Supplemental material to “ Timescales in creep and yielding of attrac - tive gels ”

In the present work, the Plexiglas cylinders used in our Couette geometry were sand-blasted in order to provide a roughness of about 1 μm which leads to significant scattering of the incident ultrasonic pulses. This results in spurious fixed echoes in the raw ultrasonic data that get mixed with the echoes backscattered by the moving particles. Such fixed echoes appear as vertical lines in the spatiotemporal diagram of Supplemental Figure 1(a) that shows the successive pressure signals p(tus, t) coded in gray levels as a function of the ultrasonic time-of-flight tus (horizontal axis) after a single pulse is sent at time t (vertical axis). Using the crosscorrelation algorithm described in Ref.1 on such raw ultrasonic data leads to a dramatic underestimation of the local velocities at the location of these fixed echoes. In order to remove the undesired fixed echoes before data analysis, we average the ultrasonic signals recorded during the systematic preshear step at γ̇p = 103 s−1 (see Sect. 2.3) and subtract this average to each raw pressure signal recorded subsequently during the actual experiment. The result is shown in Supplemental Figure 1(b). In the averaging process at large shear rate, all contributions from acoustic scatterers within the sheared fluid cancel out and one is left with the static spurious signal. Subtracting this signal to the raw data appears as a very efficient way to remove the spurious contributions of the ultrasonic waves scattered off by the surface roughness of the outer fixed wall. t (s ) (a)