Extended Data Fig. 3: Point-source model of the squared visibilities.

Panel a corresponds to 12 July (two observing sequences), panel b to 19 July (two observing sequences) and panel c to 21 July (one observing sequence). Data from 12 July were collected with medium VLTI baselines (longest baseline: 90m), while data from 19 and 21 July were obtained with large baselines (longest baseline: 128m and 126m respectively), which also set the maximum angular resolution of the observations. The leftmost plots (in gray shades) display the collected data in the Einstein u_Ev_E-plane (in \({\theta }_{E}^{-1}\) units), i.e. six VLTI baselines per observing sequence split into six effective wavelengths, marked by rainbow-colored dots. The point-source squared-visibility patterns are shown in gray scale ranging from ∣V_E∣² = 0 (black) to 1 (white). The inner dashed circle marks the typical angular resolution, and the outer circle twice the typical resolution. The plots in the middle show the data with their 1−σ error bars (in rainbow colors) as a function of B/λ (in \({\theta }_{E}^{-1}\) units), where B is the baseline length and λ the observing wavelength. The black lines correspond to the point-source model (shown only in the vicinity of the data points). The rightmost plots show the geometry of the microlensing event. The microlens (black cross, undetected by the interferometer) is set fixed in the center of the North-East reference frame (both axes are in θ_E units). The red and blue dots are respectively the major and minor point-like images of the source (which position is marked by a yellow dot), aligned with the microlens along the dotted line. The trajectory of the source relative to the microlens is indicated by the bold straight line with arrow, while the curved arrow indicates the direction of rotation of the arcs. This figure shows that a point-source model is not a valid approximation here.