Please cite this article as: Malgorzata Peszynska, Anna Trykozko, Gabriel Iltis, Steffen Schlueter, Dorthe Wildenschild, Biofilm growth in porous media: experiments, computational modeling at the porescale, and upscaling, Advances in Water Resources (2015), Highlights 1 • We use 3D imaging with a barium-based contrasting agent to obtain 2 porescale geometries filled with biofilm 3 • We simulate the flow in the porescale geometries with and without 4 biofilm, and upscale the results to the conductivities which compare 5 • well with the experimental values, and which show the dependence of 6 the degree of bioclogging on the flow rates 7 • We simulate biomass growth and transport coupled to the flow and 8 obtain morphologies similar to those in the experiment 9 • We show several reduced models for conductivities and their depen-10 dence on the biofilm growth Abstract 21 Biofilm growth changes many physical properties of porous media such 22 as porosity, permeability and mass transport parameters. The growth de-23 pends on various environmental conditions, and in particular, on flow rates. 24 Modeling the evolution of such properties is difficult both at the porescale 25 where the phase morphology can be distinguished, as well as during up-26 scaling to the corescale effective properties. Experimental data on biofilm 27 growth is also limited because its collection can interfere with the growth, 28 while imaging itself presents challenges. 29 In this paper we combine insight from imaging, experiments, and nu-30 merical simulations and visualization. The experimental dataset is based on 31 glass beads domain inoculated by biomass which is subjected to various flow 32 conditions promoting the growth of biomass and the appearance of a biofilm 33 phase. The domain is imaged and the imaging data is used directly by a 34 computational model for flow and transport. The results of the computa-35 tional flow model are upscaled to produce conductivities which compare well 36 with the experimentally obtained hydraulic properties of the medium. The 37 flow model is also coupled to a newly developed biomass-nutrient growth 38 model, and the model reproduces morphologies qualitatively similar to those 39 observed in the experiment.40 Keywords: 41 porescale modeling, imaging porous media, microtomography, reactive 42 transport, biomass and biofilm growth, parabolic variational inequality, 43 Lagrange multipliers, coupled nonlinear system, multicomponent 44 multiphase flow and transport in porous media 45 2

A major difficulty in modeling multiphase flow in porous media is the emergence of trapped phases. Our experiments demonstrate that gas can be trapped in either single-pores, multipores, or in large connected networks. These large connected clusters can comprise up to eight grain volumes and can contain up to 50% of the whole trapped gas volume. About 85% of the gas volume is trapped by multipore gas clusters. This variety of possible trapped gas clusters of different shape and volume will lead to a better process understanding of bubble-mediated mass transfer. Since multipore gas bubbles are in contact with the solid surface through ultrathin adsorbed water films the interfacial area between trapped gas clusters and intergranular capillary water is only about 80% of the total gas surface. We could derive a significant (R 2 5 0.98) linear relationship between the gas-water-interface and gas saturation. We found no systematic dependency of the front velocity of the invading water phase in the velocity range from 0.1 to 0.6 cm/min corresponding to capillary numbers from 2 3 10 27 to 10 26 . Our experimental results indicate that the capillary trapping mechanism is controlled by the local pore structure and local connectivity and not by thermodynamics, i.e., by the minimum of the Free Energy, at least in the considered velocity range. Consistent with this physical picture is our finding that the trapping frequency (5 bubble-size distribution) reflects the pore size distribution for the whole range of pore radii, i.e., the capillary trapping process is determined by statistics and not by thermodynamics.

Natural rocks exhibit a heterogeneous wettability due to their heterogeneous mineral composition and distribution of adsorbates caused by fluid-solid interactions