Highly anisotropic transient optical response of charge density wave order in ${\mathrm{ZrTe}}_3$

Low dimensionality in charge density wave (CDW) systems leads to anisotropic optical properties, in both equilibrium and nonequilibrium conditions. Here, we perform polarized two-color pump probe measurements on a quasi-one-dimensional material ${\mathrm{ZrTe}}_3$, in order to study the anisotropic transient optical response in the CDW state. Profound in-plane anisotropy is observed with respect to the polarization of probe photons. Below $T_{\mathrm{CDW}}$ both the quasiparticle relaxation signal and amplitude mode (AM) oscillation signal are much larger, with ${\mathbf{E}}_{\mathrm{pr}}$ nearly parallel to the $a$ axis (${\mathbf{E}}_{\mathrm{pr}}\parallel a$) than for ${\mathbf{E}}_{\mathrm{pr}}$ parallel to the $b$ axis (${\mathbf{E}}_{\mathrm{pr}}\parallel b$). This reveals that the ${\mathbf{E}}_{\mathrm{pr}}\parallel a$ signal provides much a larger response to the variation of the CDW gap. Interestingly, the lifetime of the AM oscillations observed with ${\mathbf{E}}_{\mathrm{pr}}\parallel b$ is longer than ${\mathbf{E}}_{\mathrm{pr}}\parallel a$. Moreover, at high pump fluence where the electronic order melts and the AM oscillations vanish for ${\mathbf{E}}_{\mathrm{pr}}\parallel a$, the AM oscillatory response still persists for ${\mathbf{E}}_{\mathrm{pr}}\parallel b$. We discuss the possible origins that lead to such an unusual discrepancy between the two polarizations.

Figure 1

(a) and (b) Time-resolved reflectivity under different polarization conditions at room temperature and 10 K, respectively, with pump fluence $\sim 60\mu \mathrm{J}$/${\mathrm{cm}}^2$. The inset of (a) depicts the reflection geometry of the experiment, where the $a$ axis lies in the horizontal plane of incidence, and the $b$ axis is along the vertical direction.

Figure 2

(a) and (b) $\mathrm{\Delta }R/R$ as a function of time delay at different temperatures for ${\mathbf{E}}_{\mathrm{pr}}\parallel a$ and ${\mathbf{E}}_{\mathrm{pr}}\parallel b$, with pump fluence $\sim 60\mu \mathrm{J}$/${\mathrm{cm}}^2$. (c)–(f) Temperature dependence of several parameters characterizing the QP relaxation dynamics. The red and blue markers correspond to spectra obtained with ${\mathbf{E}}_{\mathrm{pr}}\parallel a$ and ${\mathbf{E}}_{\mathrm{pr}}\parallel b$, respectively. The solid lines in (c) are the fitted curves with the RT model using $T$-independent gap [26]. The dashed lines are guides to the eye.

Figure 3

(a) and (b) The oscillation part of transient reflectivity change for ${\mathbf{E}}_{\mathrm{pr}}\parallel a$ and ${\mathbf{E}}_{\mathrm{pr}}\parallel b$, with pump fluence $\sim 60\mu \mathrm{J}$/${\mathrm{cm}}^2$. Lines are vertically shifted for clarity. (c) and (d) correspond to the fast Fourier transformation of the oscillation part for ${\mathbf{E}}_{\mathrm{pr}}\parallel a$ and ${\mathbf{E}}_{\mathrm{pr}}\parallel b$, respectively. (e) The temperature dependence of the AM peak frequency obtained by fitting the peak position in the FFT spectra of (d). (f) The lifetime of AM oscillation for different probe photon polarizations, obtained by fitting the time-domain oscillation.

Figure 4

(a) and (b) The transient reflectivity change at different pump fluences at 10 K, with ${\mathbf{E}}_{\mathrm{pr}}\parallel a$ and ${\mathbf{E}}_{\mathrm{pr}}\parallel b$, respectively. The inset of (b) shows the amplitude of the slower positive exponential at different fluences. For ${\mathbf{E}}_{\mathrm{pr}}\parallel b$ we can only obtain a reasonable fitting parameter $A_{sb}$ below 179 $\mu \mathrm{J}$/${\mathrm{cm}}^2$. (c) and (d) The oscillation part of the time-domain spectra of (a) and (b), obtained by subtracting the fitted QP relaxation curve of from the raw traces in (a) and (b). Data are taken from 2 ps to avoid the residual quasiparticle relaxation signal which is not perfectly subtracted. (e) The fluence dependence of the amplitude (upper panel) and lifetime (lower panel) of the AM oscillation, obtained by fitting the time-domain oscillation spectra with Eq. (2). Details about the fitting procedure are shown in the Supplemental Material [25]. The shaded region in the upper panel expected intensity range in the condition that the existence of the AM signal at high fluence with ${\mathbf{E}}_{\mathrm{pr}}\parallel b$ is purely induced by the mismatch of penetration depth [25]. (f) and (g) The fast Fourier transformation of the time-domain oscillation signal for different probe polarizations.