Fig. 3. Scaling of the NLAHE with sample conductivity.

a Main panel: $E_z^{2\omega }/{\left(E_x^{\omega }\right)}^2$ and ${\sigma }_{xx}$ vs. temperature for $I\parallel a$ and $I\parallel b$ in 127-nm-thick sample. Inset: longitudinal resistivity ${\rho }_{xx}$ vs. temperature for $I\parallel a$ in same sample. ${\rho }_{xx}$ crosses over from linear temperature dependence to saturation below $T_L~18\mathrm{K}$. b Main panel: $E_z^{2\omega }/{\left(E_x^{\omega }\right)}^2$ vs. ${\sigma }_{xx}/{\sigma }_{xx0}$ for all four samples along a and b axes. All traces show scaling close to a power law with exponent of 2 at lower conductivity (higher temperature) and exhibits an upturn as ${\sigma }_{xx}$ approaches ${\sigma }_{xx0}$. Open circles mark the data points taken at the temperature closest to $T_L$. Inset: local scaling exponent $\upsilon$ vs. temperature $T$. $\upsilon$ continually increases above 2 for all traces below $T_L$.