & Joubert, D. From ultrasoft pseudopotentials to the projector SEM characterization. B.B performed the Raman spectroscopy and SEM characterization
augmented-wave method. Phys. Rev. B 59, 1758–1775 (1999). and assisted in sample preparation and electrical characterization. Y.W. performed
46. Blöchl, P. E. Projector augmented-wave method. Phys. Rev. B 50, DFT modelling of graphene/Ga/SiC heterostructures in consultation with V.C. with
17953–17979 (1994). input from J.Z., B.B., N.B. and J.A.R.; J.J. performed electrical measurements under the
47. Perdew, J. P., Burke, K. & Ernzerhof, M. Generalized gradient approximation direction of C.Z.C. with input from B.B and J.Z.; R.K., A.B. and C.J. performed ARPES
made simple. Phys. Rev. Lett. 77, 3865–3868 (1996). measurements under the direction of E.R.; N.N. performed graphene defect modelling
48. Perdew, J. P., Burke, K. & Ernzerhof, M. Generalized gradient approximation under the direction of A.v.D.; and K.W. performed cross-sectional STEM imaging. M.K.
made simple [Phys. Rev. Lett. 77, 3865 (1996)]. Phys. Rev. Lett. 78, and W.K. prepared the LEED instrument for EG/metal/SiC samples, and M.K. performed
1396 (1997). the LEED measurements. A.D.L.F.D. assisted with CHet and material characterization.
49. Marzari, M., Vanderbilt, D., De Vita, A. & Payne, M. C. Thermal C.D. and S.S. performed the EG synthesis under the direction of J.A.R.; J.S. assisted
contraction and disordering of the Al(110) surface. Phys. Rev. Lett. 82, in XPS data analysis. M.F., Q.Z., G.Z. and A.P.L. performed the scanning probe
3296–3299 (1999). characterization. Y.W.C. assisted with electrical measurements under the direction of J.Z.
50. Stukowski, A. Visualization and analysis of atomistic simulation data with
OVITO–the open visualization tool. Model. Sim. Mater. Sci. Eng. 18, Competing interests
015012 (2010). The authors declare no competing interests.
51. Momma, K. & Izumi, F. VESTA 3 for three-dimensional visualization of
crystal, volumetric and morphology data. J. Appl. Crystallogr. 44,
1272–1276 (2011). Additional information
52. Makov, G. & Payne, M. C. Periodic boundary conditions in calculations. Supplementary information is available for this paper at https://doi.org/10.1038/
Phys. Rev. B 51, 4014–4022 (1995). s41563-020-0631-x.
53. Neugebauer, J. & Scheffler, M. Adsorbate-substrate and adsorbate-adsorbate
Correspondence and requests for materials should be addressed to J.A.R.
interactions of Na and K adlayers on Al(111). Phys. Rev. B 46,
16067–16080 (1992). Reprints and permissions information is available at www.nature.com/reprints.

Nature Materials | www.nature.com/naturematerials