Lattice-Boltzmann simulations of a turbulent duct 1 flow have been carried out to obtain trajectories of passive 2 tracers in the conditions of a series of microgravity experi-3 ments of turbulent bubble suspensions. The statistics of these 4 passive tracers are compared to the corresponding measure-5 ments for single-bubble and bubble-pair statistics obtained 6 from particle tracking techniques after the high-speed cam-7 era recordings from drop-towers experiments. In the condi-8 tions of the present experiments, comparisons indicate that 9 experimental results on bubble velocity fluctuations are not 10 consistent with simulations of passive tracers, which points 11 in the direction of an active role of bubbles. The present 12 analysis illustrates the utility of a recently introduced exper-13 imental setup to generate controlled turbulent bubble sus-14 pensions in microgravity 15 Keywords turbulent flow · bubble dispersion · bubble 16 interactions · microgravity · drop tower · lattice-boltzmann 17 simulations 18 P. Bitlloch · J. Casademunt tion. To this end, Lattice-Boltzmann simulations of the flow 56 have been carried out. By using virtual passive tracers, these 57 simulations allowed to compare their statistics with that of 58 the real bubbles. Simulation results also enabled to compare 59 the two-point statistics of passive tracers to that from the 60 particle-tracking of bubble pairs. This gives interesting in-61 formation on the flow mixing properties and the probability 62 of bubble encounters. In particular we compared the char-63 acteristic times of separation between pairs of passive trac-64 ers in simulations and pairs of bubbles in the experiments. 65 All these information allowed to obtain an additional and 66 more accurate knowledge of the behavior of turbulent bub-67 bly flows under microgravity conditions. 68 2 Lattice-Boltzmann simulations 69 In order to characterize the structure and properties of a tur-70 bulent flow through a duct of square section we have per-71 formed 3D Lattice-Boltzmann simulations. The channel has 72 been discretized into a uniform grid of 320x80x80 liquid 73 nodes, representing a portion of 400x100x100 mm 3 of the 74 duct, with periodic conditions at its ends. After some tests 75 of various discretizations of the model, we decided for the 76 D3Q15 Lattice Bhatnagar-Gross-Krook (BGK) model with 77 mid-way wall boundary conditions for no-slip walls (Nour-78 galiev et al, 2003; Bitlloch, 2012). 79 For the sake of stability, since it is not possible to simu-80 late all scales of turbulence down to the Kolmogorov length, 81 we used the Smagorinsky coefficient for sub-grid scale fil-82 tering (Hou et al, 1994). This method is based on the cal-83 culation of the local effective viscosity that would dissipate 84 the sub-grid effects generated at each local point. Some rem-85 nant numerical instability was controlled by an additional 86 smoothing procedure that preserved mass and momentum 87 (Bitlloch, 2012). 88 The present code was parallelized and ran in the Mare 89 Nostrum su...