Top panel shows that the flux in the flux tube used to render loop 1 is indeed constant; we also show several cross sections along the tube, revealing that the cross section indeed expands and becomes squished. The middle panel shows the width of loop 1, inferred from the renderings in different projections (the loop was rendered without the second loop so there was no background). We took slices of the renderings along the loop, perpendicular to the direction of the projected axis at each point. We then fit a Gaussian profile to each slice. These inferred widths (red diamonds for the x −  z projection, green triangles for the y −  z projection, and blue squares for the x −  y projection) are substantially different from each other and are also different from the "expected" width if one assumes a circular cross section (solid line). Note also that the width of the loop is plotted in pixels on the rendered images, to demonstrate that the loop is well resolved in all directions. The bottom panel shows the actual widths of loop 1, that is, the major and minor axes of the ellipse fitted into the cross sections. Again, the width under the "circular cross section" assumption is shown for reference (black solid line).