At the end of the first column on page 5048, Ref. [1], it is written: “Therefore the straight lines drawn in the figure represent the fitting using $n_{2\text{eff}}=n_2+n_{4}I$ giving $n_2=(1.0\pm 0.2)\times {10}^{-22}{\mathrm{m}}^2/\mathrm{W}$ and $n_4=(1.1\pm 0.3)\times {10}^{-36}{\mathrm{m}}^4/{\mathrm{W}}^2$ for $\lambda =1064\mathrm{nm}$;….”

This procedure cannot be applied in the present case because the relative errors of the $n_{2\text{eff}}$ at intensities smaller than $400\mathrm{GW}/{\mathrm{cm}}^2$ were underestimated (see Fig. 4). By analyzing the results in a recent experiment, we found that the first three data points in (blue) triangles (Fig. 4) should be removed because of the poor signal-to-noise (S/N) ratio. Only the data for $I_0>400\mathrm{GW}/{\mathrm{cm}}^2$, acquired with larger S/N ratio, can be considered. In conclusion, at $\lambda =1064\mathrm{nm}$, the values should be corrected to $n_2=(0.8\pm 0.2)\times {10}^{-20}{\mathrm{m}}^2/\mathrm{W}$ and $n_4\le 1.1\times {10}^{-36}{\mathrm{m}}^4/{\mathrm{W}}^2$ [2].