Tunneling-tip-induced collapse of the charge gap in the excitonic insulator ${\mathrm{Ta}}_2{\mathrm{NiSe}}_5$

Tuning many-body electronic phases by an external handle is of both fundamental and practical importance in condensed matter science. The tunability mirrors the underlying interactions, and gigantic electric, optical and magnetic responses to minute external stimuli can be anticipated in the critical region of phase change. The excitonic insulator is one of the more exotic phases of interacting electrons, produced by the Coulomb attraction between a small and equal number of electrons and holes, leading to the spontaneous formation of exciton pairs in narrow-gap semiconductors/semimetals. The layered chalcogenide ${\mathrm{Ta}}_2{\mathrm{NiSe}}_5$ has been recently discussed as a candidate for the excitonic insulator, though the nature of the excitation gap that opens below $T_c=328\mathrm{K}$ remains hotly debated. Here, we demonstrate a drastic collapse of the excitation gap in ${\mathrm{Ta}}_2{\mathrm{NiSe}}_5$ and the realization of a zero-gap state by moving the tip of a cryogenic scanning tunneling microscope towards the sample surface by a few angstroms. We argue that the collapse of the gap is driven predominantly by the electrostatic charge accumulation at the surface induced by the proximity of the tip and the resultant carrier doping. The fragility of the gap in the insulating state of ${\mathrm{Ta}}_2{\mathrm{NiSe}}_5$ strongly suggests the gap possesses many-body character. Our results establish a reversible phase-change function based on ${\mathrm{Ta}}_2{\mathrm{NiSe}}_5$.

Figure 1

(a) Conceptual image of the transition from a narrow-gap semiconductor (left) or semimetal (right) to an excitonic insulator (middle). A many-body gap $2{\mathrm{\Delta }}_E$ opens up in the excitonic phase. (b),(c) Doping of an ordinary semiconductor or an excitonic insulator leading to a rigid shift of the chemical potential or the collapse of the energy gap, respectively.

Figure 2

(a),(b) Crystal structure of ${\mathrm{Ta}}_2{\mathrm{NiSe}}_5$ presented in two different views. The horizontal dashed line marks the cleavage plane. (c) Large-area topography of ${\mathrm{Ta}}_2{\mathrm{NiSe}}_5$ ($53\text{nm}\times 53$ nm, $I_s=100$ pA, $V_s=400$ mV, $T=77$ K). The insets show the correspondence to the crystallographic axes and the line-cut height profile through the step edge denoted by the red line. The black bar represents a length of 10 nm. (d) Small-area topography of ${\mathrm{Ta}}_2{\mathrm{NiSe}}_5$ with atomic resolution ($3.5\text{nm}\times 3.5$ nm, $I_s=1$ nA, $V_s=400$ mV, $T=77$ K). The insets show the correspondence to the crystal structure and crystallographic axes. The white bar represents a length of 1 nm. (e) Spectroscopic data taken at 4.3 K, exhibiting a charge gap and a terracelike structure at negative biases, marked by the triangle. The inset displays the spectra taken at various temperatures, tracking the gap evolution.

Figure 3

(a) The conductance $dI/dV$ spectra taken with various tip-sample distances $z$ from 0.95 to 0.50 nm. The $dI/dV$ curves are normalized by their maximum values and vertically shifted for clarity, with the horizontal black lines indicating the zero conductance for each curve. The terracelike structure discussed in the text is marked by the black triangles and the dashed lines. Three regimes—(i) stable 0.25-eV gap, (ii) gap shrinking, and (iii) stable zero gap—can be identified. (b) Schematic picture of the experimental setup. The tip-sample distance is controlled by varying the tunneling current $I_s$ while maintaining a fixed bias voltage $V_s$. The changing distance modifies the number of carriers induced by the difference in work function between the tip and the sample. (c) Experimentally determined charge gap as a function of the estimated surface carrier density and inverse distance. Different dataset are color and symbol coded: balls, $V_s=2$ V; squares, $V_s=1.5$ V.

Figure 4

(a) $dI/dV$ spectra showing the reversibility of the gap control, with 1 to 4 depicting the sequence of data acquisition. The spectra have been normalized by their maximum value and vertically offset for clarity, with the horizontal black lines marking the zero level for each curve. (b) Topography before and after the gap-control procedure.