2.1. Initial flow set-up

The direct numerical simulation (DNS) is performed in a three-dimensional (3-D) channel domain ${\mathcal {V}}$ with coordinates ${\boldsymbol {x}}=(x,y,z)\in {\mathcal {V}}$ and in the streamwise $x$-, the wall-normal $y$- and spanwise $z$-directions, respectively. The streamwise length of the channel is $L_x=5.61$, wall-normal height is $L_y=2 H$ with $H =1$ and spanwise width is $L_z=2.99$, consistent with those in the DNS in Zhao et al. (Reference Zhao, Yang and Chen2016). The dimensionless 3-D incompressible Navier–Stokes (NS) equations of the velocity ${\boldsymbol {u}}=(u,v,w)$ scaled by $H$ and the bulk velocity $U_b$ read

(2.1)\begin{equation} \left.\begin{gathered} \frac{\partial {\boldsymbol{u}}}{\partial t}+{\boldsymbol{u}} \boldsymbol{\cdot} \boldsymbol{\nabla} {\boldsymbol{u}}={-}\boldsymbol{\nabla} p+ \frac{1}{{{\it Re}}} \nabla^2 {\boldsymbol{u}}+{\boldsymbol{f}},\\ \boldsymbol{\nabla} \boldsymbol{\cdot} {\boldsymbol{u}}=0, \end{gathered}\right\} \end{equation}

where $t$ denotes the time, $p$ the pressure, $\nu$ the kinematic viscosity and ${{\it Re}}={U_b H}/{\nu }$ the Reynolds number. The flow is driven by a constant flux rate with $U_b= 1$ by exerting a homogeneous time-dependent body force ${\boldsymbol {f}}=(\tau _w,0,0)$, where

(2.2)\begin{equation} \tau_w\equiv\frac{1}{{\it Re}}{\frac{\mathrm{d} \langle u\rangle}{\mathrm{d} y}} \end{equation}

denotes the wall shear stress and $\langle {\cdot }\rangle$ the volume average.

As sketched in figure 1(a), periodic boundary conditions are applied in $x$- and $z$-directions, and the no-slip boundary conditions are applied at the upper boundary $y = -1$ and lower boundary $y = 1$. The initial flow field contains a vortex ring embedded in the laminar Poiseuille flow. The vortex ring serves as a controllable initial disturbance to trigger transition. The Poiseuille base flow has a velocity profile

(2.3)\begin{equation} u_l=U_0(1-y^2) \end{equation}

with the centreline velocity $U_0=1.5$. If not otherwise specified, the Reynolds number is set as ${{\it Re}}= 3333$ (Kim, Moin & Moser Reference Kim, Moin and Moser1987). If the flow undergoes a transition, this ${\it Re}$ is equivalent to the friction Reynolds number ${\it Re}_\tau ={\it Re} \sqrt {\tau _w}=207$, where $\tau _w$ is obtained in the fully developed turbulent stage. The vorticity ${\boldsymbol {\omega }} = \boldsymbol {\nabla } \times {\boldsymbol {u}}$ of the base flow only has a spanwise component

(2.4)\begin{equation} \omega_l=3y. \end{equation}

Figure 1.

(a) Schematic of the initial flow set-up for the vortex-ring-induced transition. The grey plane denotes the solid-wall boundary of the channel flow. The vortex ring (light blue surface) is placed at the centre of the computation domain. The possible vortex bursting is located at $(x_{br},y_{br},z_{br})$, and it has a strong induction effect near the upper boundary (marked by dark grey patch). (b) Schematic of the vortex ring. The dash-dotted line denotes the centreline $\mathcal {C}$ of the vortex tube (light blue surface), and the red line denotes a typical twisted vortex line attached on the vortex tube. Local cylindrical coordinates $(s,\rho,\theta )$ inside the vortex tube are marked in a cross section.

The vortex ring, represented by the light blue surface in figure 1(a), is placed at the centre of the computation domain, the farthest position from the wall. Note that the vortex ring can be placed anywhere in the outer layer. In general, the induction effect of the vortex ring on flow transition grows with the initial distance between the vortex-ring centre and the nearest wall.

The embedded vortex ring is set up using the method in Shen et al. (Reference Shen, Yao, Hussain and Yang2023). The geometry of vortex lines attached on the vortex tube is sketched in figure 1(b). The vorticity of the vortex ring is specified as

(2.5)\begin{equation} {\boldsymbol{\omega}}(s,\rho,\theta)=\varGamma f(\rho) \left(\frac{\rho\eta}{1-\kappa \rho \mathrm{cos}\theta}{\boldsymbol{e}}_{\theta}+{\boldsymbol{e}}_{s}\right), \end{equation}

where $(s,\rho,\theta )$ are curved cylindrical coordinates, $({\boldsymbol {e}}_{s},{\boldsymbol {e}}_{\rho },{\boldsymbol {e}}_{\theta })$ denote axial, radial and azimuthal basis, respectively, $\kappa$ is the constant curvature of the centreline $\mathcal {C}$ of the vortex ring, $\varGamma$ is the vorticity flux of the vortex ring and $f(\rho )=\exp (-{\rho ^2}/{2 \sigma ^2})/({2 {\rm \pi}\sigma ^2})$ is a Gaussian function parameterised by the standard deviation $\sigma$. Moreover, the local twist rate $\eta$ measures the twisting number of vortex lines along the vortex centreline inside the vortex ring.

In the numerical construction, ${\boldsymbol {\omega }}(s,\rho,\theta )$ in (2.5) in curved cylindrical coordinates is mapped to ${\boldsymbol {\omega }}(x, y, z)$ in Cartesian coordinates using the algorithm in appendix A of Shen et al. (Reference Shen, Yao, Hussain and Yang2023). Various closed vortex tubes (Shen et al. Reference Shen, Yao, Hussain and Yang2022) can be constructed with a given parametric equation for the centreline and the function $f(\rho )$ for the vorticity-flux distribution.

2.2. Internal structures of the vortex ring

The configuration parameters of the vortex ring are set based on the DNS result of the same channel flow in Zhao et al. (Reference Zhao, Yang and Chen2016). As illustrated by the isosurface of the swirling strength $\lambda _{ci}$ (Zhou et al. Reference Zhou, Adrian, Balachandar and Kendall1999) in figure 2, the head of the hairpin-like vortex tube forms a ring-like structure consisting of helical vortex lines in the late stage of the natural transition. In particular, the twist rate of vortex lines is not uniformly distributed on the vortex ring. The chiralities of the helical lines are opposite on the two halves of the ring.

Figure 2.

Modelling an external vortex in flow transition using a vortex ring. (a) Isosurface of the swirling strength $\lambda _{ci}=2$ (Zhou et al. Reference Zhou, Adrian, Balachandar and Kendall1999) at $t =110$ in the late stage of the natural transition in channel flow reported in Zhao et al. (Reference Zhao, Yang and Chen2016). The isosurface is colour-coded by the helicity density $h={\boldsymbol {\omega }}\boldsymbol {\cdot }{\boldsymbol {u}}$. The hairpin head consisting of helical vortex lines is sketched in the closed-up view on the left. The orange arrows denote axial flows inside the vortex tube induced by the helical vortex lines with opposite chiralities. (b) Isosurface of the $\lambda _{ci}=2$ (colour-coded by $h$) in the present vortex-ring-induced transition. The helical vortex lines with opposite chiralities near the vortex ring are plotted in the closed-up view on the left. The possible bursting location is marked by the asterisk at $s={\rm \pi} R/2$ in the local cylindrical coordinates.

Thus, we set the differential twist, i.e. the non-uniform local twist rate $\eta =A\cos (s/R)$ along $\mathcal {C}$, with the constant $A=120$ (Shen et al. Reference Shen, Yao, Hussain and Yang2023). For comparison, we also set cases with the uniform twist $\eta =2A/{\rm \pi}$ and zero twist $\eta =0$. The initial twist is crucial in determining whether the vortex ring will burst during its evolution (Shen et al. Reference Shen, Yao, Hussain and Yang2023). In addition, the vorticity flux $\varGamma$, vortex ring radius $R = 0.25$ and ring thickness $\sigma = 0.0125$ are comparable to the DNS result in Zhao et al. (Reference Zhao, Yang and Chen2016) at $t=110$ in the late transition. Note that the induction effect of the vortex ring on flow transition strongly depends on $A$, which plays a similar role as ${\it Re}$, whereas it weakly depends on $R$ and $\sigma$ (Shen et al. Reference Shen, Yao, Hussain and Yang2022,  Reference Shen, Yao, Hussain and Yang2023).

2.3. DNS

The NS equations (2.1) are solved by DNS using the same spectral method in Kim et al. (Reference Kim, Moin and Moser1987). The velocity in the $x$- and $z$-directions is decomposed into Fourier modes, and ${\boldsymbol {u}}$ in the $y$-direction is solved by the Helmholtz equations for each mode. The time stepping is varied to ensure the Courant–Friedrichs–Lewy number less than $0.6$. The initial vorticity field, the superposition of laminar background (2.4) and vortex ring (2.5), is decomposed into Fourier–Chebyshev modes. To satisfy the boundary conditions for a channel, the initial velocity field is calculated from the vorticity via the Helmholtz equation in the Fourier–Chebyshev space. A boundary correction is then applied under the divergence-free constraint, which is detailed in Appendix A. In the calculation of flow evolution, the two-thirds truncation method is applied for dealiasing (Canuto et al. Reference Canuto, Hussaini, Quarteroni and Zang1988). The low-storage third-order semi-implicit Runge–Kutta method (Philippe, Robert & Michael Reference Philippe, Robert and Michael1991; Yang & Pullin Reference Yang and Pullin2011) is used for time marching.

We performed convergence tests on four grids with $N_x\times N_y\times N_z =768\times 769 \times 768$, $576\times 577 \times 576$, $384\times 385 \times 384$ and $192\times 193 \times 192$ grid points. After examining the convergence of the maximal vorticity magnitude for vortex evolution at $t\leqslant 1.5$ and ${\it Re}_\tau$ for flow transition at $t\leqslant 70$, we chose the grid of $384\times 385 \times 384$ for the simulation at ${\it Re} = 3333$. We conducted an additional DNS at ${\it Re}=8090$ on the grid of $768\times 769 \times 768$ to study the effect of ${\it Re}$.

3.1. Stages in transition process

In this section, we demonstrate that the initially embedded vortex ring can induce transition in wall flows at moderate ${\it Re}$, and the critical magnitude of the initial disturbance is highly dependent on the initial twist of the vortex ring. For the vortex rings with three different $\eta$, we first examine the critical vorticity flux $\varGamma = \varGamma ^*$ to induce transition. Here, $\varGamma ^*$ is determined by assessing whether the flow will undergo a transition through a series of simulations conducted for various values of $\varGamma$.

As listed in table 1, we set up four DNS cases for the vortex-ring-induced transition to illustrate the influence of the internal structure inside the vortex ring. In Case 1, the initial vortex ring with chirality-opposite helical structures mimics the hairpin head in the natural transition, and it has the minimum $\varGamma ^*$. As shown in figure 2, the isosurface of $\lambda _{ci}$ colour-coded by the helicity density $h={\boldsymbol {u}}\boldsymbol {\cdot }{\boldsymbol {\omega }}$ shows positive and negative $h$ in the two halves of the vortex ring. In this case, we set $\varGamma \approx 1.2 \varGamma ^*$ to quickly trigger transition.

Table 1.

Parameters $\eta$ and $\varGamma$ of the vortex ring in four representative cases of the vortex-ring-induced transition of channel flow. The threshold vorticity flux $\varGamma ^*$ for triggering transition, $\varGamma ^*$ normalised by $\varGamma ^*_{\eta =0}$ ($\varGamma ^*$ in Case 4) and the volume-averaged disturbance energy $E_t$ are listed. Note that Cases 1, 3 and 4 have the transition with $\varGamma \approx 1.2 \varGamma ^*$.

In Case 2, the vortex ring is the same except that the twist is uniform, and $\varGamma ^*$ is more than twice of that in Case 1. Thus, Case 2 has no transition with $\varGamma < \varGamma ^*$. Note that Cases 1 and 2 have roughly the same disturbance energy

(3.1) \begin{equation} E_t=\frac{1}{V}\int_{\mathcal{V}}\frac12((u-u_l)^2+v^2+w^2)\,\mathrm{d} {\mathcal{V}}, \end{equation}

where $V$ denotes the volume of the channel. Therefore, the different twist distributions in Cases 1 and 2 only have a minor effect on $E_t$, but they have an effect on the transition process. Case 3 is the same as Case 2 except $\varGamma > \varGamma ^*$, so it has transition. In Case 4, the vortex ring with $\eta = 0$ and $\varGamma > \varGamma ^*$ also induces transition.

The evolution of ${{\it Re}}_\tau$, indicating the variation of the wall friction, in the four cases is plotted in figure 3. It can be roughly divided into three stages. Before a characteristic time $t_1=1$, the major dynamics is only within the vortex ring, including the vortex bursting in Case 1. From $t_1$ to $t_2 = 10$, the vortex ring breaks up, and ${{\it Re}}_\tau$ grows mildly. From $t_2$ to $t_3 = 40$, ${{\it Re}}_\tau$ starts to surge and the transition occurs in the cases with $\varGamma > \varGamma ^*$, whereas ${{\it Re}}_\tau$ remains at the laminar value for $\varGamma < \varGamma ^*$. The flow evolution in the three stages is elaborated on in the following.

Figure 3.

Temporal evolution of the wall-friction Reynolds number in various cases (listed in table 1).

3.1.1. Stage 1: dynamics inside the vortex ring

Figure 4 shows the main dynamics inside the vortex ring around $t_1$. In Case 1, an initially helical vortex ring with differential twist has the azimuthal vorticity $\omega _\theta$, which generates the axial velocity

(3.2)\begin{equation} u_s(s, \rho, \theta)=\frac{1}{1-\kappa \rho \cos \theta} \frac{\varGamma }{2{\rm \pi}}\exp\left(-\frac{\rho^2}{2\sigma^2}\right)\eta \end{equation}

along the vortex-ring centreline (see Appendix B for a detailed derivation). The two twist vortex waves with opposite chiralities collide at the symmetrical plane at $s={{\rm \pi} R}/{2}$. This leads to vortex bursting (Shen et al. Reference Shen, Yao, Hussain and Yang2023): a part of the vortex tube is flattened on the symmetrical plane, forming a disc-like structure with highly spiral vortex lines.

Figure 4.

Evolution of the isosurface of $\lambda _{ci}$ during the vortex bursting in Case 1 with two Reynolds numbers. The surface is colour-coded by the helicity density: (a) $t=0.3$ and $\lambda _{ci}=2.4$; (b) $t=0.9$ and $\lambda _{ci}=2.4$;(c) $t=1.5$ and $\lambda _{ci}=2.1$.

Using the vortex-ring model in Shen et al. (Reference Shen, Yao, Hussain and Yang2023), the starting and ending times of the vortex bursting in Case 1 are defined as

(3.3)\begin{equation} t_{bs}=\frac{2 {\rm \pi}R}{\varGamma A }\approx 0.3 \end{equation}

and

(3.4)\begin{equation} t_{be}=\frac{ 10 {\rm \pi}^2 R}{3 \varGamma A }\approx 1.5, \end{equation}

respectively. Figure 4(a,c) shows the vortical structures at $t_{bs}$ and $t_{be}$.

Moreover, the vortices for ${\it Re}=3333$ and 8090 are nearly identical in figure 4(a), because the viscosity dependence is weak at early times, consistent with the model in (3.2). In figure 4(c), the ${\it Re}$-effect becomes noticeable because the model in (3.2) breaks down near the bursting site with strong viscous vortex reconnection. As discussed in Shen et al. (Reference Shen, Yao, Hussain and Yang2023), increasing ${\it Re}$ is equivalent to increasing an effective $\eta$. A larger $\eta$ in (3.2) suggests a larger $u_s$, leading to stronger vortex bursting with generating larger disc-like structures. By contrast, there is no bursting in Cases 2–4 for initial vortex rings with uniform or zero twist.

3.1.2. Stage 2: breakup of the vortex ring

Figure 5(a–c) illustrate the breakup dynamics of the vortex ring from $t_1$ to $t_2$ in Cases 1–3, respectively. At $t=t_1$, the initially helical vortex ring splits into primary and secondary rings due to the axial flow (Cheng et al. Reference Cheng, Lou and Lim2010). The primary vortex ring undergoes further transition, and the secondary rings dissipate before $t_2$. By contrast, the vortex with $\eta =0$ in Case 4 retains its ring shape at $t=t_1$ in figure 5(d).

Figure 5.

Evolution of the isosurface of $\lambda _{ci}$ in the second stage of transition in (a) Case 1 with $\eta =A\cos {\xi }$ and $\varGamma =0.047$, (b) Case 2 with $\eta =2A/{\rm \pi}$ and $\varGamma =0.047$, (c) Case 3 with $\eta =2A/{\rm \pi}$ and $\varGamma =0.097$ and(d) Case 4 with $\eta =0$ and $\varGamma =0.278$. The isocontour values in each row are $\lambda _{ci}=2.1$ at $t=1$ and $\lambda _{ci}=1$ at $t=5$ and 9. The isosurface is colour-coded by the helicity density. The blue arrows in (a) denote the direction of the mean flow. Black lines denote vortex lines.

From $t_1$ to $t_2$, the vortex rings with and without the initial twist evolve very differently. For $\eta >0$, the vortex rings experience dramatic deformation in figure 5(a–c). The uniformly twisted vortex ring folds and elongates under the mean shear in figure 5(b,c). This process from $t=1$ to $t=5$ in Case 2 is further illustrated in the side view in figure 6. The vortex ring is persistently stretched along the streamwise direction until $t=9$, and eventually breaks up into several pairs of streamwise vortices. Although the evolutions of the vortices are similar in Cases 2 and 3, the larger $\varGamma$ in Case 3 induces a transition whereas the flow relaminarises in Case 2. The ring with differential twist has a similar evolution in figure 5(a). Since its internal structure is already unstable at $t=t_1$, the breakup process in Case 1 is faster than in Cases 2 and 3, which agrees with the different growth rates of ${{\it Re}}_\tau$ in figure 3.

Figure 6.

Evolution of the isosurface of $\lambda _{ci}$ from $t=1$ to $t=5$ in Case 2 with $\eta =2A/{\rm \pi}$ and $\varGamma =0.047$ (side view). The isosurface is colour-coded by the helicity density.

The vortex ring with $\eta = 0$ does not break up in stage 2. Its aspect ratio oscillates in figure 5(d) due to the elliptic instability (Kerswell Reference Kerswell2002) with the ring deformation in the $x$-direction and the curvature instability (Fukumoto & Hattori Reference Fukumoto and Hattori2005; Blanco-Rodríguez & Dizès Reference Blanco-Rodríguez and Dizès2017) with the variation of the ring curvature.

3.1.3. Stage 3: late transition

Cases 1–3 at $t_3$ in stage 3 of late transition are very similar. In figure 7 for Case 1, the inverse hairpin-like structure in figure 7(a) evolves into streamwise vortices in figure 7(b), which characterise a quasi-stable TS wave (Boiko et al. Reference Boiko, Westin, Klingmann, Kozlov and Alfredsson1994). Then the flow evolves as in the natural transition (Zaki Reference Zaki2013; Zhao et al. Reference Zhao, Yang and Chen2016), i.e. from streaks to the $\varLambda$- or hairpin-like structure in figure 7(c), and finally it breaks down into turbulence in figure 7(d). Moreover, the transition follows an almost identical route at a larger ${\it Re}=8090$, whereas the rise of ${\it Re}_\tau$ occurs much earlier than that at ${\it Re}=3333$ (not shown).

Figure 7.

Evolution of the isosurface of $\lambda _{ci}$ after the dissipation of the vortex ring in Case 1. The surface is colour-coded by the wall distance. Note that the channel is flipped in the $y$-direction for clearly showing the vortical structure: (a) $t=10$ and $\lambda _{ci}=1.3$; (b) $t=20$ and $\lambda _{ci}=0.5$; (c) $t=30$ and $\lambda _{ci}=1$; (d) $t=40$ and $\lambda _{ci}=1.3$.

In particular, the evolution and generation of the hairpin vortices are depicted in figure 8. The signature of hairpin vortices aligns with that in Adrian, Meinhart & Tomkins (Reference Adrian, Meinhart and Tomkins2000) in terms of the streamwise length, wall-normal location and angle between hairpin head and neck. From $t=31$ to $t=34$, the primary hairpin vortex elongates streamwise, and the secondary hairpin vortex forms at the ridge of the primary one (Zhou et al. Reference Zhou, Adrian, Balachandar and Kendall1999). Then, more hairpin-like structures emerge at the ridges of the primary and secondary hairpins at $t=35$, forming a coherent packet of hairpins. These observations are consistent with the auto-generation mechanism (Zhou et al. Reference Zhou, Adrian, Balachandar and Kendall1999; Adrian Reference Adrian2007; Kim et al. Reference Kim, Adrian, Balachandar and Sureshkumar2008).

Figure 8.

Evolution of the isosurface of $\lambda _{ci}=1$ during the generation and evolution of hairpin vortices in Case 1. The isosurface is colour-coded by the helicity density. Note that the channel is flipped in the $y$-direction for clearly showing the vortical structure.

Note that the route of late transition induced by the vortex ring with large vorticity flux and zero-twist in Case 4 is distinguished from the others induced by the twisted vortex ring. The initial vortex ring gradually dissipates and disappears in Case 4, without forming any other vortical structures at $t=10$ in figure 9(a). It evolves into streamwise vortices near the outer layer at very late times in figure 9(b,c) and finally breaks down into turbulence in figure 9(d).

Figure 9.

Evolution of the isosurface of $\lambda _{ci}$ in Case 4: (a) $t=10$ and $\lambda _{ci}=1$; (b) $t=20$ and $\lambda _{ci}=0.7$;(c) $t=30$ and $\lambda _{ci}=0.5$; (d) $t=50$ and $\lambda _{ci}=2$. The surface is colour-coded by the wall distance.

3.2. Transition induced by vortex bursting

The internal dynamics of the vortex ring around $t_1= 1$ is much faster than the disturbance growth in the boundary layer around $t_3=40$, so these two processes can be decoupled. The initial amplitude of the disturbance is determined by the dynamics of the vortex ring before $t_1$, whereas the subsequent linear amplification mechanism is mainly influenced by the background shear.

According to the rapid distortion theory (Sreenivasan & Narasimha Reference Sreenivasan and Narasimha1978; Savill Reference Savill1987; Nazarenko, Kevlahan & Dubrulle Reference Nazarenko, Kevlahan and Dubrulle1999; Deissler Reference Deissler2004), the amplitude of the disturbance is low in the early stage of transition, so the disturbance growth can be treated as inviscid. Moffatt (Reference Moffatt1967) applied the rapid distortion theory to the viscous sublayer (Saric et al. Reference Saric, Reed and Kerschen2002) at $y^+<5$, where

(3.5)\begin{equation} y^+={(1-y)}{\it Re}_\tau \end{equation}

is the normalised distance from the upper wall at $y=1$, and the wall-normal length scale is much smaller than the streamwise one.

We consider the near-wall velocity ${\boldsymbol {u}}_B=(u_B,v_B,w_B)$ with

(3.6a–c)\begin{equation} u_B=u(x,y^+,z),\quad v_B=v(x,y^+,z),\quad w_B=w(x,y^+,z),\quad y^+ < 5. \end{equation}

Moffatt (Reference Moffatt1967) showed that when the disturbance amplitude is small, the streamwise disturbance mode $\hat {u}_B$ grows linearly with the wall-normal disturbance mode $\hat {v}_B$ as

(3.7)\begin{equation} \hat{u}_B(k_x,k_z)=\hat{u}_{B0}(k_x,k_z)-\mathcal{S}\hat{v}_B(k_x,k_z) t, \end{equation}

where $\hat {\cdot }$ denotes the Fourier transform of a quantity in the $x$–$z$ plane with wavenumbers $k_x$ and $k_z$, subscript $0$ denotes the initial value and $\mathcal {S}= -2U_0y$ is the shear rate of the Poiseuille flow. From (3.7), $\hat {v}_B$ is determined by the state around $t=t_1$, $\mathcal {S}$ is constant for a given $y$ and $\hat {u}_B$ depends on $\hat {v}_B$, e.g. $\hat {u}_B$ increases with $t$ for positive $-\mathcal {S}\hat {v}_B$.

We examine the evolution of $\hat {v}_B(k_{xz},y^+)$ during the period of vortex bursting from $t_{bs}$ to $t_{be}$, with $k_{xz}=\sqrt {k_x^2+k_z^2}$. Based on the Parseval identity, we use

(3.8)\begin{equation} \overline{v_B^2}(y^+)=\frac{1}{A_{xz}}\iint v_B^2({x,y^+,z})\,\mathrm{d}\kern1pt x\,\mathrm{d} z = \sum_{k_{xz}}\hat{v}_B^2(k_{xz},y^+) \end{equation}

to quantify the averaged amplitude of $\hat {v}_B$, where $A_{xz}$ denotes the area of the $x$–$z$ plane. Figure 10 plots the evolution of $\overline {v_B^2}(y^+=1)$ in Cases 1–3 in the very early stage when (3.7) is valid. Note that the evolution of $\overline {v_B^2}(y^+=5)$ is qualitatively the same (not shown). Case 2 has the lowest amplitude, and it has no transition. Cases 1 and 3 have larger amplitudes, suggesting the higher disturbance growth rate from (3.7), and both cases have transition. At early times, $\overline {v_B^2}(y^+=1)$ decays in Cases 2 and 3 due to the viscous effect, and its decay rate depends on the initial $\varGamma$. By contrast, $\overline {v_B^2}(y^+=1)$ grows by 15 % from $t_{bs}$ to $t_{be}$ in Case 1. The growth of $v_B$ during the vortex bursting will be further discussed in § 4.

Figure 10.

Evolution of $\overline {v_B^2}(y^+=1)$, as metric for the averaged amplitude of $\hat {v}_B(k_{xz},y^+=1)$, from $t_{bs}$ to $t_{be}$ during the vortex bursting in Cases 1–3.

The transition is induced by the growth of $v_B$. According to (3.7), the streamwise disturbance of $u_B$ grows with the vortex-ring induced $v_B$ (Landahl Reference Landahl1975,  Reference Landahl1980). The streaks are lifted by $v_B$ in figure 7(a) and then elongated by the mean shear in figure 7(b), resulting in streamwise streaks downstream (Brandt & Henningson Reference Brandt and Henningson2002). A hairpin vortex (Zhou et al. Reference Zhou, Adrian, Balachandar and Kendall1999) emerges from the streaks in figure 7(c), and then it evolves into a hairpin packet by the auto-generation mechanism (Adrian et al. Reference Adrian, Meinhart and Tomkins2000; Adrian Reference Adrian2007; Kim et al. Reference Kim, Adrian, Balachandar and Sureshkumar2008). The vortices break down and form a turbulent spot in figure 7(d), which eventually leads to a fully developed turbulent state (Wu Reference Wu2023).

In summary, the threshold vorticity flux $\varGamma ^*$ is reduced from $0.231$ in Case $4$ to $0.0394$ in Case $1$ (see table 1) due to the induction of $v_B$ in the vortex-ring evolution. In particular, the vortex bursting in Case 1 produces strong $v_{B}$ near the boundary, enhancing the growth of $u_B$ and then triggering a transition.

4.1. Introduction of the di-vorticity

Next we elucidate the induction mechanism of $v_B$ in the near-wall region by the large di-vorticity generated in the outer layer during the vortex bursting. The di-vorticity ${{\boldsymbol {\varrho }}}=\boldsymbol {\nabla }\times {\boldsymbol {\omega }}$ measures the local rotational intensity of the vorticity, and large ${{\boldsymbol {\varrho }}}$ may correspond to the local helical geometry of vortex lines. The identity $\nabla ^2 {\boldsymbol {u}}=\boldsymbol {\nabla }(\boldsymbol {\nabla }\boldsymbol {\cdot }{\boldsymbol {u}})-{\boldsymbol {\varrho }}$ suggests the Poisson equation

(4.1)\begin{equation} \nabla^2 {\boldsymbol{u}}={-}{\boldsymbol{\varrho}} \end{equation}

for the incompressible flow.

The di-vorticity has a very intermittent distribution in the transition. In Case 1, the maximum $| {\boldsymbol {\varrho }}|$ can be hundreds of times the volume-averaged $| {\boldsymbol {\varrho }}|$ during the vortex bursting. For example, at $t=0.9$ in Case 1, the region with $| {\boldsymbol {\varrho }}| \geqslant 1000$ enclosed by the isosurface of $| {\boldsymbol {\varrho }}|=1000$ occupies only $0.2\,\%$ of the volume of the entire channel in figure 11(a), but ${\boldsymbol {\varrho }}$ in this negligibly small region can reconstruct the vortical structure very close to the original in figure 11(b). Therefore, the highly concentrated ${\boldsymbol {\varrho }}$ near the site of vortex bursting can drive the near-wall flow dynamics in a finite time period after vortex bursting.

Figure 11.

Reconstruction of the flow field using a small potion of the di-vorticity in Case 1 at $t=0.9$.(a) Isosurface of $\lambda _{ci}=2$ (blue) in the DNS, and isosurface of $| {\boldsymbol {\varrho }}|=1000$ (red). (b) Isosurface of $\lambda _{ci}=2$ in the flow field reconstructed from the di-vorticity in the region enclosed by the isosurface of $| {\boldsymbol {\varrho }}|=1000$ (red surface in (a)). Note that this subdomain occupies only $0.2\,\%$ of $V$.

With higher-order velocity derivatives, ${\boldsymbol {\varrho }}$ has conservation constraints under given boundary conditions. The solid-wall boundary conditions are

(4.2)\begin{equation} \left.\begin{gathered} {\boldsymbol{u}}=0,\quad {\boldsymbol{u}}\in{\partial}{\mathcal{V}},\\ \frac{{\partial} v}{{\partial} y} = 0,\quad v\in{\partial}{\mathcal{V}}, \end{gathered}\right\} \end{equation}

which imposes the constraint

(4.3)\begin{equation} \int_{\mathcal{V}}y^n\varrho_y\mathrm{d} {\mathcal{V}}=0,\quad n\in\mathbb{N} \end{equation}

on the di-vorticity. This constraint implies that the wall can have a response near the boundary to counter ${\boldsymbol {\varrho }}$ generated in a vortex-dynamics event remote from the wall. For $n=0$, (4.3) suggests that the flow evolution only redistributes $\varrho _y$ rather than changing the volume integration of $\varrho _y$. The implication of (4.3) will be further discussed in § 4.3.

4.2. Generation of the di-vorticity during vortex bursting

We demonstrate that large ${\boldsymbol {\varrho }}$ can be generated during the bursting of a vortex ring. The governing equation of ${\boldsymbol {\varrho }}$ is

(4.4)\begin{equation} \frac{\mathrm{D} {\boldsymbol{\varrho}}}{\mathrm{D} t}=\boldsymbol{\nabla}\times\boldsymbol{\nabla}\times{\boldsymbol{a}}+{\boldsymbol{\omega}}\boldsymbol{\cdot}\boldsymbol{\nabla}{\boldsymbol{\omega}}+{\boldsymbol{L}}, \end{equation}

where ${\boldsymbol {a}} = -\boldsymbol {\nabla } p+\nabla ^2 {\boldsymbol {u}}/{\it Re}$ denotes the acceleration of a fluid particle, and

(4.5)\begin{equation} {\boldsymbol{L}}\equiv(\boldsymbol{\nabla}{\boldsymbol{\omega}}\boldsymbol{\cdot}\boldsymbol{\nabla}{\boldsymbol{u}}-\boldsymbol{\nabla}{\boldsymbol{u}}\boldsymbol{\cdot}\boldsymbol{\nabla}{\boldsymbol{\omega}}):{\boldsymbol{\varepsilon}} \end{equation}

has a form similar to the Lamb vector ${\boldsymbol {\omega }}\times {\boldsymbol {u}}$, with the Levi–Civita symbol ${\boldsymbol {\varepsilon }}$. On the right-hand side, the term $\boldsymbol {\nabla }\times \boldsymbol {\nabla }\times {\boldsymbol {a}}$ for the viscous effect can be neglected for very large ${\it Re}$, and then ${\boldsymbol {\omega }}\boldsymbol {\cdot }\boldsymbol {\nabla }{\boldsymbol {\omega }}$ and ${\boldsymbol {L}}$ dominate the evolution of ${\boldsymbol {\varrho }}$.

Given the initial vortex ring in (2.5), the evolution of ${\boldsymbol {\varrho }}$ is estimated in curved cylindrical coordinates $(s,\rho,\theta )$. The vorticity-stretching and Lamb-like terms become (see Appendix B for a detailed derivation)

(4.6)\begin{equation} {\boldsymbol{\omega}}\boldsymbol{\cdot}\boldsymbol{\nabla}{\boldsymbol{\omega}}=\left(0,0,\omega_s \frac{1}{h_s} \frac{{\partial} \omega_\theta}{{\partial} s} + \omega_\theta \frac{1}{\rho} \frac{{\partial} \omega_\theta}{{\partial} \theta}\right) \end{equation}

and

(4.7)\begin{equation} \left.\begin{gathered} {\boldsymbol{L}}=\left(\frac{1}{\rho}\frac{{\partial} \omega_\theta}{{\partial} \rho} \frac{{\partial} u_\theta}{{\partial} \theta}-\frac{1}{h_s}\frac{{\partial} u_s}{{\partial} \rho} \frac{{\partial} \omega_\theta}{{\partial} s}-\frac{1}{\rho}\frac{{\partial} u_\theta}{{\partial} \rho} \frac{{\partial} \omega_\theta}{{\partial} \theta}\right., \\ \frac{1}{\rho^2}\frac{{\partial} \omega_\theta}{{\partial} \theta}\frac{{\partial} u_s}{{\partial} \theta}- \frac{1}{h_s\rho}\frac{{\partial} \omega_\theta}{{\partial} s}\frac{{\partial} u_\theta}{{\partial} \theta}+ \frac{1}{h_s^2}\frac{{\partial} u_s}{{\partial} s}\frac{{\partial} \omega_\theta}{{\partial} s},\\ \left.-\frac{1}{h_s}\frac{{\partial} \omega_s}{{\partial} \rho}\frac{{\partial} u_s}{{\partial} s}- \frac{1}{\rho}\frac{{\partial} \omega_\theta}{{\partial} \rho}\frac{{\partial} u_s}{{\partial} \theta}\right), \end{gathered}\right\} \end{equation}

respectively, where $h_s=1-\kappa \rho \cos \theta$ is the Lamé coefficient in the $s$-direction.

We analyse the variation of ${\boldsymbol {\varrho }}$ using (4.4) before $t_1$ when the vortex bursting occurs (see figure 4). In the model of $\eta$ for a bursting vortex ring (Shen et al. Reference Shen, Yao, Hussain and Yang2023), the initial twist distribution $\eta (s,t=0)=A\cos (s/R)$ evolves into

(4.8)\begin{equation} \eta(s,\rho,t)= A\cos\left(\frac{s-B(\rho)\eta(s,\rho,t) t}{R}\right) \end{equation}

with $A=120$ and $B={\varGamma }\exp (-{\rho ^2}/{2\sigma ^2})/(2{\rm \pi} h_s)$, and (4.8) can be solved numerically using the explicit Newton method. Substituting (2.5) and (B2) into (4.8), we find that

(4.9a,b)\begin{equation} \frac{{\partial} \omega_\theta}{{\partial} s}\sim \frac{{\partial} \eta}{{\partial} s}\quad \textrm{and}\quad \frac{{\partial} u_s}{{\partial} s}\sim \frac{{\partial} \eta}{{\partial} s} \end{equation}

are dominant terms, where $\sim$ denotes an estimate of the order of magnitude. Other derivatives of ${\boldsymbol {u}}$ and ${\boldsymbol {\omega }}$ only depend on the initial configuration and remain constant during vortex bursting. With $ {\partial } \eta / {\partial } s\gg {\partial } \eta / {\partial } \theta$ and $ {\partial } \eta / {\partial } s \gg {\partial } \eta / {\partial } \rho$, we rank the magnitude of terms in (4.6) and (4.7) by $ {\partial } \eta / {\partial } s$. By only keeping the dominant terms, (4.4) is approximated by

(4.10)\begin{equation} \frac{\mathrm{D} {\boldsymbol{\varrho}}}{\mathrm{D} t}\approx\frac{1}{h_s}\left(-\frac{{\partial} u_s}{{\partial} \rho} \frac{{\partial} \omega_\theta}{{\partial} s}, -\frac{1}{\rho}\frac{{\partial} \omega_\theta}{{\partial} s} \frac{{\partial} u_\theta}{{\partial} \theta}+\frac{1}{h_s}\frac{{\partial} u_s}{{\partial} s} \frac{{\partial} \omega_\theta}{{\partial} s},\omega_s \frac{{\partial} \omega_\theta}{{\partial} s}- \frac{{\partial} \omega_s}{{\partial} \rho}\frac{{\partial} u_s}{{\partial} s}\right). \end{equation}

Here, the most influential term

(4.11)\begin{equation} I=\frac{1}{h_s^2}\frac{{\partial} u_s}{{\partial} s}\frac{{\partial} \omega_\theta}{{\partial} s}\sim \left(\frac{{\partial} \eta}{{\partial} s}\right)^2 \end{equation}

contains the product of two dominant terms. Finally, substituting (4.8) into (4.11), we estimate the order of (4.4) in terms of $\eta$ as

(4.12)\begin{equation} \frac{\mathrm{D} {\boldsymbol{\varrho}}}{\mathrm{D} t}\approx I{\boldsymbol{e}}_{\rho}\sim \left(\frac{{\partial} \eta}{{\partial} s}\right)^2 {\boldsymbol{e}}_{\rho}. \end{equation}

In figure 12, the distribution of $({ {\partial } \eta }/{ {\partial } s})^2$ along the $s$-direction shows that large $| {\boldsymbol {\varrho }}|$ is generated during vortex bursting and the di-vorticity is concentrated near the vortex bursting site at $s={\rm \pi} R/2$. The extremely twisted vortex lines, with the surge of $ {\partial } \eta / {\partial } s$ in (4.12), manifest as the disc structure with large $|{\boldsymbol {\varrho }}|$ at $s={\rm \pi} R/2$ in figure 11(a).

Figure 12.

Evolution of $({ {\partial } \eta }/{ {\partial } s})^2$ in terms of the axial coordinate $s$ along the centreline of the vortex ring during vortex bursting in Case 1.

4.3. Induction of $v_B$

From (3.7), we investigate the effect of $\varrho _y$, the most important component in ${\boldsymbol {\varrho }}$, on inducing $v_B$ in the early transition. The Poisson equation (4.1) implies that although the large ${\boldsymbol {\varrho }}$ generated from vortex bursting is remote from the wall, it can have a strong induction effect in the entire domain, especially in the near-wall region.

To analyse $v_{B}$, we focus on (4.1) in the wall-normal direction

(4.13)\begin{equation} \left.\begin{gathered} \nabla^2 v={-}\varrho_y,\quad v \in {\mathcal{V}},\\ v=0\quad \textrm{and}\quad \frac{{\partial} v}{{\partial} y}=0,\quad v\in {\partial} {\mathcal{V}}. \end{gathered}\right\} \end{equation}

This Poisson equation has a redundant boundary condition. We first consider only the Dirichlet boundary condition as

(4.14)\begin{equation} \left.\begin{gathered} \nabla^2 v ={-}\varrho_y,\quad v\in {\mathcal{V}},\\ v=0,\quad v\in {\partial} {\mathcal{V}}, \end{gathered}\right\} \end{equation}

and then treat the Neumann boundary condition as a boundary response to the vortices remote from the wall.

We express the solution (see details in Appendix C)

(4.15)\begin{equation} v({\boldsymbol{x}})=\int_{\mathbb{R}^3}\varrho_y(\tilde{{\boldsymbol{x}}} )\varPhi({\boldsymbol{x}}-\tilde{{\boldsymbol{x}}})\,\mathrm{d} \tilde{{\boldsymbol{x}}} \end{equation}

to (4.14) in terms of the fundamental solution

(4.16)\begin{equation} \varPhi({\boldsymbol{x}})=\frac{1}{4{\rm \pi}| {\boldsymbol{x}}|}, \end{equation}

where $\tilde {{\boldsymbol {x}}}$ and $\varrho _y(\tilde {{\boldsymbol {x}}})$ are extended quantities in $\mathbb {R}^3$ (see their definitions in Appendix C). As sketched in figure 13, (4.15) extends the integration region from the channel domain ${\mathcal {V}}$ to the entire 3-D space $\mathbb {R}^3$. As discussed in § 3, the vortex bursting triggers transition near the upper wall by inducing finite $v_B$. Without loss of generality, we consider the induction effect of $\varrho _y$ near the upper wall at ${\boldsymbol {x}}_B=(x,y\to 1,z)$ based on (4.15).

Figure 13.

Schematic for extending the integration region of the Green function from the channel domain to the entire 3-D space. The extended domain consists of the original channel domain ${\mathcal {V}}$ and its mirror domains ${\mathcal {V}}_i$. The induction effect of $\varrho _y$ in ${\mathcal {V}}\cup {\mathcal {V}}_1$ is shown on the left. The region with concentrated $\varrho _y$ is coloured in green, including ${\mathcal {V}}_{br}$ near the site of vortex bursting and ${\mathcal {V}}_B$ near the upper boundary. The red and blue arrows denote twist vortex waves with opposite chiralities within the vortex ring in Case $1$. The vortex bursts around ${\boldsymbol {x}}_{br}\in {\mathcal {V}}_{br}$ and generates large ${\boldsymbol {\varrho }}$ in the direction ${\boldsymbol {e}}_{\rho }$.

The solution (4.15) to the Poisson equation can be simplified. The fundamental solution (4.16) decays at the rate of $| {\boldsymbol {x}}_B-\tilde {{\boldsymbol {x}}}|^{-1}$, so the induction effect can be neglected remote from the upper boundary except for the vortex-bursting disc ${\mathcal {V}}_{br}$ (marked in figure 13) where $\varrho _y$ is concentrated. We approximate that only domains ${\mathcal {V}}$ and ${\mathcal {V}}_1$ in figure 13 contribute to the induction near ${\boldsymbol {x}}_B$ as

(4.17)\begin{equation} v_B({\boldsymbol{x}})\approx\int_{{\mathcal{V}}\cup {\mathcal{V}}_1}\varrho_y(\tilde{{\boldsymbol{x}}}) \varPhi({\boldsymbol{x}}-\tilde{{\boldsymbol{x}}})\,\mathrm{d} \tilde{{\boldsymbol{x}}},\quad {\boldsymbol{x}}\to{\boldsymbol{x}}_B\enspace \textrm{and}\enspace \tilde{{\boldsymbol{x}}}\in {\mathcal{V}}\cup {\mathcal{V}}_1, \end{equation}

where ${\mathcal {V}}_1$ is the image domain adjacent to the upper boundary of ${\mathcal {V}}$.

Thus, the induction (4.15) from $\varrho _y$ can be divided into two parts as

(4.18)\begin{equation} v_B \approx v_{B}^{B}+v_{B}^{br} \end{equation}

with

(4.19)\begin{equation} v_{B}^{br} = \int_{{\mathcal{V}}_{br}\cup{\mathcal{V}}_{1,br}} \varrho_y(\tilde{{\boldsymbol{x}}})\varPhi({\boldsymbol{x}}-\tilde{{\boldsymbol{x}}})\,\mathrm{d} \tilde{{\boldsymbol{x}}} = 2\int_{{\mathcal{V}}_{br}} \varrho_y(\tilde{{\boldsymbol{x}}}) \varPhi({\boldsymbol{x}}-\tilde{{\boldsymbol{x}}})\,\mathrm{d} \tilde{{\boldsymbol{x}}} \end{equation}

contributed from the bursting site and

(4.20)\begin{equation} v_{B}^{B}=\int_{{\mathcal{V}}_B \cup{\mathcal{V}}_{1,B}} \varrho_y(\tilde{{\boldsymbol{x}}})\varPhi({\boldsymbol{x}}-\tilde{{\boldsymbol{x}}})\,\mathrm{d} \tilde{{\boldsymbol{x}}}=2\int_{{\mathcal{V}}_B }\varrho_y(\tilde{{\boldsymbol{x}}}) \varPhi({\boldsymbol{x}}-\tilde{{\boldsymbol{x}}})\,\mathrm{d} \tilde{{\boldsymbol{x}}}, \end{equation}

from the near-wall region ${\mathcal {V}}_B$. The two regions are coloured in green in figure 13.

First, we consider the induction effect in region ${\mathcal {V}}_{br}$ with the centre of vortex bursting at ${\boldsymbol {x}}_{br}=(x_{br},y_{br},z_{br})$. The multipole expansion (John Reference John1998) is applied to (4.17), which is detailed in Appendix D. The approximation $(\tilde {{\boldsymbol {x}}}-{\boldsymbol {x}}_{br})/({\boldsymbol {x}}_B-{\boldsymbol {x}}_{br})\to {\boldsymbol {0}}$ holds in ${\mathcal {V}}_{br}$, satisfying the multipole assumption.

With the multipole expansion, the induction of

(4.21)\begin{equation} v_{B}^{br}=2\int_{{\mathcal{V}}_{br}}\varrho_y(\tilde{{\boldsymbol{x}}})\varPhi({\boldsymbol{x}},\tilde{{\boldsymbol{x}}})\,\mathrm{d} \tilde{{\boldsymbol{x}}}=\frac{1}{2 {\rm \pi}}\left(\frac{q_y}{| {\boldsymbol{x}}-{{\boldsymbol{x}}_{br}}|}+ \frac{({\boldsymbol{x}}-{{\boldsymbol{x}}_{br}})\boldsymbol{\cdot} {\boldsymbol{p}}_y }{| {\boldsymbol{x}}-{{\boldsymbol{x}}_{br}}|^3}+\cdots\right) \end{equation}

from $\varrho _y$ in (4.19) is decomposed into the monopole

(4.22)\begin{equation} q_y =\int_{{\mathcal{V}}_{br}}\varrho_y(\tilde{{\boldsymbol{x}}})\,\mathrm{d}\tilde{{\boldsymbol{x}}}, \end{equation}

the dipole

(4.23)\begin{equation} {\boldsymbol{p}}_y=\int_{{\mathcal{V}}_{br}} (\tilde{{\boldsymbol{x}}}-{\boldsymbol{x}}_{br} )\varrho_y(\tilde{{\boldsymbol{x}}})\,\mathrm{d} \tilde{{\boldsymbol{x}}} \end{equation}

and high-order terms. Since $q_y$ is concentrated in ${\mathcal {V}}_{br}$ and conserved in (4.3),

(4.24)\begin{equation} q_y \approx\int_{{\mathcal{V}}}\varrho_y(\tilde{{\boldsymbol{x}}})\,\mathrm{d}\tilde{{\boldsymbol{x}}}=0 \end{equation}

vanishes in (4.21). Then, (4.21) is simplified into

(4.25)\begin{equation} v_{B}^{br}(y,d)=\frac{1}{2 {\rm \pi}}\frac{({\boldsymbol{x}}-{{\boldsymbol{x}}_{br}})\boldsymbol{\cdot} {\boldsymbol{p}}_y }{| {\boldsymbol{x}}-{{\boldsymbol{x}}_{br}}|^3}=\frac{1}{2 {\rm \pi}} \frac{(y-{y_{br}}) {p}_{y,y} }{(d^2+(y-y_{br})^2)^{3/2}} \end{equation}

with the only non-vanishing component ${\boldsymbol {p}}_y = (0,p_{y,y},0)$ and the distance $d=((x-x_{br})^2+(z-z_{br})^2)^{1/2}$ from the vortex-bursting site.

In the near-wall region ${\mathcal {V}}_B$, we re-express (4.25) in terms of $y^+$ as

(4.26)\begin{equation} v_{B}^{br}(y^+,d)=\frac{1}{2 {\rm \pi}}\frac{(1-y^+{/}{\it Re}_\tau-{y_{br}}) {p}_{y,y} }{(d^2+(1-y^+{/}{\it Re}_\tau-y_{br})^2)^{3/2}}. \end{equation}

In the inner layer with $y^+\ll {{\it Re}}_\tau$, (4.26) can be expanded around $y^+=0$ and linearised into

(4.27)\begin{equation} v_{B}^{br}(y^+,d)=\frac{1}{2{\rm \pi}{\it Re}_{\tau}} \frac{y_{br} p_{y,y}}{(y_{br}^2+d^2)^{3/2}}\left(\frac{3y_{br} }{d^2+y_{br}^2}-\frac{1}{y_{br}}\right)y^+. \end{equation}

Thus, the value of $v_B^{br}$ in (4.27) grows when the vortex bursts more closely to the upper wall, and it grows with $y^+$ linearly in the inner layer. The location $(x_{br},y_{br},z_{br})$ in (4.27) is determined by the initial vortex-ring disturbance, and the dipole strength $p_{y,y}$ is determined by the vortex evolution, which is discussed in § 4.2. Our DNS results show that larger ${\it Re}_\tau$ leads to smaller $v_B$ (not shown), consistent with (4.27), and larger $p_{y,y}$, as implied by the larger disc in figure 4(c).

In addition to $v_B^{br}$, the wall-induced velocity $v_B^B$ in (4.18) also contributes to $v_B$. In the evaluation of $v_B^B$ in (4.20), the DNS result exhibits a very thin layer of $\varrho _y$ in ${\mathcal {V}}_B$ as a response to the Neumann boundary condition (4.2). The averaged $v_B^{B}$ is around a quarter of the averaged $v_B^{br}$.

The contours of $v_{B}$ from the DNS and $v_{B}^{br}$ from the model (4.27) are compared in figure 14. The approximation of $v_{B}^{br}$ using (4.27) agrees with the DNS result of $v_B$, except that the averaged amplitude of $v_{B}^{br}$ is $30\,\%$ smaller than $v_B$ in the DNS. This discrepancy is primarily attributed to neglecting $v_B^B$ in the model, with a minor contribution from neglecting the high-order terms in the multipole decomposition in (4.21). Therefore, (4.27) captures the dominant induction effect near the wall from the vortex evolution remote from the wall.

Figure 14.

Contour of (a) $v_B$ in the DNS and (b) $v_{B}^{br}$ from the model in (4.27) in Case 1 in the $x$–$z$ plane at $y^+=1$ and at $t=1$ in Case 1. The asterisk marks the position of vortex bursting projected on the $x$–$z$ plane.