Direct STM Measurements of R- and H-type Twisted MoSe2/WSe2 Heterostructures

When semiconducting transition metal dichalcogenides heterostructures are stacked the twist angle and lattice mismatch leads to a periodic moir\'e potential. As the angle between the layers changes, so do the electronic properties. As the angle approaches 0- or 60-degrees interesting characteristics and properties such as modulations in the band edges, flat bands, and confinement are predicted to occur. Here we report scanning tunneling microscopy and spectroscopy measurements on the band gaps and band modulations in MoSe2/WSe2 heterostructures with near 0 degree rotation (R-type) and near 60 degree rotation (H-type). We find a modulation of the band gap for both stacking configurations with a larger modulation for R-type than for H-type as predicted by theory. Furthermore, local density of states images show that electrons are localized differently at the valence band and conduction band edges.

Stacking two monolayer van der Waals materials with either a lattice mismatch or a twist angle between them creates a moiré superlattice which changes the electronic structure of the heterostructure [1][2][3][4][5][6][7][8] . For example, in twisted bilayer graphene the twist angle controls the emergence of charged ordered states 9 , unconventional superconductivity 7 , correlated insulating states 8 and magnetic states 10 , which are the product of the formation of moiré potentials in the structure. In a case of a lattice mismatch between the layers such as graphene and boron nitride heterostructures, new Dirac cones and the Hofstadter butterfly pattern appear even at zero twist angle 2,4-7 . Novel quantum phenomenon have also been predicted and observed in twisted transition metal dichalcogenide (TMD) heterostructures including homobilayers 11 and heterobilayers [12][13][14][15][16][17][18] . Near R-or H-type stacking, both the electronic and topographic structure are modified leading to flat bands, confinement, and atomic reconstruction 1,[11][12][13][14][19][20][21][22][23] . TMD-TMD bilayers are of interest due to the larger range of angles at which flat bands can form and TMD-TMD heterostructures can more accurately represent the Hubbard model than graphene [24][25][26] .
One way to measure the electronic properties of these structures is through scanning tunneling microscopy (STM) and spectroscopy (STS), which unlike far-field optical measurements can provide direct measurements of the magnitude and wavelength of the moiré potential as well as the electronic structure. STM uses a sharp metal tip to directly probe the surface by applying a bias voltage between the tip and the sample. Since STM probes the electronic properties of the surface, this requires that the surface of the devices be exposed and clean which is challenging for TMDs. However, recently a few STM, conductive atomic force microscopy (cAFM), transmission electron microscopy (TEM), and scanning TEM (STEM) observations have been conducted. STM measurements on twisted bilayer WSe2 found flat bands near both 0 and 60 degrees and that the flat bands are localized differently for the two cases 20 . Other experiments on heterobilayer TMDs include observations of a modulation in the moiré potential at a moiré length of 8.3 nm in chemical vapor deposition grown MoS2/WSe2 heterostructures 19 and variations in the spectral peaks' sharpness between 80 K and 5 K 1 . Experimental results for STM measurements on WSe2/WS2 heterobilayers reveal flat bands near the edge of the valence edge band at low temperature 21 .
Lastly, for MoSe2/WSe2 heterostructures recent STM, cAFM, and TEM measurements at room temperature mapped the variation in the energy of the valence and conduction band of the H-type (60 degree) MoSe2/WSe2 structure for wavelengths ranging from 6 nm to 17 nm and obtained tunneling spectroscopy at the high symmetry stacking locations. These measurements found that there was a decrease in the magnitude of the modulation as the moiré wavelength decreased 27 .
STEM and TEM measurements on MoSe2/WSe2 explored atomic reconstruction of both types at room temperature for CVD grown structures for twist angles less than 1 degree 22,23 .
In this work, we measure both R-type (near 0 degree) and H-type (near 60 degree) MoSe2/WSe2 heterostructures to understand modulations in the band structure with respect to the moiré pattern. Using STS measurements, we map the band edges and modulation of the band gap for a range of twist angles at 77 K and 4.5 K. A schematic of a typical STM device and the experimental setup is shown in Fig. 1a. Optical microscope images of the R-type and H-type devices are shown in Supplementary Figs. 2a and b respectively. For the experiments conducted in this work the WSe2 is the top layer and thus the tungsten STM tip preferentially tunnels into this layer. One of the challenges of performing the STM measurements of TMDs at low-temperature is making good electrical contact to the device. We found that placing a graphene layer under the TMDs allowed for reliable STM measurements over a broad range of tunnel voltages. Without the graphene layer when probing energies near or inside the band gap, the tip could crash into the sample destroying the sample and the tip in the process. To address the challenges of obtaining clean TMD heterostructures, devices were annealed in vacuum and AFM cleaned before beginning STM measurements (see the Methods section for more details). The heterostructures were then scanned in the STM to find clean, defect-free areas before taking spectroscopy measurements.
When TMD heterostructures such as MoSe2/WSe2 are twisted, the atomic stackings between the layers vary periodically in space. There are three main high symmetry points that are of interest as shown in Figs. 1d and e along with the corresponding atomic registry stackings and moiré lattices. For the R-type stacking the three high symmetry atomic stacking orders are tungsten on molybdenum ( ′ ) or equivalently selenium on selenium ( ′ ), tungsten on selenium ( ′ ), and selenium on molybdenum ( ′ ), where M is the metal atom, X is the chalcogen atom, the apostrophe denotes the bottom layer, and the R indicates the R-type configuration. The H-type also has three high symmetry stacking configurations ′ ( ′ ), ′ , and ′ labeled in Fig. 1e. In Fig. 1b, we show STM topography for a 5 nm moiré in R-type heterostructure at 4.5 K and Fig. 1c shows a 12 nm moiré in an H-Type device. The R-type heterostructure has a hexagonal moiré pattern while the H-type has a triangular pattern in agreement with previous results and calculations 22,23,27 .
To compare the electronic properties of the R-and H-type stacking configurations, we perform scanning tunneling spectroscopy measurements at the high symmetry locations of the moiré unit cells. In Fig. 2 which can also be seen in Fig 2a. In Fig. 1f, we see the alignment between the WSe2 and MoSe2 is a type II band alignment such that the MoSe2 is the conduction band edge and the WSe2 is the valence band edge. Therefore, for R-type devices the oscillation in the band gap mainly arises from the valence band edge of the WSe2. By taking the difference in the global minima and maxima in the valence band we find the maximum amplitude in oscillation is 260 meV and the difference between the two maxima to be 60 meV. Using the constant value from the conduction band edge, we find that the band gaps for each of the three high symmetry points are 2.28 +/-0.01 eV, 2.07 +/-0.01 eV, and 2.01 +/-0.01 eV respectively. We note that the band gaps are comparable with the spectroscopy taken at another site on the sample. In both the point spectroscopy and the line spectroscopy the maximum band modulation occurs at a different stacking configuration compared to the DFT calculations which predict a maximum at ′ 28 .
However, these calculations do not consider lattice relaxations effects. It has been shown for Htype stacking the inclusion of relaxation effects in the DFT model changes which stacking order has the largest band gap 27 . Further, reports have shown that strain and interlayer distance can affect the band modulation 29 .
In Figs. 3d-f, LDOS images for a R-type device at 4.5 K are shown. At energies near the valence band edge around -1.3 eV the electronic wave functions are localized between the black and green sites (Fig. 3d). As the energy decreases to -1.7 eV the electrons become localized on the black site (Fig. 3e) and as the energy decreases farther to -1.8 eV the electrons localize on the red site (Fig. 3f). This is in good agreement with the line spectroscopy in Fig. 3b. From the color scale in the line spectroscopy, we note that the dI/dV values are higher at the black and green sites at higher energy and as the energy decreases the dI/dV becomes highest on the red site as verified by the LDOS images.
Next, we focus more closely on the H-type device. In Fig. 4b, the constant current line spectroscopy at 77 K for the H-type device is shown. The valence band varies with stacking configuration, but unlike the R-type device the conduction band edge also varies with position.
By taking the difference between the conduction and a valence band edge at each of the lines in In conclusion, our spectroscopic measurements of the R-type stacking show that the modulation in the band between the maximum and minima is larger than previous predictions and the largest modulation in the bands occurs at the ′ site. The modulation in the bands is dominated by shifts in the valence band for the R-type while both bands are strongly modulated for H-type stacking. The R-type also has overall larger band gaps than the H-type. Our H-type results agree with previous measurements taken at room temperature with signs of localization near the conduction and valence band edges of the H-type device.

Methods
Device Fabrication. Monolayer MoSe2 and WSe2, 10 to 30 nm thick hexagonal boron nitride, and monolayer or bilayer graphene were obtained through mechanical exfoliation techniques from bulk crystals 30 . The alignment between the MoSe2 and WSe2 was determined through second harmonic generation (SHG) 31

Data Availability
Data that support the plots within this paper and other findings of this study are available from the corresponding author upon reasonable request.

Supplementary Material
See