betaFeedback

WorksPhysics and Astronomy

Role of boundary conditions and wall orientation on the transport of settling inertial particles in wall-bounded flows

Abstract

Gravitational settling affects particle transport in turbulent flows in two ways; explicitly, by introducing a finite settling velocity, and implicitly, by modifying how the particles interact with the flow field. For wall-bounded flows, when the wall is horizontal (gravity perpendicular to the wall) both the explicit and implicit effects of settling impact the particle transport towards the wall, whereas when the wall is vertical (gravity parallel to the wall) only the implicit effect plays a role. Surprisingly, it was recently demonstrated that even when the settling parameter SvSv is very small, settling can play a very significant role in controlling the near-wall transport in a horizontal channel. In this paper, we use direct numerical simulations to explore how this finding is affected by the particle boundary conditions and whether it also occurs in vertical channels where only the implicit effect of settling plays a role. We show that the sensitivity of the particle transport to SvSv depends upon the particle boundary conditions, with elastic-collisions (that generate a zero mean particle flux) and absorbing-wall conditions (that generate a negative mean particle flux) exhibiting qualitatively and quantitatively different sensitivities to SvSv. It is found that for vertical channels the impact of settling on the particle transport is in fact negligible if SvSv is small, and only becomes significant when Sv≥O(1)Sv\geq O(1). Finally, we examine the physical mechanisms governing the settling velocity of particles in vertical channel flows, and find that low-speed streaks can play a significant role in suppressing the settling velocity of the particles near the walls, in contrast to the core region where the preferential sweeping mechanism leads to an enhancement of their settling velocity.

Introduction

The transport of small, heavy particulate matter in wall-bounded turbulent flows occurs in many environmental multiphase flows, such as volcanic ash transport [1, 2], dust storms [3, 4] and pollen transport [5, 6]. In these flows, the gravitational settling of the particles plays an important role in their transport towards and away from the surface of the Earth. If the particles are modeled as being inertialess, then under the assumption that they are sufficiently small, the particles will settle on average at their Stokes settling velocity. When their inertia is not negligible, as can be the case for particle transport in the lower portion of the atmospheric boundary layer [7], then their settling velocity depends on their inertia through the role of multiple competing mechanisms [8, 9]. Many open questions remain regarding understanding and modeling the transport of inertial, settling particles in wall-bounded flows. Discrepancies between observations and modeling of dust size distributions in the planetary boundary layer have been reported [10, 11, 12, 13], and it is possible that some of these discrepancies are due to the models ignoring a number of the physical mechanisms that arise due to the combined effects of particle inertia and settling, as discussed in [8].

The simplest context for understanding how the combined effects of settling and inertia impact particle transport in turbulence is that of the transport of small (compared to the Kolmogorov scale), heavy (compared to the fluid density) particles in homogeneous turbulence, which can occur far from flow boundaries [14]. Pioneering studies on this problem are those of Maxey [15] and Wang and Maxey [16], which revealed that inertial particles preferentially sample flow regions where the strain-rate dominates the local velocity gradients compared with the rotation-rate. Gravitational settling introduces an asymmetry into how the particles interact with these regions, leading to a preference for particles to sample downward moving, strain dominated regions of the flow, an effect referred to as the preferential sweeping mechanism. This mechanism leads to an enhancement of the mean particle settling velocity, and is often thought to be strongest when the settling parameter SvSv and Stokes number StSt (both based on the Kolmogorov scales) are O(1)O(1). Tom and Bragg [17] also developed a multiscale version of the preferential sweeping mechanism to explain in detail the role that different sized eddies play in this process, and how the enhanced settling depends on StSt, SvSv and the flow Taylor Reynolds number ReλRe_\lambda.

Bragg, Richter and Wang [9] showed that in the context of settling in a horizontal channel flow, the near-wall accumulation of particles is still highly sensitive to SvSv even when Sv≪1Sv \ll1. Based on an analysis of the near-wall asymptotics of the particle statistics in the regime Sv≪1Sv \ll1, they showed that this occurs because near to the wall, the other mass flux contributions arising due to particle inertia (which dominate further away from the wall when Sv≪1Sv \ll1) become sub-leading compared with the Stokes settling flux due to the turbulent fluctuations vanishing at the wall. Therefore, the widespread assumption that settling can be ignored when Sv≪1Sv \ll1 is not in general valid, and can therefore play a potentially significant role in wall deposition rates.

In many problems involving wall-bounded particle transport, for example in atmospheric flows over sloped terrain, or for particle transport through an inclined channel, the wall is not perpendicular to gravity. In the limit of a vertical wall, the explicit effect of gravity on the wall-normal particle transport vanishes. Nevertheless, gravity can still strongly influence the wall-normal transport due to its implicit effect on the way the particles interact with the turbulent flow. Another open question is the extent to which the findings in [9] depend on the particular choice of particle boundary conditions. For example, that study considered an absorbing wall boundary condition, but how would the sensitivity to SvSv change if elastic particle-wall collisions were considered? The purpose of this paper is to explore these open questions.

Theory

Mechanisms governing wall-normal particle transport

In this work, we are interested in small (compared with the Kolmogorov length scale), heavy particles with particle-to-fluid density ratio ρp/ρf≫1\rho_p/\rho_f \gg1, which are subject to linear drag and gravitational forces [18]. When all variables are expressed in non-dimensional wall-units, the equation reads

x¨p(t)≡v˙p(t)=1St(up(t)−vp(t))+SvSteg,(1)\ddot{\boldsymbol{x}}^p(t) \equiv\dot{\boldsymbol{v}}^p(t) = \frac{1}{St}\left(\boldsymbol{u}^p(t)-\boldsymbol{v}^p(t)\right) + \frac{Sv}{St}\boldsymbol{e}_g, \tag*{(1)}

where xp(t)\boldsymbol{x}^p(t) and vp(t)\boldsymbol{v}^p(t) are the position and velocity of the particle, respectively, up(t)\boldsymbol{u}^p(t) denotes the fluid velocity at the particle position, and eg\boldsymbol{e}_g denotes the unit vector in the direction of gravity. In the equation, St≡τp/τ∗St \equiv\tau_p/\tau_\ast is the particle Stokes number based on the fluid friction timescale τ∗\tau_\ast, and Sv≡τpg/u∗Sv \equiv\tau_p g/u_\ast is the particle settling number based on the fluid friction velocity u∗u_\ast.

The contribution (Sv/St)eg(Sv/St)\boldsymbol{e}_g in the equation of motion corresponds to the explicit gravitational effect which acts in the direction eg\boldsymbol{e}_g. However, even for particle motion in the plane orthogonal to eg\boldsymbol{e}_g, gravitational settling still impacts the particle motion implicitly through up(t)\boldsymbol{u}^p(t), since up(t)\boldsymbol{u}^p(t) depends on the three-dimensional trajectory of the particle xp(t)\boldsymbol{x}^p(t) which is itself dependent on gravity. This impact of gravity on up(t)\boldsymbol{u}^p(t) corresponds to the implicit effect of gravity discussed above.

An equation for the mean particle momentum in the wall-normal direction (assuming the flow is homogeneous in the other directions) can be derived from (1) using a phase-space PDF analysis [8]:

StDt⟨wp(t)⟩z=⟨up(t)⟩z−⟨wp(t)⟩z−St(∇zS+Sϱ∇zϱ)+Sveg⋅ez,(2)StD_t\langle w^p(t)\rangle_z = \langle u^p(t)\rangle_z-\langle w^p(t)\rangle_z-St\left(\nabla_z S+\frac{S}{\varrho}\nabla_z\varrho\right)+Sv\boldsymbol{e}_g\cdot\boldsymbol{e}_z, \tag*{(2)}

where zp(t)z^p(t), wp(t)w^p(t) and up(t)u^p(t) are the projections of xp(t)\boldsymbol{x}^p(t), vp(t)\boldsymbol{v}^p(t) and up(t)\boldsymbol{u}^p(t) in the ez\boldsymbol{e}_z (wall-normal) direction, respectively, Dt≡∂t+⟨wp(t)⟩z∇zD_t \equiv\partial_t+\langle w^p(t)\rangle_z\nabla_z, ϱ\varrho is the particle concentration PDF, ⟨⋅⟩z\langle\cdot\rangle_z denotes an ensemble average conditioned on zp(t)=zz^p(t)=z, and S≡⟨[wp(t)−⟨wp(t)⟩z]2⟩zS\equiv\langle[w^p(t)-\langle w^p(t)\rangle_z]^2\rangle_z is the variance of the wall-normal particle velocity.

In (2), −St∇zS-St\nabla_zS corresponds to the turbophoretic drift mechanism that drives the particles towards the wall in the near-wall region. The term −St(S/ϱ)∇zϱ-St(S/\varrho)\nabla_z\varrho represents an inertial diffusion mechanism that arises due to the diffusive effect of the decoupling between the particle and fluid velocities. The term ⟨up(t)⟩z\langle u^p(t)\rangle_z denotes the mean wall-normal fluid velocity sampled by the particles, and this is non-zero due to the particles preferentially sampling the flow. As discussed in [9], for a horizontal channel flow, in the quasi-homogeneous region of a boundary layer, this term will act to drive the particles towards the wall and is associated with the preferential sweeping mechanism. Close to the wall this term drives the particles away from the wall, due to the diffusive effect of the turbulent fluctuations on the particle transport. Indeed, for a zero mass flux configuration where ⟨wp(t)⟩z=0\langle w^p(t)\rangle_z = 0, the positivity of ⟨up(t)⟩z\langle u^p(t)\rangle_z is required near the wall to balance the drift towards the wall due to the turbophoretic drift (and gravitational settling for the case of a horizontal channel). A more detailed discussion of each term appearing in (2) can be found in [9].

Explicit and implicit gravitational effects in the zero mass flux case

Corresponding to (2), the equation governing the particle concentration is [8]

∂tϱ+∇z(ϱ⟨wp(t)⟩z)=0,(3)\partial_t \varrho+ \nabla_z\left(\varrho\langle w^p(t)\rangle_z\right) = 0, \tag*{(3)}

such that for a statistically stationary system we have ϱ⟨wp(t)⟩z=Φ\varrho\langle w^p(t)\rangle_z = \Phi, where Φ\Phi is the constant mass flux that is determined by the boundary conditions for the particles. When the mass flux is zero Φ=0\Phi= 0, (2) reduces to

0=⟨up(t)⟩z−St(∇zS+Sϱ∇zϱ)+Sveg⋅ez.(4)0 = \langle u^p(t)\rangle_z - St\left(\nabla_z S + \frac{S}{\varrho}\nabla_z\varrho\right) + Sv e_g\cdot e_z. \tag*{(4)}

For a vertical wall, eg⋅ez=0e_g\cdot e_z = 0 and the explicit gravitational settling contribution vanishes in the wall-normal direction. The implicit effect of settling on the remaining terms persists, however. For a horizontal wall, eg⋅ez=−1e_g\cdot e_z = -1 and both the explicit and implicit effects of gravitational settling influence the particle transport.

Rearranging (4) and integrating over the region z≤H−dp/2z \le H-d_p/2 (where HH is the height of the flow, and dpd_p is the particle diameter) leads to the following formal solution

ϱ(z)=ϱ0exp⁡(−eg⋅ezSvSt∫zH−dp/21S dq)exp⁡(−1St∫zH−dp/21S[⟨up(t)⟩q−St∇qS] dq)≡ϱ0exp⁡(A)exp⁡(B),(5)\begin{aligned} \varrho(z) = \varrho_0\exp\left(-e_g\cdot e_z\frac{Sv}{St}\int_z^{H-d_p/2}\frac{1}{S}\,dq\right)\exp\left(-\frac{1}{St}\int_z^{H-d_p/2}\frac{1}{S}\left[\langle u^p(t)\rangle_q-St\nabla_q S\right]\,dq\right) \\ &\equiv\varrho_0\exp(A)\exp(B), \tag*{(5)} \end{aligned}

where ϱ0≡ϱ(H−dp/2)\varrho_0 \equiv\varrho(H-d_p/2) which must be chosen such that ∫dp/2H−dp/2ϱ(z) dz=1\int_{d_p/2}^{H-d_p/2}\varrho(z)\,dz = 1, where , and we have introduced A(z)≡−eg⋅ezSvSt∫zH−dp/21S dqA(z)\equiv-e_g\cdot e_z\frac{Sv}{St}\int_z^{H-d_p/2}\frac{1}{S}\,dq and B(z)≡−1St∫zH−dp/21S(⟨up(t)⟩q−St∇qS) dqB(z)\equiv-\frac{1}{St}\int_z^{H-d_p/2}\frac{1}{S}\left(\langle u^p(t)\rangle_q-St\nabla_q S\right)\,dq for simplicity of discussion. A(z)A(z) is associated with the gravitational settling contribution in (4) while B(z)B(z) is associated with the difference between the preferential sampling and turbophoretic drift terms, ⟨up(t)⟩z−St∇zS\langle u^p(t)\rangle_z-St\nabla_z S. We will now consider the contributions from A(z)A(z) and B(z)B(z) to the behavior of ϱ\varrho.

When Sv=0Sv = 0, A(z)=0A(z) = 0 and Sikowsky [19] showed that ϱ\varrho exhibits a power-law singularity in the limit z→0z \to0 (except when StSt is very large, in which case ϱ\varrho is constant in this limit), and that this comes from the term involving B(z)B(z). To understand the role of A(z)A(z) we can consider the regime Sv≪1Sv \ll1 and suppose that in this regime the effect of settling is perturbative. Then, the quantities SS and ⟨up(t)⟩z\langle u^p(t)\rangle_z can be expanded as a regular perturbation series in SvSv, with the leading-order terms independent of SvSv. Let us denote these leading-order contributions using the superscript [0][0]. According to the analysis of [19, 20], S[0]∼az4S^{[0]} \sim az^4 and ⟨up(t)⟩z[0]∼bz3\langle u^p(t)\rangle_z^{[0]} \sim bz^3 for St≪1St \ll1 and z≪1z \ll1, where a,ba, b are positive and depend on StSt but not zz. Using these, the asymptotic behavior of (5) for z≪1z \ll1 (with z≪Hz \ll H) and for St≪1St \ll1, Sv≪1Sv \ll1 is

ϱ(z)∼ϱ0(zH)b−4aStaStexp⁡(−Sv3aSteg⋅ezz−3).(6)\varrho(z) \sim\varrho_0 \left(\frac{z}{H}\right)^{\frac{b-4aSt}{aSt}} \exp\left(-\frac{Sv}{3aSt}e_g \cdot e_z z^{-3}\right). \tag*{(6)}

When St≥O(1)St \ge O(1) and inertial effects dominate the particle motion (but not so large that the particles move ballistically through the boundary layer), the analysis of [19] shows that S[0]∼azγS^{[0]} \sim az^\gamma with γ(St)∈[3,4]\gamma(St) \in[3, 4], and the contribution from ⟨up(t)⟩q\langle u^p(t)\rangle_q in (5) is sub-leading. In this case, the asymptotic behavior of (5) for z≪1z \ll1 (with z≪Hz \ll H) and for St≥O(1)St \ge O(1), Sv≪1Sv \ll1 is

ϱ(z)∼ϱ0(zH)−γexp⁡(−Sva(1−γ)Steg⋅ezz1−γ).(7)\varrho(z) \sim\varrho_0 \left(\frac{z}{H}\right)^{-\gamma} \exp\left(-\frac{Sv}{a(1-\gamma)St}e_g \cdot e_z z^{1-\gamma}\right). \tag*{(7)}

The power-law contributions in (6) and (7) correspond to those discussed [19, 20] for the case of non-settling particles, and describe the near-wall particle accumulation due to the turbophoretic mechanism that causes the particles to drift towards the wall. This contribution is independent of the orientation of the wall in the regime Sv≪1Sv \ll1. The exponential contributions in (6) and (7) correspond to the near-wall accumulation driven by gravitational settling when eg⋅ez∈[−1,0)e_g \cdot e_z \in[-1, 0). These contributions only remain finite when the wall is not vertical, i.e. when eg⋅ez≠0e_g \cdot e_z \ne0. Since these contributions grow as exponentials of power-law functions of zz, they increase rapidly in the limit z→0z \to0 and quickly dominate over the power-law contribution to ϱ(z)\varrho(z) in this limit. Hence, even if Sv≪1Sv \ll1, there will always exist a region sufficiently close to the wall where the gravitational contribution to ϱ(z)\varrho(z) dominates over the power-law contribution due to inertia, and hence in this region, settling is not in fact a perturbative contribution to the particle motion (contrary to the naive assumption that this follows due to Sv≪1Sv \ll1). This is the essence of the argument put forth in [9] for the particular case of a horizontal wall where eg⋅ez=−1e_g \cdot e_z = -1.

It is important to note that although the analysis above only corresponds to the near wall region z≪1z \ll1, it also indirectly implies something about the behavior of ρ(z)\rho(z) further away from the wall. This is because as a PDF it must satisfy the integral constraint ∫dp/2H−dp/2ρ(z) dz=1\int_{d_p/2}^{H-d_p/2} \rho(z)\,dz = 1. Hence, if settling strongly impacts ρ(z)\rho(z) near the wall even when Sv≪1Sv \ll1, it must also affect it further away from the wall in order to preserve ∫dp/2H−dp/2ρ(z) dz=1\int_{d_p/2}^{H-d_p/2} \rho(z)\,dz = 1.

Explicit and implicit gravitational effects in the finite mass flux case

For the case where the mass flux Φ\Phi is finite, e.g. due to an absorbing wall boundary condition, the formal solution to (2) is

ρ(z)=ρ0exp⁡(A)exp⁡(B)(1+ΦSt∫zH−dp/21Sexp⁡[−A(q)−B(q)] dq).(8)\rho(z) = \rho_0 \exp(\mathcal{A})\exp(\mathcal{B})\left(1+\frac{\Phi}{St}\int_z^{H-d_p/2}\frac{1}{S}\exp[-\mathcal{A}(q)-\mathcal{B}(q)]\,dq\right). \tag*{(8)}

When Φ=0\Phi= 0, this result reduces to (5), but in general it contains an additional contribution due to Φ\Phi. The value of Φ\Phi and its dependence on the system parameters is determined by the boundary conditions for the particles. One case of interest is particle transport in a high Reynolds number turbulent boundary layer (e.g. the atmospheric boundary layer) where HH is so large that at z=O(H)z = O(H) the effect of particle inertia negligible and ⟨wp(t)⟩H−dp/2=Sveg⋅ez\langle w^p(t)\rangle_{H-d_p/2} = Sv e_g \cdot e_z, leading to

ρ(z)=ρ0exp⁡(A)exp⁡(B)(1+ρ0SvSteg⋅ez∫zH−dp/21Sexp⁡[−A(q)−B(q)] dq).(9)\rho(z) = \rho_0 \exp(\mathcal{A})\exp(\mathcal{B})\left(1+\rho_0\frac{Sv}{St}e_g \cdot e_z\int_z^{H-d_p/2}\frac{1}{S}\exp[-\mathcal{A}(q)-\mathcal{B}(q)]\,dq\right). \tag*{(9)}

In this case we clearly see that the presence of a finite mass flux due to settling introduces an additional dependence upon SvSv into the solution for ρ(z)\rho(z) compared with the zero-flux case. For this reason, the sensitivity of ρ(z)\rho(z) to SvSv for a vertical channel may differ from that for a horizontal channel (the case which was the case explored in [9]). Since 1Sexp⁡[−A(q)−B(q)]≥0\frac{1}{S}\exp[-\mathcal{A}(q)-\mathcal{B}(q)] \ge0 then the contribution ρ0SvSteg⋅ez∫zH−dp/21Sexp⁡[−A(q)−B(q)] dq\rho_0\frac{Sv}{St}e_g \cdot e_z\int_z^{H-d_p/2}\frac{1}{S}\exp[-\mathcal{A}(q)-\mathcal{B}(q)]\,dq will be negative when eg⋅ez∈[−1,0)e_g \cdot e_z \in[-1,0). Hence the contribution to ρ(z)\rho(z) arising from the finite flux Φ\Phi will act to reduce the near-wall accumulation of the particles compared to the zero-flux case when eg⋅ez∈[−1,0)e_g \cdot e_z \in[-1,0). The impact of settling on ρ(z)\rho(z) may therefore be very different depending on Φ\Phi, e.g. due to the additional contribution of settling to ρ(z)\rho(z) through ρ0SvSteg⋅ez∫zH−dp/21Sexp⁡[−A(q)−B(q)] dq\rho_0\frac{Sv}{St}e_g \cdot e_z\int_z^{H-d_p/2}\frac{1}{S}\exp[-\mathcal{A}(q)-\mathcal{B}(q)]\,dq in the case of a high Reynolds number boundary layer. Intuitively, this makes sense because for elastic collisions that generate Φ=0\Phi= 0, sufficiently strong settling may cause the particles to become trapped near the wall, since only very strong turbulent fluctuations will be able to resuspend them into the flow. Hence, settling will lead to a strong accumulation of the particles at the wall, stronger than for the Sv=0Sv = 0 case. On the other hand, for an absorbing horizontal wall that generates Φ<0\Phi< 0 in the presence of settling, strong settling will simply cause the particles to rapidly fall through the viscous sublayer before being absorbed by the wall. In this case settling suppresses the ability of the particles to linger in the viscous sublayer where the turbulent fluctuations are weak, which is precisely what allows for their strong accumulation in the Sv=0Sv = 0 case.

Direct Numerical Simulations

To investigate the influence of boundary conditions and the explicit and implicit effects of settling, we used DNS to track the motion of settling, inertial particles in an incompressible closed channel flow. For the fluid phase, the momentum equation together with the continuity equation were solved

∂tu+(u⋅∇)u=−1ρf∇p+ν∇2u,(10)\partial_t u + (u \cdot\nabla)u = -\frac{1}{\rho_f}\nabla p + \nu\nabla^2u, \tag*{(10)}
∇⋅u=0,(11)\nabla\cdot u = 0, \tag*{(11)}

where uu is the fluid velocity, pp is pressure, and ν\nu is the kinematic viscosity of the fluid. The fully developed channel flow is driven by a constant mean pressure gradient and is periodic in the streamwise and spanwise directions, with the spatial coordinates in these directions denoted by xx and yy, respectively. The wall-normal spatial coordinate is zz, and at z=0z = 0 (lower wall) and z=Lz=2H1/2z = L_z = 2H_{1/2} (upper wall, where H1/2H_{1/2} is the half channel height), no-slip conditions are imposed for the velocity field. The domain has a size of Lx×Ly×Lz=2π×π×1L_x \times L_y \times L_z = 2\pi\times\pi\times1. A pseudospectral method is used in the periodic directions while a 2nd-order finite difference method is used in the wall-normal direction. Time stepping of the fluid velocity and pressure fields is achieved using a 3rd-order Runge-Kutta method. The channel flows simulated all have a friction Reynolds number of Reτ≡uτH1/2/ν=315Re_\tau\equiv u_\tau H_{1/2}/\nu= 315.

Lagrangian point particles are tracked in the flow by solving equation (1) using a backward Euler method. The fluid velocity at the particle positions is computed with a sixth-order Lagrangian interpolation method. The particle loading is assumed to be small enough to neglect two-way coupling and particle-particle collisions for consistency with the previous study [8] that this work builds on and compares with.

Three closed channel flow configurations are considered: a horizontal channel with reflecting walls for the particles (i.e. elastic particle-wall collisions), a vertical channel with reflecting side walls for the particles, and a horizontal channel with absorbing upper and lower walls for the particles (see Figure 1). In each case, periodic boundary conditions are applied to the particles in the streamwise and spanwise directions.

Three configurations: horizontal channel with $\Phi = 0$, vertical channel with $\Phi = 0$, and horizontal channel with $\Phi < 0$

Figure 1. The three configurations (a) a horizontal channel with Φ=0\Phi= 0 (HZF case), (b) a vertical channel with Φ=0\Phi= 0 (VZF case), and (c) a horizontal channel with Φ<0\Phi< 0 (HNF case). In these schematics, the shaded planes denote the surfaces where the particles rebound elastically, while the non-shaded planes denote the surfaces where either periodic or absorbing wall boundary conditions apply.

For all three configurations, the particles are initially released with a uniform distribution and a velocity equal to their Stokes settling velocity. In the horizontal and vertical channel flows with two reflecting walls (see Figure 1 (a) & (b)), the particles remain in the domain throughout the entirety of the simulation. At steady state, these cases generate a zero particle mean mass flux Φ=0\Phi= 0. For the horizontal channel with absorbing walls (see Figure 1 (c)), new particles must be introduced to replace those that are absorbed by the walls in order to maintain a steady state. These new particles are released randomly at the centerline of the domain with a velocity equal to their Stokes settling velocity plus the local fluid velocity. At steady state, the horizontal channel with absorbing walls generates a negative particle mean mass flux Φ<0\Phi< 0 due to settling. Hereafter, the horizontal and vertical channel cases with Φ=0\Phi= 0 will be referred to by HZF and VZF, respectively, while the horizontal channel cases with Φ<0\Phi< 0 will be referred to by HNF.

Note that in [8, 9] the case with an absorbing wall for the particles used an open channel with an elastically rebounding boundary condition for the particles at the top of the domain. However, for consistency with the other cases explored in this work, a closed channel is used here for the cases with absorbing walls (i.e. the HNF cases). Nevertheless, as we will see, the results for the particle statistics near the lower wall (the region we focus on) using the closed channel are very similar to those in [8, 9] for an open channel.

In the DNS, four values of StSt corresponding to weak to intermediate particle inertia were considered St∈[0.93,46.5]St \in[0.93, 46.5], consistent with the values considered in [8, 9]. For reasons that will be explained in the next section, different values of SvSv in the range Sv∈[0,1]Sv \in[0,1] were explored for the different cases, with smaller values considered for the HZF cases. The particle statistics are computed using a stretched grid with refined near-wall resolution, with grid points Nx×Ny×Nz=128×128×180N_x \times N_y \times N_z = 128 \times128 \times180 corresponding to spacing (in wall-units) Δx+×Δy+×Δz+=30.95×15.48×0.25\Delta_x^+ \times\Delta_y^+ \times\Delta_z^+ = 30.95 \times15.48 \times0.25 (wall), 12.34 (center).

Results and Discussions

Statistically stationary regimes

The near-wall particle concentration in a turbulent channel flow can take a significant amount of time to reach a statistically stationary regime. Since the impact of settling when Sv≪1Sv \ll1 is most important in the near-wall region, we therefore first check that the results do indeed correspond to a statistically stationary regime.

In Figure 2, two HZF cases [(a) & (b)] and two HNF cases [(c) & (d)] are shown. For all StSt and SvSv numbers tested, the maximum particle concentration ϱmax⁡=max⁡[ϱ]\varrho_{\max} = \max[\varrho] is used to test for statistical stationarity. Since the particles accumulate most strongly near the lower wall in the horizontal channel configurations, at steady state ϱmax⁡\varrho_{\max} is associated with ϱ\varrho near or at the lowest grid point. As the plots illustrate, the DNS is run for very long times to ensure that ϱmax⁡\varrho_{\max} reaches an approximate steady state. Once ϱmax⁡\varrho_{\max} has reached its approximate steady state value (which generally occurs by t∼O(105τ∗)t \sim O(10^{5}\tau_{*})), we continue to run the DNS for at least another 1×105τ∗1 \times10^{5}\tau_{*} before the particle statistics are constructed by averaging over many snapshots of the particle positions and velocities in this steady state. The second-order particle statistics ⟨[wp(t)]2⟩z\langle[w^{p}(t)]^{2}\rangle_{z} and ⟨[up(t)]2⟩z\langle[u^{p}(t)]^{2}\rangle_{z} over this same time range are also in a steady state.

Figure 2. Fluid and particle statistics during the simulation (red curve corresponds to the friction velocity uτu_{\tau} as a function of time, shown for reference). (a) ϱmax⁡=max⁡[ϱ]\varrho_{\max} = \max[\varrho] at St=4.65St = 4.65, Sv=10−5Sv = 10^{-5} in HZF case, (b) ϱmax⁡\varrho_{\max} at St=46.5St = 46.5, Sv=10−5Sv = 10^{-5} in HZF case, (c) ϱmax⁡\varrho_{\max} at St=4.65St = 4.65, Sv=10−4Sv = 10^{-4} in HNF case, (d) ϱmax⁡\varrho_{\max} at St=46.5St = 46.5, Sv=10−4Sv = 10^{-4} in HNF case.

Influence of boundary conditions on concentrations in a horizontal channel

In Figure 3 the particle concentration profiles for HNF cases are shown for different StSt, SvSv values. When raising SvSv from Sv=10−4Sv = 10^{-4} to Sv=10−3Sv = 10^{-3}, both of which correspond to the weak-settling regime, the near-wall concentration ϱ\varrho is significantly affected by the change in SvSv, and decreases as SvSv is increased. This is consistent with the analysis in Bragg et al. [9] that demonstrated that the near-wall transport of inertial particles can be strongly affected by settling even when SvSv is very small. In the near wall region for the HNF cases, settling enables the particles to move almost ballistically through the near-wall region, leading to a flattened concentration profile in this region.

Figure 3. Particle concentration ϱ\varrho at different SvSv for (a) St=0.93St = 0.93, (b) St=4.65St = 4.65, (c) St=9.3St = 9.3, (d) St=46.5St = 46.5 in the HNF case.

In Figure 4 the particle concentration profiles for HZF cases are shown. For this configuration we find that SvSv plays an even more significant role when Sv≪1Sv \ll1 compared to the HNF cases just considered in Figure 3. Indeed, we find that we had to use values of SvSv that are in general an order of magnitude smaller than those for the HNF cases in Figure 3 since otherwise the particle statistics do not reach a statistically steady state (instead they all slowly accumulate at the wall). In Figure 4(a) only ϱ\varrho at Sv=0Sv = 0 and 10−510^{-5} are shown, because even at Sv=10−4Sv = 10^{-4}, settling is significant enough to prevent the statistics from reaching a steady state. This issue is highlighted in Figure 5 which shows ϱmax⁡\varrho_{\max} as a function of time for St=0.93St = 0.93 and (a) Sv=10−5Sv = 10^{-5} and (b) Sv=10−4Sv = 10^{-4}. Whereas the case with Sv=10−5Sv = 10^{-5} reaches a steady state after approximately t=1×105τ∗t = 1 \times10^{5}\tau_{*}, for the case with Sv=10−4Sv = 10^{-4}, ϱmax⁡\varrho_{\max} continues to slowly increase with time and does not reach a steady state. This is associated with all of the particles slowly drifting towards the wall and remaining trapped there, which would appear to be the fate of them all for t→∞t \to\infty.

Figure 4. Particle concentration ϱ\varrho at different SvSv for (a) St=0.93St = 0.93, (b) St=4.65St = 4.65, (c) St=9.3St = 9.3, (d) St=46.5St = 46.5 in the HZF case.

Figure 5. Maximum particle concentration ϱmax⁡\varrho_{\max} as a function of time tt for St=0.93St = 0.93 in a HZF case (a) Sv=10−5Sv = 10^{-5}, (b) Sv=10−4Sv = 10^{-4}.

In Figure 4 (b), (Figure 4c) & (Figure 4d) three small SvSv numbers (Sv∈[0,10−5,10−4]Sv \in[0, 10^{-5}, 10^{-4}]) are tested, for which it is shown that the concentration is significantly affected by SvSv when going from 10−510^{-5} to 10−410^{-4}. The sensitivity to SvSv is relatively weak for the case with St=46.5St = 46.5, however. This reflects the fact that when the particle inertia is sufficiently large, inertial effects dominate in the very near wall region, and settling would only play a large role in that region for larger SvSv. The difference between the results for Sv=0Sv = 0 and 10−510^{-5} is in general negligible, however. This is reassuring since the particle equation of motion is regularly perturbed with respect to SvSv and therefore although the particle motion might be sensitive to SvSv even when SvSv is very small, this sensitivity must vanish in the limit Sv→0Sv \to0. The results suggest that the particle motion is extremely sensitive to SvSv, however, and that values as low as 10−510^{-5} are required for the effect of settling to be truly negligible. We again stress that this is not only an issue for the near-wall region, but for the particle concentration throughout the entire flow, because the integral constraint on ϱ\varrho means that any effects of SvSv on ϱ\varrho in the near-wall region are also indirectly felt further away from the wall. This is confirmed in Figure 4 (b), (Figure 4c) & (Figure 4d), with the modification of ϱ\varrho in the near-wall region due to settling also causing ϱ\varrho to be modified further away from the wall so that the integral constraint ∫dp/2H−dp/2ϱ(z) dz=1\int_{d_p/2}^{H-d_p/2} \varrho(z)\,\mathrm{d}z = 1 is preserved.

Whereas for Φ<0\Phi< 0, increasing SvSv led to a decrease in the near-wall particle concentration, for Φ=0\Phi= 0 the opposite occurs. This is in agreement with the observations made earlier regarding equation (9). Clearly, according to the comparison between the cases with Φ=0\Phi= 0 and Φ<0\Phi< 0, the effect of SvSv on ϱ\varrho depends strongly on the particle boundary conditions which determine Φ\Phi, with the effect being different not only quantitatively, but also qualitatively. This agrees with the discussion based on the theoretical result in equation (8).

It is also worth noting that the results in Figure 3 & Figure 4 show that for some cases, the highest particle concentration occurs at a small distance from the wall, including for the cases with Sv=0Sv = 0. This has also been observed in previous DNS of non-settling inertial particles in turbulent channel flow [21]. It is possible that this effect is physical, and is caused by the inertial particle velocities being too large relative to the fluid velocities very close to the wall for the particles to accumulate there, and so they accumulate at a small distance from the wall where the drag forces are more effective in capturing them. On the other hand, it could be due to insufficient statistical convergence very close to the wall and/or a finer grid being needed close to the wall. This may be the case since the asymptotic results in Sikovskiy [19] suggest that for z→0z \to0 and Sv=0Sv = 0 the concentration ϱ\varrho should grow as a power law.

Explicit and implicit effect of settling on particle concentrations

As discussed in Section II, settling can impact ϱ\varrho either explicitly or implicitly, depending on the orientation of the wall relative to gravity. In order to assess the relative importance of these effects, we compare the results from both the HZF (where both the explicit and implicit effects of settling play a role) and VZF (where only the implicit effect of settling plays a role) cases. The results are shown in Figure 6, and while the HZF results were already shown in Figure 4, they are included here again for direct comparison with the VZF cases. In contrast with the HZF cases where ϱ\varrho is very sensitive to settling even when SvSv is small, for the same SvSv numbers the effect of settling on ϱ\varrho is negligible for the VZF cases. This suggests that in the small SvSv regime the implicit effect of settling is negligible, and that the explicit effect is likely the dominant cause of the sensitivity of ϱ\varrho to SvSv in both the HZF and HNF cases. This is consistent with the analysis in [9] which predicts that it is the explicit effect of settling that will cause the particle motion in horizontal flows to be sensitive to SvSv even when SvSv is small (but finite).

Figure 6. DNS results of ϱ\varrho at at (a,b) St=0.93St = 0.93, (c,d) St=4.65St = 4.65, (e,f) St=9.3St = 9.3 and (g,h) St=46.5St = 46.5 in log-log scale, (a),(c),(e),(g) for HZF and (b),(d),(f),(h) VZF cases.

The VZF cases in Figure 6 also include results for Sv=1Sv = 1 in order to assess how large SvSv has to be in vertical channels before the implicit effect of settling starts to impact ϱ\varrho. The results show noticeable differences between the Sv=0Sv = 0 and Sv=1Sv = 1 cases, with the greatest differences occurring away from the wall and closer to the channel core. A practical implication of these results is that

while the neglect of settling in models and simulations of particle transport in horizontal wall-bounded flows is not justified even when Sv≪1Sv \ll1 (unless SvSv is extremely small, e.g. ≤O(10−5)\le O(10^{-5})), it is reasonable for vertical wall-bounded flows unless Sv≥O(1)Sv \ge O(1).

Mechanisms Governing Settling in a Vertical Channel

In [8] we explored in detail the mechanisms governing the settling velocity of inertial particles in HNF cases, where the settling is in the wall-normal direction. We now consider this for our VZF cases where the settling is in the streamwise direction. For the VZF cases, the exact steady-state equation (corresponding to the equation of motion in equation (1)) for the mean particle streamwise velocity can be derived from the phase-space PDF equation and is [8]

⟨wxp(t)⟩z=−StϱSxz∇zϱ−St∇zSxz+⟨uxp(t)⟩z+Ux+Sv(12)\langle w_x^p(t)\rangle_z = -\frac{St}{\varrho}S_{xz}\nabla_z\varrho- St\nabla_z S_{xz} + \langle u_x^p(t)\rangle_z + U_x + Sv \tag*{(12)}

where wxp(t)w_x^p(t) is the streamwise component of the particle velocity, Sxz≡⟨[wxp(t)−⟨wxp(t)⟩z][wp(t)−⟨wp(t)⟩z]⟩zS_{xz} \equiv\langle[w_x^p(t)-\langle w_x^p(t)\rangle_z][w^p(t)-\langle w^p(t)\rangle_z]\rangle_z is the covariance of the streamwise and wall-normal particle velocity, uxp(t)u_x^p(t) is the fluctuating streamwise fluid velocity at the particle position, and UxU_x is the mean streamwise fluid velocity evaluated at the particle position. Far from the wall where gradients in the turbulent statistics are weak, the equation reduces to ⟨wxp(t)⟩z∼⟨uxp(t)⟩z+Ux+Sv\langle w_x^p(t)\rangle_z \sim\langle u_x^p(t)\rangle_z + U_x + Sv (unless StSt is very large, i.e. large enough for the particles to move ballistically through the channel core) and in this region we would expect ⟨uxp(t)⟩z>0\langle u_x^p(t)\rangle_z > 0 due to the preferential sweeping mechanism. The results in Figure 7 confirm this expectation. Close to the wall, however, the results show that ⟨uxp(t)⟩z<0\langle u_x^p(t)\rangle_z < 0. For Sv=0Sv=0 this is known to be associated with the accumulation of the inertial particles in low-speed streaks in the boundary layer [22, 23, 24], and our results show that this continues to be the case even for Sv=O(1)Sv=O(1). The implication of this is that for particles settling near a vertical boundary, the preferential sampling of low speed streaks acts to suppress their settling speed, in contrast to their settling far from boundaries where preferential sampling of downward moving strain-dominated regions of the flow enhances their settling speed through the preferential sweeping mechanism.

Figure 7. Plots of the fluctuating streamwise fluid velocity averaged at the particle positions ⟨uxp(t)⟩z\langle u_x^p(t)\rangle_z at (a) St=0.93St = 0.93, (b) St=4.65St = 4.65, (c) St=9.3St = 9.3, (d) St=46.5St = 46.5

The results in Figure 8 show that close to the wall where the concentration gradients are strong, the diffusive contribution −(St/ϱ)Sxz∇zϱ-(St/\varrho)S_{xz}\nabla_z\varrho in (12) can be significant for larger StSt. This contribution is also negative, hindering the settling of the particles close to the wall together with ⟨uxp(t)⟩z\langle u_x^p(t)\rangle_z. The results in Figure 9 imply that close to the wall the contribution −St∇zSxz-St\nabla_zS_{xz} in (12) can also be significant for larger StSt (due to noise in the data, the figure shows SxzS_{xz} rather than its gradient). This term is by contrast positive close to the wall, and therefore enhances the settling velocity of the particle. The physical interpretation of this term is similar to the standard turbophoretic drift velocity (see [8] for a detailed discussion), except that the standard turbophoretic drift involves exclusively wall-normal particle velocity fluctuations, whereas −St∇zSxz-St\nabla_z S_{xz} involves wall-normal and streamwise particle fluctuating velocity components. In particular, it is associated with the particles transferring streamwise momentum in the wall-normal direction due to their ability to remember their interaction with the flow in regions where the streamwise fluctuations were on average different (due to turbulence inhomogeneity) than those at their current location, thereby transferring streamwise momentum to their current location.

Figure 8. Plots of the diffusion term −(St/ϱ)Sxz∇zϱ-(St/\varrho)S_{xz}\nabla_z\varrho at (a) St=0.93St=0.93, (b) St=4.65St=4.65, (c) St=9.3St=9.3, (d) St=46.5St=46.5

Figure 9. Plots of the covariance SxzS_{xz} at (a) St=0.93St = 0.93, (b) St=4.65St = 4.65, (c) St=9.3St = 9.3, (d) St=46.5St = 46.5

Finally, we consider the overall modification to the particle settling velocity in the vertical channel, ⟨wxp⟩−Ux−Sv\langle w_x^p\rangle- U_x - Sv, which corresponds to the first three terms on the rhs of (12). As indicated earlier, further away from the wall where gradients in the particle statistics are weak, turbulence leads to an enhancement of the particle settling velocity through the preferential sweeping mechanism and ⟨wxp⟩−Ux−Sv>0\langle w_x^p\rangle- U_x - Sv > 0, although the enhancement is negligible unless Sv≥O(1)Sv \ge O(1). Near the wall, however, turbulence significantly suppresses the settling of the inertial particles, and the results suggest that the preferential sampling of the low speed streaks through the sampling term ⟨uxp(t)⟩z\langle u_x^p(t)\rangle_z makes an important contribution to this suppression process.

Figure 10. Plots of the turbulence modification to the particle settling velocity ⟨wxp⟩−Ux−Sv\langle w_x^p\rangle- U_x - Sv in a vertical channel at (a) St=0.93St = 0.93, (b) St=4.65St = 4.65, (c) St=9.3St = 9.3, (d) St=46.5St = 46.5

Conclusions

In this work, we have considered how wall orientation and particle boundary conditions impact the transport of settling inertial particles in wall-bounded turbulent flows. This work was in part motivated by our recent study that found that, contrary to usual expectations, the concentrations of inertial particles could be strongly affected by settling even when the the settling parameter is Sv≪1Sv \ll1 [9]. That study was for particles in a horizontal channel flow with absorbing walls, and in this paper we have explored whether this same surprising behavior remains true for horizontal channel flows with elastic particle-wall collisions, and also whether it still occurs when the channel is vertical. When the channel is vertical, settling does not explicitly affect the wall-normal particle transport and concentration profiles, but it does affect them implicitly through the way that it influences how the particles interact with the turbulent flow field. These questions were examined both theoretically, by analyzing the exact phase-space PDF transport equations governing the particles, and numerically, using DNS. In each case, the particles were assumed to be point particles, governed by drag and gravitational forces. For the horizontal channels, it was found that the type of boundary condition strongly affects the sensitivity of the particle concentrations to SvSv. In particular, the concentrations were far more sensitive to SvSv for elastic walls than absorbing walls, with the concentrations growing in the near wall region with increasing SvSv for the former, and decreasing for the latter. For the elastic walls, even when Sv≪1Sv \ll1, the effect of settling on the transport was so strong that the concentration did not even reach a steady state for some of the Stokes numbers. Furthermore, we found that values as small as Sv=O(10−5)Sv = O(10^{-5}) are required in order for the effect of settling on the particle concentrations to be truly negligible.

We then compared the effect of settling on particle concentrations in horizontal and vertical channels (both having elastic particle-wall collisions). The results showed that in contrast to the horizontal case, the effect of settling is negligible when Sv≪1Sv \ll1 for the vertical case, and only becomes significant when Sv≥O(1)Sv \ge O(1). By implication, this means that the strong effect of settling in the horizontal cases even when Sv≪1Sv \ll1 must be due to the explicit effect of settling, with only a subleading contribution from the implicit effect. Unlike the horizontal channels, in the case of vertical channels, it is safe to neglect settling provided that Sv≪1Sv \ll1.

Finally, we examined the mechanisms governing the settling velocity of the inertial particles in the case of a vertical channel, just as had been done for horizontal channels in [8]. Far from the wall, the turbulent contribution to the settling velocity is associated with the preferential sweeping mechanism which enhances the settling speed of the particles, just as it does for the horizontal channel. As the wall is approached, and the turbulence inhomogeneity becomes significant, the particles no longer preferentially sample downward moving strain dominated regions of the flow but instead preferentially sample regions where the fluctuating streamwise fluid velocity is negative. This has previously been observed for the Sv=0Sv = 0 case where is was associated with the preferential sampling of low speed streak structures in the boundary layer, and the same holds true for Sv=O(1)Sv = O(1). The preferential sampling of such regions of the flow hinders the settling of the particles. Near the wall a diffusive mechanism associated with gradients in the particle concentration also acts to hinder the settling of the particles, while a mechanism analogous to the turbophoretic drift acts to enhance the settling velocity of the particles, and is associated with the transport of streamwise momentum in the wall-normal direction by the particles.

Acknowledgements

Support from the Army Research Office (Award No. W911NF-22-2-0222) is gratefully acknowledged. We also thank Andrew P. Grace for help with using the DNS code.

References

  1. [1]T. Koyaguchi, K. Ochiai, and Y. J. Suzuki, The effect of intensity of turbulence in umbrella cloud on tephra dispersion during explosive volcanic eruptions: Experimental and numerical approaches, Journal of Volcanology and Geothermal Research 186, 68 (2009).DOI
  2. [2]M. Bursik, Q. Yang, A. Bear-Crozier, M. Pavolonis, and A. Tupper, The development of volcanic ash cloud layers over hours to days due to atmospheric turbulence layering, Atmosphere 12, 285 (2021).DOI
  3. [3]R. W. Zurek and L. J. Martin, Interannual variability of planet-encircling dust storms on Mars, J. Geophys. Res. 98, 3247 (1993).DOI
  4. [4]M. J. Strausberg, H. Wang, M. I. Richardson, S. P. Ewald, and A. D. Toigo, Observations of the initiation and evolution of the 2001 Mars global dust storm, J. Geophys. Res. 110, E02006 (2005).DOI
  5. [5]B. J. Viner and R. W. Arritt, Increased pollen viability resulting from transport to the upper boundary layer, Field Crops Res. 119, 195 (2010).DOI
  6. [6]D. Timerman, D. F. Greene, J. Urzay, and J. D. Ackerman, Turbulence-induced resonance vibrations cause pollen release in wind-pollinated plantago lanceolata l. (plantaginaceae), Journal of The Royal Society Interface 11, 20140866 (2014).DOI
  7. [7]D. H. Richter and M. Chamecki, Inertial effects on the vertical transport of suspended particles in a turbulent boundary layer, Boundary Layer Meteorol. 167, 235 (2018).DOI
  8. [8]A. D. Bragg, D. H. Richter, and G. Wang, Mechanisms governing the settling velocities and spatial distributions of inertial particles in wall-bounded turbulence, Phys. Rev. Fluids 6, 064302 (2021).DOI
  9. [9]A. D. Bragg, D. H. Richter, and G. Wang, Settling strongly modifies particle concentrations in wall-bounded turbulent flows even when the settling parameter is asymptotically small, Physical Review Fluids 6, 124301 (2021).DOI
  10. [10]C. L. Ryder, E. J. Highwood, P. D. Rosenberg, J. Trembath, J. K. Brooke, M. Bart, A. Dean, J. Crosier, J. Dorsey, H. Brindley, J. Banks, J. H. Marsham, J. B. McQuaid, H. Sodemann, and R. Washington, Optical properties of saharan dust aerosol and contribution from the coarse mode as measured during the fennec 2011 aircraft campaign, Atmos. Chem. Phys. 13, 303 (2013).DOI
  11. [11]P. D. Rosenberg, D. J. Parker, C. L. Ryder, J. H. Marsham, L. Garcia-Carreras, J. R. Dorsey, I. M. Brooks, A. R. Dean, J. Crosier, J. B. McQuaid, and R. Washington, Quantifying particle size and turbulent scale dependence of dust flux in the sahara using aircraft measurements, J. Geophys. Res. Atmos. 119, 7577 (2014).DOI
  12. [12]A. A. Adebiyi and J. F. Kok, Climate models miss most of the coarse dust in the atmosphere, Science Advances 6, 10.1126/sciadv.aaz9507 (2020).DOI
  13. [13]A. T. Evan, C. Flamant, S. Fiedler, and O. Doherty, An analysis of aeolian dust in climate models, Geophysical Research Letters 41, 5996 (2014).DOI
  14. [14]S. B. Pope, Turbulent Flows (Cambridge University Press, New York, 2000).DOI
  15. [15]M. R. Maxey, The gravitational settling of aerosol particles in homogeneous turbulence and random flow fields, J. Fluid Mech. 174, 441 (1987).DOI
  16. [16]L. P. Wang and M. R. Maxey, Settling velocity and concentration distribution of heavy particles in homogeneous isotropic turbulence, J. Fluid Mech. 256, 27 (1993).DOI
  17. [17]Tom, J., & Bragg, A. D. (2019). Multiscale preferential sweeping of particles settling in turbulence. J. Fluid Mech., 871, 244.
  18. [18]M. R. Maxey and J. J. Riley, Equation of motion for a small rigid sphere in a nonuniform flow, Phys. Fluids 26, 883 (1983).DOI
  19. [19]D. P. Sikovsky, Singularity of inertial particle concentration in the viscous sublayer of wall-bounded turbulent flows, Flow, Turbulence and Combustion 92 (2014).DOI
  20. [20]P. L. Johnson, M. Bassenne, and P. Moin, Turbophoresis of small inertial particles: theoretical considerations and application to wall-modelled large-eddy simulations, J. Fluid Mech. 883, A27 (2020).DOI
  21. [21]C. Marchioli, A. Soldati, J. Kuerten, B. Arcen, A. Tanière, G. Goldensoph, K. Squires, M. Cargnelutti, and L. Portela, Statistics of particle dispersion in direct numerical simulations of wall-bounded turbulence: Results of an international collaborative benchmark test, Int. J. Multiph. Flow 34, 879 (2008).arxiv.org/abs/0801.2349
  22. [22]M. Rashidi, G. Hetsroni, and S. Banerjee, Particle-turbulence interaction in a boundary layer, International Journal of Multiphase Flow 16, 935 (1990).DOI
  23. [23]J. K. Eaton and J. R. Fessler, Preferential concentration of particles by turbulence, Int. J. Multiphase Flow 20, 169 (1994).DOI
  24. [24]C. Marchioli and A. Soldati, Mechanisms for particle transfer and segregation in a turbulent boundary layer, J. Fluid Mech. 468, 283 (2002).DOI

Paper details

vlnc:W_bwmx6hw8
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents