This work presents the GPU acceleration of the open-source code CaNS for very fast massively-parallel simulations of canonical fluid flows. The distinct feature of the many-CPU Navier–Stokes solver in CaNS is its fast direct solver for the second-order finite-difference Poisson equation, based on the method of eigenfunction expansions. The solver implements all the boundary conditions valid for this type of problems in a unified framework. Here, we extend the solver for GPU-accelerated clusters using CUDA Fortran. The porting makes extensive use of CUF kernels and has been greatly simplified by the unified memory feature of CUDA Fortran, which handles the data migration between host (CPU) and device (GPU) without defining new arrays in the source code. The overall implementation has been validated against benchmark data for turbulent channel flow and its performance assessed on a NVIDIA DGX-2 system (16 T V100 32Gb, connected with NVLink via NVSwitch). The wall-clock time per time step of the GPU-accelerated implementation is impressively small when compared to its CPU implementation on state-of-the-art many-CPU clusters, as long as the domain partitioning is sufficiently small that the data resides mostly on the GPUs. The implementation has been made freely available and open source under the terms of an MIT license.

Polymer additives are commonly utilized to manipulate bubbly flows in various applications. Here we investigate the effects of clean and contaminated bubbles driven upward (upflow) in Newtonian and viscoelastic turbulent channel flows. Interface-resolved direct numerical simulations are performed to examine sole and combined effects of soluble surfactant and viscoelasticity using an efficient three-dimensional finite-difference-front-tracking method. The incompressible flow equations are solved fully coupled with the FENE-P viscoelastic model and the equations governing interfacial and bulk surfactant concentrations. The latter coupling is accomplished by a nonlinear equation of state that relates the surface tension to the surfactant concentration. For Newtonian turbulent bubbly flows, the effects of Triton X-100 and 1-pentanol surfactant are examined. It is observed that the sorption kinetics highly affect the dynamics of bubbly flow. A minute amount of Triton X-100 is found to be sufficient to prevent the formation of bubble clusters restoring the single-phase behavior while even two orders of magnitude more 1-pentanol surfactant is not adequate to prevent the formation of layers. For viscoelastic turbulent flows, it is found that the viscoelasticity promotes formation of the bubble wall-layers and thus the polymer drag reduction is completely lost for the surfactant-free bubbly flows, while the addition of small amount of surfactant (Triton X-100) in this system restores the polymer drag reduction resulting in 25% drag reduction for the Wi = 4 case.

We present a numerical method for interface-resolved simulations of evaporating two-fluid flows based on the volume-of-fluid (VoF) method. The method has been implemented in an efficient EFT-based two-fluid Navier-Stokes solver, using an algebraic VoF method for the interface representation, and extended with the transport equations of thermal energy and vaporized liquid mass for the single-component evaporating liquid in an inert gas. The conservation of vaporizing liquid and computation of the interfacial mass flux are performed with the aid of a reconstructed signed-distance field, which enables the use of well-established methods for phase change solvers based on level-set methods. The interface velocity is computed with a novel approach that ensures accurate mass conservation, by constructing a divergence-free extension of the liquid velocity field onto the entire domain. The resulting approach does not depend on the type of interface reconstruction (i.e. can be employed in both algebraic and geometrical VoF methods). We extensively verified and validated the overall method against several benchmark cases, and demonstrated its excellent mass conservation and good overall performance for simulating evaporating two-fluid flows in two and three dimensions.

The rupture of a liquid film due to coflowing turbulent shear flows in the gas phase is studied using a volume-of-fluid method. To simulate this multiphase problem, we use a simplified numerical setup where the liquid film is "sandwiched" between two fully developed boundary layers from a turbulent channel simulation. The film deforms and eventually ruptures within the shear zone created by the coflows. This efficient setup allows systematic variation of physical parameters to gauge their role in the aerodynamically driven deformation and rupture of a liquid film under fully developed sheared turbulence. This work presents a detailed study of the developing pressure field over the deforming film and related aerodynamic effects, as previously suggested by other authors, in particular the role of the inviscid lift and drag forces. A cumulative lift force is introduced to capture the effect of the alternating pressure minima and maxima forming over the film, which amplify and eventually rupture the film. A velocity scale derived from the lift-induced drag force reflects the state of the turbulent boundary layer over the film and collapses the temporal development of this cumulative lift force as well as the amplitude of film deformation with some success for the different film thicknesses and Reynolds numbers.

The rupture of a liquid film due to co-flowing turbulent shear flows in the gas phase is studied using a volume-of-fluid method. To simulate this multiphase problem, we use a simplified numerical setup where the liquid film is 'sandwiched' between two fully developed boundary layers from a turbulent channel simulation. The film deforms and eventually ruptures within the shear zone created by the co-flows. This efficient setup allows systematic variation of physical parameters to gauge their role in the aerodynamically-driven deformation and rupture of a liquid film under fully developed sheared turbulence. The developing pressure field over the deforming film and related aerodynamic effects is studied, the importance of which was previously suggested by other authors, and in particular the role of the inviscid lift and drag forces. A cumulative lift force is introduced to capture the effect of the alternating pressure minima and maxima forming over the film which amplify and eventually rupture the film. A velocity scale derived from the lift-induced drag force reflects the state of the turbulent boundary layer over the film and collapses the temporal development of this cumulative lift force as well as the amplitude of film deformation with some success for the different film thicknesses and Reynolds numbers.

Equation (2.10) in Costa, Brandt & Picano (2020) for the lift force model used in the point-particle direct numerical simulations (DNS), and which is derived from the classical lift force of Saffman (1965), (Equation presented) does not correspond to the force model actually used in the point-particle DNS with lift force presented in the manuscript. Instead, the following equation was used: (Equation presented) which replaces the first occurrence of the term |Us| on the right-hand-side of (1) with |ω|D. We recall that two cases were considered in the manuscript depending on the value of J in the lift force equation: J = 1 in the case denoted PP-Saffman; and J given by (Equation presented) with ϵ = √|ω|ν/|Us|, in the case denoted PP-McLaughlin. Also, equation (2.13) of the manuscript - describing the perfectly elastic hard-sphere rebound - is incorrect; the term D/2 should be D: (Equation presented) Despite the lapse in the manuscript, equation (4) was implemented correctly (Costa et al. 2020). The results from the point-particle DNS with the model reported in (2.10) of Costa et al. (2020) ((1) above) differ from those reported in the manuscript, and are shown (Figure presented) in figure 1 (cf. figures 7 and 8 of Costa et al. (2020)). The statistics presented here have been collected in the fully developed state from 600 samples over a time interval of 250h/Ub, which ensured statistical convergence of the results. The results from the point-particle cases presented in the original manuscript are also reproduced here with this (higher) statistical sampling, and show very minor differences with respect to figures 7 and 8 of Costa et al. (2020). In light of these results, the conclusions drawn from the results in the last section of § 3 of the manuscript must be therefore reformulated: (i) The Saffman lift model does not correctly predict the near-wall statistics of the interface-resolved DNS very close to the wall, including the near-wall concentration peak. (ii) The equation proposed by Mei (1992) that fits the model of McLaughlin (1991) shows results similar to those reported in the original manuscript for this model. That is, it predicts well the near-wall concentration peak, and fails to predict the other observables near the wall. (iii) Equation (2) for Fl presented above, with J = 1, predicts very well all the observables in figure 1. We have therefore accidentally discovered that the expression (2) for Fl predicts the observed particle statistics very well. Still, the reason for the strikingly good agreement remains elusive to us. We hope that this result can be further exploited for the improvement lift force models for point-particle simulations of wall-bounded turbulent flows. 

We present a direct comparison between interface-resolved and one-way-coupled point-particle direct numerical simulations (DNS) of gravity-free turbulent channel flow laden with small inertial particles, with high particle-to-fluid density ratio and diameter of approximately three viscous units. The most dilute flow considered, solid volume fraction O(10(-5)), shows the particle feedback on the flow to be negligible, whereas differences with respect to the unladen case, notably a drag increase of approximately 10 %, are found for a volume fraction O(10(-4)). This is attributed to a dense layer of particles at the wall, caused by turbophoresis, flowing with large particle-to-fluid apparent slip velocity. The most dilute case is therefore taken as the benchmark for assessing the validity of a widely used point-particle model, where the particle dynamics results only from inertial and nonlinear drag forces. In the bulk of the channel, the first- and second-order moments of the particle velocity from the point-particle DNS agree well with those from the interface-resolved DNS. Close to the wall, however, most of the statistics show major qualitative differences. We show that this difference originates from the strong shear-induced lift force acting on the particles in the near-wall region. This mechanism is well captured by the lift force model due to Saffman (J. Fluid Mech., vol. 22 (2), 1965, pp. 385-400), while other widely used, more elaborate, approaches aiming at extending the lift model for a wider range of particle Reynolds numbers can actually underpredict the magnitude of the near-wall particle velocity fluctuations for the cases analysed here.

We use interface-resolved direct numerical simulations to study the dynamics of a single sediment particle in a turbulent open channel flow over a fixed porous bed. The relative strength of the gravitational acceleration, quantified by the Galileo number, is varied so as to reproduce the different modes of sediment transport - resuspension, saltation and rolling. The results show that the sediment dynamics at lower Galileo numbers (i.e. resuspension and saltation) are mainly governed by the mean flow. Here, the regime of motion can be predicted by the ratio between the gravity and the shear-induced boundary force. In these cases, the sediment particle rapidly takes off when exposed to the flow, and proceeds with an oscillatory motion. Increasing the Galileo number, the frequency of these oscillations increases and their amplitude decreases, until the transport mode switches from resuspension to saltation. In this case, the sediment travels by short successive collisions with the bed. Further increasing the Galileo number, the particle rolls without detaching from the bed. Differently from the previous modes, the motion is triggered by extreme turbulent events, and the particle response depends on the specific initial conditions, at fixed Reynolds number. The results reveal that close to the onset of sediment motion, only turbulent sweeps can effectively trigger the particle motion by increasing the stagnation pressure upstream. We show that for the parameters in this study, a criterion based on the streamwise flow-induced force can successfully predict the incipient movement.

Interface-resolved direct numerical simulations are performed to examine the combined effects of soluble surfactant and viscoelasticity on the structure of a bubbly turbulent channel flow. The incompressible flow equations are solved fully coupled with the FENE-P viscoelastic model and the equations governing interfacial and bulk surfactant concentrations. The latter coupling is achieved through a non-linear equation of state which relates the surface tension to the surfactant concentration at the interface. The two-fluid Navier-Stokes equations are solved using a front-tracking method, augmented with a very efficient FFT-based pressure projection method that allows for massively parallel simulations of turbulent flows. It is found that, for the surfactant-free case, bubbles move toward the wall due to inertial lift force, resulting in formation of wall layers and a significant decrease in the flow rate. Conversely, a high-enough concentration of surfactant changes the direction of lateral migration of bubbles, i.e., the contaminated bubbles move toward the core region and spread out across the channel. When viscoelasticity is considered, viscoelastic stresses counteract the Marangoni stresses, promoting formation of bubbly wall-layers and consequently strong decrease in the flow rate. The formation of bubble wall-layers for combined case depends on the interplay of the inertial and elastic, and Marangoni forces. 

During the last decade, many approaches for resolved-particle simulation (RPS) have been developed for numerical studies of finite-size particle-laden turbulent flows. In this paper, three RPS approaches are compared for a particle-laden decaying turbulence case. These methods are, the Volume-of-Fluid Lagrangian method, based on the viscosity penalty method (VoF-Lag); a direct forcing Immersed Boundary Method, based on a regularized delta function approach for the fluid/solid coupling (IBM); and the Bounce Back scheme developed for Lattice Boltzmann method (LBM-BB). The physics and the numerical performances of the methods are analyzed. Modulation of turbulence is observed for all the methods, with a faster decay of turbulent kinetic energy compared to the single-phase case. Lagrangian particle statistics, such as the velocity probability density function and the velocity autocorrelation function, show minor differences among the three methods. However, major differences between the codes are observed in the evolution of the particle kinetic energy. These differences are related to the treatment of the initial condition when the particles are inserted in an initially single-phase turbulence. The averaged particle/fluid slip velocity is also analyzed, showing similar behavior as compared to the results referred in the literature. The computational performances of the different methods differ significantly. The VoF-Lag method appears to be computationally most expensive. Indeed, this method is not adapted to turbulent cases. The IBM and LBM-BB implementations show very good scaling.