betaFeedback

WorksPhysics and Astronomy

Tessellation-based analysis of impurity clustering in the edge plasma of tokamaks

Abstract

Confinement quality in fusion plasma is significantly influenced by the presence of heavy impurities, which can lead to radiative heat loss and reduced confinement. This study explores the clustering of heavy impurity, i.e., Tungsten in edge plasma, using high-resolution direct numerical simulations of the Hasegawa–Wakatani equations. We use Stokes number to quantify the inertia of impurity particles. It is found that particle inertia will cause spatial intermittency in particle distribution and the formation of large-scale structures, i.e., the clustering of particles. The degrees of clustering are influenced by Stokes number. To quantify these observations, we apply a modified Voronoi tessellation, which assigns specific volumes to impurity particles. By determining time changes of these volumes, we can calculate the impurity velocity divergence, which allows to assess the clustering dynamics. To quantify the clustering statistically, several approaches are applied, such as probability density function (PDF) of impurity velocity divergence and joint PDF of volume and divergence.

Introduction

Fusion reactors aim to achieve fusion reactions by confining and heating a plasma mixture of ions and electrons at extremely high temperatures. Impurities in fusion plasmas can profoundly affect the performance and stability. The impurities primarily originate from the interaction between the intense heat of the plasma and the tokamak’s walls, Their presence can lead to significant radiation losses and make it challenging to maintain the requisite temperatures for fusion reactions [34]. These impurities can also dilute the primary fusion fuel, thereby reducing the overall fusion reaction rate. Additionally, strong radiation on the plasma’s edge can drastically reduce electrical conductivity, leading to disruptions in the current profile inside the q=2q = 2 surface [33]. Moreover, before reaching density limits, plasmas often develop MARFEs (multifaceted asymmetric radiation from the edge) which produce localized radiation, usually at the inner wall or the X-point, caused by strong impurity radiation near the edge [33].

The behavior of plasma flow within reactors, specifically in the edge region of a tokamak, is influenced by drift-wave turbulence and zonal flows. Our high-resolution simulations to understand this flow are based on the Hasegawa–Wakatani model [11], which provides insights into cross-field transport caused by electrostatic drift waves. The electric field perpendicular to magnetic field lines, is particularly significant because it strongly drives cross-field fluxes, and influences edge pressure profiles and overall stability [37]. The Hasegawa–Wakatani equations are particularly useful for studying plasma turbulence, which is a major cause of energy loss in tokamak devices. They capture the essential physics of drift-wave turbulence. Despite their simplicity, the Hasegawa–Wakatani equations have proven to be a powerful tool for understanding the complex dynamics of plasma behavior in tokamak devices [12].

For the particles, the Lagrangian perspective has been of interest in recent years[10, 14, 2]. This perspective delves into transport properties by analyzing the trajectories of multiple tracer particles. Through numerical simulations, we solve equations about these particle trajectories within particular plasma flow velocity fields, like the E×BE \times B field. This method highlights the significant impact of coherent structures on transport. The interplay between eddy trapping and zonal shear flows results in non-diffusive transport [17, 4, 25, 18, 16]. Both experimental and numerical findings confirm the presence of these coherent structures in edge turbulence [17].

In fusion plasma research, existing studies have focused on passive flow tracers in edge plasma, using the Hasegawa-Wakatani model, without taking inertial effects into account [7, 6]. In the work by Priego et al. [30], inertial effect is considered and a fluid model for impurities is used, wherein the impurity particle velocity is considered to be the sum of the E×BE \times B and polarization drifts. The polarization drift, accounting for impurity particle inertia, is a higher-order correction to the E×BE \times B drift velocity and introduces compressible effects. In contrast to fluid model, we track each impurity particle individually. In addition, we assume that heavy impurity particles may “lag behind” plasma flow velocity due to the significant inertia. Indeed, the role of inertia in particle advection by turbulent flows is well-documented in fluid dynamics, known to result in clustering around vortical structures [31]. Our study aims to analyze possible clustering effects for impurities in magnetized plasma.

A key element in understanding these heavy impurities is their tendency for self-organization. This self-organization is seen in clusters of impurities and void regions [20] (Monchaux et al. 2010). It can be mathematically quantified by deviation from Poissonian statistics [28, 24]. The goal is to investigate the statistical signature of clustering and void regions. Recent results for hydrodynamics turbulence are promising [23]. Finite-time measures to quantify divergence and the rotation of the particle velocity by determining respectively the volume change rate of the Voronoi cells and their rotation were proposed in Oujia et al. [28]; Maurel-Oujia et al. [24] and applied in the context of hydrodynamic turbulence. Here we apply them to characterize the dynamics of the self-organization of heavy impurities in plasma and assess their clustering and void formation.

The paper is structured as follows: § 2 outlines the theoretical basis used for simulation, tessellation-based analysis method and numerical setup. § 3 details our numerical simulation results and tessellation-based analysis results. Finally, § 4 summarizes our findings and suggests potential directions for future investigations.

Models for simulation and analysis method

Hasegawa–Wakatani model

The model for our numerical simulations is based on Hasegawa–Wakatani (HW) equations for plasma edge turbulence driven by drift-wave instability [11]. Our focus in this study is on the two-dimensional slab geometry of the HW model, see e.g. Kadoch et al. [14]. The 2D HW model is a representative paradigm for understanding the nonlinear dynamics of drift-wave turbulence. Figure 1 depicts a representation of the flow configuration.

Image not extracted

Figure 1. Illustration of the two-dimensional slab geometry. In the tokamak edge, the 2D slab geometry flow configuration is depicted using the Hasegawa–Wakatani system. Here, the radial direction is represented by xx, while yy is the poloidal direction. There is an imposed mean plasma density gradient ∇n0\nabla n_0 in the radial direction. The 2D flow is computed within a square domain measuring 64 Larmor radii (ρs\rho_s) on each side, as indicated by the green outline. On the right, a vorticity field is displayed.

The magnetic field lines are assumed to be straight and perpendicular to the slab. Ions are considered cold, ignoring temperature gradient effects. The key variables are normalized as detailed in the work done by Bos et al. [2] and Futatani et al. [9],

x/ρs→x,ωcit→t,eϕ/Te→ϕ,n1/n0→nx/\rho_s \to x,\quad\omega_{ci}t \to t,\quad e\phi/T_e \to\phi,\quad n_1/n_0 \to n

ρs\rho_s is the ion Larmor radius at electron temperature TeT_e. It is defined as ρs=Te/miωci\rho_s = \frac{\sqrt{T_e/m_i}}{\omega_{ci}}, where ωci\omega_{ci} is the ion cyclotron frequency. Here n1n_1 and n0n_0 represent plasma density fluctuation and equilibrium density, while ϕ\phi indicates electrostatic potential fluctuation. The HW model consists of two equations that describe the evolution of plasma potential and fluctuating plasma density, respectively:

(∂∂t−μν∇2)∇2ϕ=[∇2ϕ,ϕ]+c(ϕ−n),(1)\left(\frac{\partial}{\partial t}-\mu_\nu\nabla^2\right)\nabla^2\phi= [\nabla^2\phi,\phi] + c(\phi-n), \tag*{(1)}
(∂∂t−μD∇2)n=[n,ϕ]−κ∂ϕ∂y+c(ϕ−n),(2)\left(\frac{\partial}{\partial t}-\mu_D\nabla^2\right)n=[n,\phi]-\kappa\frac{\partial\phi}{\partial y}+c(\phi-n), \tag*{(2)}

where μD\mu_D is cross-field diffusion coefficient and μν\mu_\nu is kinematic viscosity. The term κ\kappa defined as κ=−∂xln⁡(n0)\kappa=-\partial_x\ln(n_0), is a measure of the plasma density gradient. Here κ\kappa is assumed constant, implying n0∝exp⁡(−κx)n_0\propto\exp(-\kappa x). The Poisson bracket is defined as: [A,B]=∂A∂x∂B∂y−∂A∂y∂B∂x[A,B]=\frac{\partial A}{\partial x}\frac{\partial B}{\partial y}-\frac{\partial A}{\partial y}\frac{\partial B}{\partial x}. In these equations, the electrostatic potential ϕ\phi is the stream-function for the E×BE\times B velocity, represented by u=∇⊥ϕ\mathbf{u}=\nabla^\perp\phi, where ∇⊥=(−∂∂y,∂∂x)\nabla^\perp=\left(-\frac{\partial}{\partial y},\frac{\partial}{\partial x}\right). So, ux=−∂ϕ∂yu_x=-\frac{\partial\phi}{\partial y} and uy=∂ϕ∂xu_y=\frac{\partial\phi}{\partial x} with vorticity ω=∇2ϕ\omega=\nabla^2\phi.

In the equations, cc is so-called adiabaticity parameter which measures the parallel electron response. It is defined as:

c=Tek∥2e2n0ηωci(3)c=\frac{T_e k_\parallel^2}{e^2 n_0 \eta\omega_{ci}} \tag*{(3)}

With η\eta as electron resistivity and k∥k_\parallel as the effective parallel wavenumber, parameter cc determines the phase difference between electrostatic potential and plasma density fluctuations. The model described above is the classical Hasegawa–Wakatani model (cHW). For c≫1c\gg1 (adiabatic limit), the model reduces to the Hasegawa–Mima equation, where electrons follow a Boltzmann distribution. At c≪1c\ll1 (hydrodynamic limit), the system reduces to a form that is analogous to two-dimensional Navier–Stokes equation, where density fluctuations are passively influenced by the E×BE\times B drift. Notably, for c∼1c\sim1 (quasi-adiabatic regime), a phase shift between the potentials and densities is observed, enabling particle transport and reflecting a complex interplay between them. To obtain zonal flow, a revised version of the model, known as the modified Hasegawa–Wakatani model (mHW), can be considered. This modification was introduced in Pushkarev et al. [32] which consists in setting the coupling term c(ϕ−n)c(\phi-n) to zero for modes ky=0k_y=0. In this paper, we will focus on the quasi-adiabatic regime, i.e., c=0.7c=0.7 (cHW), which is relevant to the edge plasma of tokamaks [2]. For readers interested in the results of other regimes, we refer to Appendix A.

Impurity particles model

Earlier research [7, 6, 13, 30] assumed that impurity particles, due to no inertia, act as "passive tracers" aligning closely with fluid flow. Yet, for heavier particles, this assumption might not always hold, particularly when the particles have significant mass, leading to noticeable inertial effects. To address this, the Stokes number (StSt) can be introduced, see e.g. Oujia et al. [28], a dimensionless parameter that quantifies particle inertia in fluid flow. The Stokes number is defined as:

St=τpτη,(4)St=\frac{\tau_p}{\tau_\eta}, \tag*{(4)}

where τp\tau_p denotes the impurity particle’s relaxation time, it measures the time the particle takes to adapt to fluid flow alterations. A smaller τp\tau_p means the particle adapts faster. On the other hand, τη\tau_\eta is the characteristic timescale for turbulence, indicating how fast the fluid flow changes. A smaller τη\tau_\eta suggests more rapid turbulence changes. A high Stokes number means that a particle’s movement is primarily driven by inertia, causing it to keep its direction despite fluid changes. In previous studies [7 b,] 2009), the impurity particles are treated as passive tracers of the flow, indicating τp=0\tau_p = 0. However, here we consider heavy impurity particles with τp>0\tau_p > 0, for instance Tungsten.

The heavy impurity particles are modeled as charged point particles. They are considered as test particles, which means that there is no impurity-impurity interaction and impurities have no impact on the plasma dynamics, while the plasma flow velocity affects the motion of impurities (one-way coupling). Impurities in our model are assumed to be cold, just as the treatment of ions in bulk plasma. Thus thermal motion of impurities could be neglected for simplicity. The impurity particles experience not only the Lorentz force due to the electric and magnetic fields but also drag force resulting from momentum transfer with plasma ions. Positive plasma ions transfer momentum to heavy charged impurity particles by interacting with electrostatic potential in the vicinity of the impurity particle. Electrons transferring momentum to impurities is negligible due to their small mass. Coulomb scattering theory provides, macroscopically, an approximation of a drag force FdragF_{\mathrm{drag}} dependent on the relative velocity between the macroscopic velocity of the plasma ion and the velocity of the impurity particle (Kilgore et al. 1993): Fdrag=KmtnimivrF_{\mathrm{drag}} = K_{\mathrm{mt}}n_im_iv_r, where vrv_r is the relative velocity between the macroscopic velocity of the plasma ions and the impurity particles, KmtK_{\mathrm{mt}} is the momentum transfer coefficient, nin_i is the plasma ion density, and mim_i is the ion mass of the plasma. This drag force tends to reduce the velocity difference, causing the impurity particles to relax to the plasma flow velocity. The relaxation time of the impurity particle τp\tau_p, is the period during which impurity particles adjust to the plasma flow due to the drag force. In our model, we approximate this drag force as mp(up−vp)/τpm_p(u_p-v_p)/\tau_p, which is proportional to the relative velocity between the plasma flow and impurities. The relaxation time can be expressed as: τp=mpKmtnimi\tau_p = \frac{m_p}{K_{\mathrm{mt}}n_im_i}. This relation highlights that τp\tau_p is inversely proportional to the plasma ion density and momentum transfer coefficient, while directly proportional to the impurity mass. Given this relationship, as the ion density nin_i increases, the relaxation time decreases, leading to faster coupling of the impurity particles to the plasma flow. Conversely, larger impurity masses result in longer relaxation times, reflecting the slower response of heavier particles to the plasma flow. According to Newton’s second law:

mpdvp,jdt=mp(up,j−vp,j)τp+Ze(E(xp,j)+vp,j×B)(5)m_p\frac{d v_{p,j}}{d t} = \frac{m_p(u_{p,j}-v_{p,j})}{\tau_p} + Ze\left(\mathbf{E}(\mathbf{x}_{p,j})+\mathbf{v}_{p,j}\times\mathbf{B}\right) \tag*{(5)}
E(xp,j)=−∇ϕ(xp,j)(6)\mathbf{E}(\mathbf{x}_{p,j})=-\nabla\phi(\mathbf{x}_{p,j}) \tag*{(6)}

Equation 2.5 describes the forces exerted on each heavy impurity particle, denoted by jj (j=1,…,Npj = 1,\ldots,N_p, with NpN_p being the total number of particles). ZZ is charge state of the impurity particle and ee is the elementary charge. The variable vp,jv_{p,j} is the jj-th particle’s velocity, while up,ju_{p,j}, E(xp,j)\mathbf{E}(\mathbf{x}_{p,j}) and ∇ϕ(xp,j)\nabla\phi(\mathbf{x}_{p,j}) representing the fluid velocity, Electric field and gradient of potential at the particle’s location xp,j\mathbf{x}_{p,j}, respectively. Let us note that when τp=0\tau_p = 0, we recover the situation studied in previous works (Futatani et al. 2008a,b, 2009). In this case, the impurity particles are considered as passive tracers of the flow, i.e. vp,j≡up,jv_{p,j} \equiv u_{p,j}. Applying the same normalization as those used for the HW equation, we obtain:

dvp,jdt=(up,j−vp,j)τp+α(−∇ϕ(xp,j)+vp,j×b)(7)\frac{d v_{p,j}}{d t} = \frac{(u_{p,j}-v_{p,j})}{\tau_p} + \alpha\left(-\nabla\phi(\mathbf{x}_{p,j})+\mathbf{v}_{p,j}\times\mathbf{b}\right) \tag*{(7)}

where α=Zmimp\alpha= \frac{Zm_i}{m_p}, mim_i is the ion mass of the plasma, and b\mathbf{b} represents the unit vector along the direction of the magnetic field. Combining with the equation:

dxp,jdt=vp,j(8)\frac{d\mathbf{x}_{p,j}}{d t} = \mathbf{v}_{p,j} \tag*{(8)}

it is thus possible to compute the trajectory of the impurity particles.

Modified Voronoi tessellation

Impurity particles satisfy the conservation equation:

Dtnp=−np∇⋅vp,(9)D_t n_p = -n_p \nabla\cdot\mathbf{v}_p, \tag*{(9)}

where DtD_t is the Lagrangian derivative, npn_p is the impurity particle number density and vp\mathbf{v}_p the particle velocity. From this equation, the divergence of particle velocity ∇⋅vp=−Dtnpnp\nabla\cdot\mathbf{v}_p = -\frac{D_t n_p}{n_p} is crucial to understand. A positive divergence (∇⋅vp>0)(\nabla\cdot\mathbf{v}_p > 0) leads to density decrease, whereas a negative divergence, i.e. convergence (∇⋅vp<0)(\nabla\cdot\mathbf{v}_p < 0) causes an increase in density.

To quantify the divergence, we employ the modified version of the classical Voronoi tessellation technique, which uses the center of gravity, instead of the circumcenter of the Delaunay cell to define Voronoi cell vertices. The purpose of this method is to assign a volume to each impurity particle, enabling the subsequent calculation of the divergence. More details about the classical and modified Voronoi tessellation and their difference can be found in Oujia et al. [28]; ); Maurel-Oujia et al. [24]. The modified Voronoi tessellation divides a plane into regions according to the positions of impurity particles, which allows to assign a cell, called "modified Voronoi cell", and thus a corresponding volume VV, to each particle. Each cell can be considered as the region of influence of an impurity particle. In our two-dimensional scenario, the volume of the cell is determined by calculating its area. Figure 2 shows the modified Voronoi tessellation. In the figure, clustered particles are represented by small cells and void regions are represented by large cells.

Image not extracted

Figure 2. (a) Electric potential ϕ\phi (stream function) and 10410^{4} superimposed impurity particles for St=1St = 1 in the case c=0.7c = 0.7 (cHW). (b) A magnified view with modified Voronoi tessellation. Large cells denote void regions, small cells represent clusters.

To compute the divergence of impurity particle velocity DD we analyze particle distributions at two time instants tkt^k and tk+1=tk+Δtt^{k+1} = t^k + \Delta t, where Δt\Delta t is the time step of

AARRΔt\Delta tNpN_pμD\mu_Dμν\mu_\nuκ\kappacc
64 ×\times 641024 ×\times 10245×10−45 \times10^{-4}10610^65×10−35 \times10^{-3}5×10−35 \times10^{-3}10.7

caption: tbl:1: Simulation parameters for the flow. AA: Domain area; RR: Grid resolution; Δt\Delta t: Time step; NpN_p: Number of impurity particles; μD\mu_D: Diffusion coefficient; μν\mu_\nu: Kinematic viscosity; κ≡−∂xln⁡(n0)\kappa\equiv-\partial_x \ln(n_0), is a measure of the plasma density gradient; cc: Adiabaticity parameter

:::

the simulation and kk is the discrete time index. Each particle distribution snapshot is assessed using modified Voronoi tessellation. The inverse of volume, 1/V1/V, approximates the local particle number density npn_p in the discrete setting. It is proved that [28, 24]:

D(vp)=−1npDtnp=2ΔtVk+1−VkVk+1+Vk+O(1NpΔt)(10)\mathcal{D}(v_p)=-\frac{1}{n_p}D_t n_p=\frac{2}{\Delta t}\frac{V^{k+1}-V^k}{V^{k+1}+V^k}+O\left(\frac{1}{N_p}\Delta t\right) \tag*{(10)}

Thus by measuring the volume change of each particle between tkt^k and tk+1t^{k+1}, we can determine the discreet divergence of its velocity D(vp)\mathcal{D}(v_p). The velocity divergence of impurity particles measures the rate of volume change for a particle group over time, indicating how much the particle velocity field is spreading out or converging at a particular point in space.

Numerical setup

The simulation was conducted within a domain that spans an area AA of 64×6464 \times64 with periodic boundary condition. The domain was discretized to a resolution RR of 1024×10241024 \times1024 grid points. The time step Δt\Delta t was 5×10−45 \times10^{-4}. The number of impurity particles NpN_p was 10610^6. The adiabaticity parameter cc is 0.7. The values of specific physical parameters are listed in Table 1. The characteristic time scale of turbulence, τη\tau_\eta, is defined as 1/2Zms1/\sqrt{2Z_{\mathrm{ms}}}, with ZmsZ_{\mathrm{ms}} denoting one half of the mean-square vorticity. In the simulation, τη=0.35\tau_\eta= 0.35 in the statistically steady flow. The simulations are based on the work of Kadoch et al. [14] where the flow is initialized with Gaussian random fields. Here we start the simulations with a flow already being in the statistically steady state (i.e. after the long transition phase of drift-wave instabilities, see Kadoch et al. [14]) and one million uniformly distributed heavy impurity particles with random velocity are injected.

For the impurity particles, we will focus on Tungsten 74184W20+{}^{184}_{74}\mathrm{W}^{20+}, a heavy one that can be made from the material that is exposed to the high heat flux in ASDEX-Upgrade (Krieger et al. [19]) and EAST (Yao [36]), expecting the same situation for ITER (Pitts [29]). For 74184W20+{}^{184}_{74}\mathrm{W}^{20+}, α=Zmi/mp=0.22\alpha= Zm_i/m_p = 0.22. As for Stokes number St=τp/τηSt = \tau_p/\tau_\eta, τη=0.35\tau_\eta= 0.35 in the statistically steady state of of flow for c=0.7c = 0.7 (cHW), but the exact value of τp\tau_p for 74184W20+{}^{184}_{74}\mathrm{W}^{20+} is unknown. To resolve this, we explore a range of τp\tau_p values spanning several orders of magnitude, resulting in Stokes numbers: St=0,0.05,0.5,1,5,10St = 0, 0.05, 0.5, 1, 5, 10 and 50. This parametric approach allows us to systematically explore the dynamics of Tungsten impurity, from particles that are tightly coupled to the flow (St≪1St \ll1) to those that are essentially independent of the flow (St≫1St \gg1). The simulation parameters are listed in Table 2.

The system of HW equations, specifically Eq. (1) and Eq. (2) are solved using a classical pseudospectral method (Canuto et al. [3]). The equations governing the motion of the impurity particles, namely Eq. (7) and Eq. (8), are solved using a

StSt00.050.250.512550

caption: tbl:2: Stokes number used in the simulation

:::

second-order Runge–Kutta (RK2) scheme and linear interpolation is used for computing the fluid velocity at the particle position. The code of this study builds upon that used in the research done by Kadoch et al. [14].

Results

Simulation results

The analyzed data set consists of the electric potential ϕ\phi (stream function), particle position and velocity data generated by direct numerical simulation (DNS). The contour curves of constant ϕ\phi represent the streamlines of the flow field and provide a clear visualization of fluid flow in two dimensions. Closed contour curves indicate the presence of vortices. Figure 3 illustrates the ϕ\phi fields and the behavior of impurity particles in statistically steady state within the quasi-adiabatic regime (c=0.7c = 0.7, cHW), where impurity particles are considered as passive tracers without considering inertial effect (St=0St = 0). As shown by Figure 3 impurity particles are distributed randomly. Figure 4 shows that with increasing Stokes numbers (St=0.05,0.5,1St = 0.05, 0.5, 1), particle inertia begins to dominate, causing them to deviate from the fluid streamlines due to Coriolis force induced by the vortices. This deviation causes them to concentrate where the vorticity is minimal, typically at vortex peripheries. For even higher Stokes numbers (St=5,10,50St = 5, 10, 50), the impurity particles tend to maintain their trajectories, moving more ballistically and exhibiting reduced clustering. From Eqn. (7), when the coefficient of the flow effect term, 1/τp=1/(τηSt)1/\tau_p = 1/(\tau_\eta St), is much larger than the coefficient term of the Lorentz force α\alpha, the particle dynamics are primarily governed by the flow, while the Lorentz force plays a relatively minor role. For St=0.05,0.5,1St = 0.05, 0.5, 1, 1/τp=1/(0.35St)≫α=0.221/\tau_p = 1/(0.35St) \gg\alpha= 0.22, the particles are more strongly influenced by the fluid flow than by the electromagnetic forces. However, as the Stokes number increases to values such as St=10,50St = 10, 50, the particles are less coupled to the fluid flow and the electromagnetic forces start to play an important role. As shown in Figure 4, for case St=10St = 10 and St=50St = 50, the positively charged impurity particles are mostly located in negative potential regions (red region), as expected. The effects of different α\alpha values are further explored in Appendix B.

Image not extracted

Figure 3. Electric potential fields ϕ\phi (stream function) superimposing 10410^{4} impurity particles (out of 10610^{6}) for St=0St=0 in statistically steady state within the quasi-adiabatic regime (c=0.7c = 0.7, cHW).

Image not extracted

Figure 4. Electric potential fields ϕ\phi (stream function) superimposing 10410^{4} impurity particles (out of 10610^{6}) for various Stokes numbers in statistically steady state within the quasi-adiabatic regime (c=0.7c = 0.7, cHWc_{\mathrm{HW}}).

Let us note that in regions of impurity clustering, where the impurity density may locally increase, there is a risk that clustering could lead to deviations from the assumption of quasi-neutrality that is underlying the derivations of the HW model (Naulin et al. [27]). This strict condition, which requires that the impurity density remains smaller than the bulk current divergence on the characteristic drift-wave time scale, cannot always be guaranteed in regions of intense clustering. Nevertheless, as noted in Naulin et al. [27], considering impurities as test particles is still valuable for understanding tendencies in impurity transport within a given flow field, and the clustering effects observed in our simulations provide insights into impurity dynamics. We further acknowledge that resolving this discrepancy would require a more sophisticated model that accounts for the back-reaction of impurity densities on the plasma fields and their contribution to quasi-neutrality. Such a model would need to self-consistently address the plasma profile problem in the presence of impurities, which is beyond the scope of the present work, but remains a valuable direction for future research.

Tessellation-based analysis results

We calculate the volume VV for each impurity particle, and then plot the Probability Density Function (PDF) of V/V‾V/\overline{V}, where V‾\overline{V} is the average volume, calculated as V‾=642/Np\overline{V}=64^{2}/N_{p} (64264^{2} is the domain area, NpN_{p} is the total number of particles). Figure 5(a) shows the PDF of V/V‾V/\overline{V}.

Image not extracted

Figure 5. (a) PDF of volume normalized by the mean for different Stokes numbers against the gamma distribution. (b) PDF of the divergence of impurity particle velocity for different Stokes numbers for 74184W20+{}^{184}_{74}\mathrm{W}^{20+} (α=0.22\alpha= 0.22) in quasi-adiabatic regime (c=0.7c = 0.7, cHWc_{\mathrm{HW}}).

A random variable XX that is gamma-distributed with shape kk and scale parameter θ\theta is denoted X∼Γ(k,θ)X \sim\Gamma(k,\theta). The corresponding PDF is f(x)=Γ(k)−1θ−kxk−1exp⁡(−x/θ)f(x)=\Gamma(k)^{-1}\theta^{-k}x^{k-1}\exp(-x/\theta). For particles that are randomly distributed, the PDF of the volumes follows a gamma distribution [5]. For St=0St=0, using the maximum likelihood PDF estimation for fitting, we find that the PDF of the volumes aligns well with the gamma distribution, specifically we have Γ(1.72,0.58)\Gamma(1.72,0.58). This suggests, as expected, that the impurity particles are randomly distributed when St=0St=0, indicating no clustering of particles. Let us note that the PDF of the volumes for St=0St=0 obtained using classical Voronoi tessellation follows a gamma distribution Γ(7/2,2/7)\Gamma(7/2,2/7) [5]. Since we use a different method, the modified Voronoi tessellation, the fitting gives different values, i.e. Γ(1.72,0.58)\Gamma(1.72,0.58). The PDF of the volume is typically used to classify "cluster cells" and "void cells" (Monchaux et al. 2010). A cell below a certain threshold is considered a cluster cell, whereas a cell above this threshold is categorized as a void cell. In our study, the threshold is V/V‾∼0.5V/\overline{V}\sim0.5. As shown by Figure 5(a), when StSt increases, we observe an increase in the number of cluster cells (with V/V‾<0.5V/\overline{V}<0.5). However, once StSt exceeds St=1St=1, the number of cluster cells begins to decrease. This is also highlighted by Figure 4 which shows that the particles are densely packed for St=1St=1.

The PDF of the divergence of impurity particle velocity is displayed in Figure 5(b). As shown in Figure 5(b), the divergence of the fluid velocity (represented by the divergence of the impurity particle velocity with St=0St = 0 is small but not zero. However, as we know, in an incompressible continuous fluid, the divergence of the fluid velocity should be zero. The reason for this discrepancy lies in our computational method, which segments the fluid into separate pieces known as "modified Voronoi cells" as mentioned before. The deformation of a modified Voronoi cell does not correspond to

the deformation of a fluid volume in a continuous context. The figure implies that the divergence is attributable to discretization errors. A negative divergence in a particle velocity field indicates that particles converge towards a point, thus increasing density locally. Conversely, positive divergence shows that particles spread out from a point, reducing the local density. As shown in Figure 5(b), for increasing Stokes number, the PDF of divergence D\mathcal{D} widens, reflecting stronger particles convergence and divergence activities. The symmetry observed in Figure 5(b) suggests that positive and negative divergence values are comparable. This balance is due to the conservation of impurity particles.

The joint PDFs of the divergence D\mathcal{D} and normalized volume V/V‾V/\overline{V} show more insight into cluster formation (cf. Figure 6). There is a high probability along D=0\mathcal{D} = 0, indicating groups of particles move as an incompressible flow. High negative/positive divergence probability in cluster cells (V/V‾<0.5V/\overline{V} < 0.5) may be due to the fact that the probability of finding particles in clusters is higher than in void region as illustrated by Figure 5(a). As the Stokes number increases, particles deviate more from the fluid flow, increasing their relative velocity. Therefore, high negative/positive divergence could also be from ’caustics’ where the multi-valued particle velocities at a given position cause large velocity differences between neighboring particles, leading to extreme divergence [35]. Figure 7 visually supports this by coloring particles based on their divergence D\mathcal{D}. From Figure 7, we observe that particles with large negative/positive divergence values are densely packed in clusters, indicating simultaneous particles convergence and divergence.

Image not extracted

Figure 6. Joint PDF of the volume in log scale and divergence in linear scale for (a) St=0.05St = 0.05, (b) 0.5, (c) 1, (d) 5, (e) 10 and (f) 50 in quasi-adiabatic regime (c=0.7c = 0.7, cHWc_{\mathrm{HW}}).

Image not extracted

Figure 7. Spatial distribution of one million particles colored with the divergence D\mathcal{D} for (a) St=0.05St = 0.05, (b) 0.5, (c) 1, (d) 5, (e) 10 and (f) 50 in quasi-adiabatic regime (c=0.7c = 0.7, cHWc_{\mathrm{HW}}).

Conclusions

In this work, we studied the spatial distribution of impurity particles in the edge plasma of tokamaks. High-resolution simulations of the plasma flow were based on the paradigmatic model in edge plasma of tokamaks, namely Hasegawa–Wakatani model, while the impurity particles were tracked using Lagrangian method. Assuming that impurity particles have a non-zero relaxation time (τp≠0\tau_p \ne0), which means that the particles do not immediately adapt to fluctuations in plasma flow, we observed the phenomenon of impurity clustering within the plasma. To gain deeper insights into the self-organization of impurity particles, we employed a modified Voronoi tessellation technique to compute the divergence of impurity particle velocity D\mathcal{D}. The main results are:

(i) The impact of StSt

  • As StSt increases, impurities cluster in low-vorticity regions due to their inability to adapt to rapid changes in plasma flow. At high StSt, impurities move randomly with less clustering.

  • With an increase in StSt, there is a higher probability of observing higher values of negative or positive divergence (for α=0.22\alpha= 0.22). This implies that the impurity particle velocity field is converging or diverging faster. These large values of negative or positive divergence are more likely to be found in cluster cells where (V/Vˉ<0.5V/\bar{V} < 0.5).

(ii) The impact of α\alpha:

  • The influence of α\alpha becomes significant only when StSt is large, which tends to drive the impurity particles into negative electric potential regions.

For future work, it could be interesting to extend the analysis to multi-resolution techniques that would provide a better understanding of the clustering behavior of impurity particles at different scales, see e.g. Matsuda et al. [22]. This approach could reveal more details about the spatial distribution and dynamics of clusters that are not apparent at a single scale of observation. Another interesting direction is to explore the transport dynamics of heavy impurity particles for different Stokes number by tracking them for a long time. Finally, the limitation of this work is that we assumed that impurities do not affect plasma flow. While this simplification allowed for a focused study on impurity particle dynamics, it may not capture the full complexity of the interaction between impurities and plasma flow. A possible extension in future studies is to explore the impact of impurity particles on the plasma flow based on the work of Benkadda et al. [1].

Acknowledgements

ZL acknowledges Philipp Krah for the insightful discussions. Centre de Calcul Intensif d’Aix-Marseille is acknowledged for providing access to its high performance computing resources.

Funding

This work was supported by I2M (Z.L., T.M.O., K.S.); the French Federation for Magnetic Fusion Studies (FR-FCM) and the Eurofusion consortium, funded by the Euratom Research and Training Programme (Z.L., T.M.O., K.S., S.B., grant number 633053); and the Agence Nationale de la Recherche (ANR), project CM2E (Z.L., T.M.O., B.K., K.S., grant number ANR-20-CE46-0010-01). The views and opinions expressed herein do not necessarily reflect those of the European Commission.

Declaration of interests

The author reports no conflict of interest.

Appendix A

The behavior of 74184W20+{}^{184}_{74}\mathrm{W}^{20+} in different flow regimes

Hydrodynamic and adiabatic regime

We explored the behavior of 74184W20+{}^{184}_{74}\mathrm{W}^{20+} in hydrodynamic (c=0.01c = 0.01, cHWc_{\mathrm{HW}}) and adiabatic regime (c=2c = 2, cHWc_{\mathrm{HW}}). Figure 8 to Figure 9 show potential fields and impurity behavior in hydrodynamic and adiabatic regimes. We have the same observation as quasi-adiabatic case (c=0.7c = 0.7, cHWc_{\mathrm{HW}}): at St=0St = 0, impurities act as passive tracers, uniformly distributed along fluid streamlines; as St=0.05St = 0.05, particles start to cluster in areas of low vorticity and clustering is clear at St=1St = 1; at St=50St = 50, there is less clustering and particles are mostly presented in negative potential areas. Similar to the quasi-adiabatic case, As shown in Figure 10(a) and Figure 11(a), as StSt increases, the number of cluster cells (V/V‾<0.5V/\overline{V} < 0.5) initially increases until St=1St = 1. Figure 10(b) and Figure 11(b) demonstrate that the PDF of divergence D\mathcal{D} broadens with increasing Stokes number.

Image not extracted

Figure 8. Electric potential fields ϕ\phi (stream function) superimposing 10410^{4} impurity particles (out of 10610^{6}) for various Stokes numbers in statistically steady state within the hydrodynamic regime (c=0.01c = 0.01, cHW).

Image not extracted

Figure 9. Electric potential fields ϕ\phi (stream function) superimposing 10410^{4} impurity particles (out of 10610^{6}) for various Stokes numbers in statistically steady state within the adiabatic regime (c=2c = 2, cHW).

Image not extracted

Figure 10. (a) PDF of volume normalized by the mean (V/V‾V/\overline{V}) (b) PDF of impurity velocity divergence D\mathcal{D} for different Stokes numbers for 74184W20+(α=0.22){}^{184}_{74}\mathrm{W}^{20+}(\alpha= 0.22) in hydrodynamic regime (c=0.01c = 0.01, cHW)

Image not extracted

Figure 11. (a) PDF of volume normalized by the mean (V/V‾)(V/\overline{V}) (b) PDF of impurity velocity divergence D\mathcal{D} for different Stokes numbers for 74184W20+{}^{184}_{74}\mathrm{W}^{20+} (α=0.22)(\alpha= 0.22) in hydrodynamic regime (c=2,cHW)(c = 2, \mathrm{cHW})

Zonal flows

In the modified Hasegawa-Wakatani model (mHW) with c=2c = 2, zonal flows are present. Figure 12 shows the zonal flows and superimposing impurity particles. While it may not be clearly visible that particles with St=1St = 1 are primarily presented in regions of low vorticity, as indicated by Figure 12, it is evident in the vorticity field (Figure 13). For St=50St = 50, particles are predominantly found in regions of negative potential. As shown in Figure 14a, there are fewer cluster cells (V/V‾<0.5)(V/\overline{V} < 0.5) compared to other regimes where zonal flows are absent (c=0.01,0.7,2,cHW)(c = 0.01, 0.7, 2, \mathrm{cHW}). Figure 14b illustrates that, in the presence of zonal flows, the likelihood of large negative or positive divergence is reduced compared to other regimes, suggesting that zonal flows tend to prevent the convergence and divergence of particles.

Image not extracted

Figure 12. Electric potential fields ϕ\phi superimposing 10410^{4} impurity particles (out of 10610^{6}) for various Stokes numbers in statistically steady state within zonal flows (c=2,mHW)(c = 2, \mathrm{mHW}).

Figure 13. Vorticity field superimposing 10410^{4} impurity particles (out of 10610^{6}) for various Stokes numbers in statistically steady state within zonal flows (c=2c = 2, mHW).

Appendix B

Analysis of Tungsten impurity with different charge state in quasi-adiabatic regime (c=0.7,cHW)(c = 0.7, \mathrm{cHW}).

The clustering of two other Tungsten impurities with different charge state is analyzed in this appendix. We consider 74184W1+{}^{184}_{74}\mathrm{W}^{1+} (α=0.01)(\alpha= 0.01) and 74184W44+{}^{184}_{74}\mathrm{W}^{44+} (α=0.48)(\alpha= 0.48) [21]. Applying modified Voronoi tesselation, the PDF of volume normalized by the mean

PDF of volume normalized by the mean

(a)

PDF of the impurity velocity divergence

(b)

Figure 14. (a) PDF of volume normalized by the mean (V/V‾V/\overline{V}) (b) PDF of the impurity velocity divergence D\mathcal{D} for different Stokes numbers for 74184W20+{}^{184}_{74}\mathrm{W}^{20+} (α=0.22\alpha= 0.22) in zonal flows (c=2c = 2, mHW).

(V/V‾V/\overline{V}) and the PDF of impurity velocity divergence D\mathcal{D} for different Stokes numbers are plotted in Figure 15 and Figure 16.

PDFs of normalized volume and impurity velocity divergence

Figure 15. (a) PDF of volume normalized by the mean (V/V‾V/\overline{V}) (b) PDF of impurity velocity divergence D\mathcal{D} for different Stokes numbers for 74184W1+(α=0.01){}^{184}_{74}\mathrm{W}^{1+}(\alpha= 0.01) in quasi-adiabatic regime (c=0.7c = 0.7, cHW).

PDFs of normalized volume and impurity velocity divergence

Figure 16. (a) PDF of volume normalized by the mean (V/V‾V/\overline{V}) (b) PDF of the impurity velocity divergence D\mathcal{D} for different Stokes numbers for 74184W44+(α=0.48){}^{184}_{74}\mathrm{W}^{44+}(\alpha= 0.48) in quasi-adiabatic regime (c=0.7c = 0.7, cHW).

Figure 15(a) and Figure 16(a) show that for both 74184W1+{}^{184}_{74}\mathrm{W}^{1+} and 74184W44+{}^{184}_{74}\mathrm{W}^{44+}, at St=0St = 0, the impurity particles exhibit a random distribution, as evidenced by the alignment of the curve with the gamma distribution. This indicates the absence of inertial effects. As StSt increases, the number of cluster cells (V/V‾<0.5V/\overline{V} < 0.5) increases and then decreases after exceeding St=1St = 1. From Figure 15(b), for 74184W1+{}^{184}_{74}\mathrm{W}^{1+}, it is observed that the PDFs of divergence broaden with increasing StSt, then the PDFs begin to narrow from St=10St = 10 to St=50St = 50. From Figure 15(b), for 74184W44+{}^{184}_{74}\mathrm{W}^{44+}, it is observed that the PDFs of divergence widen as StSt increases.

Table 3 and Table 4 provide a quantitative comparison of how the divergence characteristics change with varying α\alpha values, which are indicative of the charge state of the Tungsten ions (α=Zmi/mp\alpha= Zm_i/m_p). Given that negative and positive divergence values are comparable, the overall average divergence approaches zero. So we calculate the mean of positive divergence D‾+\overline{\mathcal{D}}_{+}, which, by implication, mirrors the behavior observed for negative divergence.

StSt0.050.5151050
D‾+(α=0.01)\overline{D}_{+}(\alpha= 0.01)0.060.41.59.215.19.0
D‾+(α=0.22)\overline{D}_{+}(\alpha= 0.22)0.060.41.59.517.439.8
D‾+(α=0.48)\overline{D}_{+}(\alpha= 0.48)0.060.41.410.924.6963.0

Table 3. Mean of positive divergence D‾+\overline{D}_{+} for different Stokes numbers and different α\alpha values.

StSt0.050.5151050
σD2(α=0.01)\sigma_D^2(\alpha= 0.01)0.0091.110.7110631531087
σD2(α=0.22)\sigma_D^2(\alpha= 0.22)0.0091.010.61214412615902
σD2(α=0.48)\sigma_D^2(\alpha= 0.48)0.0091.09.31632752931319

Table 4. Variance of divergence σD2\sigma_D^2 for different Stokes numbers and different α\alpha values.

For low Stokes numbers (St=0.05,0.5,1St = 0.05, 0.5, 1), the mean of positive divergence (D‾+\overline{\mathcal{D}}_{+}) remains relatively consistent regardless of the ionization state, suggesting that the charge state has a minimal impact on the divergence behavior of the impurities in this range. However, as we examine higher Stokes numbers (St=5,10,50St = 5, 10, 50), a clear trend emerges: both the variance of divergence (σD2\sigma_{\mathcal{D}}^{2}) and the mean of positive divergence (D‾+\overline{\mathcal{D}}_{+}) increase with higher α\alpha values. This observation implies that at higher StSt values, the divergence behavior of the impurities is influenced by their charge state.

References

  1. [1]Benkadda, S., Gabbai, P., Tsytovich, V.N. & Verga, A. 1996 Nonlinearities and instabilities of dissipative drift waves in dusty plasmas. Phys. Rev. E. 53 (3), 2717.
  2. [2]Bos, W. J. T., Kadoch, B., Neffaa, S. & Schneider, K. 2010 Lagrangian dynamics of drift-wave turbulence. Physica D 239, 1269–1277.DOI
  3. [3]Canuto, C., Hussaini, Y.M., Quarteroni, A. & Zang, T.A. 2007 Spectral methods: fundamentals in single domains. Springer Science & Business Media.DOI
  4. [4]del Castillo-Negrete, D., Carreras, B. A. & Lynch, V. E. 2004 Fractional diffusion in plasma turbulence. Phys. Plasmas 11, 3854–3864.
  5. [5]Ferenc, J. S. & Néda, Z. 2007 On the size distribution of Poisson Voronoi cells. Physica A 385 (2), 518–526.arxiv.org/abs/cond-mat/0406116
  6. [6]Futatani, S., Benkadda, S. & del Castillo-Negrete, D. 2009 Spatiotemporal multiscaling analysis of impurity transport in plasma turbulence using proper orthogonal decomposition. Phys. Plasmas 16 (4), 042506.DOI
  7. [7]Futatani, S., Benkadda, S., Nakamura, Y. & Kondo, K. 2008a Multiscaling dynamics of impurity transport in drift-wave turbulence. Phys. Rev. Lett. 100 (2), 025005.pubmed.ncbi.nlm.nih.gov/18232881
  8. [9]Futatani, S., Bos, W. J. T., del Castillo-Negrete, D., Schneider, K., Benkadda, S. & Farge, M. 2011 Coherent vorticity extraction in resistive drift-wave turbulence: Comparison of orthogonal wavelets versus proper orthogonal decomposition. C. R. Phys. 12, 123–131.DOI
  9. [10]Gheorghiu, T., Militello, F. & Rasmussen, J. J. 2024 On the transport of tracer particles in two-dimensional plasma edge turbulence. Phys. Plasmas 31 (1), 013901.DOI
  10. [11]Hasegawa, A. & Wakatani, M. 1983 Plasma edge turbulence. Phys. Rev. Lett. 50, 682–686.DOI
  11. [12]Horton, W. 2012 Turbulent transport in magnetized plasmas. World Scientific.
  12. [13]Kadoch, B., Bos, W.J. & Schneider, K. 2010 Origin of Lagrangian intermittency in drift-wave turbulence. Phys. Rev. Lett. 105 (14), 145001.DOI
  13. [14]Kadoch, B., del Castillo-Negrete, D., Bos, Wouter, J. T. & Schneider, K. 2022 Lagrangian conditional statistics and flow topology in edge plasma turbulence. Phys. Plasmas 29 (102301).DOI
  14. [16]Krasheninnikov, S. I. 2024 On anomalous transport of multi-species plasma associated with the resistive ballooning and resistive drift waves driven turbulence. Phys. Plasmas 31 (5).DOI
  15. [17]Krasheninnikov, S. I., D’Ippolito, D. A. & Myra, J. R. 2008 Recent theoretical progress in understanding coherent structures in edge and SOL turbulence. J. Plasma Phys. 74, 679–717.DOI
  16. [18]Krasheninnikov, S. I. & Smirnov, R. D. 2024 Anomalous transport of multi-species edge plasma with the generalized hasegawa–wakatani model and the lr effects. Phys. Plasmas 31 (5).DOI
  17. [19]Krieger, K., Maier, H. & Neu, R. 1999 Conclusions about the use of Tungsten in the divertor of ASDEX Upgrade. J. Nucl. Mater. 266-269, 207–216.DOI
  18. [20]Lin, Z., Maurel-Oujia, T., Kadoch, B., Krah, P., Saura, N., Benkadda, S. & Schneider, K. 2024 Synthesizing impurity clustering in the edge plasma of tokamaks using neural networks. Phys. Plasmas 31 (3), 032505.DOI
  19. [21]Maget, P. et al. 2020 Natural poloidal asymmetry and neoclassical transport of impurities in tokamak plasmas. Plasma Phys. Control. Fusion. 62, 025001.DOI
  20. [22]Matsuda, K., Schneider, K., Oujia, T., West, J., Jain, S. & Maeda, K. 2022 Multiresolution analysis of inertial particle tessellations for clustering dynamics. Center for Turbulence Research, Proceedings of the Summer Program 2022.
  21. [23]Matsuda, K., Schneider, K. & Yoshimatsu, K. 2021 Scale-dependent statistics of inertial particle distribution in high Reynolds number turbulence. Phys. Rev. Fluids 6, 064304.DOI
  22. [24]Maurel-Oujia, T., Matsuda, K. & Schneider, K. 2023 Computing differential operators of the particle velocity in moving particle clouds using tessellations. J. Comput. Phys. p. 112658.
  23. [25]van Milligen, B. P., Sanchez, R. & Carreras, B. A. 2004 Probabilistic finite-size transport models for fusion: Anomalous transport and scaling laws. Phys. Plasmas 11, 2272–2285.DOI
  24. [27]Naulin, V., Garcia, O. E., Priego, M. & Rasmussen, J. J. 2006 The application of passive tracers for investigating transport in plasma turbulence. Physica Scripta T122, 129–134.DOI
  25. [28]Oujia, T., Matsuda, K. & Schneider, K. 2020 Divergence and convergence of inertial particles in high Reynolds number turbulence. J. Fluid Mech. 905, A14.
  26. [29]Pitts, R. A. et al. 2009 Status and physics basis of the ITER divertor. Phys. Scr. T138, 014001.DOI
  27. [30]Priego, M., Garcia, O. E., Naulin, V. & Rasmussen, J. J. 2005 Anomalous diffusion, clustering, and pinch of impurities in plasma edge turbulence. Phys. Plasmas 12 (6).DOI
  28. [31]Provenzale, A. 1999 Transport by coherent barotropic vortices. Ann. Rev. Fluid Mech. 31 (1), 55–93.DOI
  29. [32]Pushkarev, A. V., Bos, W. J. T. & Nazarenko, S. V. 2013 Zonal flow generation and its feedback on turbulence production in drift wave turbulence. Phys. Plasmas 20, 042304.DOI
  30. [33]Stangeby, P.C 2000 The Plasma Boundary of Magnetic Fusion Devices. CRC Press.
  31. [34]Wesson, J. & Campbell, D.J. 2011 Tokamaks. Oxford University Press.
  32. [35]Wilkinson, M. & Mehlig, B. 2005 Caustics in turbulent aerosols. Europhys. Lett. 71 (2), 186–192.DOI
  33. [36]Yao, D. M. et al. 2016 Design, R&D and commissioning of EAST Tungsten divertor. Phys. Scr. T167, 014003.DOI
  34. [37]Zhang, Y., Guo, Z. B. & Diamond, P. H. 2020 Curvature of radial electric field aggravates edge magnetohydrodynamics mode in toroidally confined plasmas. Phys. Rev. Lett. 125 (25).DOI

Paper details

vlnc:W_edpnpkwa
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