https://doi.org/10.

1038/s41567-021-01174-7

Supplementary information

Deep moiré potentials in twisted transition

metal dichalcogenide bilayers
In the format provided by the
authors and unedited
 Supplementary Information for
Deep moiré potentials in twisted transition metal dichalcogenide bilayers

Sara Shabani,1 Dorri Halbertal,1 Wenjing Wu,2 Mingxing Chen,3 Song Liu,4 James
Hone,4 Wang Yao,5 Dmitri N. Basov,1 Xiaoyang Zhu,2 and Abhay N. Pasupathy1
1
Department of Physics, Columbia University, New York, NY, USA
2
Department of Chemistry, Columbia University, New York, NY, USA
3
School of Physics and Electronics, Hunan Normal University,
Key Laboratory for Matter Microstructure and Function of Hunan Province,
Key Laboratory of Low-Dimensional Quantum Structures and
Quantum Control of Ministry of Education, Changsha, Hunan, China
4
Department of Mechanical Engineering, Columbia University, New York, NY, USA
5
Department of Physics and Center of Theoretical and Computational Physics, University of Hong Kong, Hong Kong, China

S1. DETAILS OF CALCULATION OF FIGURE 1H

Figure 1h presents a mechanical relaxation calculation aiming to reproduce different non-uniform related strain
features that were measured in figure 1g. Beyond the intrinsic system properties, namely the generalized stacking
fault energy function (GSFE) lattice constants and elastic properties for this system (see methods section), the
X
simulation uses as initial and boundary conditions the locations and stacking registry of selected HX stacking points.
Here we provide further details about this process. Figure S1 presents the initial and boundary conditions used to
construct figure 1h. The white circles mark locations of dislocations appearing in the measurement of figure 1g. The
dislocations are lattice defects, and are not directly supported by the model. However, locally the simulation can still
account for the strain maps and stacking configurations. Therefore, the calculation was divided into several separate
calculations, surrounding the dislocations from different orientations (region marked by dashed lines with different
X
colors). In each region, the colored dots mark the positions of points where HX stacking configurations were forced
for a given simulation. The false-color shows the stacking energy density of the initial configuration, used as a starting
point for the simulations, which was generated as an interpolation between the forced stacking configurations. The
solution of figure 1h is a result of stitching these 5 calculations.

FIG. S1: Initial and boundary conditions for the simulation presented in figure 1h. The white circles mark locations of
dislocations appearing in the measurement of figure 1g. The simulation area was divided into 5 regions, marked by dashed
X
lines with different colors. In each region, the colored dots mark the positions of points where HX stacking configurations
2
were forced. The false-color shows the stacking energy density (in meV /nm ) of the initial configuration, used as a starting
point for the simulations. The solution of figure 1h is a result of stitching these 5 calculations.
 2

S2. TRANSITION FROM TRIANGULAR TO HONEYCOMB STRUCTURES IN THE MOIRÉ PATTERN

M M
At moiré wavelengths of close to 10 nm near H-stacking, the moiré unit cell shows regions of HX and HM stacking
which are approximately equal in area, resulting in a triangular lattice moiré pattern as shown in figure 1d of the main
M
text. As the wavelength increases beyond ∼ 15 nm, the HM regions shrink in size to soliton lines, with nearly the
M
entire moiré unit cell occupied by the HX stacked regions. The moiré pattern then resembles a honeycomb lattice.
The transition between the triangular and honeycomb regions is seen in figure S2.

FIG. S2: Honeycomb lattices formed by the shear strain instead of triangular lattice seen at small angle (in pm).
 3

S3. BAND EDGE ANALYSIS

In two dimensions, the density of states ρS (E) of an ideal semiconductor with a single valence and conduction band
at T=0 features sharp band edges with a constant value of the DOS beyond the band edge. Our experiments are
carried out at room temperature, which results in a broadening of the edges due to the Fermi distribution of electrons
in the tip and sample at non-zero temperature. The current in the STM tunnel junction is given by[1]:

Z +∞
4πe
I(V ) = (f (E + eV ) − f (E))ρT (E + eV )ρS (E)M (z, V ) dE (1)
~ −∞

where f (E) is the Fermi-Dirac distribution at the temperature of the measurement. In our experiment, we are
primarily interested the positions of the band edges, where ρS (E) is a strong function of energy. Under the assumption
that the tip density of states ρT and the matrix element M are weak functions of energy in this energy range, we
approximate:

Z +∞
4πe
I(V ) = (f (E + eV ) − f (E))ρT (E + eV )ρS M dE (2)
~ −∞

This gives the expression for the differential conductance of the tip-sample junction:

+∞
4πe2 ρT M
Z
dI(V ) df (E + eV )
= ρS (E) dE (3)
dV ~ −∞ dE

The result of this process is shown in figure S3a for a hypothetical semiconductor with a valence band edge at -0.5 eV
and conduction band edge at 1.0 eV. The sharp band edge develops a finite slope due to temperature broadening at
non-zero temperature. A simple practical method that we adopt to define the band edges is also shown on this plot,
by drawing intersecting straight lines below and above the gap edge. The use of this method is shown in practice
on real data in figures S3b and S3c for the valence and conduction bands respectively. We note that the accuracy of
this process is not limited by the temperature broadening of the spectrum - the determination of the band edge is
ultimately set by the signal to noise ratio of the measurement.

FIG. S3: a Theoretical calculation of the temperature-broadened differential conductance of a semiconductor with valence
band edge at -0.5 V and conduction band edge at +1.0 V. The dashed lines are linear fits to the spectrum above and below
the band edges, and the crossing points mark the position of the band edges. b,c Determination of the valence and
conduction band edges from typical experimental spectra following the procedure outlined in a.
 4

S4. ANALYSIS LEADING TO FIGURES 3D AND 4A OF THE MAIN TEXT

In order to extract band edges as a function of moiré wavelength, we start by defining the moiré unit cells, based
X
upon the positions of the HX positions from the topograph in figure 3a of the main text (see figure S4a). We avoid
moiré unit cells that have aspect ratios larger than 2. We then use Delaunay triangulation to define the moiré unit
X
cells from the HX positions shown in figure S4a. For each Delaunay triangle, we then find the centroid, which defines
M M
the position of HM and HX stacking respectively. We use the spectroscopic data of figure 3b and 3c to find the band
edges at these points, and set this to the valence and conduction band edge values for each given triangle. The result
of this procedure for the valence band edge is shown in figure S4b. We then take the area of the moiré unit cell and
convert it to a moiré wavelength by assuming it to be an equilateral triangle in order to generate the plot in figure
3d of the main text. This necessarily introduces scatter into the data, since triangles with different aspect ratios have
different spectroscopic properties. However, it allows us to represent the entire data set of band edges as a function
of moiré wavelength.
In order to then find the moiré potential for each moiré unit cell, we take the difference in band edges between each
M M
HX region and the three neighboring HM regions, as shown in figure S4c. This data is used to generate the plot in
figure 4a of the main text.

X
FIG. S4: a HX points of the moiré lattice shown as black markers, superposed on the spectroscopic map of valence band
M M X
edges (in eV ). b Determination of valence band edges of each HX and HM region after Delaunay triangulation of the HX
points in a (in eV ). c Determination of the moiré potential. The moiré potential is defined as the absolute value of the
difference between a given triangle in b (shown in red) and the average of its three nearest neighbors shown in blue.
 5

S5. SPECTROSCOPY ON A R-STACKED HETEROBILAYER

M X M
Figure S5 shows the differential conductance spectra on RX (RM ) and RM of the R-stacked sample shown in the
M X
main text. STM measurements alone could not distinguish between the RX and RM configurations at this moiré
wavelength, and the displayed spectra is an average over both of these two sites. The moire wavelength in this sample
is ∼ 5nm. The moiré potential is too weak to quantify accurately from these measurements at this wavelength.

M X M
FIG. S5: dI/dV measurements on RX (RM ) and RM , corresponding to the topograph shown in Figure 1 of the main text.
The set point is I = 100 pA and V = -2 V .
 6

S6. STRAIN ANALYSIS OF DIFFERENT MOIRE WAVELENGTHS

In general, the moiré pattern for H-stacked samples can be schematically represented as triangular regions of
M M X
HM and HX stacking, with the vertices being the regions of HX stacking. In the absence of external strain,
each of the triangular regions is an equilateral triangle. In reality, due to the external strain, the three sides of
each triangle are not equal, giving rise to triangle aspect ratios different from unity (the aspect ratio is defined as
R = abc/(8(s − a)(s − b)(s − c)) where a,b,c are the three sides of the triangle and s = (a + b + c)/2). We plot below
M M
the band edges in the regions of HM and HX stacking as a function of the aspect ratio of every moire unit cell we
have measured in the sample. If the band edges do not depend on the external strain, we should see no correlation
between aspect ratio and band edge positions. In reality, we do see a small but non-negligible dependence of the band
edge positions on the aspect ratio at small aspect ratios (ie, close to equilateral triangles). The magnitude of this
dependence on aspect ratio is fairly small (40-80 meV for the valence bands and 20-40 meV for the conduction band),
so it does not negate any of the main conclusions of our manuscript.

FIG. S6: Dependence of the (a) valence and (b) conduction band edges as a function of aspect ratio of individual moiré
triangles. The scatter in the plot at a given value of aspect ratio arises from the dependence of the moiré potential on triangle
size, as described in the main text.
 7

S7. ORBITAL MIXING

M
Figure S7a shows the orbital mixing for the equilibrium layer distance (6.66 A) of the HM configuration. Figure S7b
is the corresponding orbital mixing for an artificially reduced layer distance (6.46 A). The weights were obtained by
projecting the wavefunctions of the heterostructure onto the orbitals of the constituents by means of the k-projection
method. Contributions of each monolayer to the electronic bands were calculated using the layer k-projection method
as implemented in the program KPROJ[2]. The trend is the same for all the three stacking configurations.

M
FIG. S7: Orbital content of the band structure for the HM configuration with (a) equilibrium interlayer separation (6.66 A)
and (b) reduced interlayer separation (6.46 A).
 8

S8. COHESIVE ENERGY

The cohesive energy for the various stacking orders are calculated as shown in figure S8. We assumed the cohesive
energy is the binding energy between the two layers. Essentially, the DFT-D3 and vdw-DF methods produce the same
trend in Eb, such that, it has the largest value for MX and the smallest value for XX. This trend can be understood
since MX has the smallest layer distance, whereas XX has the largest layer distance. The energy difference between
the two lowest energy stackings is 22.1 meV/f.c. and 28.5 meV/f.c. based on vdW-DFT (optB88) and DFT-D3
respectively. These values are reasonably consistent with the 22.34 meV per formula unit obtained by the GSFE
functional, used in our structural relaxation calculations.

FIG. S8: Cohesive energy for different stacking configurations using DFT-D3 and vdW-DF(optB88) methods
 9

[1] H. M. Hill, A. F. Rigosi, K. T. Rim, G. W. Flynn, and T. F. Heinz, Nano letters 16, 4831 (2016).
[2] M. Chen and M. Weinert, Physical Review B 98, 245421 (2018).