Exploring the Charge Density Wave Phase of $1T\text{−}{\mathrm{TaSe}}_2$: Mott or Charge-Transfer Gap?

$1T\text{−}{\mathrm{TaSe}}_2$ is widely believed to host a Mott metal-insulator transition in the charge density wave (CDW) phase according to the spectroscopic observation of a band gap that extends across all momentum space. Previous investigations inferred that the occurrence of the Mott phase is limited to the surface only of bulk specimens, but recent analysis on thin samples revealed that the Mott-like behavior, observed in the monolayer, is rapidly suppressed with increasing thickness. Here, we report combined time- and angle-resolved photoemission spectroscopy and theoretical investigations of the electronic structure of $1T\text{−}{\mathrm{TaSe}}_2$. Our experimental results confirm the existence of a state above $E_F$, previously ascribed to the upper Hubbard band, and an overall band gap of $\sim 0.7\mathrm{eV}$ at $\overline{\mathrm{\Gamma }}$. However, supported by density functional theory calculations, we demonstrate that the origin of this state and the gap rests on band structure modifications induced by the CDW phase alone, without the need for Mott correlation effects.

Figure 1

(a) Pictorial view of the starlike lattice reconstruction in the CDW phase of $1T\text{−}{\mathrm{TaSe}}_2$. (b) Surface-projected Brillouin zone (BZ) of the undistorted “normal” state. The red dashed lines mimic the Fermi surface and the blue solid line indicates the experimental path through the BZ as measured by TR-ARPES. (c) Sketch of the TR-ARPES experiment with laser photon energies.

Figure 2

(a) ARPES maps along the $\overline{M}\text{−}\overline{\mathrm{\Gamma }}$ direction at selected pump-probe delays. Dashed ellipses highlight the location of the CDW gap. (b) EDCs for a range of $k_{\parallel }$ from $\overline{\mathrm{\Gamma }}$ to $\overline{M}$ at $-1\mathrm{ps}$ delay. (c) Double peak fitting procedure (see Supplemental Material C [27]) and (d) temporal evolution of the CDW gap at $k_{\parallel }\simeq 0.85{\AA }^{-1}$. The gray shaded area is the uncertainty, the brown line is a phenomenological fit. The inset shows the VB dynamics (orange dots, right axis) and the electronic temperature (black dots, left axis) at $k_{\parallel }\simeq 0.21{\AA }^{-1}$.

Figure 3

(a) ARPES spectra along the $\overline{(M)}\text{−}\overline{\mathrm{\Gamma }}\text{−}\overline{C}$ path at various pump-probe delays, as indicated. (b) EDCs integrated over $0.6<k_{\parallel }<0.9{\AA }^{-1}$ ($\overline{\mathrm{\Gamma }}\text{−}\overline{C}$ direction) at delays ranging from $-0.05$ to 0.85 ps. (c) Same as (b) after highlighting the transiently populated state (TS) above $E_F$. (d) Population dynamics of the TS and (e) binding energy dynamics of TS and VB. The gray shaded areas in panels (d) and (e) are the uncertainties resulting from the peak fitting procedure.

Figure 4

(a) In-plane view of the starlike lattice reconstruction consisting of three different types of Ta atoms. (b) PDOS of each Ta atom type in the CDW phase. Calculated bands of (c) undistorted, (d) $1/2$ PLD and (e) full PLD structures along the $M\text{−}\mathrm{\Gamma }\text{−}C$ path. The size of the dots in (d) and (e) represents the spectral weight after band unfolding on the undistorted BZ, while blue- and red-colored bands represent the contributions (projections) of Ta atoms type $B$ and $C$, respectively. (f) Ta PDOS of the undistorted structure. (g),(h) The Ta-dominant band of the undistorted structure, properly shifted, matches the split bands close to $E_F$ in both $1/2$ PLD and full PLD cases.