Giant magnetic splitting inducing near-unity valley polarization in van der Waals heterostructures

Monolayers of semiconducting transition metal dichalcogenides exhibit intriguing fundamental physics of strongly coupled spin and valley degrees of freedom for charge carriers. While the possibility of exploiting these properties for information processing stimulated concerted research activities towards the concept of valleytronics, maintaining control over spin–valley polarization proved challenging in individual monolayers. A promising alternative route explores type II band alignment in artificial van der Waals heterostructures. The resulting formation of interlayer excitons combines the advantages of long carrier lifetimes and spin–valley locking. Here, we demonstrate artificial design of a two-dimensional heterostructure enabling intervalley transitions that are not accessible in monolayer systems. The resulting giant effective g factor of −15 for interlayer excitons induces near-unity valley polarization via valley-selective energetic splitting in high magnetic fields, even after nonselective excitation. Our results highlight the potential to deterministically engineer novel valley properties in van der Waals heterostructures using crystallographic alignment.

far. Of central consequence is the resulting field-induced valley polarization of the longlived charge carriers, even though both valleys in the two constituent materials are initially equally populated. The degree of polarization, reaching near-unity values arises entirely due to the strong degeneracy lifting under magnetic fields. It emerges as a major aspect of the TMDC heterostructures, advancing both our understanding of the fundamental phenomena and the application potential of pre-designed atomically-thin systems.
The heterostructure under study (shown in Fig. 1a) consists of a monolayer of WSe 2 transferred on top of a MoSe 2 monolayer, exfoliated onto a SiO 2 /Si substrate. During the transfer process, the well-cleaved axes of the monolayers are deterministically aligned parallel to each other, ensuring a twist angle of either nearly 0 • (AA-stacking) or 60 • (AB stacking) degrees. The relative angle of the stacking configuration is confirmed through spatially-resolved second harmonic generation (SHG) spectroscopy [27]. Figure 1b illustrates the intensity of the parallel component of the SHG signal of the two individual monolayer materials in a polar plot. By fitting the data with a cos 2 (3θ) function we directly obtain a relative stacking angle θ of either 6 ± 1 • or 54 ± 1 • (the two possibilities stem from the phase-insensitivity of the SHG intensity measurement on a single layer). To clarify the precise alignment, we perform a spatial scan of the resulting heterostructure where the total SHG intensity is recorded at each position, as shown in Fig. 1c. In the overlapping region of the two layers denoted by the white framed area in Fig. 1a, we clearly observe a pronounced destructive interference of the SHG signal with respect to the individual monolayers, consistent with a nearly 60 • stacking configuration [27]. Thus, we conclude that the sample has an AB-like stacking configuration with a relative angle of 54 ± 1 • .
A characteristic photoluminescence (PL) spectrum of the heterostructure recorded at 4 K is shown in Fig. 1d. In line with recent reports, it consists of two separate spectral regions [11,12]: The intralayer transitions between 1.6 eV and 1.  Using the definition of the splitting as contrast to experimentally determined g factors of excitons in individual TMDCs, found to be around −4 in most cases [14][15][16][17][18][19]. In monolayers, the magnitude of the Zeeman coupling is often understood in terms of a simplified semi-quantitative model, including three contributions, namely the spin, the atomic orbitals, and the valley magnetic moment [13][14][15][16][17]. Since the optical transitions are spin-conserving between conduction and valence band, the net contribution from the spin to the energy splitting of the respective resonances is zero. On the other hand, only the valence bands carry a non-zero magnetic moment µ l from the atomic orbitals with µ l = 2 for the K+ valley and µ l = −2 for the K− valley leading to an overall splitting between the valley-selective transitions of −4µ B B. The third contribution, the valley magnetic moment µ k , arises from the self-rotation of the Bloch wavepackets [28]. It is defined by ±µ c k = m 0 /m e for the conduction band and ±µ v v = m 0 /m h for the valence band in the K+/K− valley, respectively. Assuming that optical transitions in a monolayer take place between valleys of the same index, these contributions cancel out almost entirely and the total field-induced Zeeman shift for a monolayer TMDC can then be in many cases. The deviations from this value are attributed both to nonequivalent effective masses of electrons and holes as well as to the complexities of the orbital contributions to the Zeeman shift beyond the simplified model [17].
In an AB-stacked heterostructure, however, we encounter a markedly different situation for optically bright transitions. Second-harmonic generation spectroscopy SHG measurements were carried out at room temperature with a Ti:sapphire laser (pulse length 100 fs, central wavelength 810 nm) focused on the sample via a 40x microscope objective. The signal was coupled into a grating spectrometer and detected with a CCD camera.
For polarization-dependent measurements, the laser light was linearly polarized and the reflected light was analyzed by the same polarizer, thereby selecting the parallel signal component of the SHG. The sample was rotated by a mechanical stage in order to obtain angle resolution. For mapping of the total SHG intensity, the sample was excited using circularly polarized light without any polarization analysis in the detection. The sample was scanned under the microscope using a motorized x-y stage and the total SHG intensity was recorded for each sample position.

Magneto-PL spectroscopy
The sample was placed on a x-y-z piezoelectric stage and cooled down to 4.2 K in a cryostat filled with liquid helium. Magnetic fields up to 30 T were applied by means of a resistive magnet in Faraday configuration. For static PL measurements, laser light at an energy of 1.94 eV was focused onto the sample with a microscope objective resulting in a spot size of ∼4 µm. The polarization of the PL was analyzed with a quarter-wave plate and a linear polarizer. Using a nonpolarizing beam splitter, the backscattered PL was guided to the spectrometer and detected with a liquid-nitrogen-cooled CCD. Time-resolved PL measurements were carried out with a pulsed diode laser (laser energy 1.80 eV, repitition rate 2.5 Mhz) which was synchronized to an avalanche photodiode. The PL from the interlayer exciton was spectrally selected with a longpass filter.