Atomically Stacked MoS2 Bilayers Regain the Direct Band Gap Monolayers Own

For more than a decade, one of the most celebrated properties of molybdenum disulfide has also been one of its most fragile. When this two-dimensional semiconductor is shaved down to a single atomic layer, its electronic bands rearrange themselves so that electrons can emit light efficiently, a hallmark known as a direct band gap. Stack a second layer on top, however, and the useful optical behaviour largely evaporates: the bands shift, the gap becomes indirect, and photoluminescence collapses by orders of magnitude. A team led by researchers at National Taiwan Normal University, working with collaborators at MIT, National Taiwan University, National Yang Ming Chiao Tung University and the National Synchrotron Radiation Research Center, now reports in Nature that this long-accepted trade-off can be defeated simply by changing how the two layers are stacked. By growing bilayer MoS2 with a rarely achieved 1H stacking geometry through a two-step chemical vapour deposition process, they produced bilayers that behave optically like monolayers while retaining the electronic advantages of two layers.

The distinction between stacking geometries sounds arcane, but it is the entire story. In the most common bilayer arrangement, known as 2H stacking, the top sulfur plane sits rotated by 180 degrees relative to the bottom layer, which restores inversion symmetry and drives the valence band maximum away from the valleys where the conduction band minimum resides. The 3R, or rhombohedral, arrangement keeps the layers aligned in the same orientation and breaks that symmetry, enabling ferroelectric behaviour through interlayer sliding. The 1H geometry, the one the Taiwanese-led team targeted, places the second layer in perfectly commensurate vertical alignment with the first in a specific relative orientation that keeps the band extrema aligned in momentum space. Achieving this configuration in a scalable, crystal-growth setting rather than by manually transferring and stacking exfoliated flakes is what makes the new work remarkable.

The researchers accomplished it with a modified two-step chemical vapour deposition scheme in which growth temperature is modulated between the nucleation of the first layer and the growth of the second. Their Extended Data analysis traces the mechanism to the diffusion behaviour of molybdenum adatoms landing on the completed monolayer. Depending on temperature, arriving adatoms follow either edge diffusion pathways, hopping along equivalent crystallographic directions to attach at the flake boundary, or surface diffusion pathways, migrating across the terrace of the underlying monolayer. An Arrhenius analysis of the hopping-rate ratio shows how the balance between these two regimes selects the resulting stacking polytype: 1H, 3R or 2H. Under the conditions favouring 1H growth, molybdenum adatoms form ribbons along preferred crystallographic directions that branch and eventually coalesce into seamless bilayer triangles, a process confirmed by bright-field and dark-field transmission electron microscopy showing fully stitched, unidirectional bilayer nanoribbons.

Structural verification was exhaustive. High-angle annular dark-field scanning transmission electron microscopy, cross-sectional STEM prepared by focused ion beam milling, and selected-area diffraction patterns were compared against simulated diffraction patterns for each candidate polytype. The examined regions consistently showed the 1H stacking arrangement with atomic-scale uniformity, and interlayer spacings measured at ten separate locations in the cross-sectional images matched the expected commensurate geometry. Second-harmonic generation microscopy and atomic force microscopy provided additional, wafer-scale confirmation that the resulting flakes were structurally distinct from the 3R and 2H bilayers grown under the alternative temperature conditions of the same process.

With the structure nailed down, the team turned to the electronic structure. Angle-resolved photoemission spectroscopy performed with a photoelectron momentum microscope at the Taiwan Photon Source probed the valence bands directly, and the experimental spectra were compared with density functional theory calculations. In conventional bilayer MoS2, the valence band maximum at the K valley sits measurably higher in energy than at the Gamma point only in monolayers; in bilayers the Gamma point wins and the gap becomes indirect. For the 1H bilayers, the measured energy separation between the valence band at K and at Gamma remained consistent with a direct gap, and the extracted value for a monolayer reference in the same experiment, 146 millielectronvolts, agreed with published literature, providing an internal calibration for the measurement.

The optical consequences followed immediately. Photoluminescence mapping across the 1H bilayer flakes revealed intensified excitonic emission, and, crucially, the spectra lacked the low-energy indirect-gap emission features that usually betray bilayer character. In ordinary 2H bilayers, most electron-hole recombination funnels through the indirect transition and emits weakly at longer wavelengths; the 1H bilayers showed no such signature, indicating that radiative recombination proceeds through the direct, momentum-conserving channel. In other words, the extra layer adds carrier capacity and mobility without exacting the usual optical penalty.

Perhaps the most surprising result concerns valley physics. Monolayer MoS2 owes its valleytronics credentials to broken inversion symmetry, which couples the spin and valley degrees of freedom and allows circularly polarized light to selectively populate one of two inequivalent valleys. Bilayers with 2H stacking restore inversion symmetry and destroy this valley contrast. The 1H bilayers, however, not only preserved valley-selective circular polarization but actually exhibited stronger valley polarization than monolayers under both resonant and nonresonant excitation. The authors attribute this enhancement primarily to suppressed intervalley scattering in the top layer of the stack, meaning that once excitons are injected into a valley they are less likely to relax into the opposite valley before recombining. Robust polarization under off-resonant excitation is particularly valuable for practical devices, since it relaxes the demanding requirement for exactly resonant optical pumping.

The implications reach across several device families. Bilayer transition metal dichalcogenides are already attractive for next-generation transistors because they offer higher carrier mobility than monolayers and, in suitably stacked forms, electrically switchable polarity. They also underpin emerging sliding ferroelectric devices, in which an interlayer displacement toggles a polarization state. The demonstration that a specific, growth-accessible stacking order can restore a direct band gap means engineers may no longer have to choose between the electronic merits of two layers and the optical merits of one. Light-emitting transistors, valleytronic logic and integrated optoelectronic circuits built on MoS2 all become more plausible when the same material platform supports charge transport, light emission and valley polarization simultaneously.

There are also broader lessons for the rapidly growing field of stacking-engineered quantum materials. The explosion of interest in twisted and commensurate bilayers of graphene and transition metal dichalcogenides has shown that interlayer registry, not chemistry alone, dictates electronic behaviour. Most such studies, however, rely on mechanical assembly of exfoliated flakes, which is artisanal, slow and poorly suited to manufacturing. The two-step CVD approach reported here shows that a desired polytype can be selected during synthesis by tuning adatom diffusion kinetics, and that adjacent bilayer ribbons can stitch together seamlessly into larger crystals. If the same kinetic control can be extended to other materials and other stacking targets, it would move stacking engineering closer to wafer-scale production.

Caveats and open questions remain, as with any single study. The photoemission measurements probed multiple crystallographic orientations of the grown flakes within the probing area, requiring careful angular slicing of the momentum images, and the reported direct-gap character rests on the agreement between experiment and density functional theory rather than on a direct measurement of the conduction band. Whether the enhanced valley polarization survives at elevated temperatures, in encapsulated device structures, and across wafer-scale films will need to be established. Still, the core message stands: the indirect band gap of bilayer MoS2 is not an immutable fact of nature but a consequence of stacking geometry, and that geometry can now be grown on demand. As the authors put it in their abstract, the results establish 1H MoS2 as a model system for stacking-engineered quantum materials, and they underscore its potential for valleytronic and optoelectronic applications that the field has been pursuing since monolayer MoS2 first announced its direct gap sixteen years ago.

Subject of Research: Stacking-controlled electronic and optical properties of CVD-grown bilayer molybdenum disulfide

Article Title: Stacking-induced direct band gap in CVD-grown 1H MoS2 bilayers

Article References: Yang, T. H., Chen, I.-T., Zhang, M.-J., Huang, J.-Y., Kuo, T.-H., Chen, S.-Y., Li, H.-Y., Chao, Y.-C., Hennighausen, Z. B., Dien, V. K., Wei, H.-W., Wu, M.-C., Yen, H.-W., Chuang, T.-H., Wei, D.-H., Kong, J., Lu, T.-H., Lin, K.-I., & Lan, Y.-W. (2026). Stacking-induced direct band gap in CVD-grown 1H MoS2 bilayers. Nature. https://doi.org/10.1038/s41586-026-11069-3

Image Credits: AI Generated

DOI: 10.1038/s41586-026-11069-3

Keywords: MoS2, transition metal dichalcogenides, two-dimensional materials, band gap engineering, chemical vapour deposition, stacking polytypes, photoluminescence, angle-resolved photoemission spectroscopy, valleytronics, excitons, optoelectronics, density functional theory