Robust Excitonic‐Insulating States in Cu‐Substituted Ta2NiSe5

Excitonic insulators exhibit intriguing quantum phases that further attract numerous interests in engineering the electrical and optical properties of Ta2NiSe5. However, tuning the electronic properties such as spin‐orbit coupling strength and orbital repulsion via pressure in Ta2NiSe5 are always accompanied with electron‐hole pair breaking, which is a bottleneck for further applications. Here, the robust excitonic‐insulating states invariant with electron‐doping concentrations in Ta2NiSe5 are demonstrated. The electron doping is conducted by substituting Cu into Ni site (Ta2Ni1‐xCuxSe5). The majority carrier of pristine sample is a hole‐type and is converted to electron‐type with a doping concentration over x = 0.01, whose carrier density can be controlled by varying the Cu concentration. The excitonic transition temperature (Tc) does not significantly alter with electron‐doping concentrations, which is stark contrast with the declining Tc as the hole‐type dopant of Fe or Co increases. The optical conductivity data also demonstrate the invariant excitonic‐insulating states in Cu‐doped Ta2NiSe5. The findings of invariant excitonic‐insulating states in n‐type Cu‐substituted Ta2NiSe5 can be utilized for further electronic device applications by using excitons.


Introduction
Excitonic insulator, where conduction band electrons and valence band holes spontaneously form excitons by Coulomb interaction and condensed in semiconductors with low carrier density, is an intriguing quantum phase of solids due to the possibility for realizing various quantum physical insulator. [9] The excitonic phase transition in Ta 2 NiSe 5 has been demonstrated by previous measurements, [12][13][14][15][16] including scanning tunneling spectroscopy (STS), [14] electrical transport, [15] angle-resolved photoemission spectroscopy (ARPES), [12] and optical spectroscopy. [13,16] Recently, several studies have demonstrated to tune the electrical properties of the excitonic-insulating phase of Ta 2 NiSe 5 . [13,[15][16][17][18][19][20][21][22][23] Excitonic-insulating phase was suppressed by applying an external pressure of ≈3 GPa accompanied with phase transformation from excitonic insulator to semimetal. [17] Moreover, superconductivity emerged at a high pressure of ≈8 GPa, originating from the enhanced electron-lattice coupling strength. [18] Excitonic insulator-semimetal transformation was also induced by light excitation, which is confirmed by laser ARPES. [19,20] Resistivity measurements show that the T c was reduced with chalcogen substitutions [13] and thinning the thickness of Ta 2 NiSe 5 as well. [21] Furthermore, the doping effects in 2D materials such as carrier concentration changes and Fermi level shifts play a key role for further applications, realizing highly efficient 2D transistors, [24] photoemitters, [25] and catalysts. [26] While electron-doping effects in Ta 2 NiSe 5 have been realized in terms of K doses adsorbed on the surface, the exciton gaps are shrunken to finally transform into semimetal state. [22,23] The effect of K atoms is still limited to the surface, not into bulk. Therefore, it is prerequisite to explore the relationship between carrier density and ground state of the system in bulk.
In this study, we substituted Cu atom as a dopant into Ni sites in Ta 2 NiSe 5 to understand electron-doping effects on the excitonic-insulating states. X-ray diffractometry and Raman spectrometry measurements confirmed the crystal structures of Ta 2 Ni 1-x Cu x Se 5 . We clearly demonstrate the robustness of excitonic-insulating states by performing and analyzing electrical transport and optical conductivity measurements.

Results and Discussion
The crystal structure of Ta 2 NiSe 5 is presented in Figure 1a. Ta and Ni atoms are alternatively chemically bonded within the layers, which are sandwiched by the slab layers of Se atoms with a separation distance of ≈3.9 Å via van der Waals interaction along b axis, similar to graphene or transition metal dichalcogenides. [27,28] At high temperature above T c = 327 K, the orthorhombic phase is thermodynamically stable with almost zero gap. [13,14,21] Since the exciton binding energy (E ex ) is greater than the bandgap, electrons and holes in conduction and valance bands combine to excitons, which are condensed into the excitonic-insulating state with nonparabolic, i.e., flattened band edge below T c (Figure 1b). The excitonic phase transition in Ta 2 NiSe 5 is mainly contributed primarily from strong exciton-binding energy, while the structural transformation from orthorhombic to monoclinic is a side effect. [13,14,29] The excitonic condensation was confirmed by STS data, which show the sharp peaks within conduction and valance bands below T c (Figure 1c), similar to the previous report. [14] Optical conductivity increases sharply at the band edge (Figure 1d), that implying emergence of exciton condensation below T c . [13,16] The excitonic phase transformation was clearly observed by temperature-dependent resistivity (ρ) (Figure 1e). The excitonic phase transition from zero-gap semiconductor to excitonic insulator was clearly manifested at T c = 327 K by the distinct appearance of anomalous kink from the activation energy (E ρ ) (inset). We next explored the existence of condensation at a Cu concentration of x = 0.1 (Figure 1e). While the conductivity was increased due to electron doping as expected, the similar excitonic-insulating feature was observed with a slightly shifted transition temperature to 321 K. This Cu invariant excitonicinsulating feature is rather striking, stark contrast with the stereotype of charge-doping effect in excitonic insulator with the exciton pairs to be broken with the charge dopants. [30] To prove the excitonic-insulating states invariant with doping concentration in Ta 2 Ni 1-x Cu x Se 5 , we performed comprehensive electrical transport measurements in a wide range of doping concentrations (x = 0-0.1). The reduction of resistivity with temperature indicates the semiconducting behavior for x ≤ 0.1 (Figure 2a). As the doping concentration increases, the resistivity decreases at x ≤ 0.04 and slightly increases at x > 0.04 due to ion scattering at high doping concentration (inset), similar to other materials such as doped Si and other semiconductors. [31][32][33][34] A step-like feature of activation energy is clearly manifested at T c , regardless of doping concentrations ( Figure 2b and Figure S3, Supporting Information). [13,21] The excitonicinsulating gap at T c is 0.35 eV at x = 0 and slightly increases to 0.38 eV at x = 0.01 (inset). At higher doping concentration, the exciton-insulating gap is not appreciably changed. At temperature (400 K) above T c , the bandgap is as small as ≈60 meV.
We next conducted Hall measurements to extract the carrier density ( Figure S4, Supporting Information). Cu dopant can be considered as an n-type dopant in Ta 2 NiSe 5 of electron configurations of guest Cu atoms with [Ar]3d 10 4s 1 compared with host Ni atoms with [Ar]3d 8 4s 2 , thus having excess electrons in Cu atoms. [35][36][37][38] The intrinsic carrier is p-type in Ta 2 NiSe 5 with a hole density of ≈1.3 × 10 16 cm -3 at T = 50 K, ≈1.3 × 10 17 cm -3 at T = 100 K, and ≈4.5 × 10 18 cm -3 at T = 200 K (Figure 2c). At x = 0.01, the carrier type is converted from p-type to n-type with an electron density of ≈4.6 × 10 16 cm -3 at 50 K, larger than intrinsic hole density. The electron density further increases to x = 0.03 and then saturates at higher Cu-doping concentration. The saturated carrier density beyond x = 0.03 originates from the defects generated by Fermi level shift or low impurity ionization energy of dopant in Si and other insulating materials. [31][32][33][34] To explicitly investigate the external p-dopant dependence, we injected the p-type dopant using Co and Fe atoms in host Ta 2 NiSe 5 . The centimeter-scale sample sizes were grown, while retaining sing crystallinity using the same chemical vapor transport method as n-dopant ( Figure S5, Supporting Information). The conductivity was reduced at higher Co-doping concentration except low-temperature region, while being enhanced at higher Fe-doping concentration over a given entire temperature range (Figure 2d, see Figure S6, Supporting Information for ρ). The similar excitonic-insulating features with a sharp peak in E ρ at T c still appeared with Co and Fe dopant at x = 0.01 and 0.03 (inset). The hole density with Co doping at high regime was much larger by two orders of magnitude than that of Fe-doping (Figure 2e, see Figure S7, Supporting Information for Hall data). The conductivity σ = N·e·µ, where N is carrier density of electron or hole, e is elementary charge of 1.602 × 10 -19 C, and www.advmatinterfaces.de µ is mobility of charge carrier. [31] Such conductivity reduction at high Co-doping concentration is explained by dominant impurity scattering to consequently reduce mobility. [31][32][33][34] Figure 2f summarizes T c versus doping concentration. T c is inversely proportional to hole-doping concentration. In contrast, T c is nearly constant (320 K) over a range of electron-doping concentration of 0 < x ≤ 0.1. This implies that Coulomb interaction is strong enough to preserve electron-hole pair condensation in a wide range of doping concentration.
We also performed the crystal structure analysis (Figure 3). To study the crystal structure, we further analyzed the powder X-ray diffraction (XRD) patterns. The 2θ XRD patterns of Ta 2 Ni 1-x Cu x Se 5 were well matched with those of monoclinic phase (i.e., excitonic-insulating phase) (Figure 3a). [15] Such XRD peak positions were not significantly changed with varying Cu-doping concentration of 0 ≤ x ≤ 0.1 (Figure 3a). The XRD peaks of Cu (bottom of inset) were not observed in our powder sample, indicating the absence of Cu aggregates (inset). The peak positions with corresponding d spacings and full width at half maximum (FWHM) of (020), (004), and (111) planes for various x concentrations were plotted in Figure 3b,c. The (020) plane distance, including van der Waals gap, around 6.45 Å with its FWHM of ≈0.11° was independent of Cu-doping concentration. This implies that Cu sites are not intercalated within the layer. Lattice spacings from (004) and (111) planes as well as corresponding FWHMs were rather stochastic with Cu-doping concentrations, indicating no appreciable strains involved within the lattice. Since atomic radius of Cu is 128 pm, much smaller than that of Ta (147 pm), [38] we rule out the possibility of Cu substitution into Ta. Moreover, atomic radii of Cu and Ni (125 pm) Figure 1. a) Crystal structure of monoclinic Ta 2 NiSe 5 . b) Schematic illustration of phase transition from nearly zero gap semiconductor (orthorhombic) to excitonic insulator (monoclinic) with corresponding crystal structure transformation. c) Normalized conductance ((dI/dV)/(I/V)) obtained from STS measurement at 77 K. d) Optical conductivity (σ 1 ) spectrum at T > T c (red curve) and T < T c (blue curve) obtained from reflectance measurement. e) T-dependent ρ of Ta 2 NiSe 5 (black line) and Ta 2 Ni 0.9 Cu 0.1 Se 5 (purple line). The ρ anomalies are marked with arrows by the sharp peak of E ρ at T c (inset).

www.advmatinterfaces.de
are similar to each other, we conclude that Cu atoms are substituted into Ni sites without involving strain. This is confirmed by performing X-ray photoelectron spectroscopy (XPS) analysis (Figure 3d-g): Ta 4f 7/2 and Se 3d peaks did not show significant changes with Cu concentrations with a pass energy of 50 eV (Figure 3d,g). Meanwhile, the intensity of Ni 2p 3/2 peak slightly   (Table S1, Supporting Information). The quantitative ratio of Ta:Ni:Se presented ≈2:1:5, matched with nominal ratio, whereas the Cu concentration was smaller than the nominal x, but gradually increased with x.
Raman spectroscopy measurements were conducted for identifying the room temperature phase of single-crystal Ta 2 Ni 1-x Cu x Se 5 samples (Figure 4a and Figure S9, Supporting Information). Several A g and B 2g peaks at x = 0 are congruent with those of monoclinic (excitonic insulating) phase, which is clearly distinguished from high temperature (above T c ) orthorhombic phase (nearly zero gap semiconductor). [21,39] Such Raman peak positions, for example, strong A g peaks near 96 and 122 cm -1 , did not alter regardless of Cu-doping concentration, indicating that all measured single-crystal Ta 2 Ni 1-x Cu x Se 5 features the monoclinic phase, implying the excitonic-insulating state at room temperature ( Figure 4b). Meanwhile, the Raman peak intensities stochastically varied with Cu-doping concentrations due to random orientation of anisotropic samples from the polarizer ( Figure S10, Supporting Information). Furthermore, the excitonic-insulating phase transformation was clearly demonstrated by using T-dependent optical measurements (Figure 4c and Figure S11, Supporting Information). We analyzed infrared (IR) modes in reflectance spectroscopy with a beam polarized along the c-axis where the optical phonon modes are more clearly visualized ( Figure S11, Supporting Information). The IR modes near 160 cm −1 above T c (Figure 4c, marked with red arrows in 350 K data), are spitted into two IR modes at lower temperature below T c (green arrows in 80-300 K data). The IR mode splitting with temperature below T c was again persistent to manifest the structural phase transformation from the orthorhombic to monoclinic phase (Figure 1b).
We further analyzed the T-dependent optical conductivities for samples of x = 0, 0.05, and 0.1, which were obtained from the reflectance measurements with a beam polarized along the a-axis where the excitons are mostly condensed [13,16] (Figure 5 and Figure S12, Supporting Information). The exciton-binding energy of ≈0.3-0.4 eV was clearly observed with a maximum absorption peak at 80 K and further red-shifted with increasing temperature (Figure 5a-c). As the excitons are condensed with lowering temperature, the excitonic peaks become sharpened (Figure 5a-c and Figure S13, Supporting Information). Also, the absorption edge approach lower photon energy with increasing T and finally disappeared at above T c (Figure 5a-c). The full widths at half maximum (FWHMs) of exciton peak are widened above T c , whereas FWHMs are narrowed with lowering temperature below T c (Figure 5d). The exciton-binding energy from peak position and bandgap (E g ) extracted from absorption edge in terms of temperature and Cu-doping concentration are plotted in Figure 5e. The exciton-binding energy did not change appreciably with Cu doping, featuring the robust excitonic-insulating states invariant with Cu concentration. Meanwhile, the bandgap is reduced with elevating temperature and approaches to zero bandgap near T c . This clearly demonstrates a E gap -T dome shape regardless of Cu-doping concentration (Figure 5f).

Conclusion
In summary, we have successfully demonstrated the robust excitonic-insulating state in electron-doped Ta 2 NiSe 5 and provided insight in electron-doping effects by substitution of Cu atoms into Ni sites. Our results provide a robust excitonicinsulating state in electron-doped samples, stark contrast with conventional charge-doping effect with broken exciton pairs in K-dosed excitonic insulator. [22,23] While Ta 2 NiSe 5 invariant with electron-doping concentration are appealing to reveal the electron-hole pair condensations at T c (≈320 K) above room temperature, which seems to be an ideal material in solids and hence seems to be ideal material platform for quantum electronic devices, realization of excitonic superfluid condensations above room temperature is yet to be explored. Nevertheless, our findings offer an opportunity to use excitons for quantum electronic devices by using robust excitonic-insulating state in Ta 2 Ni 1-x Cu x Se 5 .

Experimental Section
To synthesize Ta 2 Ni 1-x Cu x Se 5 single crystals, the solid-state powder samples were first prepared. Amounts of Ta, Ni, Cu, and Se powders were stoichiometrically mixed and placed in evacuated quartz tube.
The reaction was conducted at 800 °C for 72 h with a heating rate of 100 °C h −1 . The samples were then pulverized to the powder for structural characterization. The powder XRD patterns (Cu K a as a radiation source) were well matched with that of the monoclinic Ta 2 NiSe 5 phase for x between 0 and 0.1. The samples x > 0.1 due to generation of the secondary phases ( Figure S1, Supporting Information) were excluded. The centimeter-scale single crystals of Ta 2 Ni 1-x Cu x Se 5 were grown by chemical vapor transport with solid-reacted powders of Ta 2 Ni 1-x Cu x Se 5 as a source and iodine as a transport agent ( Figure S2, Supporting Information). The ingredients were inserted in evacuated quartz tube and placed in the two-zone furnace. The samples were kept at 950 °C (sample side) and 800 °C (growth zone) for 2 weeks and then quenched to room temperature by turning off the furnace.
Electrical transport properties were measured for the temperature range of 50-400 K by using a physical property measurements system (PPMS Dynacool, Quantum Design) with a 4-probe configuration along the a-axis to exclude the contact resistance. The carrier density of Cu (Co and Fe)-doped samples was obtained from Hall measurement within a magnetic field range of ±9 T (±3 T) at T of 50, 100, and 200 K. The activation energy (E ρ ) was extracted from the temperature-dependent resistivity (ρ) data by using the equation of thermal activation of ρ (ρ = ρ 0 exp(E ρ /k B T)). The T c values could be determined from both kink of ρ and jump of E ρ , calculated from the ρ (E ρ = -k B T 2 (dlnρ/dT)). [13] Photoemission spectroscopy measurement (UHV ESCA system, PREVAC and AXIS Nova, Kratos) was performed to investigate chemical states of Cu-substituted Ta 2 NiSe 5 . To remove surface contaminants, the (0l0) plane of single crystals was cleaved prior to sample loading. Photoemission measurement was carried out under 5 × 10 −10 Torr by www.advmatinterfaces.de using monochromatic Al Kα as an X-ray source (1486.7 eV) with a pass energy of 50 eV. The core level spectra intensities were normalized by using the Ta 4f 7/2 for comparison with other samples. The C 1s level (284.6 eV) was used as an internal standard for calibrating binding energy. A pass energy of 160 eV was used to detect a minute composition of Cu.
Scanning tunneling spectroscopy was performed in ultrahigh vacuum chamber with a base pressure of ≈2.0 × 10 −11 Torr at T = 77 K by using a commercial LT-STM (Omicron, Germany). Electrochemically etched W tips were used after the removal of surface oxides by electron bombardment in ultrahigh vacuum chamber. Tips were calibrated by measuring reference spectra on the HOPG substrate to avoid the tip artifacts. A tunneling bias was applied to the sample. In the STS measurements, a conventional lock-in technique with a voltage modulation of 6-10 mV rms at 919 Hz was used. The relative density of states from STS data was obtained from differential conductance (dI/dV) of current (I)-bias voltage (V) curve.
The wide range (16 cm 1 -24 000 cm −1 ) of reflectance (R(ω)) was measured with an incident angle of 10 o by using a vacuum-type Fouriertransform infrared (FTIR) spectrometer (Bruker Vertex 80v) and a continuous-cold finger-type ARS optical cryostat with a commercial temperature controller (Lakeshore 325) at temperatures between 80 and 350 K. The single crystals larger than 3 × 2 × 0.2 mm 3 dimensions were used for measurements. A linearly polarized beam aligned along the a-and c-axes was used, where excitons were primarily condensed along the a-axis and phonon mode splitting caused by the structure phase transition was clearly visible along the c-axis. An in-situ Au or Al evaporation method was used [40] to obtain accurate and reliable R(ω) of samples. The optical conductivity was obtained from the measured reflectance spectrum (R(ω)) by using the Kramers-Kronig (KK) relation between the amplitude ( R( ) ω ) and the phase (φ(ω)) of the reflection coefficient with the Fresnel equation and relations between optical constants. [41] For the KK analysis, data from 0 to ∞ were extrapolated by using following methods: To extrapolate spectral range from 16 cm −1 to zero, the Hagen-Rubens relation (metallic behavior) of R 1 ( ) ω ω − ∝ was used for data at a high temperature above the structural phase transition temperature (T c ) and R(ω) = constant (insulating behavior) for low temperatures below T c . To extrapolate 24 000 to ∞, data were first adopted from published manuscript of undoped Ta 2 NiSe 5 [42] for the spectral range from 24 000 to 51 500 cm −1 . Spectra were then extrapolated by using power law of R(ω) ∝ ω −1 for 51 000-10 6 cm −1 and R(ω) ∝ ω −4 (free electron behavior) for 10 6 cm −1 -∞.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.