Spin-orbit engineering in transition metal dichalcogenide alloy monolayers

Abstract

Binary transition metal dichalcogenide monolayers share common properties such as a direct optical bandgap, spin-orbit splittings of hundreds of meV, light-matter interaction dominated by robust excitons and coupled spin-valley states. Here we demonstrate spin-orbit-engineering in Mo(1-x)WxSe2 alloy monolayers for optoelectronics and applications based on spin- and valley-control. We probe the impact of the tuning of the conduction band spin-orbit spin-splitting on the bright versus dark exciton population. For MoSe2 monolayers, the photoluminescence intensity decreases as a function of temperature by an order of magnitude (4-300 K), whereas for WSe2 we measure surprisingly an order of magnitude increase. The ternary material shows a trend between these two extreme behaviours. We also show a non-linear increase of the valley polarization as a function of tungsten concentration, where 40% tungsten incorporation is sufficient to achieve valley polarization as high as in binary WSe2.

Figure 1. Highly crystalline Mo_(1−x)W_xSe₂ alloy monolayers for spin-orbit engineering.

(a) Simple band structure scheme in the K⁺ valley (at the K point of the Brillouin zone) to indicate the different signs and magnitudes of the valence and conduction band spin splittings when going from MoSe₂ to WSe₂ monolayers. Optically bright (red arrows) and dark (grey arrows) A-exciton transitions are indicated. (b) Nano-resolution x-ray photoelectron (nano-XPS) measurements on Mo_(1−x)W_xSe₂ alloys showing gradual composition change with varying x, where orange, green, blue, red, black correspond to x=0%, 30%, 40%, 90 and 100%, respectively. For increasing x (W) content, W (Mo) content increases (decreases), whereas selenium ratio remains at the same values without any significant single or double (V_Se and V_2Se) vacancy formation. (c) E_2g Raman peak position shift as a function of composition x. (d) Low temperature photoluminescence (PL) spectroscopy is a very simple and efficient technique to probe the material quality. Impurities and defects will trap optically excited carriers, resulting in emission below the optical bandgap. PL spectrum at T=4 K of Mo_0.7W_0.3Se₂ alloy monolayer showing very sharp emission of the charged exciton (trion T) and the neutral A-exciton (A) and negligible defect-related emission. Inset: representation of the alloy monolayer, the order of magnitude of the Bohr radius a_B of an electron–hole pair (exciton) is shown. The narrow PL linewidth therefore indicates high quality alloy material on a nano-scopic scale. (e) PL spectra at 4 K of monolayers for tungsten (W) composition from x=0 to 100% in Mo_(1−x)W_xSe₂. The dominant, sharp A-exciton emission is indicated.

Figure 2. Tuning the spin-orbit splitting in Mo_(1−x)W_xSe₂ monolayers.

(a) The splitting between A- and B-excitons is measured by PL excitation spectroscopy (PLE) (red squares), where error bars correspond to laser energy step size and reflectivity (open circles), with error bars from multi-Lorentzian fits. The sum of the valence band and conduction band spin splittings calculated by DFT is shown for comparison, for individual values and computational details see Methods and Supplementary Note 1. (b) PL spectra of A-exciton in monolayer Mo_0.7W_0.3Se₂ for different laser excitation energies. We uncover a clear maximum when the laser energy is in resonance with the B-exciton. (c) Reflectivity spectra using a white light source, uncovering in addition to the A-exciton also the B-exciton spectral position that can be tuned by varying the alloy composition. Dotted lines are a guide to the eye to indicate the energy shifts. (d) Same measurements as b but for all samples, the A-exciton PL intensity is plotted as a function of laser energy. These PLE measurements allow determining the B-exciton energy with very high precision.

Figure 3. Temperature dependence and valley polarization engineering in Mo_(1−x)W_xSe₂ monolayers.

(a) Comparing the global PL emission intensity of several samples is challenging, as the set-up has to compensate thermal expansion/movement. For the measurements of the PL emission intensity as a function of temperature, we have exfoliated several monolayer flakes of different materials in close proximity onto the same substrate, which is mounted on a three-axis attocube nano-positioner. Therefore PL emission for different samples are measured under identical conditions, that is, same detection and laser spot size. Only highly reproducible in-plane movement is needed to change sample. (b) The PL spectra of monolayer MoSe₂, Mo_0.3W_0.7Se₂ and WSe₂ are shown for different temperatures, the relative intensities can be directly compared. (c) Using the spectra from b, we integrate the total number of counts including A-exciton and trion (MoSe₂ and Mo_0.3W_0.7Se₂) and in addition the localized states (WSe₂). We compare the total number of counts for the three monolayer materials as a function of temperature, see Supplementary Note 2 for details (d) The measured valley polarization, that is, circular PL polarization degree P_c is plotted as a function of tungsten (W) content in the sample. P_c is defined as , where I⁺ and I⁻ are the σ⁺ and σ⁻ polarized PL components, respectively. We observe a highly non-linear increase in the valley polarization as more tungsten is incorporated. For the measurement, for each sample the laser energy is 140 meV above the A-exciton. The error bars correspond to the polarization resolution of our set-up. (e) Using σ⁺ circularly polarized laser excitation, we detect the A-exciton emission in σ⁺ (black) and σ⁻ (red) polarization. For x=0.3, we detect no polarization, as for binary MoSe₂. Surprisingly, for x=0.4 we detect up to 40% PL polarization. The results for x=0.9 also show high polarization. (f) The circular PL polarization P_c is plotted as a function of the excitation laser energy to find optimal valley polarization conditions. While for x≤0.3 the valley polarization remains low, we demonstrate for x⩾0.4 a wide range of laser excitation energies that can be used for valley index initialization.