Transition metal dichalcogenides (TX₂) belong to a family of two-dimensional layered materials, where adjacent layers are weakly bound by van der Waals forces, like in graphite. These materials offer a large variety of properties, from insulating/semiconducting to metallic ones, depending on the combination of T and X. Among them, we have been focusing on semiconducting MoSe₂.

We grew wafer-scale MoSe₂ thin films on Se terminated GaAs(111)B substrates using the molecular beam epitaxy (MBE) method. The Scanning transmission electron microscope (STEM) image in Fig. 1 shows that two-monolayer (2L)-thick MoSe₂ is grown in a layer-by-layer manner. It also indicates that the MoSe₂ layers can climb over the monolayer step of GaAs. Though Fig. 1 shows an STEM image for a double-layer MoSe₂, it should be noted that the layer number can be varied with growth time. We carried out Raman spectroscopy measurements and investigated layer number dependence. Figure 2 shows Raman spectra around the A_1g peak for single-layer (1L), double-layer (2L), and more-than-four-layer (> 4L) MoSe₂ thin films. The spectrum for Se-terminated GaAs is also shown as a reference. According to a previous report on MoSe₂ mechanically cleaved from bulk crystal [1], the A_1g peak shifts to a higher wavenumber with an increasing number of layers because of multiple interlayer interactions. The A_1g peaks in our MoSe₂ thin films also show the blue shift (Fig. 2) when the layer number increases. The layer number of our films estimated from the positions of A_1g peaks, using reported values for cleaved samples as references [1], well coincides with that directly determined by STEM observation. On the other hand, splitting of the A_1g peak is not clearly observed in our film samples, which is in contrast to the cleaved ones with thicknesses of > 3L [1]. The peak splitting originates from the minute difference in vibration frequencies between the in-phase and out-of-phase vibrations and is very sensitive to the layer number [1]. Therefore, the absence of the peak splitting in our films likely stems from a fluctuation of the layer number (±1 layer) due to partial formation of growth nuclei on atomically flat surfaces.

Our results indicate that wafer-scale MoSe₂ thin films with a controlled layer number can be prepared by MBE, which is widely used for growth of high-quality thin films of conventional semiconductors. Application of this technique to other TX₂ materials may allow fabrication of heterostructures consisting of various TX₂ materials.