Time Evolution of the Electronic Structure of $1T\mathrm{\text{−}}{\mathrm{TaS}}_2$ through the Insulator-Metal Transition

Femtosecond time-resolved photoemission is used to investigate the time evolution of electronic structure in the Mott insulator $1T\mathrm{\text{−}}{\mathrm{TaS}}_2$. A collapse of the electronic gap is observed within 100 femtoseconds after optical excitation. The photoemission spectra and the spectral function calculated by dynamical mean field theory show that this insulator-metal transition is driven solely by hot electrons. A coherently excited lattice displacement results in a periodic shift of the spectra lasting for 20 ps without perturbing the insulating phase. This capability to disentangle electronic and phononic excitations opens new directions to study electron correlation in solids.

Figure 1

(a) The in-plane CDW distortion produces “stars” of Ta atoms in which the per atom electronic density is larger towards the center (yellow) and smaller towards the periphery (green). (b) Sketch of the electronic states with the uppermost cluster orbital (UCO) band in the metallic ($T_l=300\mathrm{K}$) and Mott phase ($T_l=30\mathrm{K}$). (c) Normal emission spectra acquired at 300 K (red curve) and $T_l=30\mathrm{K}$ (black curve). (d) Phase diagram of a Mott transition in the $U/W$, $T_e/W$ plane as obtained from the Hubbard model [Eq. (1)]; $U$ is the Coulomb repulsion, $W$ is the bandwidth. Notice that in thermal equilibrium $T_e=T_l$ whereas out of equilibrium $T_e\gg T_l$. The shaded area is a crossover phase between the metallic and Mott phase whereas the red line delimits a first order transition. Red and black circles indicate $U/W$ values for $T_l=300\mathrm{K}$ and $T_l=30\mathrm{K}$, respectively. The black arrow shows the large increase of $T_e$ induced by the absorption of the pump pulse.

Figure 2

(a) Spectra acquired in the metallic phase ($T_l=300\mathrm{K}$) at different pump-probe delays. (b) Photoelectron intensity map as function of pump-probe delay and binding energy ($T_l=300\mathrm{K}$).

Figure 3

(a) Spectra acquired at $\tau =100\mathrm{fs}$ (green curve) and $\tau =550\mathrm{fs}$ (blue curve) are compared to the spectrum collected before the arrival of the pump pulse (black curve). Although $T_e$ reaches high values, the lattices temperature remains roughly at 30 K. (b) Time-resolved ARPES intensity of the Hubbard peak normalized to the equilibrium value. The exponential fit (black line) provides the decay time ${\tau }_H=680\mathrm{fs}$. (c) Spectral function of the Hubbard model calculated at the center of the Brillouin zone for $U=0.4$ and $W=0.29\mathrm{eV}$. The electronic temperature and filling factor are $T_e=1350\mathrm{K}$, $n=0.48$ (green curve), $T_e=1100\mathrm{K}$, $n=0.49$ (blue curve), $T_e=60\mathrm{K}$, $n=0.5$ (black curve).

Figure 4

(a) Photoelectron intensity map measured in the MI phase ($T_l=30\mathrm{K}$) as a function of pump-probe delay and binding energy. (b) Two spectra acquired on a maximum ($\tau =1.4\mathrm{ps}$) and minimum ($\tau =1.6\mathrm{ps}$) of one CDW oscillation period display a rigid shift with respect to each other. (c) Measured shift of the Hubbard peak (diamonds) and the calculated $gX(\tau )$ (solid line). The inset displays the Fourier transform of the oscillations as a function of frequency.