betaFeedback

WorksPhysics and Astronomy

Lattice dynamics of the excitonic insulator Ta2_2Ni(Se1−x_{1-x}Sx_x)5_5

Abstract

Recently, we employed electronic polarization-resolved Raman spectroscopy to reveal the strongly correlated excitonic insulator (EI) nature of Ta2NiSe5, Volkov et al. [arXiv:2007.07344], and also showed that for Ta2_2Ni(Se1−x_{1-x}Sx_x)5_5 alloys the critical excitonic fluctuations diminish with sulfur concentration x exposing a cooperating lattice instability that takes over for large x, Volkov et al. [arXiv:2104.07032]. Here we focus on the lattice dynamics of this EI family. We identify all Raman-active optical phonons of fully symmetric and ac-quadrupole-like symmetries and study their evolution with temperature and sulfur concentration. We demonstrate the change of selection rules at temperatures below the orthorhombic-to-monoclinic transition at Tc(x) that is related to the EI phase. We find that Tc(x) decrease monotonically from 328 K for Ta2NiSe5 to 120 K for Ta2NiS5 and that the magnitude of lattice distortion also decreases with the sulfur concentration x. For x < 0.7, the two lowest-frequency B2g phonon modes show strongly asymmetric lineshapes at high temperatures due to Fano interference with the broad excitonic continuum present in a semimetallic state. Within the framework of extended Fano model, we develop a quantitative description of the interacting exciton-phonon excitation lineshape, enabling us to derive the intrinsic phonon parameters and determine the exciton-phonon interaction strength, that affects the transition temperature Tc(x). We also observe signatures of the acoustic mode scattered assisted by the structural domain walls formed below Tc. Based on our results, we additionally present a consistent interpretation of the origin of oscillations observed in time-resolved pump-probe experiments.

† [email protected]

‡ [email protected]

Introduction

In a narrow-gap semiconductor or semimetal, if the exciton binding energy exceeds the band gap, the Coulomb attraction between electrons and holes favors spontaneous formation of a macroscopic number of excitons. These bosonic quasiparticles then form a coherent state, opening an interaction-induced gap and leading to a insulating phase, named excitonic insulator [1, 2, 3]. If the electron and hole states belong to bands of different symmetry, the resulting excitonic insulator state is expected to break the crystal lattice symmetry.

Consider the case of a semimetal. If the band gap is indirect, i.e. the valence band maximum and the conduction band minimum are located at different places in the kk-space, condensation of excitons leads to a charge-density-wave (CDW) state and accompanying crystal distortion which breaks translational symmetry [3]. The CDW wavevector is then given by the spanning vector that connects the valence band maximum to the conduction band minimum. One such example is semimetal 1T-TiSe2_2, which has been proposed to hold an excitonic insulating state below 190 K [4, 5, 6, 7]. In contrast, if the band gap is direct, excitonic condensation only results in lattice distortion which breaks point-group symmetry. In this situation, the size of the unit cell does not change but its shape changes. Ta2_2NiSe5_5 is a candidate for excitonic insulator of this kind.

Ta2_2NiSe5_5 shows a second-order structural phase transition at Tc=326T_c = 326 K [8, 9]. In the high temperature phase, this material has an orthorhombic structure (point group D2hD_{2h}); in the low-temperature phase, the structural is monoclinic (point group C2hC_{2h}) with the angle β\beta between aa and cc directions deviating from 90°, Figure 1. Below TcT_c, the valence-band top flattens as discovered by angle-resolved photoemission spectroscopy (ARPES) studies [10, 11, 12]. Band flatness at low temperature is viewed as a characteristic feature of the excitonic insulator ground state. Moreover, studies of nonequilibrium dynamics have demonstrated photoinduced enhancement of excitonic order [13, 14], photoinduced multistage phase transitions [15], ultrafast reversal of excitonic order [16], and coherent order-parameter oscillations [17].

Figure 1. Crystal structure of Ta2_2NiSe5_5. The frame in (a) indicates the unit cell.

The lattice degrees of freedom play an important role in the physics of Ta2_2NiSe5_5. In particular, the orthorhombic-to-monoclinic structural change can be induced via electron-phonon coupling [18] even if the origin of the transition is excitonic [19]. Moreover, a recent study combining ultrafast experiments with calculations proposes that the phase transition in Ta2_2NiSe5_5 is structural in its nature [20]. Thus, understanding the underlying physics of excitonic insulator candidates requires the lattice dynamics to be examined in a detailed way. Interestingly, substitution of Se with S results in a whole family of compounds Ta2_2Ni(Se1−x_{1-x}Sx_x)5_5, where the orthorhombic-to-monoclinic transition has been suggested to be suppressed with xx [8]. While it is known, that S substitution leads to an enhanced direct band gap [21], which is detrimental to the EI state, its effects on the lattice modes has not been systematically studied.

Because the inversion symmetry of Ta2_2Ni(Se1−x_{1-x}Sx_x)5_5 is preserved across TcT_c, the soft mode of this zero-wavevector structural transition, if any, is Raman active. Polarization resolved low frequency Raman spectroscopy, therefore, is particularly suitable for studying the physics of the phase transition, and for evaluating the contribution of the lattice vibrations to the transition. This experimental method offers both the high energy resolution and the ability to selectively probe bosonic excitations in different symmetry channels [22].

In this work, we present a systematic study of the lattice dynamics for the family of Ta2_2Ni(Se1−x_{1-x}Sx_x)5_5 (x=0x = 0, 0.25, 0.67, and 1) pristine crystals and alloys to quantitatively explore the spectral parameters of the phonons. We find the change of selection rules, indicating an orthorhombic-to-monoclinic structural transition at Tc(x)T_c(x), which is finite for all xx. The transition temperature Tc(x)T_c(x) and the magnitude of lattice distortion both decrease with sulfur concentration xx.

For Ta2_2Ni(Se1−x_{1-x}Sx_x)5_5 with x≤0.67x \le0.67, the two lowest-energy symmetry-breaking B2gB_{2g} phonon modes exhibit strong asymmetric lineshape above the transition temperature, resulting from the coupling between these modes and an excitonic continuum of the same symmetry. We develop an extended model to analyze the physics of this Fano-type interference effect. Using this model, we disentangle the excitonic and phononic contributions to the Raman response, and illustrate the effect of the coupling between them: apart from renormalizing the bare phononic and excitonic responses, this effect also results in an additional interference term in the total Raman response. We find that the B2gB_{2g} phonon modes exhibit no intrinsic softening on cooling towards the transition temperature. In contrast, the excitonic response shows clear soft-mode behavior [23, 24]. Additionally, many modes exhibit an anomalous increase in linewidth above TcT_c, suggesting the closing of the EI gap. For Ta2_2NiSe5_5 below TcT_c, we also observe signatures of the acoustic excitations with finite momenta, enabled by the formation of a quasi-periodic structure of domain walls.

For Ta2_2NiS5_5, on the contrary, the phonon modes in the symmetry-breaking channel maintain symmetric lineshape, but their frequency anomalously decreases upon cooling. Because the structural change for Ta2_2NiS5_5 is very weak, no anomaly of phonon frequency or width is detected around its structural transition temperature.

For all compositions, we find that most phonons exhibit strong dependence of their intensity on temperature, and present arguments in favor of this being a result of a coupling to electronic excitations.

Additionally, we perform density functional theory (DFT) calculations of the lattice dynamics for Ta2_2NiSe5_5 and Ta2_2NiS5_5. The phonon frequencies agree well with the experiment, and the obtained displacement patterns of the optical modes allow to explain the differences of the couplings of the three non-symmetric phonons to the excitonic continuum.

Finally, we demonstrate that our results are consistent with the modes observed in time-resolved experiments. We show however, that the signatures of the Fano-shaped phonons and the excitonic continuum above TcT_c may be hard to observe in that type of experiments.

The rest of this paper is organized as follows. In Sec. II we describe the sample preparation and experimental setup. In Sec. III we show an overview of the phonon spectra of Ta2_2Ni(Se1−x_{1-x}Sx_x)5_5 family, and compare the measured phonon frequencies at low temperatures with the calculated values. In § 4 from the temperature dependence of the phonon intensity we obtain phase transition temperatures, which are compared with the resistance measurements. In § 5 we analyze the Fano interference feature of B2gB_{2g}-symmetry Raman response. In § 10 we present the temperature dependence of phonon-mode parameters: the results for B2gB_{2g}-symmetry modes are given in SubSec. VIA, and those for AgA_g-symmetry modes are given in SubSec. VIB. In § 13 we convert Raman response from frequency domain to time domain, and discuss relevant time-resolved studies on Ta2_2NiSe5_5. In § 14 we provide a summary of the observations and their implications. Further discussion is provided in appendices: the methods used for the estimation of laser heating rate in Appendix A; the mathematical formalism for the generalized Fano model in Appendix § 6.1, and more illustration of this model in Appendix § 7; the fitting results of the anharmonic decay model for the phonon modes in Appendix § 7.1.

Experimental

Single crystals of Ta2_2Ni(Se1−x_{1-x}Sx_x)5_5 family were grown by chemical vapor transport method. Elemental powders of tantalum, nickel, selenium and sulfur were mixed with stoichiometric ratio and then sealed in an evacuated quartz ampule with a small amount of iodine as the transport agent. The mixture was placed in the hot end of the ampule (∼950∘\sim950^{\circ}C) under a temperature gradient of about 10∘10^{\circ}C/cm. After about a week mm-sized needle-like single crystals were found at the cold end of the ampule. The crystals are shiny and cleave easily. We used x-ray diffraction and electron dispersive X-ray spectroscopy to verify the exact composition of the crystals and their uniformity. The samples for resistance and Raman measurements are from the same batch.

The resistance is measured along aa axis in a four-probe configuration using a Quantum Design PPMS system.

The TEM images of domain structures are taken on a FEI Titan Themis G2 system using a Gatan double-tilt cryo-stage.

For Raman measurements, the samples were cleaved in ambient conditions to expose the acac plane; the cleaved surfaces were then examined under a Nomarski microscope to find a strain-free area. Raman-scattering measurements were performed in a quasi-back-scattering geometry from the samples mounted in a continuous helium-gas-flow cryostat.

For acquisition of the low frequency Raman response required for this study, we used a custom fast f/4f/4 high resolution 500/500/660 mm focal lengths triple-grating spectrometer with 1800 mm−1^{-1} master holographic gratings comprised of (i) aberration corrected subtractive stage providing a reliable 12 orders-of-magnitude stray light rejection at as low frequency as 4 cm−1^{-1} from the elastic line, (ii) a third stage monochromator, and (iii) a liquid-nitrogen-cooled charge-coupled device (CCD) detector (Princeton Instruments). All the acquired data were corrected for the spectral response of the spectrometer. Three slit configurations were used: 100 μ\mum slit width providing 0.19 meV spectral resolution; 50 μ\mum slit width rendering 0.10 meV spectral resolution; and, for high resolution data, 25 μ\mum slit width rendering 0.06 meV spectral resolution.

For polarization optics, a Glan-Taylor polarizing prism (Melles Griot) with a better than 10−510^{-5} extinction ratio to clean the laser excitation beam and a broad-band 50 mm polarizing cube (Karl Lambrecht Corporation) with an extinction ratio better than 1:500 for the analyzer was used. Two polarization configurations were employed to probe excitations in different symmetry channels. The relationship between the scattering geometries and the symmetry channels [22] is given in Table I.

Scattering GeometrySymmetry Channel (D2hD_{2h} group)Symmetry Channel (C2hC_{2h} group)
aaAgA_gAgA_g
acB2gB_{2g}AgA_g

Table I. The Raman selection rules in the high-temperature orthorhombic (point group D2hD_{2h}) and low-temperature monoclinic (point group C2hC_{2h}) phases. Upon the reduction of symmetry from D2hD_{2h} to C2hC_{2h}, the AgA_g and B2gB_{2g} irreducible representations of D2hD_{2h} group merge into the AgA_g irreducible representation of C2hC_{2h} group.

The 647 nm line from a Kr+^{+} ion laser was used for excitation. Incident light was focused to an elongated along the slit direction 50×\times100 μ\mum2^2 spot. For data taken below 310 K, laser power of 8 mW was used. To reach temperature above 310 K, we kept the environmental temperature at 295 K and increased laser power to reach higher sample temperature in the excitation spot. All reported data were corrected for laser heating in two mutually consistent ways [xref?]. (i) by Stokes/anti-Stokes intensity ratio analysis, based on the principle of detailed balance, (ii) by checking laser power that is inducing the phase transition. In addition to that, we performed a thermoconductivity model calculation that suggests a linear scaling of the heating rate with the beam spot size.

The first-principle calculations were performed using DFT within Projector Augmented-Wave (PAW) formalism [25] and Perdew-Burke-Ernzerhof (PBE) parametrized exchange-correlation energy functional [26]. For Ta2_2NiSe5_5, we used the implementation of the Quantum ESPRESSO package [27], using the pseudopotentials generated by Dal Corso [28] and with grimme-d3 Van-der-Waals correction [29] included. For Ta2_2NiS5_5, the implementation of the Vienna Ab initio Simulation Package (VASP) code [30, 31] was used. The phonon frequencies for both cases were calculated by finite displacement method as implemented in PHONOPY [32]. The initial structures subject to relaxation for Ta2_2NiSe5_5 and Ta2_2NiS5_5 have been taken from experiments on the high-temperature orthorhombic phase, Ref.[33] and Ref.[34], respectively.

Overview

In this section we show an overview of the phonon spectra of Ta2_2Ni(Se1−x_{1-x}Sx_x)5_5 family, and compare the measured phonon frequencies at low temperatures with the calculated values.

The phonon spectra of Ta2_2Ni(Se1−x_{1-x}Sx_x)5_5 family (xx=0, 0.25, 0.67, and 1) are summarized in Figure 2. For stoichiometric compositions at high temperature, we observe 8 modes in aaaa geometry and 3 modes in acac geometry, in agreement with the 8 AgA_g and 3 B2gB_{2g} Raman-active phonon modes expected in the high-temperature orthorhombic phase (space group CmcmCmcm): the AgA_g modes appear in aaaa scattering geometry and the B2gB_{2g} modes in acac scattering geometry [Table. I]. Except for the Ta2NiS5{\mathrm{Ta}}_2{\mathrm{NiS}}_5 sample, the two lowest-energy B2gB_{2g} modes, B2g(1)B_{2g}^{(1)} and B2g(2)B_{2g}^{(2)}, exhibit strongly broadened and asymmetric lineshapes at high temperature. This anomalous broadening is almost absent for Ta2NiS5{\mathrm{Ta}}_2{\mathrm{NiS}}_5 (note the logarithmic intensity scale in Figure 2).

Raman response spectra in the $aa$ and $ac$ scattering geometries for ${\mathrm{Ta}}_2{\mathrm{NiSe}}_{5-x}{\mathrm{S}}_x$

Figure 2. Raman response χ′′\chi^{\prime\prime} in the aaaa and acac scattering geometries for Ta2NiSe5−xSx{\mathrm{Ta}}_2{\mathrm{NiSe}}_{5-x}{\mathrm{S}}_x family, plotted in semi-log scale. The phonon modes are classified by the AgA_g and B2gB_{2g} irreducible representations of D2hD_{2h} point group. The AgA_g and B2gB_{2g} modes of Ta2NiSe5{\mathrm{Ta}}_2{\mathrm{NiSe}}_5 are labeled by red and green colors, respectively; the AgA_g and B2gB_{2g} modes of Ta2NiS5{\mathrm{Ta}}_2{\mathrm{NiS}}_5 are labeled by purple and blue colors, respectively. For the samples with x=0.25x=0.25 and 0.670.67, the modes are labeled by the color scheme of Ta2NiSe5{\mathrm{Ta}}_2{\mathrm{NiSe}}_5 if its energy is close to that of Ta2NiSe5{\mathrm{Ta}}_2{\mathrm{NiSe}}_5 modes, or the color scheme of Ta2NiS5{\mathrm{Ta}}_2{\mathrm{NiS}}_5 if its energy is close to that of Ta2NiS5{\mathrm{Ta}}_2{\mathrm{NiS}}_5 modes. The spectral resolution is 0.19 meV for the acac spectra at high temperature, and 0.06 meV for the acac spectra at 35 K and all aaaa spectra.

At low temperature, a transition into a monoclinic phase (space group C2/cC2/c) occurs. In the low-temperature phase, because of the mirror symmetry breaking, all Raman-active phonon modes appear in both aaaa and acac scattering geometries [Table. I]. All of the expected Raman-active modes were observed: they are identified by dashed lines in Figure 2; the additional weak spectral features result from the second-order scattering. For example, the weak feature at 34.1 meV in aaaa scattering geometry for Ta2NiSe5{\mathrm{Ta}}_2{\mathrm{NiSe}}_5 corresponds to a second-order scattering feature of the 16.8 meV B2g(2)B_{2g}^{(2)} mode.

ModeTa2NiSe5{\mathrm{Ta}}_2{\mathrm{NiSe}}_5 Exp.Ta2NiSe5{\mathrm{Ta}}_2{\mathrm{NiSe}}_5 Calc.Ta2NiS5{\mathrm{Ta}}_2{\mathrm{NiS}}_5 Exp.Ta2NiS5{\mathrm{Ta}}_2{\mathrm{NiS}}_5 Calc.
Ag(1)A_g^{(1)}4.224.225.014.82
B2g(1)B_{2g}^{(1)}8.667.616.507.66
Ag(2)A_g^{(2)}12.2811.8716.1315.18
B2g(2)B_{2g}^{(2)}15.2510.1416.4116.20
Ag(3)A_g^{(3)}16.7615.5319.1318.27
B2g(3)B_{2g}^{(3)}18.3819.3932.9633.23
Ag(4)A_g^{(4)}22.1721.9133.8233.82
Ag(5)A_g^{(5)}24.1424.0936.5036.36
Ag(6)A_g^{(6)}27.0226.3440.0740.13
Ag(7)A_g^{(7)}29.3929.0143.1043.07
Ag(8)A_g^{(8)}36.3237.4749.4949.62

Table II. TABLE II. Comparison of the experimentally measured phonon energies at 35 K and the calculated values for Ta2NiSe5{\mathrm{Ta}}_2{\mathrm{NiSe}}_5 and Ta2NiS5{\mathrm{Ta}}_2{\mathrm{NiS}}_5. Unit is meV; modes are identified by their irreducible representations in the orthorhombic phase, where 8 AgA_g modes and 3 B2gB_{2g} modes are expected.

The measured phonon frequencies for Ta2NiSe5{\mathrm{Ta}}_2{\mathrm{NiSe}}_5 and Ta2NiS5{\mathrm{Ta}}_2{\mathrm{NiS}}_5 at 35 K generally match well with the calculated values [Table. II]. As the calculations were carried out in the orthorhombic phase, this suggests that monoclinicity by itself does not strongly affect the phonon frequencies. Larger than 10% discrepancies between the measured and calculated values appear only for the B2g(1)B_{2g}^{(1)} and B2g(2)B_{2g}^{(2)} modes of Ta2NiSe5{\mathrm{Ta}}_2{\mathrm{NiSe}}_5 and the B2g(1)B_{2g}^{(1)} mode of Ta2NiS5{\mathrm{Ta}}_2{\mathrm{NiS}}_5. As discussed in § 5, these modes exhibit an anomalous behavior that can be attributed to their strong coupling to the excitons.

For the alloy compositions (x=0.25x=0.25 and 0.670.67), the phonon modes show larger linewidth compared with that for the stoichiometric compositions. Moreover, the B2g(3)B_{2g}^{(3)} and Ag(3)A_g^{(3)}-Ag(8)A_g^{(8)} modes exhibit two distinct frequencies: one close to the frequency of that mode in Ta2NiSe5{\mathrm{Ta}}_2{\mathrm{NiSe}}_5, and the other close to the frequency in Ta2NiS5{\mathrm{Ta}}_2{\mathrm{NiS}}_5. Such behavior is commonly observed for alloys in which the frequencies of the same phonon mode in the two end-point materials differ substantially [35, 36, 37]. The doping dependence of the phonon frequencies measured at 35 K is presented in Figure 3. The frequencies of the phonon modes for Ta2NiSe5{\mathrm{Ta}}_2{\mathrm{NiSe}}_5 are consistent with the recent experimental studies [38, 39, 40, 41, 42] and the calculated values [43].

Phonon energy versus sulfur content for Raman-active phonons

Figure 3. The doping dependence of the Raman-active phonon frequencies at 35 K for Ta2_2Ni(Se1−x_{1-x}Sx_x)5_5 family. For the doped samples, the B2g(3)_{2g}^{(3)} and Ag(3)_g^{(3)}-Ag(8)_g^{(8)} phonons exhibit two-mode behavior (i.e. two peaks in the spectrum), characteristic of alloys [35, 36, 37].

Phase Transition

In this section we deduce the phase transition temperatures from the observed change in selection rules in Raman response and compare the results with the resistance measurements.

Below TcT_c, the 11 phonon modes are allowed by symmetry to appear in both scattering geometries [Table. I]; thus, 8 additional modes appear in acac geometry and 3 in aaaa. We call the appearance of phonon modes below TcT_c in the orthogonal to allowed above TcT_c scattering geometry as phonon-mode ”leakage”. As shown in Figure 2, all four samples exhibit phonon-mode ”leakage” at low temperature, indicating presence of a point group symmetry breaking structural phase transition.

In Figure 4 we show the temperature dependence of the ”leakage” of Ag(1)_g^{(1)} and B2g(1)_{2g}^{(1)} modes for Ta2_2Ni(Se1−x_{1-x}Sx_x)5_5 family. The ”leakage” of other modes is given in Sec. VI. For each sample, the ”leakage” of different modes appear below the same temperature.

Figure 4. Temperature dependence of the integrated intensity of Ag(1)_g^{(1)} and B2g(1)_{2g}^{(1)} modes in the ’forbidden’ scattering geometry (acac and aaaa, respectively) for (a) Ta2_2NiSe5_5, (b) Ta2_2NiS5_5, (c) Ta2_2Ni(Se0.75_{0.75}S0.25_{0.25})5_5, and (d) Ta2_2Ni(Se0.33_{0.33}S0.67_{0.67})5_5.

We assign the temperature of the ”leakage” onset as

Sulfur Content xxTransition Temp. (K)
0328±5
0.25260±10
0.67170±10
1120±10

Table 3. The structural phase transition temperature of Ta2_2Ni(Se1−x_{1-x}Sx_x)5_5 family determined by observing phonon-mode ”leakage” in Raman spectra.

the structural phase transition temperature TcT_c. The phase transition temperatures for Ta2_2Ni(Se1−x_{1-x}Sx_x)5_5 family are given in Table III. The transition temperatures Tc(x)T_c(x) decrease with sulfur doping xx. Moreover, for larger sulfur content xx the phonon intensities in the ’forbidden’ scattering geometry also decrease. This quantity is proportional to the square of the order parameter (OP), which can be represented by, e.g., the deviation of the angle β\beta between a and c axes [Figure 1 (b)] from 90°. While the general monotonic growth of the leakage intensity on cooling is consistent with these expectations, the details of temperature dependence can vary for different modes [Figure 4(a)]. The decreasing trend of the phonon intensity in the ’forbidden’ scattering geometry with sulfur content xx therefore indicates that the structural distortion becomes weaker with higher sulfur concentration.

A complementary way to determine the phase transition temperature was demonstrated in [8], where anomalies in the temperature dependence of resistance have been detected at Tc(x)T_c(x). In Figure 5 we show the resistance data analysis for the samples from the same batch as used for the Raman measurements. For Ta2_2NiSe5_5 and Ta2_2Ni(Se0.75_{0.75}S0.25_{0.25})5_5, the temperature at which temperature derivative of resistance displays a kink and that of transport activation gap shows a peak, coincides with the Raman-determined transition temperature. For Ta2_2Ni(Se0.33_{0.33}S0.67_{0.67})5_5 and Ta2_2NiS5_5, though, the features in resistance data are less pronounced. The latter can be anticipated from their much more insulating character, potentially masking a small change of resistance due to transition on top of a large background resistance.

Temperature dependence of resistance and transport activation gap for Ta₂Ni(Se₁−xSx)₅

Figure 5. Temperature dependence of the resistance along aa axis for Ta₂Ni(Se₁−xSx)₅ family. (a) The resistance, plotted in semi-log scale. (b) The temperature derivative of resistance, plotted in semi-log scale. (c) The temperature derivative of the transport activation gap Eg=kBTlog⁡10[R(T)/R(360K)]E_g = k_B T \log_{10}[R(T)/R(360K)]. The transition temperatures determined from Raman measurements are marked by dashed lines for comparison.

Fano interference in the B2g_{2g} symmetry channel

In this section we analyze the asymmetric lineshape of the B2g_{2g}-symmetry phonon modes observed for x<0.7x < 0.7 in the high temperature phase. We describe the asymmetric lineshape and its temperature dependence in Sub-Sec. V A. To fit the spectra, we develop a generalized Fano model, discussed in SubSec. V B, in which the asymmetric broadening results from the coupling of the phonon modes to an excitonic continuum. Besides the optical phonon modes, the excitonic continuum also couples to the acoustic phonon mode. We discuss how such coupling enhances the transition temperature Tc(x)T_c(x), and leads to softening of the B2gB_{2g}-symmetry acoustic mode in SubSec. V C. The appearance of a quasi-periodic domain wall structure below TcT_c introduces additional complexity, and, in SubSec. V D, we discuss how the fitting model must be modified to account for low-energy spectral features which are absent above TcT_c. The fitting results for Ta₂NiSe₅, relevant to the excitonic continuum, are presented and interpreted in SubSec. V E (those related to the intrinsic phonon properties are shown later in § 10). It is demonstrated that the strengths of the interactions between individual phonons and excitonic continuum are consistent with the displacement patterns deduced from ab initio calculations. For comparison, the fitting results for the alloy compositions are given in SubSec. V F.

Temperature dependence of asymmetric Fano lineshapes

The temperature dependence of the spectra measured in acac scattering geometry for Ta₂NiSe₅ is shown in Figure 6. Above TcT_c, the B2g(1)B_{2g}^{(1)} and B2g(2)B_{2g}^{(2)} modes have noticeably asymmetric lineshapes, while the lineshape of the B2g(3)B_{2g}^{(3)} mode remains quite symmetric. Upon approaching TcT_c on cooling, the spectral weight of low-frequency continuum displays an enhancement, and the apparent linewidths of the B2g(1)B_{2g}^{(1)} and B2g(2)B_{2g}^{(2)} modes increase. Below TcT_c, the modes recover the conventional Lorentzian lineshapes.

Temperature dependence of Raman response

Figure 6. Temperature dependence of Raman response χ′′\chi'' in the acac scattering geometry for Ta₂NiSe₅. The data is shown by black dots, along with the Fano fits [Eq. (B1)] shown as red curves. The spectral resolution is 0.19 meV for panels (a-e), 0.10 meV for panels (f-g), and 0.06 meV for panel (h). The Fano model describes the coupled dynamics of the B2g(1)B_{2g}^{(1)}-B2g(3)B_{2g}^{(3)} phonon modes and the overdamped excitonic continuum. The bare phonon modes [Eq. (1)] are shown by green curves, and the bare excitonic continuum [Eq. (2)] is shown by blue curves. Below TcT_c, especially at 300 K, additional spectral weight is observed at low energies, resulting from the emergence of finite-energy acoustic modes due to scattering off structural domain walls (see Figure 9). The purple curves show the coupled response of the acoustic modes and the excitonic continuum. As the excitonic continuum is suppressed on cooling below TcT_c, the lineshapes of the phonons become essentially symmetric.

An asymmetric lineshape arises due to Fano interference when a sharp phonon interacts with a broad continuum. In this case, the continuum is due to the overdamped exciton fluctuations [23, 24], and the exciton-phonon interaction is caused by modulation of electronic bands near the Fermi surface by the B2gB_{2g}-symmetry lattice vibrations.

The generalized Fano model for data above TcT_c

We describe the Fano model that includes the coupling between the B2g(1−3)B_{2g}^{(1-3)} phonon modes and the continuum; more details can be found in Appendix § 6.1.

We first discuss the Raman response of the phonons. The Raman scattering intensity is related to the fluctuation spectrum of the phonon coordinate ⟨QωQ−ω⟩\langle Q_\omega Q_{-\omega}\rangle [44], which is given by the imaginary part of the phonon coordinate susceptibility χp′′(ω)\chi_p^{\prime\prime}(\omega) times the Bose factor. The phononic Raman response function consequently has the form:

χp′′(0)(ω)=4tp2ωωpγp(ω2−ωp2)2+2γp2(ω2+ωp2)+γp4(1)\chi_p^{\prime\prime(0)}(\omega)=\frac{4t_p^2\omega\omega_p\gamma_p}{(\omega^2-\omega_p^2)^2+2\gamma_p^2(\omega^2+\omega_p^2)+\gamma_p^4} \tag*{(1)}

where tpt_p is the light-scattering vertex for the phonon, ωp\omega_p is the mode’s bare frequency, and γp\gamma_p is the half width at half maximum (HWHM) which is related to the oscillator lifetime.

An exciton in a gapped system (a semiconductor or insulator) is expected to have a sharply peaked response at the exciton energy described by a bosonic response function similar to the Eq. (1). The most important difference for a semimetal is the presence of a gapless continuum of interband particle-hole excitations enabling the decay of the q=0q=0 exciton into unbound particle-hole pairs (Landau damping). To describe the continuum in data, we find it sufficient to assume a purely relaxational dynamics (corresponding to strong overdamping) for the excitonic response. This leads to the form

χe′′(0)(ω)=te2ω(ωe2/γe)2+ω2(2)\chi_e^{\prime\prime(0)}(\omega)=\frac{t_e^2\omega}{(\omega_e^2/\gamma_e)^2+\omega^2} \tag*{(2)}

where tet_e controls the overall intensity determined by the light-scattering vertex and the excitonic density of states, ωe\omega_e is the frequency of the over-damped excitation, and γe\gamma_e is the relaxation rate, typically much larger than ωe\omega_e. Note that for the purpose of the fitting description below Ωe(T)=ωe2/γe\Omega_e(T)=\omega_e^2/\gamma_e is a single parameter denoting the temperature-dependent peak frequency in the broad response function χe′′(ω)\chi_e^{\prime\prime}(\omega). For a purely excitonic transition this peak represents the soft mode: Ωe(Teel)=0\Omega_e(T_e^{\mathrm{el}})=0 is the condition for bare excitonic transition temperature at TeelT_e^{\mathrm{el}}, where the static excitonic susceptibility χe(0)\chi_e(0) diverges (see Eq. (54)).

In the absence of the exciton-phonon interaction between the modes the respective phononic and excitonic responses Eqs. (1) and (2) simply sum up. The exciton-phonon interaction, however, couples these dynamical responses: in particular, a bilinear coupling of the exciton and the B2gB_{2g} phonon coordinates is expected [18] (note that the bare phonons are diagonal normal modes and thus no bilinear interaction between them is present). The resulting response can be obtained by solving the coupled equations of motion Appendix § 6.1. Each of the responses χp′′(ω)\chi_p^{\prime\prime}(\omega) and χe′′(ω)\chi_e^{\prime\prime}(\omega) get renormalized by the interaction, and in addition, the response of the phonon coordinate to the exciton one, χint′′(ω)\chi_{\mathrm{int}}^{\prime\prime}(\omega), becomes finite. The latter leads to an essential interference term in the total Raman response function [45, 46, 47].

To appreciate the effect of renormalization, we first consider a simplified case in which the excitonic continuum couples to a single sharp phonon mode (γp≪ωp\gamma_p\ll\omega_p) Appendix § 7. Then, solving the coupled dynamical equations, one obtains the full Raman response function

χ′′(ω)=ω[2tpvωp−te(ω2−ωp2)]2(Ωe2+ω2)(ω2−ωp2)2+4Ωe2vωp(ω2−ωp2)+4v2ωp2(3)\chi^{\prime\prime}(\omega)=\frac{\omega[2t_pv\omega_p-t_e(\omega^2-\omega_p^2)]^2}{(\Omega_e^2+\omega^2)(\omega^2-\omega_p^2)^2+4\Omega_e^2v\omega_p(\omega^2-\omega_p^2)+4v^2\omega_p^2} \tag*{(3)}

could be broken down into a sum of three contributions:

χ′′(ω)=χp′′(ω)+χe′′(ω)+χint′′(ω),(4)\chi^{\prime\prime}(\omega)=\chi_p^{\prime\prime}(\omega)+\chi_e^{\prime\prime}(\omega)+\chi_{\mathrm{int}}^{\prime\prime}(\omega), \tag*{(4)}

where the first two terms correspond to the phonon response proportional to square of light coupling vertex tp2t_p^2 and the excitonic continuum proportional to te2t_e^2, respectively, while the third one, that is proportional to the tptet_pt_e combination, is the interference term appearing due to the exciton-phonon coupling with strength vv [47].

We first discuss the phonon response in the presence of the coupling with the excitonic continuum. As we will show now, the coupling to the continuum of overdamped excitonic excitations renormalizes the bare linewidth to

γpv=v2ωpωp2+Ωe2,(5)\gamma_p^v=\frac{v^2\omega_p}{\omega_p^2+\Omega_e^2}, \tag*{(5)}

thus, the bare γp\gamma_p can be neglected altogether [48] The renormalized phonon response is given by

χp′′(ω)=4tp2ωωpγpv[ω2−(ωp2−2Ωeγpv)]2+4ωp2(γpv)2+(ω2−ωp2)3ωp2+Ωe2(6)\chi_p^{\prime\prime}(\omega)=\frac{4t_p^2\omega\omega_p\gamma_p^v}{\left[\omega^2-\left(\omega_p^2-2\Omega_e\gamma_p^v\right)\right]^2+4\omega_p^2(\gamma_p^v)^2+\frac{(\omega^2-\omega_p^2)^3}{\omega_p^2+\Omega_e^2}} \tag*{(6)}

Comparing Eq. (6) to Eq. (1) for the bare phonon, the phonon frequency renormalization is ωp2→ωp2−2Ωeγpv\omega_p^2\to\omega_p^2-2\Omega_e\gamma_p^v [49]. On increasing coupling strength vv, the phononic response maximum shifts to ωp,max⁡=ωp−v2(2ωpΩe−v2)2v2Ωe+ωp(Ωe2+ωp2)\omega_{p,\max}=\omega_p-\frac{v^2(2\omega_p\Omega_e-v^2)}{2v^2\Omega_e+\omega_p(\Omega_e^2+\omega_p^2)}.

For the excitonic response χe′′(ω)\chi_e^{\prime\prime}(\omega) the renormalization effects are most significant. Those are best illustrated by the low-frequency slope of the excitonic response function ∂χe′′(ω)∂ω∣ω→0\left.\frac{\partial\chi_e^{\prime\prime}(\omega)}{\partial\omega}\right|_{\omega\to0}. In the absence of exciton-phonon interaction the low frequency slope for the bare response is equal to te2/Ωe2t_e^2/\Omega_e^2 (see Eq. (2)), while the interaction vv is rapidly enhancing the slope as

∂χe′′(ω)∂ω∣ω→0=te2[Ωe(T)−2v2ωp2]2,(7)\left.\frac{\partial\chi_e^{\prime\prime}(\omega)}{\partial\omega}\right|_{\omega\to0}=\frac{t_e^2}{\left[\Omega_e(T)-\frac{2v^2}{\omega_p^2}\right]^2}, \tag*{(7)}

leading to a critical value at vcr=Ωeωp/2v_{\rm cr}=\sqrt{\Omega_e\omega_p/2}, when the excitonic spectral weight is pushed to the lowest frequencies and thus causing the divergence in the static susceptibility

χ(ω→0,T)=te2−4tetpvωp+2tp2Ωe(T)ωpΩe(T)−2v2ωp,(8)\chi(\omega\to0,T)=\frac{t_e^2-4t_et_p\frac{v}{\omega_p}+2t_p^2\frac{\Omega_e(T)}{\omega_p}}{\Omega_e(T)-\frac{2v^2}{\omega_p}}, \tag*{(8)}

signifying the enhancement of the excitonic phase transition temperature, caused by coupling to the optical phonon.

Finally, the mutual response of the exciton and phonon coordinates appearing due to the coupling vv between them, leads to the sign-changing interference term

χint′′(ω)=4tetpvωωpωp2−ω2Ωe2+ωp2[ω2−(ωp2−2Ωeγp′)]2+4ω2γp′2+(ω2−ωp2)3ωp2+Ωe2.(9)\chi_{\rm int}^{\prime\prime}(\omega)=\frac{4t_et_pv\omega\omega_p\frac{\omega_p^2-\omega^2}{\Omega_e^2+\omega_p^2}}{\left[\omega^2-\left(\omega_p^2-2\Omega_e\gamma_p^\prime\right)\right]^2+4\omega^2\gamma_p^{\prime2}+\frac{(\omega^2-\omega_p^2)^3}{\omega_p^2+\Omega_e^2}}. \tag*{(9)}

The sign of this term depends on phase difference between the exciton and phonon oscillators. Because the phase of driven by light phononic oscillator is flipping to the opposite one at the resonant frequency, the sign of this term changes close to the bare phonon frequency ωp\omega_p. Thus, this term is chiefly responsible for a skewed, asymmetric shape of the resulting Fano feature. Depending on the sign of vv the peak is skewed to the left or to the right of the original phonon frequency. Note that even for weak coupling v≪ωpv\ll\omega_p this term can have an appreciable magnitude close to the phonon energy.

To analyze the actual measured low-frequency data shown in Figure 6, we use the model with three B2gB_{2g} optical phonon modes, all interacting with the excitonic continuum, see for details Appendix § 6.1. As above, the total Raman response χ′′(ω)\chi^{\prime\prime}(\omega) can be similarly decomposed into three contributions, Eq. (4): the first two describing the renormalized phononic and excitonic responses, respectively, with the third one arising due to the interference effects. In Figure 7, an example of such decomposition is shown for the data taken at 380 K, along with the deduced bare responses (without the effect of coupling), χp′′(0)(ω)\chi_{p}^{\prime\prime(0)}(\omega) and χe′′(0)(ω)\chi_{e}^{\prime\prime(0)}(\omega). The renormalized phonon features become broader due to the interaction with the continuum and their frequencies shift. Most importantly, the coupling noticeably increases the excitonic response at low frequencies. The sign-changing interference term enhances the asymmetric lineshape of the combined response.

Figure 7. The effect of coupling on Raman spectrum, relevant to the data measured for Ta2_2NiSe5_5 at 380 K. The Raman response functions are shown for bare (dashed lines) and renormalized by interaction (solid lines), χe,p′′(0)(ω)\chi_{e,p}^{\prime\prime(0)}(\omega) and χe,p′′(ω)\chi_{e,p}^{\prime\prime}(\omega) correspondingly. The interference term χint′′(ω)\chi_{\rm int}^{\prime\prime}(\omega) is shown by solid orange line. The solid red line corresponds to the total sum of the solid blue, solid green, and solid orange lines, Eq. (4). The black dots represent the experimentally measured spectrum.

To fit the acac spectra, first, we determine the tet_e parameter of the excitonic continuum (Eq. (2)) by fitting the phonon free region of spectra between 25 and

40 meV. As expected, this quantity remains temperature-independent [Fig. 10(b)]. Second, we perform a global fitting of the spectral region below 22 meV for all measured temperatures. We assume (consistently with the expectation due to anharmonic effects, see Eq. (16)-(17) below) the linewidth of the three B2gB_{2g}-symmetry phonon modes to be a linear function of temperature above TcT_c. For spectra below TcT_c, we fit the “leaked” phonon modes with the Lorentzian lineshapes, and account for their contribution in the global fitting.

The coupling between excitonic continuum and acoustic modes

Besides the optical phonon modes, the excitonic continuum also couples to the acoustic phonon modes. The acoustic branch has a linear dispersion ωs(q)=csq\omega_s(q)=c_s q (csc_s is sound velocity) at small wavevectors qq; its coupling to light and to the excitonic continuum both vanish in the long-wavelength limit [50]: ts=τsqt_s=\tau_s\sqrt{q} and vs=βsqv_s=\beta_s\sqrt{q} (τs\tau_s and βs\beta_s are constants). However, the acoustic mode has important contribution to the enhancement of the phase-transition temperature [51]. The diverging condition for static Raman susceptibility χ(ω→0,T)\chi(\omega\to0,T) is defined by coupling to all modes, optical and acoustic:

Ω~e(Tc)−2βs2cs=0,(10)\widetilde{\Omega}_e(T_c)-\frac{2\beta_s^2}{c_s}=0, \tag*{(10)}

in which

Ω~e(T)=Ωe(T)−∑i=132vi2ωpi2.(11)\widetilde{\Omega}_e(T)=\Omega_e(T)-\sum_{i=1}^{3}\frac{2v_i^2}{\omega_{pi}^2}. \tag*{(11)}

This equation determines the symmetry-breaking phase transition temperature TcT_c.

Because the spectroscopic Fano feature associated with the coupling between the long-wavelength longitudinal acoustic mode and the excitonic continuum lies at frequencies far below accessibility of Raman experiment, the constant βs\beta_s cannot be derived directly from an interference feature of a typical Raman spectra. Nevertheless, the coupling constant βs\beta_s has observable consequences already above TcT_c. In particular, the acoustic mode has been shown to soft on cooling towards TcT_c [52]. This softening does not result from intrinsic instability of the acoustic modes; rather, it is caused by the coupling of the acoustic mode to the softening excitonic excitations [Eq. (2)].

To demonstrate this point, we solve for the lowest-energy pole of the Green’s function for the interacting phononic and excitonic excitations assuming an infinitesimal qq. The energy of the pole gives the renormalized (experimentally-measured) sound velocity c~s\tilde{c}_s above TcT_c [Eq. (B23) of Appendix B]:

c~s2=cs2−2β2csΩ~e(T)(12)\tilde{c}_s^2=c_s^2-\frac{2\beta^2c_s}{\widetilde{\Omega}_e(T)} \tag*{(12)}

in which csc_s is the bare sound velocity in the absence of coupling to the excitons. Eq. (10) enables us to rewrite Eq. (12) as

c~s(T)=cs1−Ω~e(Tc)Ω~e(T)(13)\tilde{c}_s(T)=c_s\sqrt{1-\frac{\widetilde{\Omega}_e(T_c)}{\widetilde{\Omega}_e(T)}} \tag*{(13)}

From Eq. (13), it is clear that the sound velocity softens to zero at TcT_c in the absence of any intrinsic ferroelastic instability. The renormalization of the sound velocity above TcT_c can be estimated by using Eq. (13). Using the low-temperature value of the sound velocity csc_s [52], we can then calculate c~s(T)\tilde{c}_s(T) with the parameters obtained from the Fano fits, i.e. without free parameters. The result yields c~s(T=400K)/cs≈0.65\tilde{c}_s(T=400\mathrm{K})/c_s\approx0.65, which is in remarkable agreement with the experimental ratio c~s,exp⁡(T=400K)/cs,exp⁡≈0.65\tilde{c}_{s,\exp}(T=400\mathrm{K})/c_{s,\exp}\approx0.65. The agreement between the estimated and experimentally-determined values suggests that in Ta2_2NiSe5_5, the softening of the acoustic mode results solely from the coupling to the softening excitonic excitations.

The generalized Fano model for data below TcT_c

Below TcT_c, especially around 300 K, this model appears to be insufficient to account for an additional low-frequency spectral weight, see Figure 6(e) and Figure 8. To understand the origin of this spectral feature below a few meV, we recall that below TcT_c formation of quasi-periodic structural domains have been observed by transmission electron microscopy (TEM) [9, 15]. Such quasi-periodic structure can take a recoil of quasi-momenta, enabling Raman coupling to acoustic mode at finite momentum qd=2πdq_d=\frac{2\pi}{d}, where dd is the periodicity of the domain pattern, causing the appearance of the additional spectral feature at frequency ωd≈csqd\omega_d\approx c_s q_d. Therefore, fitting this ultra-low-frequency spectral feature provides an independent approach to determine the coupling constant νd=βsqd\nu_d=\beta_s\sqrt{q_d} between the acoustic lattice excitations and the excitonic continuum, and hence define the constant βs\beta_s that controls the enhancement of the transition temperature to TcT_c due to coupling between excitons and longitudinal strain fields of the same symmetry. This low-energy feature due to recoil on the quasi-periodic structure of domain walls is best seen about 30 K below TcT_c, when the excitonic continuum is sufficiently suppressed but still adequate to provide a Fano-interference feature.

Raman spectrum showing acoustic, excitonic, and coupled components

Figure 8. The contribution from the acoustic modes to low-frequency Raman spectrum in acac polarization, relevant to the data measured for Ta2_2NiSe5_5 at 300 K. The data is shown by black dots; the total Fano fit is shown as red curve; the bare excitonic and phononic components are presented in blue and green, respectively; the coupled response of the acoustic and excitonic components is shown in purple.

In Figure 9 we compare two TEM images from a plane of Ta2_2NiSe5_5 measured above and below TcT_c. The domains have stripe shape, aligned parallel to the aa-axis. The average spacing dˉ\bar{d} between these stripes along cc-axis direction is on the order of 200 Å. For the purpose of spectral analysis, we will assume that the spacing between stripes ii in units of the crystal unit cell constant cc follows the Poisson distribution

Two images of the Ta$_2$NiSe$_5$ recorded from the $ac$ plane by transmission electron microscopy

Figure 9. Two images of the Ta2_2NiSe5_5 recorded from the acac plane by transmission electron microscopy at (a) 350 K and (b) 300 K.

P(i;μ=dˉc)=μie−μi!,(14)P\left(i;\mu=\frac{\bar{d}}{c}\right)=\frac{\mu^i e^{-\mu}}{i!}, \tag*{(14)}

where μ\mu is average inter-domain distance in the number of cc-direction unit cells.

We deduce the unrenormalized speed of sound from the dispersion of the acoustic mode measured by inelastic x-ray scattering at temperatures far away from TcT_c: cs≈30 meV A˚c_s\approx30\ \mathrm{meV}\,\text{\AA}, see Ref. [52]. We also note that at high temperatures the acoustic excitations away from long-wavelength limit are significantly broadened [52].

Recognizing that the low-frequency feature is inhomogeneously broadened by random distribution of the stripe distances, for the fitting procedure to the measured response function we perform summation over the distribution function:

χ′′(ω)=∑i=1∞P(i;μ)χi′′(ω),(15)\chi''(\omega)=\sum_{i=1}^{\infty}P(i;\mu)\chi_i''(\omega), \tag*{(15)}

in which each χi′′(ω)\chi_i''(\omega) is the full response function containing coupling to acoustic mode for domain walls ii lattice constants apart, see Eq. (51) in Appendix. As a reminder, for each individual component, the frequency and HWHM are proportional to qiq_i; the light-scattering vertex and coupling are proportional to qi\sqrt{q_i}.

The fitting results for Ta2_2NiSe5_5

The fits to the data, as well as the phononic and excitonic components, are shown in Figure 6. For the 300 K and 265 K spectra below TcT_c we also include the acoustic contribution due to coupling via quasi-periodic structure of domain walls. In Figure 8 we show the acoustic contribution at 300 K. The coupling of the excitonic continuum to the acoustic components enhances the low-energy response. The parameter β=7.7 meV A˚12\beta= 7.7\,\mathrm{meV}\,\text{\AA}^{\frac{1}{2}} is consistent with Eq. (10) for TcT_c.

The spectral parameters: frequency, FWHM and integrated intensity, −- of the three B2gB_{2g} phonon modes are shown in SubSection. VIB. The rest of the fitting parameters are given in Figure 10.

Temperature dependence of the Fano fitting parameters for the $ac$ Raman spectra of Ta$_2$NiSe$_5$

Figure 10. Temperature dependence of the Fano fitting parameters for the acac Raman spectra of Ta2_2NiSe5_5. (a) The coupling strength νi\nu_i (i=1,2,3i=1,2,3) in Eq. (38). (b) The light-scattering vertex tet_e, and the exciton energy at which the excitonic continuum has maximum intensity Ωe(T)=ωe2/γe\Omega_e(T)=\omega_e^2/\gamma_e, Eq. (2). The blue line is a linear fit to the Ωe(T)\Omega_e(T) data above TcT_c. The dashed lines indicate transition temperature TcT_c and bare excitonic transition temperature TcexT_c^{ex}.

The temperature dependence of the coupling strength between the individual phonon modes and the excitonic continuum are detailed in Figure 10(a). The ratios of the coupling strength to the phonon frequency for the B2g(1)B_{2g}^{(1)} and B2g(2)B_{2g}^{(2)} modes above TcT_c are 0.32 and -0.34, respectively. These ratios are an order of magnitude larger than the typical values for stable systems with similar phonon frequency [47, 53, 54]. On the contrary, the exciton-phonon interaction with the B2g(3)B_{2g}^{(3)} mode is weak. The signs of the coupling for the B2g(1)B_{2g}^{(1)} and B2g(2)B_{2g}^{(2)} modes are opposite. For these two modes, the magnitude of coupling is temperature independent above TcT_c, but decreases and saturates

on cooling below TcT_c. The temperature dependence of the coupling strength is influenced mainly by two factors: the screening effect of free carriers, reduced below TcT_c due to the emergence of a pronounced spectral gap [23, 55], and the change of the crystal structure below TcT_c.

The difference in magnitude of the exciton-phonon coupling for B2g(1,2)B_{2g}^{(1,2)} and B2g(3)B_{2g}^{(3)} modes can be readily understood from the corresponding displacement patterns obtained from the DFT calculations for Ta2_2NiSe5_5 and shown in Figure 11. The calculated patterns are consistent with those reported in other studies [20, 52]. While both the B2g(1,2)B_{2g}^{(1,2)} modes involve displacements of Ta and Se, the B2g(3)B_{2g}^{(3)} mode involve a displacement of Se almost exclusively (the Ni atoms are at the inversion centers and thus do not contribute to the Raman-active phonon modes). As the electronic bands in Ta2_2NiSe5_5 close to the Fermi level are believed to be predominantly formed by Ta and Ni ele-

The calculated vibrational patterns corresponding to the three $B_{2g}$-symmetry phonon modes of Ta$_2$NiSe$_5$

Figure 11. The calculated vibrational patterns corresponding to the three B2gB_{2g}-symmetry phonon modes of Ta2_2NiSe5_5.

trons [12, 18, 19], with much smaller contribution of Se, it appears natural that the B2g(3)B_{2g}^{(3)} mode does not couple to the low-energy electronic degrees of freedom. Furthermore, the vibration of the Ta atoms for B2g(1)B_{2g}^{(1)} and B2g(2)B_{2g}^{(2)} displacements are in anti-phase, which can explain the opposite signs of the exciton-phonon coupling for these modes.

The temperature dependence of the light-scattering vertex tet_e and the frequency Ωe\Omega_e at which the continuum has maximum intensity are shown in Figure 10(b). Above TcT_c, the quantity Ωe\Omega_e linearly decreases on cooling; below TcT_c, it rapidly increases and saturates at low temperature.

The fitting results for Ta2_2Ni(Se1−x_{1-x}Sx_x)5_5 with x=0.25x=0.25 and 0.670.67

To compare with the results for Ta2_2NiSe5_5, we also fit the acac polarization spectra for the alloy compositions (x=0.25x=0.25 and 0.670.67) above their transition temperature. We are not able to perform a Fano analysis below TcT_c for two reasons: (a) we have no knowledge of the domain structure and the sound velocity; and (b) we cannot properly subtract the leakage of the AgA_g modes. In Figure 12 we show one characteristic Fano fit for Ta2_2Ni(Se0.75_{0.75}S0.25_{0.25})5_5 and Ta2_2Ni(Se0.33_{0.33}S0.67_{0.67})5_5, respectively. The fitting parameters are given in Figure 13.

Raman response spectra for Ta2Ni(Se0.75S0.25)5 and Ta2Ni(Se0.33S0.67)5

Figure 12. Raman response χ′′\chi'', represented by black dots, in the acac scattering geometry for Ta2_2Ni(Se0.75_{0.75}S0.25_{0.25})5_5 and Ta2_2Ni(Se0.33_{0.33}S0.67_{0.67})5_5 with the Fano fits [Eq. (B2)] shown as red curves. The spectral resolution is 0.19 meV. The phonon modes [Eq. (1)] and excitonic continuum [Eq. (2)] are represented by green and blue curves, respectively.

Temperature dependence of Fano fitting parameters

Figure 13. Temperature dependence of the Fano fitting parameters for the acac Raman spectra of Ta2_2Ni(Se0.75_{0.75}S0.25_{0.25})5_5 (a-b) and Ta2_2Ni(Se0.33_{0.33}S0.67_{0.67})5_5 (c-d). (a,c) The coupling strength νi\nu_i (i=1,2,3i=1,2,3) in Eq. (38). (b,d) the light-scattering vertex tet_e, and the energy at which the excitonic continuum has maximum intensity Ωe=ωe2/γe\Omega_e=\omega_e^2/\gamma_e, of the overdamped electronic mode, Eq. (40). The blue line is a linear fit to the Ωe\Omega_e data.

The results are similar to the case of Ta2_2NiSe5_5: for the coupling strength, we find positive ν1\nu_1, negative ν2\nu_2, and negligible ν3\nu_3; the light-scattering vertex tet_e has no temperature dependence above the transition temperature; the energy at which the excitonic continuum has maxi-

mum intensity, Ωe=ωe2/γe\Omega_e=\omega_e^2/\gamma_e, decreases linearly on cooling above the transition temperature. Extrapolating to zero value of Ωe\Omega_e, we obtain 55±1855\pm18 K for Ta2_2Ni(Se0.75_{0.75}S0.25_{0.25})5_5 and −400±70-400\pm70 K for Ta2_2Ni(Se0.33_{0.33}S0.67_{0.67})5_5.

In Table IV we present the doping dependence of the relevant physical quantities above the transition temperature for the Ta2_2Ni(Se1−x_{1-x}Sx_x)5_5 family. Because the coupling strength increases with sulfur doping, determining the intrinsic phonon parameters, especially the linewidth, becomes more difficult for Ta2_2Ni(Se0.75_{0.75}S0.25_{0.25})5_5 and Ta2_2Ni(Se0.33_{0.33}S0.67_{0.67})5_5.

x=0x=0x=0.25x=0.25x=0.67x=0.67
tet_e (arb. units)7.227.367.48
ν1\nu_1 (meV)2.452.613.61
ν2\nu_2 (meV)-4.03-4.61-5.87

Table IV. The doping dependence of the scattering vertex and the coupling above the transition temperature for the interaction between the excitonic continuum and phonon modes in the B2gB_{2g} scattering geometry of Ta2_2Ni(Se1−x_{1-x}Sx_x)5_5 family. The quantities tet_e is the vertex of light scattering process for the excitonic continuum. The quantities νi\nu_i (i=1,2i=1,2) are the coupling between the excitonic continuum and the B2g(i)B_{2g}^{(i)} phonon mode.

Temperature dependence of the intrinsic phonon parameters

In this section we present the temperature dependence of intrinsic phonon-mode parameters: energies, intensities and linewidths. The results for B2gB_{2g}-symmetry modes are given in SubSec. VIA, and those for AgA_g-symmetry modes are given in SubSec. VIB.

B2g_{2g}-symmetry Phonon Modes

Ta2_2NiSe5_5

In Figure 14 we show the temperature dependence of the spectral parameters for B2g_{2g}-symmetry phonon modes in Ta2_2NiSe5_5: the bare phonon frequency, FWHM, and integrated intensity.

Temperature dependence of the spectral parameters for the B2g-symmetry phonon modes of Ta2NiSe5

Figure 14. Temperature dependence of the spectral parameters: the bare frequency, FWHM, and the integrated intensity, — for the B2g_{2g}-symmetry phonon modes of Ta2_2NiSe5_5. For each mode, the left panel presents the frequency and FWHM; and the right panel presents the integrated intensity in the allowed and ’forbidden’ scattering geometries. The solid lines represent the fits to the anharmonic decay model [Eqs. (16)–(17)].

Below TcT_{\mathrm{c}}, the temperature dependence of both frequency ωp(T)\omega_{\mathrm{p}}(T) and FWHM Γp(T)=2γp(T)\Gamma_{\mathrm{p}}(T)=2\gamma_{\mathrm{p}}(T) of the phonon modes can be accounted for by standard model assuming anharmonic decay into two phonons with identical frequencies and opposite momenta [56]:

ωp(T)=ω0−ω2[1+2eℏω0/2kBT−1],(16)\omega_{\mathrm{p}}(T)=\omega_{0}-\omega_{2}\left[1+\frac{2}{e^{\hbar\omega_{0}/2k_{\mathrm{B}}T}-1}\right], \tag*{(16)}

and

Γp(T)=Γ0+Γ2[1+2eℏω0/2kBT−1].(17)\Gamma_{\mathrm{p}}(T)=\Gamma_{0}+\Gamma_{2}\left[1+\frac{2}{e^{\hbar\omega_{0}/2k_{\mathrm{B}}T}-1}\right]. \tag*{(17)}

We note that a low value of Γ0\Gamma_{0} indicated high quality of the crystals.

Above TcT_{\mathrm{c}}, however, the slope of the FWHM increases for the three phonon modes, and the FWHM becomes larger than the value predicted by the anharmonic decay model. The appearance of additional decay channels above TcT_{\mathrm{c}} is related to the presence of gapless particle-hole pairs. Indeed, above TcT_{\mathrm{c}} Ta2_2NiSe5_5 is gapless spectroscopically [23], while below TcT_{\mathrm{c}} a gap rapidly develops at low energies, suppressing the damping due to particle-hole pairs. Because the Fano model used for fitting takes into account the damping due to exciton-phonon interaction, existence of additional damping above TcT_{\mathrm{c}} suggests the importance of the interactions beyond that model, i.e. nonlinear ones.

More noticeable is that the frequencies of the B2g(1)_{2g}^{(1)} and B2g(2)_{2g}^{(2)} modes exhibit a large increase on cooling from around 350 K to TcT_{\mathrm{c}}. The frequency of the B2g(3)_{2g}^{(3)} mode, however, does not show such anomaly. Only B2g(1)_{2g}^{(1)} and B2g(2)_{2g}^{(2)} modes exhibit this anomaly; interestingly (1) they are allowed to couple to interband scattering by symmetry selection rules and (2) their corresponding vibration involve motion of Ni atoms and hence they couple to the electronic bands near Fermi level. Although B2g(3)_{2g}^{(3)} mode can also couple to the interband scattering, its vibrational pattern almost does not contain motion of Ni atoms.

As for the integrated intensity, B2g(1)_{2g}^{(1)} and B2g(3)_{2g}^{(3)} modes have around 2-fold increase of intensity on cooling below TcT_{\mathrm{c}}, while B2g(2)_{2g}^{(2)} mode shows temperature-independent intensity. Two factors could influence the intensity. First, the interaction-induced gap opening up below TcT_{\mathrm{c}} should play a major role. The system gradually changes from a semimetal at high temperature to an insulator at low temperature. The screening effect, which reduces the light-scattering vertex, and in turn, the intensity for phonon modes, is suppressed on cooling. This factor therefore favors increase of intensity on cooling. Second, the structural change below TcT_{\mathrm{c}} could play also a role. The ion positions are shifted within the unit cell below TcT_{\mathrm{c}}, and the vibrational patterns of the three B2g_{2g} modes are modified. Therefore, the polarizability induced by lattice vibrations, which is proportional to the phonon intensity, also change with temperature. The fact that the B2g(2)_{2g}^{(2)} mode shows temperature-independent intensity might be related to its unique vibrational pattern.

The alloy compositions (x=0.25, 0.67)

In Figure 15 and Figure 16 we show the temperature dependence of the B2g_{2g}-mode spectral parameters for Ta2_2Ni(Se0.75_{0.75}S0.25_{0.25})5_5 and Ta2_2Ni(Se0.33_{0.33}S0.67_{0.67})5_5, respectively. Although we do not perform Fano analysis for the data below TcT_{\mathrm{c}}, we use Lorentzian lineshape to fit the B2g(1)_{2g}^{(1)} mode at low-enough temperatures, at which the lineshape is essentially symmetric. For the B2g(2)_{2g}^{(2)} and B2g(3)_{2g}^{(3)} modes, because they are not well separated from the leakage of Ag_g modes, it is difficult to reliably obtain their spectral parameters.

Temperature dependence of spectral parameters for Ta₂Ni(Se₀.₇₅S₀.₂₅)₅

Figure 15. Temperature dependence of the spectral parameters: frequency, FWHM, and integrated intensity, – for the B₂g-symmetry phonon modes of Ta₂Ni(Se₀.₇₅S₀.₂₅)₅. For each mode, the left panel presents the energy and FWHM; the right panel presents the integrated intensity in the allowed and ’forbidden’ scattering geometries.

Temperature dependence of spectral parameters for Ta₂Ni(Se₀.₃₃S₀.₆₇)₅

Figure 16. Temperature dependence of the spectral parameters: frequency, FWHM, and integrated intensity, – for the B₂g-symmetry phonon modes of Ta₂Ni(Se₀.₃₃S₀.₆₇)₅. For each mode, the left panel presents the energy and FWHM; the right panel presents the integrated intensity in the allowed and ’forbidden’ scattering geometries.

For the alloy compositions, the increase of the FWHM slope above TcT_{\mathrm{c}} is not as strong as in the case of Ta2_2NiSe5_5. Moreover, although for Ta2_2NiSe5_5 the large frequency increase of the B2g(1)_{2g}^{(1)} mode on cooling happens in a 20 K energy range, the frequency increase for the alloy compositions happens in a much larger energy range.

Ta₂NiS₅

Ta₂NiS₅ is a semiconductor that, in contrast to Ta₂NiSe₅, does not show signatures of an excitonic insulator like the flattening of valence band dispersion [21, 57]. Consistent with its semiconductor nature, we observe no excitonic continuum and in turn no asymmetric lineshape for the phonon modes, Figure 2(c-d). In Figure 17 we zoom in on the temperature dependence of the phonons in ac scattering geometry. The three B₂g-symmetry phonon modes exhibit conventional Lorentzian lineshapes, and the linewidth has almost 4-fold decrease on cooling from 315 K to 35 K. Moreover, the B₂g⁽¹⁾ and

Temperature dependence of B₂g phonons in ac scattering geometry

Figure 17. Temperature dependence of the B₂g phonons in the ac scattering geometry for Ta₂NiS₅. The spectral resolution is 0.19 meV for the 315 K data, and 0.06 meV for the other data.

B₂g⁽²⁾ modes show softening behavior on cooling.

In Figure 18 we show the temperature dependence of the spectral parameters for the B₂g-symmetry phonon modes for Ta₂NiS₅: The frequency, FWHM, and the integrated intensity. These spectral parameters are obtained by fitting the measured spectral features with Lorentzian lineshapes.

Temperature dependence of spectral parameters for $B_{2g}$-symmetry phonon modes

Figure 18. Temperature dependence of the spectral parameters: frequency, FWHM, and integrated intensity, – for the B2gB_{2g}-symmetry phonon modes of Ta2_2NiSe5_5. For each mode, the left panel presents the energy and FWHM; the right panel presents the integrated intensity in the allowed and ’forbidden’ scattering geometries.

The behavior of the B₂g-symmetry modes is not consistent with the anharmonic decay model: for B₂g⁽¹⁾ and B₂g⁽²⁾ modes, the frequency anomalously decreases on cooling; for B₂g⁽³⁾ mode, the decrease of FWHM on cooling is too steep to be accounted by the anharmonic decay model. Hence, we suggest that there is a change in the phonon self energy, which must affect the apparent mode frequency. For the B₂g⁽¹⁾ and B₂g⁽²⁾ modes, whose frequency is below 30 meV, the energy decreases on cooling; for the B₂g⁽³⁾ mode, whose frequency is above 30 meV, the energy increases on cooling.

The integrated intensity of B₂g⁽²⁾ mode is temperature-independent, while that of the other modes increases on cooling. However, different from the case of Ta₂NiSe₅, in which the increase of the intensity happens below TcT_c, the intensity increase of the Ta₂NiS₅ modes is through the whole measured temperature range.

AgA_g-symmetry phonon modes

In this subsection we discuss the properties of the full-symmetric phonon modes; the relevant results are shown in Figure 19 with respect to the overall spectral weight distribution in the AgA_g channel, and in Figure 20 and Figure 21 regarding the spectral parameters of the individual AgA_g modes. In Figure 21 we compare the temperature dependence of the spectral parameters (frequency, FWHM, and integrated intensity) of the AgA_g-symmetry phonon modes for Ta2_2NiSe5_5 and Ta2_2NiS5_5 crystals. The temperature dependence of both frequency and FWHM above TcT_c for these modes can be accounted by the anharmonic decay model [Eq. (16)-(17)].

Integral $I_{aa}(\omega)$ as a function of Raman shift

Figure 19. The integral Iaa(ω)=∫0ωχaa′′(ω′)ω′dω′I_{aa}(\omega)=\int_0^\omega\frac{\chi''_{aa}(\omega')}{\omega'}d\omega' as a function of the Raman shift ω\omega for Ta2_2NiSe5_5 in aaaa scattering geometry.

Temperature dependence of spectral parameters for the $A_g^{(1)}$ phonon mode

Figure 20. Temperature dependence of the spectral parameters: frequency, FWHM, and integrated intensity, – for the Ag(1)A_g^{(1)} phonon mode of Ta2_2Ni(Se1−x_{1-x}Sx_x)5_5 with x=0.25x=0.25 (a-b) and 0.67 (c-d). For each sample, the left panel presents the energy and FWHM; the right panel presents the integrated intensity. The solid lines represent the fits to the anharmonic decay model [Eqs. (16)-(17)].

Figure 21. Temperature dependence of the spectral parameters: frequency, FWHM, and integrated intensity, – for the AgA_g-symmetry phonon modes for Ta2_2NiSe5_5 and Ta2_2NiS5_5. For each mode, the left panel presents the energy and FWHM; the right panel presents the integrated intensity. The solid lines represent the fits to the anharmonic decay model Eqs. (16)–(17).

The FWHM of the AgA_g-symmetry modes (except for Ag(3)A_g^{(3)} and Ag(8)A_g^{(8)}) for Ta2_2NiSe5_5 exhibit an anomalous increase above the transition temperature, which we attribute to enhanced electron-phonon scattering rate in its semimetal phase. However, the modes of Ta2_2NiS5_5 show no anomaly of FWHM, because it is a semiconductor with a direct gap observed throughout the measured temperature range.

The intensity of most modes has more than 2-fold increase on cooling, with a few exceptions: for Ta2_2NiSe5_5, the intensity of the Ag(1)A_g^{(1)} mode is independent of temperature, while the intensity of the Ag(2)A_g^{(2)} phonon mode decreases on cooling; for Ta2_2NiS5_5, the intensities of the Ag(1)A_g^{(1)} and Ag(2)A_g^{(2)} modes are temperature independent.

The enhancement of intensity on cooling in the semimetallic samples could in principle be related to the change of electronic structure (less screening effect) below TcT_c. However, an interesting insight can be further obtained by computing the integral Iaa(ω)=∫0ωχaa′′(ω′)ω′dω′I_{aa}(\omega)=\int_0^\omega\frac{\chi''_{aa}(\omega')}{\omega'}d\omega'

(Figure 19). In the limit ω→∞\omega\rightarrow\infty this integral is proportional to the static susceptibility in the aaaa scattering geometry, which is expected to be approximately constant as a function of temperature since there is no instability in AgA_g channel. On cooling, the phonon intensity is enhanced [Figure 21], while the electronic continuum is suppressed [Figure 2(a)]. However, the integral IaaI_{aa} is essentially temperature independent at 70 meV. This conservation implies that the enhancement of the phonon intensity is balanced by the reduction of the electronic intensity. Note that these effects occur at temperatures below TcT_c, where the continuum in B2gB_{2g} symmetry is strongly suppressed, indicating the opening of a gap. On the other hand, strong correlation effects have been shown to lead to a non-zero intensity within the gap [23]. Thus, the peculiar balance between the electronic and phononic contributions at low energies implies an unconventional coupling of the electronic modes to the AgA_g phonons in the correlated excitonic insulator state.

For comparison, in Figure 20 we show the temperature dependence of the spectral parameters for the Ag(1)A_g^{(1)} mode for Ta2_2Ni(Se1−x_{1-x}Sx_x)5_5 with x=0.25x=0.25 and 0.67. Because of the difficulty caused by broad lineshape and two-frequency behavior of the phonon modes for alloy compositions, only the lowest-frequency Ag(1)A_g^{(1)} mode, which are well sep-

arated from other modes in frequency, renders reliable fitting results. For Ta2_2Ni(Se0.75_{0.75}S0.25_{0.25})5_5, some linewidth broadening is observed above TcT_c, which is weaker than for Ta2_2NiSe5_5. However, for Ta2_2Ni(Se0.33_{0.33}S0.67_{0.67})5_5 almost no broadening is observed.

Relation to time-resolved experiments

In this section we convert Raman response from frequency domain to time domain, and discuss relevant time-resolved studies on Ta2_2NiSe5_5.

We first discuss the appearance of B2g_{2g}-symmetry excitations in the time domain, obtained by inverse Fourier transform of the Raman response, see Figure 22. Above the transition temperature, the phonon-exciton interaction significantly broadens the spectral features Figure 22(a), as a result the deduced time-resolved response decays fast and essentially dies out before 5 ps Figure 22(b). For the bare phononic response, the time-domain signal beyond 4 ps is dominated by the oscillation of 2 THz mode Figure 22(c), corresponding to the B2g(1)_{2g}^{(1)} phonon. The signal of the bare excitonic response shows a pure relaxational behavior Figure 22(d). Comparing Figure 22(b) and (c), we note that it is difficult to identify the Fano interference feature from time-domain signal, because such interference exhibits no distinct characters in the time domain, except for that the oscillations are strongly damped. Indeed, in ultrafast studies, the interference nature is not revealed even after Fourier transform of the time-domain data to the frequency domain [16, 17, 20]. The likely reason for that is the high pump fluence used, causing the exciton and phonon responses decouple already on the very early time scales.

Time-domain signals for the B2g-symmetry excitations of Ta2NiSe5

Figure 22. Time-domain signal for the B2g_{2g}-symmetry excitations of Ta2_2NiSe5_5. (a) The Raman response for Ta2_2NiSe5_5 measured in aa geometry at 380 K. The Fano fit Eq. (35) is shown by the red curve; the bare phonon modes Eq. (1) are shown by the green curve; the bare excitonic continuum Eq. (2) is shown by the blue curve. The time-domain signal corresponding to the Fano fit, bare phonon modes, and bare excitonic continuum, obtained by inverse Fourier transform, are shown in (b), (c), and (d), respectively. (e) The Raman response for Ta2_2NiSe5_5 measured in ac geometry at 35 K. The corresponding time-domain signals are shown in (f), (g), and (h), respectively.

Below the transition temperature, the energy at which the continuum has maximum response moves to higher energy, and the phonon-exciton coupling strength is reduced. Consequently, the interference effect is suppressed and the combined response is essentially the same as the bare phononic response Figure 22(f-Figure 22g). The bare excitonic response still has a pure relaxational time dependence, with very short lifetime Figure 22(d) and (Figure 22h).

For comparison, in Figure 23 we present the time-domain signal for the Ag_g-symmetry excitations. Above the transition temperature, although multiple oscillations are simultaneously present at short time scale, beyond 10 ps the time-domain signal is dominated by the oscillation of 1 THz Figure 23(b), corresponding to the Ag(1)_g^{(1)} phonon mode whose spectral width is much smaller than others Figure 23(a). At low temperature, the difference between the spectral widths of various modes is reduced Figure 23(c); hence, these modes have comparable lifetime and multiple oscillations survive beyond 10 ps Figure 23(d).

Time-domain signals for the Ag-symmetry excitations of Ta2NiSe5

Figure 23. Time-domain signal for the Ag_g-symmetry excitations of Ta2_2NiSe5_5. (a) The Raman response for Ta2_2NiSe5_5 measured in aa geometry at 380 K. The time-domain signal, obtained by inverse Fourier transform of the data, is shown in (b). (c) The Raman response for Ta2_2NiSe5_5 measured in aa geometry at 35 K. The corresponding time-domain signal is shown in (d).

In relation to the ultra-fast studies of Ta2_2NiSe5_5 [15, 17, 58, 59], the Raman results are consistent with the ultra-fast data under low fluence. One ultra-fast work [17] provides evidence for coupling between the Ag(1)_g^{(1)} mode and the amplitude mode of the condensate under high-fluence pumping. We note that in our studies we can unambiguously identify the amplitude mode with the excitonic continuum we observe. Above the transition temperature, the overdamped excitonic mode softens on cooling towards TcT_c; just below the transition temperature, its energy increases on further cooling. Such temperature dependence of mode energy is characteristic of an amplitude mode while the optical B2g_{2g} phonons only harden on cooling. In Figure 24 we present the time-domain signal corresponding to the excitonic mode around the transition temperature.

Figure 24. Time-domain signal for the excitonic continuum of Ta₂NiSe₅ above and below the transition temperature, plotted in semi-log scale. The data above TcT_c are dashed for clarity. The relaxation time starts to decrease just below TcT_c. Inset: temperature dependence of the energy at which the excitonic continuum has maximum intensity, Ωe(T)\Omega_e(T).

Additionally, a mode associated with the phase of the excitonic order parameter may exist. The electron-phonon coupling and exchange interactions modify the otherwise sombrero-shape free-energy landscape, reducing the continuous U(1) symmetry to a discrete Z2\mathbb{Z}_2 symmetry, leading to a finite energy of the phase mode. A recent pump-probe microscopy study suggests that the phase-mode energy is smaller than the Ag(1)A_g^{(1)}-mode energy (around 4 meV) [58]. However, we do not observe any mode down to 0.4 meV at low temperature [Fig. 22(e)]. Furthermore, the phase oscillations of the order parameter should harden above TcT_c, which would have manifested itself in the low-energy range of the spectra. Such signatures are not observed, either [Fig. 6]. Therefore, existence of the phase mode below 4 meV is not supported by the Raman data.

Conclusions

We have employed polarization resolved Raman spectroscopy to conduct a systematic spectroscopic study of the lattice dynamics in Ta₂Ni(Se1−x_{1-x}Sx_x)₅ (x=0,…,1x = 0, \ldots, 1) family of the excitonic insulators.

We identify and classify by symmetry all the Raman active phonon modes of AgA_g and B2gB_{2g} symmetries. A change in selection rules is detected at temperature Tc(x)T_c(x), indicating the orthorhombic-to-monoclinic structural phase transition related to the excitonic insulator state [23, 24]. We find that the symmetry breaking transition persists for entire Ta₂Ni(Se1−x_{1-x}Sx_x)₅ family with Tc(x)T_c(x) monotonically decreasing from 328 K for x=0x = 0 to 120 K for x=1x = 1. Its signatures in the resistivity data, however, become weak or undetectable for large sulfur concentration xx that we attribute to the weakness of the symmetry breaking at large xx, demonstrated by the decrease of phonon intensity in the ’forbidden’ scattering geometry.

For x<0.7x < 0.7, the two lowest-frequency acac-quadrupole-symmetry B2g(1)B_{2g}^{(1)} and B2g(2)B_{2g}^{(2)} phonon modes show strongly asymmetric lineshapes at high temperatures, which is indicative of coupling between the phonons and an excitonic continuum of the same symmetry [23, 24]. Within the framework of extended Fano model, we develop a quantitative description of the observed lineshapes, enabling us to disentangle the excitonic and phononic contributions to the spectra, to derive the intrinsic phonon parameters and determine the exciton-phonon interaction strength, that affects the transition temperature Tc(x)T_c(x) [24]. The displacement patterns obtained from abab-initio calculations explain the trends in the deduced exciton-phonon coupling values. At T<TcT < T_c the remnant excitonic continuum demonstrates signatures of coupling to the finite-momentum B2gB_{2g}-symmetry acoustic phonons, allowed due to scattering on a quasi-periodic structural of domain walls. We have also shown that the coupling to the excitonic continuum explains the acoustic mode softening observed in [24, 52].

The intrinsic parameters of the optical phonons for Ta₂NiSe₅ and Ta₂NiS₅ at low temperatures are mostly in good agreement with anharmonic decay model and DFT calculations. For alloy compositions a two-mode behavior is found for most modes, where signatures appear at two frequencies, corresponding to the ones in Ta₂NiSe₅ and Ta₂NiS₅. However, several types of anomalous behavior have been observed in the temperature dependencies. For x<0.7x < 0.7, the B2g(1)B_{2g}^{(1)} and B2g(2)B_{2g}^{(2)} modes anomalously harden close to TcT_c (e.g., Figure 14), indicating an interaction with continuum beyond Fano model description [23]. For Ta₂NiS₅ (Figure 18) the B2g(1)B_{2g}^{(1)} and B2g(2)B_{2g}^{(2)} modes do not show Fano shapes, but the frequencies of these modes anomalously soften on cooling, although never going critical. Finally, the intensity of the most modes is strongly temperature dependent. Interestingly, for Ta₂NiSe₅ we observe that the sum of electronic and phononic contributions to the aaaa static susceptibility is conserved (Figure 19), pointing to an unexpected electron-phonon coupling mechanism within the excitonic insulator state.

In addition, based on our results we have provided an interpretation of recent time-resolved pump-probe experiments and discussed the signatures expected from the modes we have observed, see Figure 22, Figure 23 and Figure 24. Overall, this work provides a comprehensive study lattice dynamics in Ta₂Ni(Se1−x_{1-x}Sx_x)₅; more generally, the unconventional behaviors we observed point to the importance of the effects of electron-phonon coupling in correlated semimetals.

Acknowledgments

M.Y. and P.A.V. contributed equally to this work. We acknowledge discussions with K. Haule. We are grateful to Y. Kauffmann for help with TEM measurements. The spectroscopic work conducted at Rutgers was supported by NSF Grant No. DMR-1709161 (M.Y. and G.B.). P.A.V. acknowledges the Postdoctoral Fellowship support from the Rutgers University Center for Materials Theory. The sample growth and characterization work conducted at the Technion was supported by the Israel Science Foundation Grant No. 320/17 (H.L., I.F. and A.K.). H.L. was supported in part by a PBC fellowship of the Israel Council for Higher Education. The work at NICPB was supported by the Estonian Research Council Grant No. PRG736 an by the European Research Council (ERC) under Grant Agreement No. 885413. M.K. was supported by NSF DMRREF grant DMR-1629059.

Estimation of heating in excitation laser spot

We use three methods to estimate the heating rate, a measure of the temperature increase per unit laser power in the focused laser spot (K/mW): (i) Stokes/anti-Stokes Raman scattering intensity ratio analysis; (ii) monitoring laser power that is inducing the phase transition; and (iii) a thermoconductivity model calculation. We use data for Ta2_2NiSe5_5 as an example to illustrate these methods.

For the first method, we note that the Stokes scattering cross section ISI_S and Anti-Stokes cross section IASI_{AS} are related by the detailed balance principle [22]

nIS(ω)=(n+1)IAS(ω),(18)nI_S(\omega)=(n+1)I_{AS}(\omega), \tag*{(18)}

in which nn stands for the Bose factor

n(ω,T)=1exp⁡(ℏω/kBT)−1,(19)n(\omega,T)=\frac{1}{\exp(\hbar\omega/k_BT)-1}, \tag*{(19)}

where ℏ\hbar is the reduced Planck’s constant, ω\omega is frequency, kBk_B is the Boltzmann’s constant, and TT is the temperature in the laser spot.

Using Eq. (18)-(19), the temperature can be derived as

T=ℏωkBln⁡(IS/IAS),(20)T=\frac{\hbar\omega}{k_B\ln(I_S/I_{AS})}, \tag*{(20)}

or, numerically,

T[K]=11.605 ωln⁡(IS/IAS)[meV].(21)T[K]=\frac{11.605\ \omega}{\ln(I_S/I_{AS})}[meV]. \tag*{(21)}

We studied the Stokes and Anti-Stokes cross section relation for the acac scattering geometry at 295 K environmental temperature with various laser power. In Figure 25 we show the results measured with 8 mW laser power as an example. The Raman responses calculated from Stokes and Anti-Stokes cross sections match well with

Figure 25. Stokes-Anti-Stokes analysis of the acac spectra for Ta2_2NiSe5_5. (a) The Stokes and Anti-Stokes cross sections measured at environmental temperature TE=295T_E=295 K with laser power P=8P=8 mW. (b) The Raman response calculated from the measured spectra in (a). The laser-spot temperature, which appears in the Bose factor nn, is 307 K. The spectral resolution is 0.19 meV for panels (a-b). (c) The laser-power dependence of the laser-spot temperature. The straight line represents a linear fit.

the temperature at the laser spot being 307 K. We use Eq. (21) with the phonon intensity integrated from 7.5 to 8.5 meV to calculate the temperature at the laser spot. By a linear fit to the power dependence of the laser-spot temperature, we find the heating rate to be 1.29±\pm0.17 K/mW.

For the second method, we gradually increase laser power at 295 K environmental temperature. We find that when the laser power is in the range of 26±\pm4 mW, the domain stripes, which are visible across the sample surface under laser illumination, disappear inside the laser spot. Moreover, the temperature dependence of phonon width and intensity has a sudden change. We assume that at this laser power, the temperature at the laser spot reaches the transition temperature 328 K [8, 9], yielding the heating rate of 1.25±\pm0.20 K/mW. We note that for Ta2_2NiSe5_5, we do not observe domain stripes below its transition temperature 120 K. Its transition temperature is identified by the sudden change of phonon intensity, see Figure 18(b) for example.

For the third method, we consider a heated region shown in Figure 26 of a rectangular shape, with the dimensions given by the beam spot size and the penetration depth. The static heat equation takes the form:

The geometry of the heated volume

Figure 26. The geometry of the heated volume. The lengths lal_a and lcl_c are determined by the spot size, while lbl_b is approximated by the skin depth; the SbS_b is in the cleave plane facing the vacuum, while the other faces are in the bulk. Within the region marked by dashed lines, the heat flow is mostly along bb axis, while outside the region heat flows in all directions comparably, such that the temperature increase above the base temperature ΔT\Delta T decays as 1/r1/r, rr being the distance from the heated region.

−κx∂2T∂x2−κy∂2T∂y2−κz∂2T∂z2=P(r),-\kappa_x \frac{\partial^2 T}{\partial x^2}-\kappa_y \frac{\partial^2 T}{\partial y^2}-\kappa_z \frac{\partial^2 T}{\partial z^2}=\mathcal{P}(\mathbf{r}),
∂T(r)∂z∣y=0=0;\left.\frac{\partial T(\mathbf{r})}{\partial z}\right|_{y=0}=0;
∂T(x→±∞,y→−∞,z→±∞)=T0.(22)\partial T(x\to\pm\infty,y\to-\infty,z\to\pm\infty)=T_0. \tag*{(22)}

where P(r)\mathcal{P}(\mathbf{r}) is the power density. We ignore the cooling provided by the helium gas flow above the sample surface, so that there is no heat flow at the y=0y=0 boundary, while deep inside the bulk the temperature should reach the base temperature T0T_0. Equation (22) can be solved using Green’s function of the Laplace equation and image method to satisfy the first boundary condition. In the latter, we extend the equation formally to y>0y>0 half-space and add an ”image” power density

P(x,y,z)θ(−y)→P(x,y,z)θ(−y)+P(x,−y,z)θ(y).(23)\mathcal{P}(x,y,z)\theta(-y)\to\mathcal{P}(x,y,z)\theta(-y)+\mathcal{P}(x,-y,z)\theta(y). \tag*{(23)}

The solution is then given by:

T(r)−T0=∫dx~′dy~′dz~′ P(κxx~′,…)θ(−y~′)+y~′→−y~′4π∣r~−r~′∣.(24)T(\mathbf{r})-T_0=\int\frac{d\tilde{x}'d\tilde{y}'d\tilde{z}'\,\mathcal{P}(\sqrt{\kappa_x}\tilde{x}',\ldots)\theta(-\tilde{y}')+\tilde{y}'\to-\tilde{y}'}{4\pi\left|\tilde{\mathbf{r}}-\tilde{\mathbf{r}}'\right|}. \tag*{(24)}

where r~=(x/κx,y/κy,z/κz)\tilde{\mathbf{r}}=(x/\sqrt{\kappa_x},y/\sqrt{\kappa_y},z/\sqrt{\kappa_z}). The solution decays as 1/r1/r at large distances and thus satisfies the second boundary condition of Eq. (22). An analytical result can be obtained for

P(x,y,z)=tPlalblcθ(la/2−∣x∣)θ(lb−∣y∣)θ(lc/2−∣z∣)(25)\mathcal{P}(x,y,z)=\frac{tP}{l_al_bl_c}\theta(l_a/2-|x|)\theta(l_b-|y|)\theta(l_c/2-|z|) \tag*{(25)}

at r=0r=0 in the limit lb2/κb≪la,c2/κa,cl_b^2/\kappa_b\ll l_{a,c}^2/\kappa_{a,c}, where PP is the laser power and tt is the transmission coefficient of the sample. Performing the integral one obtains

T(0)−T0≈tP(laκaln⁡(lcκc+lc2κc+la2κala/κa)+a↔c)κbπlalc.(26)T(0)-T_0\approx\frac{tP\left(\dfrac{l_a}{\sqrt{\kappa_a}}\ln\left(\dfrac{\dfrac{l_c}{\sqrt{\kappa_c}}+\sqrt{\dfrac{l_c^2}{\kappa_c}+\dfrac{l_a^2}{\kappa_a}}}{l_a/\sqrt{\kappa_a}}\right)+a\leftrightarrow c\right)}{\sqrt{\kappa_b}\pi l_al_c}. \tag*{(26)}

Qualitatively, the result can be understood as follows. Due to the anisotropic shape of the heated region, right below it most heat is transferred along the bb axis. However, the one-dimensional heat equation would result in a linear solution T(y)T(y), depending on the cutoff scale lbeffl_b^{eff}. It can be estimated from the condition of the heat flow to the lateral direction being equal to the heat flow along bb. Approximating the temperature gradient along a,b,ca,b,c as ΔT/la,b,c\Delta T/l_{a,b,c} one gets the condition

2lbefflaκcΔTlc+2lbefflcκaΔTla=lalcκbΔTlbeff,(27)2l_b^{eff}l_a\kappa_c\frac{\Delta T}{l_c}+2l_b^{eff}l_c\kappa_a\frac{\Delta T}{l_a}=l_al_c\kappa_b\frac{\Delta T}{l_b^{eff}}, \tag*{(27)}

that results in the estimate

lbeff=la2lc2κb2la2κc+2lc2κa.(28)l_b^{eff}=\sqrt{\frac{l_a^2l_c^2\kappa_b}{2l_a^2\kappa_c+2l_c^2\kappa_a}}. \tag*{(28)}

Equating the total heat flow outside this region to the input power one obtains the estimate

ΔT≈tP8κb(la2κc+lc2κa).(29)\Delta T\approx\frac{tP}{\sqrt{8\kappa_b(l_a^2\kappa_c+l_c^2\kappa_a)}}. \tag*{(29)}

Overall, one notices that the temperature increase scales with the linear size of the spot. The laser heating power RR is then determined as ΔT/P\Delta T/P; for the actual estimate we thus need to find the transmission coefficient tt first. The transmission coefficient can be calculated from the complex index of refraction nn by the relationship t=1−∣(n−1)/(n+1)∣2t=1-| (n-1)/(n+1)|^2:

t=4n1(n1+1)2+n22,(30)t=\frac{4n_1}{(n_1+1)^2+n_2^2}, \tag*{(30)}

in which n1n_1 and n2n_2 are the real and imaginary part of nn. The complex index of refraction can be calculated from the complex dielectric constant ϵ\epsilon by the relationship ϵ=n2\epsilon=n^2:

n1=ϵ12+ϵ22+ϵ12,n2=ϵ12+ϵ22−ϵ12,(31)n_1=\sqrt{\frac{\epsilon_1^2+\epsilon_2^2+\epsilon_1}{2}},\quad n_2=\sqrt{\frac{\epsilon_1^2+\epsilon_2^2-\epsilon_1}{2}}, \tag*{(31)}

where ϵ1\epsilon_1 and ϵ2\epsilon_2 are the real and imaginary part of ϵ\epsilon. The imaginary part ϵ2\epsilon_2 can be calculated from the real part of the optical conductivity σ1\sigma_1:

ϵ2=60ωσ1,(32)\epsilon_2=\frac{60}{\omega}\sigma_1, \tag*{(32)}

in which the unit of ω\omega is cm−1\mathrm{cm}^{-1} and that of σ1\sigma_1 is Ω−1cm−1\Omega^{-1}\mathrm{cm}^{-1}. The polarization of the incoming light is along aa-axis of the sample. Because the quantities ϵ1\epsilon_1 and σ1\sigma_1 are 8.13 and 2.47 Ω−1\Omega^{-1}cm−1^{-1} respectively at the laser wavelength [55], we find t=0.64t=0.64.

The experimentally measured thermal conductivity [60] at 295 K is κa=208\kappa_a=208 mWK−1^{-1}cm−1^{-1} and κc=57.4\kappa_c=57.4 mWK−1^{-1}cm−1^{-1}.

The length scales la=5×10−3l_a=5\times10^{-3} cm and lc=1×10−2l_c=1\times10^{-2} cm is determined by the 50×10050\times100 μ\mum2^2 laser spot. The length lbl_b is the skin depth, which is calculated from the imaginary part of the index of refraction n2n_2:

lb=12πωn2,(33)l_b=\frac{1}{2\pi\omega n_2}, \tag*{(33)}

in which the unit of ω\omega is cm−1^{-1} and that of lbl_b is cm. At the laser wavelength, n2=1.5n_2=1.5, and we find lb=7×10−6l_b=7\times10^{-6} cm (70 nm).

The known values for the variables in Eq. (26) are summarized in Table V. Because κb\kappa_b is unknown, we cannot use Eq. (26) to calculate the laser heating rate. However, if the heating rate is 1.29 K/mW at 295 K, as determined from the Stokes/anti-Stokes analysis, the value for κb\kappa_b should be 1.38 mWK−1^{-1}cm−1^{-1}.

Quantity (unit)a-axisc-axis
κ\kappa (mWK−1^{-1}cm−1^{-1})20857.4
ll (10−310^{-3}cm)510

Table V. The various parameters used in calculating the laser heating rate at 295 K. The transmission coefficient is 0.64.

An estimate for the electronic contribution to thermal conductivity κ(e)\kappa^{(e)} could be calculated from the electric resistivity ρ\rho by virtue of Wiedemann-Franz law:

κ(e)=LTρ,(34)\kappa^{(e)}=\frac{LT}{\rho}, \tag*{(34)}

in which L=2.44×10−5L=2.44\times10^{-5} mWΩ\OmegaK−2^{-2} is the Lorenz number. The resistivity along b axis at 295 K is ρb=2.34×10−1\rho_b=2.34\times10^{-1} Ω\Omegacm [61], and we find κb(e)=0.0308\kappa_b^{(e)}=0.0308 mWK−1^{-1}cm−1^{-1}. The ratio of κb(e)\kappa_b^{(e)} to κb\kappa_b is consistent with the ratios obtained along a and c axes [60]. Moreover, that κb≪κa,c\kappa_b\ll\kappa_{a,c} is consistent with the quasi-2D structure of the system [62].

To illustrate that the laser heating rate is strongly temperature dependent, we use Eq. (26) to calculate the heating rate as a function of temperature. The temperature dependence of κa\kappa_a and κc\kappa_c are taken from Ref. [60]. We assume that the ratio of κb\kappa_b to κc\kappa_c is temperature independent, and fix this ratio to be 1.38/57.4 = 2.40%. In Figure 27 we show the calculated heating rate. Because of such assumption, the laser heating rate at 295 K is the same as that determined from the Stokes/anti-Stokes analysis.

Figure 27. Temperature dependence of the laser heating rate R, calculated by Eq. (26) with the thermal conductivity data taken from Ref. [60].

The fitting model for the Fano interference

Above TcT_c, the B2g(1)B_{2g}^{(1)} and B2g(2)B_{2g}^{(2)} phonon modes of Ta2_2NiSe5_5 exhibit strongly asymmetric Fano lineshapes. To analyze the physics of this Fano interference features, we propose a model describing three phonon modes individually coupled to broad excitonic continuum. The total Raman response is described in the following way:

χ′′∼TTGT,(35)\chi''\sim T^TGT, \tag*{(35)}

where TT=(TphT  te)T^T=(T_{\mathrm{ph}}^T\ \ t_e) denotes the vertices for light scattering process (the subscript “T” denotes ”Transpose”), in which TphT=(tp1  tp2  tp3)T_{\mathrm{ph}}^T=(t_{p1}\ \ t_{p2}\ \ t_{p3}). GG comprises the Green’s functions for the interacting phononic and excitonic excitations that can be obtained by solving the Dyson equation:

G=(G0−1−V)−1.(36)G=(G_0^{-1}-V)^{-1}. \tag*{(36)}

In Eq. (36)

G0=(Gph000Ge0)(37)G_0=\begin{pmatrix}G_{\mathrm{ph}}^0&0\\0&G_e^0\end{pmatrix} \tag*{(37)}

is the bare Green’s function and

V=(0Ve−phVe−phT0)(38)V=\begin{pmatrix}0&V_{e-\mathrm{ph}}\\V_{e-\mathrm{ph}}^T&0\end{pmatrix} \tag*{(38)}

is the exciton-phonon interaction. Here the bare phononic Green’s function

Gph0=diag⁡[Gpi0]=diag⁡[−(1ω−ωpi+iγpi−1ω+ωpi+iγpi)],(39)\begin{aligned} G_{\mathrm{ph}}^0=\operatorname{diag}[G_{pi}^0] \\ &=\operatorname{diag}\left[-\left(\frac{1}{\omega-\omega_{pi}+i\gamma_{pi}}-\frac{1}{\omega+\omega_{pi}+i\gamma_{pi}}\right)\right], \tag*{(39)} \end{aligned}

correspond to the B2g(i)B_{2g}^{(i)}-symmetry phonon modes (i=1,2,3),

Ge=1ωe2/γe−iω(40)G_e=\frac{1}{\omega_e^2/\gamma_e-i\omega} \tag*{(40)}

represents the excitonic continuum, and Ve−phT=(v1 v2 v3)V_{e-ph}^{T} = (v_1\ v_2\ v_3) denotes exciton interaction strength with the corresponding phonons. For phononic equation, the parameters ωpi\omega_{pi} and γpi\gamma_{pi} have the meaning of the mode frequency and the half width at half maximum (HWHM), respectively. For the excitonic equation, ωe\omega_e is the energy of the overdamped excitations and γe\gamma_e represents the relaxation rate. The quantity Ωe=ωe2/γe\Omega_e = \omega_e^2/\gamma_e denotes the energy at which the excitonic continuum has maximum intensity.

The bare phononic and excitonic responses are

χp′′(0)=STphTGph0Tph=∑i=13tpi2SGpi0(41)\chi_p^{\prime\prime(0)} = \mathcal{S}T_{ph}^{T}G_{ph}^{0}T_{ph} = \sum_{i=1}^{3} t_{pi}^{2}\mathcal{S}G_{pi}^{0} \tag*{(41)}

and

χe′′(0)=te2SGe0.(42)\chi_e^{\prime\prime(0)} = t_e^2\mathcal{S}G_e^{0}. \tag*{(42)}

These two expressions are related to Eqns. (1)-(2) in the main text.

The renormalized phononic and excitonic responses, χp′′\chi_p^{\prime\prime} and χe′′\chi_e^{\prime\prime}, as well as the interference term χint′′\chi_{int}^{\prime\prime} are calculated from the renormalized Green’s function GG:

G=(GphGe−phGe−phTGe),(43)G = \begin{pmatrix} G_{ph} & G_{e-ph} \\ G_{e-ph}^{T} & G_e \end{pmatrix}, \tag*{(43)}

where Ge−phT=(G1e G2e G3e)G_{e-ph}^{T} = (G_{1e}\ G_{2e}\ G_{3e}) represents the interference between the phononic modes and the continuum.

The phononic term is calculated from the 3x3 phononic block:

χp′′=STphTGphTph;(44)\chi_p^{\prime\prime} = \mathcal{S}T_{ph}^{T}G_{ph}T_{ph}; \tag*{(44)}

the excitonic term results from the 1x1 excitonic block:

χe′′=te2SGe;(45)\chi_e^{\prime\prime} = t_e^2\mathcal{S}G_e; \tag*{(45)}

and the interference term is given by the off-diagonal block

χint′′=te∑i=13tpiSGie.(46)\chi_{int}^{\prime\prime} = t_e\sum_{i=1}^{3}t_{pi}\mathcal{S}G_{ie}. \tag*{(46)}

These three terms, as defined, satisfy Eq. (4):

χ′′=χp′′+χe′′+χint′′.(47)\chi^{\prime\prime} = \chi_p^{\prime\prime} + \chi_e^{\prime\prime} + \chi_{int}^{\prime\prime}. \tag*{(47)}

For the acac spectra below TcT_c, especially at about 300 K, this model cannot properly account for the low frequency Raman response. This additional spectral feature is related to Raman coupling to the longitudinal acoustic excitations at finite momenta in the presence of quasi-periodic domain walls, detailed in the main text. The Green’s function of each individual acoustic mode has form similar to the phononic Green’s function:

Gs(ω,q)=−(1ω−csq+irsq−1ω+csq+irsq).(48)G_s(\omega,q) = -\left(\frac{1}{\omega-c_sq+ir_sq}-\frac{1}{\omega+c_sq+ir_sq}\right). \tag*{(48)}

Its light-scattering vertex ts(q)t_s(q) and coupling to the excitonic continuum vs(q)v_s(q) are both proportional to the square root of wavevector: ts=τsqt_s = \tau_s\sqrt{q} and vs=βsqv_s = \beta_s\sqrt{q}. The frequency ωs\omega_s and the HWHM γs\gamma_s are both proportional to the wavevector: ωs=csq\omega_s = c_sq and γs=rsq\gamma_s = r_sq.

After introducing this mode, Eqs. (37)-(38) extend to

G0′=(Gph0000Gs000Ge),(49)G_0' = \begin{pmatrix} G_{ph}^{0} & 0 & 0 \\ 0 & G_s & 0 \\ 0 & 0 & G_e \end{pmatrix}, \tag*{(49)}
V′=(00Ve−ph00βsqVe−phTβsq0),(50)V' = \begin{pmatrix} 0 & 0 & V_{e-ph} \\ 0 & 0 & \beta_s\sqrt{q} \\ V_{e-ph}^{T} & \beta_s\sqrt{q} & 0 \end{pmatrix}, \tag*{(50)}

and the vertex for light-scattering process becomes T′T=(TphT  tsq  te)T'^{T} = \left(T_{ph}^{T}\ \ t_s\sqrt{q}\ \ t_e\right).

Then the total response function containing coupling to acoustic mode for domain walls ii lattice constants apart is given by

χi′′∼ST′TG′T′.(51)\chi_i^{\prime\prime} \sim\mathcal{S}T'^{T}G'T'. \tag*{(51)}

Due to the requirement of causality, the real part of susceptibility should be an even function while the imaginary part should be an odd function. Therefore, the static Raman susceptibility is purely real. The total static Raman susceptibility is given by

χ(ω=0)∼RTTG(ω=0)T.(52)\chi(\omega=0) \sim\mathcal{R}T^{T}G(\omega=0)T. \tag*{(52)}

The bare phononic static Raman susceptibility can be obtained from

χp(0)(ω=0)∼RTphTGph0(ω=0)Tph=∑i=132tpi2ωpiωpi2+γpi2,(53)\chi_p^{(0)}(\omega=0) \sim\mathcal{R}T_{ph}^{T}G_{ph}^{0}(\omega=0)T_{ph} = \sum_{i=1}^{3}\frac{2t_{pi}^{2}\omega_{pi}}{\omega_{pi}^{2}+\gamma_{pi}^{2}}, \tag*{(53)}

Similarly, the bare excitonic static Raman susceptibility is given by

χe(0)(ω=0)∼te2RGe0(ω=0)=te2Ωe,(54)\chi_e^{(0)}(\omega=0) \sim t_e^2\mathcal{R}G_e^{0}(\omega=0) = \frac{t_e^2}{\Omega_e}, \tag*{(54)}

For temperature-independent tet_e, the temperature dependence of 1/χe1/\chi_e directly follows that of Ωe(T)=ωe2/γe\Omega_e(T) = \omega_e^2/\gamma_e. Using the Green’s function G∗G^*, which describes the interacting phononic (both optical and acoustic) and excitonic excitations, we can evaluate the renormalization of sound velocity above TcT_c. Because the relevant frequencies for the acoustic mode are much lower than that of the optical or excitonic modes, we can keep ω\omega only in the bare acoustic mode’s Green’s function:

G∗=(ωp12000−v10ωp2200−v200ωp320−v3000−ωe2−cs2q22csq−βsq−v1−v2−v3−βsqΩe)−1,(55)G^* = \begin{pmatrix} \frac{\omega_{p1}}{2} & 0 & 0 & 0 & -v_1 \\ 0 & \frac{\omega_{p2}}{2} & 0 & 0 & -v_2 \\ 0 & 0 & \frac{\omega_{p3}}{2} & 0 & -v_3 \\ 0 & 0 & 0 & -\frac{\omega_e^2-c_s^2q^2}{2c_sq} & -\beta_s\sqrt{q} \\ -v_1 & -v_2 & -v_3 & -\beta_s\sqrt{q} & \Omega_e \end{pmatrix}^{-1}, \tag*{(55)}

in which finite linewidths of phonon modes are neglected. The pole of this Green’s function is obtained by requiring the determinant of inverse G∗G^{*} to be zero:

ω2−cs2q22csq=−βs2qΩe−∑i=132vi2ωpi2(56)\frac{\omega^{2}-c_{s}^{2}q^{2}}{2c_{s}q}=-\frac{\beta_{s}^{2}q}{\Omega_{e}-\sum_{i=1}^{3}\frac{2v_{i}^{2}}{\omega_{pi}^{2}}} \tag*{(56)}

Identifying ω2\omega^{2} as c~s2q2\tilde{c}_{s}^{2}q^{2}, in which c~s\tilde{c}_{s} represents the renormalized sound velocity above TcT_{c}, we have

c~s2=cs2−2βs2csΩe−∑i=132vi2ωpi2(57)\tilde{c}_{s}^{2}=c_{s}^{2}-\frac{2\beta_{s}^{2}c_{s}}{\Omega_{e}-\sum_{i=1}^{3}\frac{2v_{i}^{2}}{\omega_{pi}^{2}}} \tag*{(57)}

Illustration of the fitting model for the Fano interference

To provide more insights into the fitting model described in Appendix § 6.1, we consider a simplified case in which only one phonon mode couples to an excitonic continuum. When exciton-phonon interaction vv is zero, the total Raman response is reduced to the sum of the excitonic component and the phononic component:

χ′′(ω)=te2ωΩe2+ω2+4tp2γpωpω(ω2−ωp2)2+2γp2(ω2+ωp2)+γp4(58)\chi^{\prime\prime}(\omega)=\frac{t_{e}^{2}\omega}{\Omega_{e}^{2}+\omega^{2}}+\frac{4t_{p}^{2}\gamma_{p}\omega_{p}\omega}{(\omega^{2}-\omega_{p}^{2})^{2}+2\gamma_{p}^{2}(\omega^{2}+\omega_{p}^{2})+\gamma_{p}^{4}} \tag*{(58)}

in which Ωe\Omega_{e} stands for ωe2/γe\omega_{e}^{2}/\gamma_{e}.

In Figure 28 we show the excitonic continuum coupling with the B2g(1)_{2g}^{(1)} and B2g(2)_{2g}^{(2)} phonon modes in an individually way to illustrate the effect of coupling. First we discuss the renormalization effect. The renormalized phonon mode shifts in energy and broadens in lineshape. If we neglect the bare phonon width γp\gamma_{p} for simplicity, the central energy of the renormalized phonon mode, ωpv\omega_{pv}, can be expressed as

Coupling of the excitonic continuum with phonon modes

Figure 28. Coupling of the excitonic continuum with (a) the B2g(1)_{2g}^{(1)} phonon mode and (b) the B2g(2)_{2g}^{(2)} phonon mode. The parameters used for plotting are obtained from fitting the 380 K spectrum of Ta2_2NiSe5_5. The Raman response χ′′(ω)\chi^{\prime\prime}(\omega) of the excitonic and phononic excitations for cases without (dashed lines) and with (solid lines) the effect of coupling are compared. The solid red lines correspond to the addition of the solid blue, solid green, and solid orange lines.

ωpv=ωp+v2(v2−2ωpΩe)2(v2Ωe+ωpΩe2+ωp3)(59)\omega_{pv}=\omega_{p}+\frac{v^{2}(v^{2}-2\omega_{p}\Omega_{e})}{2(v^{2}\Omega_{e}+\omega_{p}\Omega_{e}^{2}+\omega_{p}^{3})} \tag*{(59)}

For both the B2g(1)_{2g}^{(1)} and B2g(2)_{2g}^{(2)} phonon modes, v2v^{2} is smaller than 2ωpΩe2\omega_{p}\Omega_{e}; therefore ωpv\omega_{pv} is smaller than ωp\omega_{p}, meaning the renormalized phonon mode shifts to lower energy. Interestingly, at frequency ωpv\omega_{pv} the renormalized excitonic continuum has the same intensity as the unrenormalized continuum; below ωpv\omega_{pv} the excitonic response is enhanced while above ωpv\omega_{pv} the excitonic response is suppressed.

Second we discuss the interference effect. For the interference term, corresponding to the off-diagonal elements of the renormalized Green’s function, there is a frequency ωint\omega_{int} at which it changes the sign. The frequency ωint\omega_{int} has the following expression

ωint=2Ωeγp+γp2+ωp2.(60)\omega_{int}=\sqrt{2\Omega_{e}\gamma_{p}+\gamma_{p}^{2}+\omega_{p}^{2}}. \tag*{(60)}

We note that ωint\omega_{int} is a bit larger than ωp\omega_{p}. The shape of the interference term is controlled by sign of vv: for positive vv, below the zero-intensity point the interference term has

positive intensity while the intensity is negative above the zero-intensity point; for negative vv, the opposite is true. Far away from the resonance, the interference term decays to zero because the phonon mode has negligible intensity at such high energy.

Finally we note that when ω\omega is much larger than ωp\omega_p and γe\gamma_e, the Raman response is In Figure 29 we show how the coupling strength vv, and the energy at which the excitonic continuum has maximum intensity Ωe=ωe2/γe\Omega_e = \omega_e^2/\gamma_e influence the combined Raman response. The lineshape becomes more asymmetric with increasing v/ωpv/\omega_p [Fig. 29 (a)]. When Ωe\Omega_e is varied, we find that the lineshape becomes more asymmetric for smaller Ωe\Omega_e [Fig. 29 (b)].

Influence of the coupling strength and the frequency of maximum excitonic continuum intensity

Figure 29. The influence of the coupling strength vv, and the frequency at which the excitonic continuum has maximum intensity Ωe=ωe2/γe\Omega_{e}=\omega_{e}^{2}/\gamma_{e} on the Raman response χ′′(ω)\chi^{\prime\prime}(\omega) of the coupled excitonic and phononic modes. (a) The combined response for varying v/ωpv/\omega_{p} with fixed Ωe/ωp=1.3\Omega_{e}/\omega_{p}=1.3; (b) The combined response for varying Ωe/ωp\Omega_{e}/\omega_{p} with fixed v/ωp=0.3v/\omega_{p}=0.3. For all panels, tp=3.8t_{p}=3.8 arb. units; te=7.2t_{e}=7.2 arb. units; ωp=7.8\omega_{p}=7.8 meV; γp=0.38\gamma_{p}=0.38 meV. The ratios derived from the 380 K spectrum of Ta2_2NiSe5_5 are Ωe/ωp=1.3\Omega_{e}/\omega_{p}=1.3 and v/ωp=0.3v/\omega_{p}=0.3. The horizontal axis is normalized to the phonon frequency ωp\omega_{p}.

approximately te2/ωt_e^2/\omega. Therefore, tet_e can be determined by fitting the data at large ω\omega.

The fitted parameters of the anharmonic decay model for Ta2_2NiSe5_5 and Ta2_2NiS5_5

The temperature dependence of both frequency and FWHM for the most phonon modes of Ta2_2NiSe5_5 and Ta2_2NiS5_5 modes can be accounted by the anharmonic decay model [Eq. (16-17)]. The values for the parameters from fitting are summarized in Table VI.

Modeω0\omega_0ω2\omega_2Γ0\Gamma_0Γ2\Gamma_2
Ta2_2NiSe5_5
Ag(1)A_g^{(1)}4.235(16)0.0038(10)0.06(4)0.0014(15)
B2g(1)B_{2g}^{(1)}8.721(5)0.0290(11)0.073(15)0.019(3)
Ag(2)A_g^{(2)}12.484(12)0.0405(22)0.09(3)0.025(6)
B2g(2)B_{2g}^{(2)}15.35(4)0.070(15)0.05(10)0.09(4)
Ag(3)A_g^{(3)}17.077(12)0.191(5)0.08(5)0.114(23)
B2g(3)B_{2g}^{(3)}18.45(3)0.050(16)0.11(10)0.08(5)
Ag(4)A_g^{(4)}22.28(3)0.082(15)0.12(8)0.05(4)
Ag(5)A_g^{(5)}24.287(11)0.129(6)0.05(3)0.072(15)
Ag(6)A_g^{(6)}27.17(4)0.137(16)0.23(11)0.09(5)
Ag(7)A_g^{(7)}29.62(6)0.21(3)0.18(16)0.12(9)
Ag(8)A_g^{(8)}36.75(14)0.30(8)0.3(4)0.09(24)
Ta2_2Ni(Se0.75_{0.75}S0.25_{0.25})5_5
Ag(1)A_g^{(1)}4.277(16)0.0039(9)0.12(5)0.0025(34)
B2g(1)B_{2g}^{(1)}8.758(7)0.0322(15)0.252(21)0.028(4)
Ta2_2Ni(Se0.33_{0.33}S0.67_{0.67})5_5
Ag(1)A_g^{(1)}4.630(12)0.0048(8)0.10(2)0.0012(11)
B2g(1)B_{2g}^{(1)}8.739(19)0.032(4)0.479(26)0.022(4)
Ta2_2NiS5_5
Ag(1)A_g^{(1)}5.025(10)0.0040(7)0.068(25)0.0014(19)
Ag(2)A_g^{(2)}16.198(4)0.0610(13)0.051(9)0.0359(26)
Ag(3)A_g^{(3)}19.276(11)0.134(4)0.11(3)0.070(11)
Ag(4)A_g^{(4)}33.92(4)0.111(19)0.28(11)0.15(6)
Ag(5)A_g^{(5)}36.634(10)0.138(5)0.056(28)0.112(16)
Ag(6)A_g^{(6)}40.29(20)0.21(12)0.6(7)0.3(4)
Ag(7)A_g^{(7)}43.39(6)0.29(4)0.01(14)0.17(9)
Ag(8)A_g^{(8)}49.72(3)0.23(3)0.04(10)0.45(8)

Table VI. TABLE VI. The fitting parameters of the anharmonic decay model [Eq. (16-17)] for the Raman-active optical phonon modes of Ta2_2Ni(Se1−x_{1-x}Sx_x)5_5 family. The frequencies and FWHM below the phase transition temperature are fitted. The anharmonic decay model is not applicable for the B2gB_{2g}-symmetry modes of Ta2_2NiS5_5. The units are in meV.

References

  1. [1]W. Kohn, Excitonic phases, Physical Review Letters 19, 439 (1967).DOI
  2. [2]B. I. Halperin and T. M. Rice, Possible anomalies at a semimetal-semiconductor transition, Rev. Mod. Phys. 40, 755 (1968).DOI
  3. [3]W. Kohn and D. Sherrington, Two kinds of bosons and bose condensates, Rev. Mod. Phys. 42, 1 (1970).DOI
  4. [4]T. Pillo, J. Hayoz, H. Berger, F. Lévy, L. Schlapbach, and P. Aebi, Photoemission of bands above the Fermi level: The excitonic insulator phase transition in 1T−TiSe2, Phys. Rev. B 61, 16213 (2000).
  5. [5]H. Cercellier, C. Monney, F. Clerc, C. Battaglia, L. Despont, M. G. Garnier, H. Beck, P. Aebi, L. Patthey, H. Berger, and L. Forró, Evidence for an Excitonic Insulator Phase in 1T−TiSe2, Phys. Rev. Lett. 99, 146403 (2007).pubmed.ncbi.nlm.nih.gov/17930692
  6. [6]C. Monney, C. Battaglia, H. Cercellier, P. Aebi, and H. Beck, Exciton Condensation Driving the Periodic Lattice Distortion of 1T−TiSe2, Phys. Rev. Lett. 106, 106404 (2011).
  7. [7]A. Kogar, M. S. Rak, S. Vig, A. A. Husain, F. Flicker, Y. I. Joe, L. Venema, G. J. MacDougall, T. C. Chiang, E. Fradkin, J. van Wezel, and P. Abbamonte, Signatures of exciton condensation in a transition metal dichalcogenide, Science 358, 1314 (2017).DOI
  8. [8]Y. F. Lu, H. Kono, T. I. Larkin, A. W. Rost, T. Takayama, A. V. Boris, B. Keimer, and H. Takagi, Zero-gap semiconductor to excitonic insulator transition in Ta2NiSe5, Nat. Commun. 8, 14408 (2017).
  9. [9]F. D. Salvo, C. Chen, R. Fleming, J. Waszczak, R. Dunn, S. Sunshine, and J. A. Ibers, Physical and structural properties of the new layered compounds Ta2NiS5 and Ta2NiSe5, J. Less Common Met. 116, 51 (1986).
  10. [10]Y. Wakisaka, T. Sudayama, K. Takubo, T. Mizokawa, M. Arita, H. Namatame, M. Taniguchi, N. Katayama, M. Nohara, and H. Takagi, Excitonic Insulator State in Ta2NiSe5 Probed by Photoemission Spectroscopy, Phys. Rev. Lett. 103, 026402 (2009).pubmed.ncbi.nlm.nih.gov/19659224
  11. [11]K. Seki, Y. Wakisaka, T. Kaneko, T. Toriyama, T. Konishi, T. Sudayama, N. L. Saini, M. Arita, H. Namatame, M. Taniguchi, N. Katayama, M. Nohara, H. Takagi, T. Mizokawa, and Y. Ohta, Excitonic Bose-Einstein condensation in Ta2NiSe5 above room temperature, Phys. Rev. B 90, 155116 (2014).DOI
  12. [12]M. D. Watson, I. Marković, E. A. Morales, P. Le Fèvre, M. Merz, A. A. Haghighirad, and P. D. C. King, Band hybridization at the semimetal-semiconductor transition of Ta2NiSe5 enabled by mirror-symmetry breaking, Phys. Rev. Research 2, 013236 (2020).
  13. [13]S. Mor, M. Herzog, D. Golež, P. Werner, M. Eckstein, N. Katayama, M. Nohara, H. Takagi, T. Mizokawa, C. Monney, and J. Stähler, Ultrafast electronic band gap control in an excitonic insulator, Phys. Rev. Lett. 119, 086401 (2017).DOI
  14. [14]Y. Murakami, D. Golež, M. Eckstein, and P. Werner, Photoinduced enhancement of excitonic order, Phys. Rev. Lett. 119, 247601 (2017).DOI
  15. [15]Q. M. Liu, D. Wu, Z. A. Li, L. Y. Shi, Z. X. Wang, S. J. Zhang, T. Lin, T. C. Hu, H. F. Tian, J. Q. Li, T. Dong, and N. L. Wang, Photoinduced multistage phase transitions in Ta2NiSe5, Nature Communications 12, 2050 (2021).
  16. [16]H. Ning, O. Mehio, M. Buchhold, T. Kurumaji, G. Refael, J. G. Checkelsky, and D. Hsieh, Signatures of Ultrafast Reversal of Excitonic Order in Ta2NiSe5, Phys. Rev. Lett. 125, 267602 (2020).DOI
  17. [17]D. Werdehausen, T. Takayama, M. Höppner, G. Albrecht, A. W. Rost, Y. Lu, D. Manske, H. Takagi, and S. Kaiser, Coherent order parameter oscillations in the ground state of the excitonic insulator Ta2NiSe5, Science Advances 4, aap8652 (2018).DOI
  18. [18]T. Kaneko, T. Toriyama, T. Konishi, and Y. Ohta, Orthorhombic-to-monoclinic phase transition of Ta2NiSe5 induced by the Bose-Einstein condensation of excitons, Phys. Rev. B 87, 035121 (2013).DOI
  19. [19]G. Mazza, M. Rösner, L. Windgätter, S. Latini, H. Hübener, A. J. Millis, A. Rubio, and A. Georges, Nature of Symmetry Breaking at the Excitonic Insulator Transition: Ta2NiSe5, Phys. Rev. Lett. 124, 197601 (2020).DOI
  20. [20]E. Baldini, A. Zong, D. Choi, C. Lee, M. H. Michael, L. Windgätter, I. I. Mazin, S. Latini, D. Azoury, B. Lv, A. Kogar, Y. Wang, Y. Lu, T. Takayama, H. Takagi, A. J. Millis, A. Rubio, E. Demler, and N. Gedik, The spontaneous symmetry breaking in Ta2NiSe5 is structural in nature (2020), arXiv:2007.02909 [cond-mat.str-el].arxiv.org/abs/2007.02909
  21. [21]Y. Chiba, T. Mitsuoka, N. L. Saini, K. Horiba, M. Kobayashi, K. Ono, H. Kumigashira, N. Katayama, H. Sawa, M. Nohara, Y. F. Lu, H. Takagi, and T. Mizokawa, Valence-band insulator in proximity to excitonic instability, Phys. Rev. B 100, 254129 (2019).
  22. [22]W. Hayes and R. Loudon, Scattering of Light by Crystals (John Wiley and Sons, New York, 1978).
  23. [23]P. A. Volkov, M. Ye, H. Lohani, I. Feldman, A. Kanigel, and G. Blumberg, Critical charge fluctuations and emergent coherence in a strongly correlated excitonic insulator, npj Quant. Mater. 6, 52 (2021).
  24. [24]P. A. Volkov, M. Ye, H. Lohani, I. Feldman, A. Kanigel, and G. Blumberg, Failed excitonic quantum phase transition in Ta2Ni(Se1−xSx)5, An accompanying paper (2021), arXiv:2104.07032 [cond-mat.str-el].
  25. [25]P. E. Blöchl, Projector augmented-wave method, Phys. Rev. B 50, 17953 (1994).pubmed.ncbi.nlm.nih.gov/9976227
  26. [26]J. P. Perdew, K. Burke, and M. Ernzerhof, Generalized gradient approximation made simple, Phys. Rev. Lett. 77, 3865 (1996).pubmed.ncbi.nlm.nih.gov/10062328
  27. [27]P. Giannozzi, O. Andreussi, T. Brumme, O. Bunau, M. B. Nardelli, M. Calandra, R. Car, C. Cavazzoni, D. Ceresoli, M. Cococcioni, N. Colonna, I. Carnimeo, A. D. Corso, S. de Gironcoli, P. Delugas, R. A. DiStasio, A. Ferretti, A. Floris, G. Fratesi, G. Fugallo, R. Gebauer, U. Gerstmann, F. Gustinino, T. Gorni, J. Jia, M. Kawamura, H.-Y. Ko, A. Kokalj, E. Küçükbenli, M. Lazzeri, M. Marsili, N. Marzari, F. Mauri, N. L. Nguyen, H.-V. Nguyen, A. O. de-la Roza, L. Paulatto, S. Poncé, D. Rocca, R. Sabatini, B. Santra, M. Schlipp, A. P. Seitsonen, A. Smogunov, I. Timrov, T. Thonhauser, P. Umari, N. Vast, X. Wu, and S. Baroni, Advanced capabilities for materials modelling with quantum ESPRESSO, Journal of Physics: Condensed Matter 29, 465901 (2017).
  28. [28]A. Dal Corso, Pseudopotentials periodic table: From H to Pu, Computational Materials Science 95, 337 (2014).DOI
  29. [29]S. Grimme, Semiempirical gga-type density functional constructed with a long-range dispersion correction, J. Comput. Chem. 27, 1787 (2006).pubmed.ncbi.nlm.nih.gov/16955487
  30. [30]G. Kresse and J. Furthmüller, Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set, Computational Materials Science 6, 15 (1996).DOI
  31. [31]G. Kresse and J. Furthmüller, Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set, Phys. Rev. B 54, 11169 (1996).pubmed.ncbi.nlm.nih.gov/9984901
  32. [32]A. Togo and I. Tanaka, First principles phonon calculations in materials science, Scripta Materialia 108, 1 (2015).arxiv.org/abs/1506.08498
  33. [33]A. Nakano, K. Sugawara, S. Tamura, N. Katayama, K. Matsubayashi, T. Okada, Y. Uwatoko, K. Munkanaka, A. Nakao, H. Sagayama, R. Kumai, K. Sugimoto, N. Maejima, A. Machida, T. Watanuki, and H. Sawa, Pressure-induced coherent sliding-layer transition in the excitonic insulator Ta₂NiSe₅, IUCrJ 5, 158 (2018).
  34. [34]S. A. Sunshine and J. A. Ibers, Structure and physical properties of the new layered ternary chalcogenides tantalum nickel sulfide (Ta₂NiS₅) and tantalum nickel selenide (Ta₂NiSe₅), Inorganic Chemistry 24, 3611 (1985).DOI
  35. [35]R. J. Elliott, J. A. Krumhansl, and P. L. Leath, The theory and properties of randomly disordered crystals and related physical systems, Rev. Mod. Phys. 46, 465 (1974).DOI
  36. [36]A. S. Barker and A. J. Sievers, Optical studies of the vibrational properties of disordered solids, Rev. Mod. Phys. 47, S1 (1975).DOI
  37. [37]L. Bergman and R. J. Nemanich, Raman Spectroscopy for Characterization of Hard, Wide-Bandgap Semiconductors: Diamond, GaN, GaAlN, AlN, BN, Annu. Rev. Mater. Sci. 26, 551 (1996).
  38. [38]S. Y. Kim, Y. Kim, C.-J. Kang, E.-S. An, H. K. Kim, M. J. Eom, M. Lee, C. Park, T.-H. Kim, H. C. Choi, B. I. Min, and J. S. Kim, Layer-Confined Excitonic Insulating Phase in Ultrathin Ta₂NiSe₅ Crystals, ACS Nano 10, 8888 (2016).DOI
  39. [39]J. Yan, R. Xiao, X. Luo, H. Lv, R. Zhang, Y. Sun, P. Tong, W. Lu, W. Song, X. Zhu, and Y. Sun, Strong Electron-Phonon Coupling in the Excitonic Insulator Ta₂NiSe₅, Inorganic Chemistry 58, 9036 (2019), pMID: 31246443.DOI
  40. [40]M.-J. Kim, A. Schulz, T. Takayama, M. Isobe, H. Takagi, and S. Kaiser, Phononic soft mode behavior and a strong electronic background across the structural phase transition in the excitonic insulator Ta₂NiSe₅, Phys. Rev. Research 2, 042039 (2020).DOI
  41. [41]K. Kim, H. Kim, J. Kim, C. Kwon, J. S. Kim, and B. J. Kim, Direct observation of excitonic instability in Ta₂NiSe₅, Nature Communications 12, 1969 (2021).
  42. [42]K. Kim, H. Kim, J. Kim, C. Kwon, J. S. Kim, and B. J. Kim, Author Correction: Direct observation of excitonic instability in Ta₂NiSe₅, Nature Communications 12, 2998 (2021).DOI
  43. [43]A. Subedi, Orthorhombic-to-monoclinic transition in Ta₂NiSe₅ due to a zone-center optical phonon instability, Phys. Rev. Materials 4, 083601 (2020).DOI
  44. [44]A. S. Barker and R. Loudon, Response functions in the theory of Raman scattering by vibrational and polariton modes in dielectric crystals, Rev. Mod. Phys. 44, 18 (1972).DOI
  45. [45]U. Fano, Effects of configuration interaction on intensities and phase shifts, Phys. Rev. 124, 1866 (1961).DOI
  46. [46]M. V. Klein, Electronic Raman scattering, in Light Scattering in Solids I, Vol. 8, edited by M. Cardona (Springer-Verlag, Berlin, 1983) Chap. 4, pp. 147–202.
  47. [47]G. Blumberg, M. V. Klein, L. Börjesson, R. Liang, and W. N. Hardy, Investigation of the temperature dependence of electron and phonon Raman scattering in single crystal YBa₂Cu₃O₆.₉₅₂, J. Supercond. 7, 445 (1994).DOI
  48. [48]Note that the interaction-induced linewidth γ̂p is often significantly larger than the bare width γp, which is the case for presented data above Tc. Consequently, an analysis of the phonon lineshape that does not explicitly include the interaction with the continuum (e.g. by simply adding the intensities of Fano-shaped phonons and the continuum) would lead to an erroneous large intrinsic phonon linewidth.
  49. [49]Note that the cubic term in the denominator can be neglected as long as the frequency is not far from ωp.
  50. [50]Y. Gallais and I. Paul, Charge nematicity and electronic raman scattering in iron-based superconductors, Comptes Rendus Physique 17, 113 (2016).DOI
  51. [51]A. E. Böhmer, P. Burger, F. Hardy, T. Wolf, P. Schweiss, R. Fromknecht, M. Reinecker, W. Schranz, and C. Meingast, Nematic Susceptibility of Hole-Doped and Electron-Doped BaFe₂As₂ from Shear Modulus Measurements, Phys. Rev. Lett. 112, 047001 (2014).pubmed.ncbi.nlm.nih.gov/24580480
  52. [52]A. Nakano, T. Hasegawa, S. Tamura, N. Katayama, S. Tsutsui, and H. Sawa, Antiferroelectric distortion with anomalous phonon softening in the excitonic insulator Ta₂NiSe₅, Phys. Rev. B 98, 045139 (2018).DOI
  53. [53]H.-H. Kung, M. Salehi, I. Boulares, A. F. Kemper, N. Koirala, M. Brahlek, P. Lošt’ák, C. Uher, R. Merlin, X. Wang, S.-W. Cheong, S. Oh, and G. Blumberg, Surface vibrational modes of the topological insulator Bi₂Se₃ observed by Raman spectroscopy, Phys. Rev. B 95, 245406 (2017).DOI
  54. [54]Chauvière, L. and Gallais, Y. and Cazayous, M. and Méasson, M. A. and Sacuto, A. and Colson, D. and Forget, A., Raman scattering study of spin-density-wave order and electron-phonon coupling in Ba(Fe₁−xCoₓ)₂As₂, Phys. Rev. B 84, 104508 (2011).
  55. [55]T. I. Larkin, A. N. Yaresko, D. Pröpper, K. A. Kikoïn, Y. F. Lu, T. Takayama, Y.-L. Mathis, A. W. Rost, H. Takagi, B. Keimer, and A. V. Boris, Giant exciton Fano resonance in quasi-one-dimensional Ta₂NiSe₅, Phys. Rev. B 95, 195144 (2017).DOI
  56. [56]P. G. Klemens, Anharmonic Decay of Optical Phonons, Phys. Rev. 148, 845 (1966).DOI
  57. [57]K. Mu, H. Chen, Y. Li, Y. Zhang, P. Wang, B. Zhang, Y. Liu, G. Zhang, L. Song, and Z. Sun, Electronic structures of layered Ta₂NiS₅ single crystals revealed by high-resolution angle-resolved photoemission spectroscopy, J. Mater. Chem. C 6, 3976 (2018).DOI
  58. [58]P. Andrich, H. M. Bretscher, Y. Murakami, D. Golež, B. Remez, P. Telang, A. Singh, L. Harnagea, N. R. Cooper, A. J. Millis, P. Wern, A. K. Sood, and A. Rao, Imaging the coherent propagation of collective modes in the excitonic insulator candidate Ta₂NiSe₅ at room temperature (2020), arXiv:2003.10799 [cond-mat.str-el].DOI
  59. [59]H. M. Bretscher, P. Andrich, P. Telang, A. Singh, L. Harnagea, A. K. Sood, and A. Rao, Ultrafast melting and recovery of collective order in the excitonic insulator Ta₂NiSe₅, Nature Communications 12, 1699 (2021).
  60. [60]Y.-S. Zhang, J. A. N. Bruin, Y. Matsumoto, M. Isobe, and H. Takagi, Thermal transport signatures of the excitonic transition and associated phonon softening in the layered chalcogenide Ta₂NiSe₅ (2021), arXiv:2103.13782 [cond-mat.str-el].DOI
  61. [61]H. Arima, Y. Naito, K. Kudo, N. Katayama, H. Sawa, M. Nohara, Y. F. Lu, K. Kitagawa, H. Takagi, Y. Uw atoko, and K. Matsubayashi, Resistive anisotropy of candidate excitonic insulator Ta₂NiSe₅ under pressure, J. Phys.: Conf. Ser. 1609, 012001 (2020).DOI
  62. [62]K. Sun, M. A. Stroscio, and M. Dutta, Graphite c-axis thermal conductivity, Superlattices and Microstructures 45, 60 (2009).

Paper details

vlnc:W_sxkh3grs
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents