Hostname: page-component-7b7567c8c-fvjsw Total loading time: 0 Render date: 2026-11-16T19:31:44.960Z Has data issue: false hasContentIssue false

Asymmetric particle transport in turbulent flows within concentric annular ducts

Published online by Cambridge University Press:  15 June 2026

Taiyang Wang
Affiliation:
HEDPS, Center for Applied Physics and Technology, School of Mechanics and Engineering Science, Peking University, Beijing 100871, PR ChinaState Key Laboratory for Turbulence and Complex Systems, School of Mechanics and Engineering Science, Peking University, Beijing 100871, PR China
Chi Zhang
Affiliation:
HEDPS, Center for Applied Physics and Technology, School of Mechanics and Engineering Science, Peking University, Beijing 100871, PR ChinaState Key Laboratory for Turbulence and Complex Systems, School of Mechanics and Engineering Science, Peking University, Beijing 100871, PR China
Yaomin Zhao*
Affiliation:
HEDPS, Center for Applied Physics and Technology, School of Mechanics and Engineering Science, Peking University, Beijing 100871, PR ChinaState Key Laboratory for Turbulence and Complex Systems, School of Mechanics and Engineering Science, Peking University, Beijing 100871, PR China
*
Corresponding author: Yaomin Zhao, yaomin.zhao@pku.edu.cn

Abstract

Content of image described in text.

We present the first direct numerical simulations of particle-laden turbulent flow in concentric annuli to investigate the effects of transverse curvature over a range of Stokes numbers. The results demonstrate that transverse curvature induces asymmetric radial transport, with particles preferentially drifting towards the outer wall. Unlike canonical planar flows where turbophoresis universally drives near-wall accumulation, the present study identifies a distinct physical regime at the convex inner wall where centrifugal effect competes with turbophoresis. As a consequence, significant particle depletion is observed near the inner wall under strong curvature, and the transient concentration field exhibits a non-monotonic evolution, with the overshoot generally being more evident at higher Stokes numbers. By deriving a transport equation and applying Sturm–Liouville modal analysis, we identify the competition between asymmetric transport modes with different decay rates as the physical mechanism driving this non-monotonic evolution, and establish a reduced-order model that captures the dynamics of the particle concentration near the walls.

Information

Type
JFM Papers
Copyright
© The Author(s), 2026. Published by Cambridge University Press
Figure 0

Table 1.Parameters of the different cases. Here, Reτ$ {\textit{Re}}_\tau$ is the friction Reynolds number, and the N$N$ and Δ+$\Delta ^+$ terms denote the number of grid nodes and the grid resolution in wall units in each direction, respectively. ‘Inner’ and ‘outer’ denote quantities normalised by the viscous scales at the inner and outer walls, respectively.Table 1 long description.

Figure 1

Figure 1.Figure 1 long description.Numerical configuration of the annular duct: (a) the cross-section of the duct; (b) an overview of the computational domain. The contours of axial velocity are shown within the volume domain, and the density contours are displayed on the inner and outer walls.

Figure 2

Figure 2.Figure 2 long description.Near-wall fluid velocity statistics: (a)$(a)$ mean axial fluid velocity uz+$u_z^+$, (b)$(b)$ axial rms fluid velocity fluctuation uz,rms′+$u^{\prime +}_{z,{\textit{rms}}}$, (c)$(c)$ radial rms fluid velocity fluctuation ur,rms′+$u^{\prime +}_{r,{\textit{rms}}}$, and (d)$(d)$ azimuthal rms fluid velocity fluctuation uθ,rms′+$u^{\prime +}_{\theta ,{\textit{rms}}}$. The superscript +$+$ denotes viscous scaling, and r+$r^+$ denotes the wall-normal distance in wall units. The markers denote the data from Kim et al. (1987) (plane channel, Ref=3300$ {\textit{Re}}_{{f}}=3300$), the University of Udine (UUD) dataset of Marchioli et al. (2008) (plane channel, Ref=2100$ {\textit{Re}}_{{f}}=2100$) and the data from Chung et al. (2002) (concentric annuli, Ref=2225$ {\textit{Re}}_{{f}}=2225$).

Figure 3

Table 2.Parameters of the different cases. Here, Np$N_{{p}}$ denotes the total particle number, Φp$\varPhi _{{p}}$ denotes the particle volume fraction, and ηw$\eta _{{w}}$ denotes the smallest eddy scale near the wall.Table 2 long description.

Figure 4

Figure 3.Figure 3 long description.Near-wall particle velocity statistics: (a)$(a)$ mean axial particle velocity vz+$v_z^+$, (b)$(b)$ axial rms particle velocity fluctuation vz,rms′+$v^{\prime +}_{z,{\textit{rms}}}$, (c)$(c)$ radial rms particle velocity fluctuation vr,rms′+$v^{\prime +}_{r,{\textit{rms}}}$, and (d)$(d)$ azimuthal rms particle velocity fluctuation vθ,rms′+$v^{\prime +}_{\theta ,{\textit{rms}}}$. The superscript +$+$ denotes viscous scaling, and r+$r^+$ denotes the wall-normal distance in wall units. The markers denote the UUD dataset of Marchioli et al. (2008) (plane channel, Ref=2100$ {\textit{Re}}_{{f}}=2100$).

Figure 5

Figure 4.Figure 4 long description.Temporal evolution of the Shannon entropy parameter.

Figure 6

Figure 5.Figure 5 long description.Spatial and temporal evolution of the dimensionless particle concentration c(r,t)$c(r,t)$ near the outer wall of (a)$(a)$ case 1, (b)$(b)$ case 2, (c)$(c)$ case 3, (d)$(d)$ case 4, (e)$(e)$ case 5 and (f)$(f)$ case 6. Here, r+$r^+$ denotes the viscous-scaled radial distance, and contours represent the particle concentration c(r,t)$c(r,t)$, with black iso-lines denoting c(r,t)=1$c(r,t)=1$.

Figure 7

Figure 6.Figure 6 long description.Spatial and temporal evolution of the dimensionless particle concentration c(r,t)$c(r,t)$ near the inner wall of (a)$(a)$ case 1, (b)$(b)$ case 2, (c)$(c)$ case 3, (d)$(d)$ case 4, (e)$(e)$ case 5 and (f)$(f)$ case 6. Here, r+$r^+$ denotes the viscous-scaled radial distance, and contours represent the particle concentration c(r,t)$c(r,t)$, with black iso-lines denoting c(r,t)=1$c(r,t)=1$.

Figure 8

Figure 7.Figure 7 long description.Dimensionless mean particle concentration within the viscous sublayer (r+<5$r^+\lt 5$) near (a)$(a)$ the outer wall and (b)$(b)$ the inner wall.

Figure 9

Figure 8.Figure 8 long description.(a)$(a)$ Radial profiles of the steady-state concentration and contributions of phoresis integrals. (b)$(b)$ Biased sampling integral Ib$I_b$ and a zoomed-in view of its variations near the outer wall. (c)$(c)$ Turbophoresis integral It$I_t$ and a zoomed-in view of its variations near the outer wall. (d)$(d)$ Centrifugal integral Ic$I_c$ and a zoomed-in view of It+Ic$I_t+I_c$, where the black dashed line marks the zero level.

Figure 10

Figure 9.Figure 9 long description.Temporal evolution of the short-time-averaged particle-sampled velocity quantities associated with D1$D_1$ and D2$D_2$: radial fluid velocity ⟨ur|r⟩$\langle u_r|r\rangle$ (solid lines), radial rms particle velocity ⟨vr2|r⟩1/2$\langle v_r^2|r\rangle ^{1/2}$ (dashed lines) and azimuthal rms particle velocity ⟨vθ2|r⟩1/2$\langle v_{\theta }^2|r\rangle ^{1/2}$ (dotted lines) for (a)$(a)$ case 1, (b)$(b)$ case 2, (c)$(c)$ case 3, (d)$(d)$ case 4, (e)$(e)$ case 5 and (f)$(f)$ case 6. The red lines denote the results at statistical steady state. In the plane channel case, ⟨ur|r⟩$\langle u_r|r\rangle$, ⟨vr2|r⟩1/2$\langle v_r^2|r\rangle ^{1/2}$ and ⟨vθ2|r⟩1/2$\langle v_{\theta }^2|r\rangle ^{1/2}$ reduce to particle-sampled normal fluid velocity, radial rms particle velocity, and spanwise rms particle velocity, respectively.

Figure 11

Figure 10.Figure 10 long description.Characteristics of the modes: (a)$(a)$ the time scales normalised by τp$\tau _{{p}}$ of the slowly decaying modes (circles), rapidly decaying modes (squares), and oscillatory decaying modes (crosses); (b)$(b)$ the initial modal energy contribution Ek,n/Ek$E_{k,n}/E_{k}$; (c)$(c)$ the cumulative modal energy contribution ∑i=0nEk,i/Ek$\sum _{i=0}^{n}E_{k,i}/E_k$; (d)$(d)$ the relative amplitudes of slowly decaying modes αn=|an|/(|an|+|bn|)$\alpha _n=|a_n|/(|a_n|+|b_n|)$ (circles) and rapidly decaying modes βn=|bn|/(|an|+|bn|)$\beta _n=|b_n|/(|a_n|+|b_n|)$ (squares).

Figure 12

Figure 11.Figure 11 long description.Radial profile of initial flux Jn+(r,0)$J^+_n(r,0)$ for (a)$(a)$ the outer-wall transport mode and (b)$(b)$ the wallward transport mode.

Figure 13

Figure 12.Figure 12 long description.Reconstructed near-wall concentration cw$c_{{w}}$ for (a)$(a)$ case 1, (b)$(b)$ case 2, (c)$(c)$ case 3, (d)$(d)$ case 4, (e)$(e)$ case 5 and (f)$(f)$ case 6.

Figure 14

Figure 13.Figure 13 long description.Profiles of (a)$(a)$ mean temperature Tf$T_{{f}}$ and density ρf$\rho _{{f}}$, and (b)$(b)$ rms of the corresponding fluctuations Tf,rms′$T^\prime _{{f,rms}}$ and ρf,rms′$\rho ^\prime _{{f,rms}}$.

Figure 15

Table 3.Parameters for case 7. Here, Reτ$ {\textit{Re}}_\tau$ is the friction Reynolds number, and the N$N$ and Δ+$\Delta ^+$ terms denote the number of grid nodes and the grid resolution in wall units in each direction, respectively. ‘Inner’ and ‘outer’ denote quantities normalised by the viscous scales at the inner and outer walls, respectively.Table 3 long description.

Figure 16

Table 4.Parameters for case 7. Here, Np$N_{{p}}$ denotes the total particle number, Φp$\varPhi _{{p}}$ denotes the particle volume fraction, and ηw$\eta _{{w}}$ denotes the smallest eddy scale near the wall.Table 4 long description.

Figure 17

Figure 14.Figure 14 long description.Spatial and temporal evolution of the dimensionless particle concentration c(r,t)$c(r,t)$ near the (a)$(a)$ outer and (b)$(b)$ inner walls for case 7, and (c,d)$(c,d)$ the corresponding dimensionless mean particle concentrations within the viscous sublayer (r+<5$r^+\lt 5$). Here, r+$r^+$ denotes the viscous-scaled radial distance, and contours represent the particle concentration c(r,t)$c(r,t)$, with black iso-lines denoting c(r,t)=1$c(r,t)=1$. A zoomed-in view of inner-wall concentration cw$c_{{w}}$ is added in figure 14(d).

Figure 18

Figure 15.Figure 15 long description.Characteristics of the modes: (a)$(a)$ the time scales normalised by τp$\tau _{{p}}$ of the slowly decaying modes (circles), rapidly decaying modes (squares), and oscillatory decaying modes (crosses); (b)$(b)$ the initial modal energy contribution Ek,n/Ek$E_{k,n}/E_{k}$; (c)$(c)$ the cumulative modal energy contribution ∑i=0nEk,i/Ek$\sum _{i=0}^{n}E_{k,i}/E_k$; (d)$(d)$ the relative amplitudes of slowly decaying modes αn=|an|/(|an|+|bn|)$\alpha _n=|a_n|/(|a_n|+|b_n|)$ (circles) and rapidly decaying modes βn=|bn|/(|an|+|bn|)$\beta _n=|b_n|/(|a_n|+|b_n|)$ (squares).

Figure 19

Figure 16.Figure 16 long description.Radial profile of initial flux Jn+(r,0)$J^+_n(r,0)$ for (a)$(a)$ the outer-wall transport mode and (b)$(b)$ the wallward transport mode.

Figure 20

Figure 17.Figure 17 long description.Reconstructed near-wall concentration cw$c_{{w}}$ for case 7.