Extended Data Fig. 8: Image processing for the detection of vortices.

Each row indicates different rotation frequency and duration parameters (indicated on the left), for which images are taken following an interaction quench from the supersolid to unmodulated BEC phase. Each column is a step of the processing protocol that proceeds as follows. The data (column 1) are normalized and denoised with a Gaussian filter of size σ = 1 (column 2) and a sharpening mask is applied to magnify the presence of vortices (column 3). The reference image is built from the data image, in which all density variations are eliminated with a Gaussian filter of size σ = 3 (column 4). The residuals (column 5) are obtained from the subtraction of the data to the reference, converting the density depletions to a positive signal. The vortices (black circles) are detected with a peak-detection algorithm with threshold 0.38. The last column shows the location of the vortices on the original image data. Varying the threshold value modifies the absolute vortex count of each individual image but not the overall qualitative result (see Extended Data Fig. 9).