Substrate tuning of the structural and electronic transition in thin flakes of the excitonic insulator candidate Ta$_2$NiSe$_5$

Abstract

Ta$_2$NiSe$_5$ continues to draw interest for its 326 K phase transition, whose dual electronic and structural nature reflects a complex interplay of electron-hole (excitonic) and electron-lattice interactions. Most studies that have attempted to decipher the relative importance of these interactions, particularly through charge transfer, have been limited to bulk samples. We utilized a thin-flake approach to modify the excitonic interactions in Ta$_2$NiSe$_5$ via an underlying film of Au. Using polarized Raman spectroscopy, we found that four layers of Ta$_2$NiSe$_5$ supported on conducting Au show a transition temperature that is both reduced by over 100 K and broadened due to an interfacial charge gradient effect, manifesting the presence of excitonic interactions. In contrast, four layers of Ta$_2$NiSe$_5$ supported on insulating Al$_2$O$_3$ show nearly bulk-like properties. We also report the development of an all-dry exfoliation and transfer protocol that generalizes substrate engineering for strongly correlated van der Waals materials.

Figures (5)

• Figure 1: Preparation of ultrathin Ta$_2$NiSe$_5$ flakes. (A) Crystal structure of Ta$_2$NiSe$_5$. The unit cell comprises two Ta$_2$NiSe$_5$ layers (L). In the orthorhombic phase, $\beta = 90^{\circ}$, whereas in the monoclinic phase at 30 K, $\beta \approx 90.7^{\circ}$Nakano_PRB_2018. (B) Topography of Ta$_2$NiSe$_5$ ribbons exfoliated onto an epitaxial Au(111) film on mica, acquired with atomic force microscopy. The inset cartoon illustrates the challenge of exfoliating a rippled compound onto an atomically flat surface. (C) Exfoliation and transfer technique assisted by thermal evaporation of the substrate material and cold welding of Au. (D and E) Optical micrographs of Ta$_2$NiSe$_5$ thin flakes on amorphous Au and Al$_2$O$_3$, respectively.
• Figure 2: Raman-active phonon modes of Ta$_2$NiSe$_5$ thin flakes. (A) Parallel ($aa$) and cross ($ac$) configurations of polarizer and analyzer employed for Raman spectroscopy. (B and C) The $aa$- and $ac$-Raman susceptibilities ($\chi"$) of a 4L Ta$_2$NiSe$_5$ flake on Au, acquired at 295 K. The phonon modes $A_{\textrm{g}}$($o$1)--$A_{\textrm{g}}$($o$8) and $B_{2\textrm{g}}$($o$1)--$B_{2\textrm{g}}$($o$3) of the orthorhombic phase are labeled. The red shading indicates the widened range of the $B_{2\textrm{g}}$ modes. The signals marked by asterisks originate either from the substrate (Au/Cr), or from a slight misalignment of the polarizer and analyzer, resulting in a leakage of modes $A_{\textrm{g}}$($o$2) and $A_{\textrm{g}}$($o$3) into the $ac$ channel. (D and E) The $aa$- and $ac$-Raman susceptibilities of the same flake, acquired at 100 K. The phonon modes $A_{\textrm{g}}$($m$1)--$A_{\textrm{g}}$($m$11) of the monoclinic phase are labeled. Note that the spectra in (C) and (E) have been scaled by the indicated factors for visualization. (F and G) The $aa$- and $ac$-Raman susceptibilities of bare Au/Cr and Al$_2$O$_3$.
• Figure 3: Detection of structural transition. (A to E) $\chi"_{ac}(\omega)$ spectra at selected temperatures for Ta$_2$NiSe$_5$ flakes of various thicknesses on Al$_2$O$_3$ and Au. Note that the spectra in (C) to (E) have been scaled by the indicated factors for comparison. (F to J) Interpolated 2D color maps of $\chi"_{ac}(\omega, T)$, showing phonon modes $A_{\textrm{g}}$($o$3)* (leaked from the $aa$ channel), $A_{\textrm{g}}$($m$4), and $A_{\textrm{g}}$($m$5). The scale bars have arbitrary units. (K to O) Plots of the second partial derivative $-\partial^2 \chi"_{ac} / \partial \omega^2$ corresponding to (F) to (J). The scale bars have arbitrary units. (P to T) Peak positions of phonon modes $A_{\textrm{g}}$($o$3)*, $A_{\textrm{g}}$($m$4), and $A_{\textrm{g}}$($m$5). The width of the bars corresponds to the wavenumber resolution of the acquired data, 1.3 cm$^{-1}$. The solid triangles mark a kink in the temperature-dependent peak positions, which coincides with the onset of a monoclinic distortion, i.e., $T_{\textrm{c}}$. The open triangles mark a second kink at lower temperatures. From the temperature difference of the first and second kinks (vertical gray bars), we estimate the relative broadening of the structural transition in (S) and (T).
• Figure 4: Detection of electronic transition. (A to E) $\chi"_{ac}(\omega)$ spectra at selected temperatures for Ta$_2$NiSe$_5$ flakes on Al$_2$O$_3$ and Au. The spectra in (C) to (E) have been scaled by the indicated factors. (F to J) Interpolated 2D color maps of $\chi"_{ac}(\omega, T)$, showing phonon modes $A_{\textrm{g}}$($o$1)* (leaked from the $aa$ channel), $B_{\textrm{2g}}$($o$1), $A_{\textrm{g}}$($m$1), and $A_{\textrm{g}}$($m$2). The dashed curves in (H) and (J) are guides to the eye. The scale bars have arbitrary units. (K to O) Fano asymmetry parameter $-1/q$ extracted from fits of $\chi"_{ac}(\omega)$ to (Eq. \ref{['Eq:Fano']}) [dashed black lines in (A) to (E)]. The error bars from the fits are smaller than the circle symbols. In (N) and (O), the data from (M) are plotted for comparison (dashed gray lines). (P to T) Temperature-dependent low-energy spectra, i.e., $\chi"_{ac}(\textrm{10 cm}^{-1}, T)$ [taken along the dashed magenta lines in (F) to (J)]. The horizontal magenta bars include data points within 15% of the peak value, from which we extract $T_{\textrm{c}}$.
• Figure 5: Phase diagram of Ta$_2$NiSe$_5$ thin flakes. (A) Flakes on Al$_2$O$_3$ and (B) flakes on Au. The open circle symbols represent $T_{\textrm{c}}$ values extracted from the structural transition (Fig. \ref{['Fig3']}, P to T). The closed diamond symbols represent $T_{\textrm{c}}$ values extracted from the electronic transition (Fig. \ref{['Fig4']}, P to T). The dashed gray line represents $T_{\textrm{c}}$ = 326 K for bulk Ta$_2$NiSe$_5$. The inset cartoon in (A) shows a 4L Ta$_2$NiSe$_5$/Al$_2$O$_3$ flake in its monoclinic phase below $T_{\textrm{c}}$. The inset cartoon in (B) presents a microscopic picture for the broadened transition in 4L Ta$_2$NiSe$_5$/Au, where the layers closest to the Au substrate have a significantly lower $T_{\textrm{c}}$ due to an interface effect, whereas the surface layers have higher $T_{\textrm{c}}$.