Dimensionless falling distance vs dimensionless time , for the median (◇), first (○), and third (◻) quartiles for batch B (a) and batch E (b). The different lines indicate the best linear fits. Experiments were held in a vessel. The suspension concentration was .
Normalized cumulative sum of the square of the radius () vs dimensionless radius () for the five batches of particles: A (○), B (◻), C (▵), D (◇), and E (▿).
Dimensionless interface thickness vs dimensionless falling distance , for batch C at . The line indicates the best linear fit. The different cross sections presented are (▵), (○), and (◻).
Dimensionless interface thickness vs dimensionless falling distance . The lines indicate the best linear fits. Three different experiments were realized for each different batch of particles; A (○), B (◻), C (▵), D (◇), and E (▿).
(a) Dimensionless interface thickness vs dimensionless predicted quartile interface . The symbols correspond to the batches A (○), B (◻), C (▵), D (◇), and E (▿). (b) Dimensionless half-widths (open symbols) and (filled symbols) vs dimensionless predicted half-widths and , respectively.
Relative quartile interface thickness vs the concentration . The solid curve represents a quadratic fit to the first five concentration values. The dotted curve is explained in the conclusion.
Concentration profile of the sedimentation front vs the normalized height . The results correspond to experiments in the cross-section vessel with particles of batch B at . The symbols correspond to the different times (○), 669.1 (×), 961.8 (▵), and 1087.2 (◻).
Concentration profiles at , , , , and for a box with 4000 point-particles, Fourier modes, sedimentation velocity , and monodisperse particles . For this figure, an average was made over 80 realizations instead on the normal 20 for this number of particles.
Height of concentration quartiles, i.e., , , and , as functions of time for the same simulation as Fig. 8.
Growth of thickness of the front in time for the same monodisperse simulation as Fig. 8. Also plotted are the best linear fit (solid line) and the best fit (dotted curve).
Scaled rate of growth of the front as a function of the numerical resolution for monodisperse simulations with sedimentation velocity and (+), 1372 (×), 4000 (▵), and 6912 (◻).
(a) Scaled growth rate of the front as a function of the scaled polydispersity , with various and , and with (+), 1372 (×), 4000 (▵), and 10976 (◻). The line is the contribution of the polydispersity without hydrodynamic interactions. The dotted curve is explained in the conclusion. (b) Scaled rates of growth of the (lower data) and the (upper data) quartiles for the same simulations. The line is the contribution of the polydispersity without hydrodynamic interactions.
Concentration profiles of Fig. 8 replotted in a frame moving with the mean sedimentation speed and rescaled by the growing thickness of the front .
Dimensionless isoconcentration velocities and relative interface thicknesses deduced from the experiments and from predictions accounting solely for polydispersity for the different particle batches at . The experimental data correspond to averages over three experimental runs.
Dimensionless isoconcentration velocities and quartile interface thickness deduced from the experiments for particles of batch B at different concentrations. The experimental data correspond to averages over three experimental runs.
Article metrics loading...
Full text loading...