Abstract
Thermal effects on the perpendicular convection of seeded pressure blobs in the scrape-off layer of magnetised fusion plasmas are investigated. Our numerical study is based on a four field full-F gyrofluid model, which entails the consistent description of high fluctuation amplitudes and dynamic finite Larmor radius effects. We find that the maximal radial blob velocity increases with the square root of the initial pressure perturbation and that a finite Larmor radius contributes to highly compact blob structures that propagate in the poloidal direction. An extensive parameter study reveals that a smooth transition to this compact blob regime occurs when the finite Larmor radius effect strength, defined by the ratio of the magnetic field aligned component of the ion diamagnetic to the vorticity, exceeds unity. The maximal radial blob velocities agree excellently with the inertial velocity scaling law over more than an order of magnitude. We show that the finite Larmor radius effect strength affects the poloidal and total particle transport and present an empirical scaling law for the poloidal and total blob velocities. Distinctions to the blob behaviour in the isothermal limit with constant finite Larmor radius effects are highlighted.
Export citation and abstract BibTeX RIS
1. Introduction
In magnetised fusion plasmas the exhaust of particles and heat critically determines the lifetime of the plasma facing components. Intermittent outbreaks of spatially localised structures [1, 2] may damage the divertor plates and the surrounding wall [3–5]. These structures, often referred to as blobs or filaments, are born in the edge in the vicinity of the last closed flux surface (LCFS) and are expelled into the scrape-off layer (SOL).
The strong magnetic field separates the spatial scales of filaments to the extent that they are strongly elongated along the magnetic field but spatially localised perpendicular to it. Typically their cross-field size σ is on the order of cm, the perpendicular ion Mach number is subsonic and the relative fluctuation levels of particle density and temperature in the SOL are of order unity [6-8]. In the SOL parallel heat conduction cools electrons much faster than ions and leads to ion to electron background temperature ratios in the range of [9, 10].
The radial transport caused by blobs depends on their production rate and the blob regime. A blob regime is a parameter space, in which a particular physical mechanism dominates the dynamics of the blob. The various blob regimes strongly depend on the blobs' cross-field size and collisionality [11, 12]. In all regimes it is the magnetic field inhomogeneity that generates perpendicular energy. The mechanism for radial motion is based on the curvature and drifts, which induce an electrical field in the poloidal direction. The resulting drift moves the blob radially outward. One particular regime is the so called inertial or resistive ballooning regime, predominantly emerging when the filament is disconnected from the divertor plates due to high SOL collisionality [13, 14], high magnetic shear in the vicinity of the X-point or electromagnetic effects [15]. In the inertial regime the basic blob dynamics are captured by two dimensional models that invoke the cold ion temperature and thin-layer (Boussinesq) approximation [16-18]. The numerically observed structure of a seeded pressure blob with Gaussian shape is plume like, which is in accordance with cold ion plasma experiments [19]. The relaxation of the thin-layer approximation allowed the treatment of realistic fluctuation levels and supported the results of former studies [20-23]. However, in hot ion plasmas spatially compact blobs are observed experimentally [8, 24-26]. This blob structure can be ascribed to the finite Larmor radius (FLR) of the ions, which was demonstrated numerically by isothermal gyrofluid models for small [27] and high fluctuation amplitudes [28]. For the sheath connected regime this was confirmed by reduced drift-fluid models [29-31].
Blobs are localised pressure perturbations, which in general involve particle density and temperature variations. Due to the relatively high ion to electron background temperature ratios , ion temperature perturbations are in particular expected to have a profound influence on blob dynamics. Consequently we extend the former isothermal gyrofluid work [27, 28] to include ion and electron temperature dynamics and dynamic FLR effects, which account for variations in magnetic field and temperature. We highlight differences in blob propagation, particle transport and scaling to the isothermal model. Our model does not adopt the thin-layer approximation and is energetically consistent. Parallel dynamics and sheath effects have been studied by means of three dimensional numerical simulations [32-37] but are neglected here. Thus, our work is relevant for blobs in the inertial regime.
Our numerical investigation reports that finite temperature variations increase the maximal radial and total centre of mass blob velocity. Finite ion temperature results in a compact blob structure with significant poloidal velocity. In an extensive parameter study we find that the maximal radial centre of mass velocities are in excellent agreement with the inertial velocity scaling law . Here, and are the electron and ion pressure amplitudes, R is the radial distance to the outboard mid-plane, ne0 is the background electron density and mi is the ion mass. We quantify the FLR effect strength by the ratio of the magnetic field aligned component of the ion diamagnetic vorticity to the magnetic field aligned component of the vorticity to show that a smooth transition to a compact blob regime occurs for a ratio of unity. With the help of the FLR strength parameter an empirical poloidal and total velocity scaling law is deduced and verified. Here, is the drift scale with reference magnetic field magnitude B0 and ion particle charge e.
The remainder of the manuscript is organised as follows. We start with the derivation of the gyrofluid model equations in section 2. In this section we derive also an energy theorem and the vorticity density equation with its associated scaling laws. The described thermal gyrofluid model is solved numerically with seeded blobs as initial condition. The numerical results of our parameter study are presented in section 3. Here, we lay the focus of the discussion on thermal effects on propagation, compactness and particle transport of the blob. The results are discussed and summarised in section 4.
2. Gyrofluid model
Our approach relies on a full-F gyrofluid model [38] for subsonic flows, which emerges by taking the gyrofluid moments over the gyrokinetic Vlasov–Maxwell equations [39]. This reduces the dimensionality and consequently the computational cost drastically. No distinction is made between the dynamical background and the low-frequency fluctuations. The ratio of gradient length-scale to the ion gyro-radius may approach unity but polarisation effects are taken in the long wavelength limit (LWL) . Finite Larmor radius (FLR) effects arise naturally in gyrofluid models resulting in a rather simple set of equations. FLR-corrected drift-fluid models are recovered when the gyrofluid model is taken in the LWL [40].
In the following we derive a four field full-F gyrofluid model for an electrostatic magnetised plasma. The model is derived for a 2D slab geometry, where the magnetic field is varying in the radial direction and is perpendicular to the 2D slab. Collisional terms and parallel dynamics are discarded whereas we keep all nonlinearities in order to obey the energy theorem and to get a proper description of the dynamics even for high fluctuation amplitudes. This results in equations for a gyro-centre density N and perpendicular gyro-centre temperature T, which are coupled via the nonlinear polarisation equation. Due to the neglect of parallel dynamics our gyrofluid model applies to the inertial regime.
2.1. Polarisation equation
The polarisation equation is the gyrofluid expression of quasi neutrality and reads
where ne is the electron density, Ni is the ion gyro-centre density, ϕ is the electrostatic potential and is the ion gyro-frequency. We note here that as a consequence of the gyro-centre transformation [39] the ion gyro-centre density Ni is not to be confused with the ion particle density . The perpendicular gradient is defined by and the unit vector in the magnetic field direction by . The mass ratio between the electrons and ions is small, permitting us to neglect electron FLR and finite electron inertia effects. This casts the electron gyro-centre variables into common fluid variables . The sum of the electron density ne and the FLR corrected ion gyro-centre density Ni is referred to as the nonlinear polarisation charge density (right hand side of equation (1)). In the polarisation equation (1) no thin-layer approximation is applied and the full nonlinear polarisation charge density is retained. Here, and in all other abbreviations the subscript '0' denotes a constant background quantity. We note that FLR corrections to the polarisation density are included in delta-f gyrokinetic and gyrofluid models [27], but are omitted in full-F gyrokinetic as well as full-F gyrofluid models. Therefore, even if we invoked the thin-layer approximation to the full-F model, it would only agree with delta-f gyrofluid models in the LWL [28]. In gyrofluid models the ion polarisation drift enters via the nonlinear polarisation charge density. The gyroaveraging operators appear as the Padé approximants [41]
Note here the temperature and magnetic field dependence of the ion gyro-radius , which introduces dynamic FLR effects . In contrast to isothermal gyrofluid models with constant FLR effects [28] the operator is no longer self-adjoint and the operator appears in the gyrofluid moment equations due to spatial variations in the temperature or magnetic field
The FLR corrected ion gyro-centre density in the polarisation equation (1) is the averaged charge contribution at a position from gyro-centres whose gyro-orbits intersect . In case of constant FLR effects the mean gyro-orbit is constant and these gyro-centres are placed on a circle with constant radius at position . On the contrary, for dynamic FLR effects the mean gyro-orbit is spatially varying, so that these gyro-centres are placed on an arbitrary curve. For the simple case of a Gaussian ion gyro-centre pressure blob this curve reduces to a circle with radius or a deformed circle at position , which is shown in figure 1. As a consequence the curves sample different ion gyro-centre densities for constant or dynamic FLR effects. This, in comparison to constant FLR effects, effectively increases or decreases the polarisation charge density in the blob centre or edge respectively.
Due to the neglect of electron FLR effects the gyroaveraging operators for electrons reduce to and .
2.2. Gyrofluid moment equations
The time evolution of a gyro-centre density N is governed by the zeroth gyrofluid moment over the gyrokinetic Vlasov equation
Here, the gyrofluid drifts (denoted with capital U) account for the drive by the drift , the interchange drive by the drift and FLR corrections to the gyro-averaged electric potential due to variations in the temperature or magnetic field by :
Note that as a consequence of the choice of our slab geometry, which is introduced later in this section, no curvature drift appears in the gyrofluid moment equations. In equations (5) and (7) we introduced the gradient of temperature and magnetic field , the generalised potential and the drift velocity :
Here, q is the particle charge. Since the gyrokinetic Lagrangian is taken in the LWL, polarisation effects are retained in the generalised potential ψ of equation (9) only via the energy term [38]. The dissipative term in the gyrofluid moment equation (4) is a hyperdiffusive term of second order . The second gyrofluid moment yields the partial differential equation for a perpendicular gyro-centre pressure P. The equation is rewritten with the help of the zeroth gyrofluid moment equation (4) into an evolution equation for a perpendicular gyro-centre temperature T:
Here, the gyrofluid drift captures FLR corrections to the drift , which emerge due to spatial variations in the gyro-centre temperature T or the magnetic field B:
In the second gyrofluid moment equation (11) a hyperdiffusive term of second order is added, which together with ensures numerical stability.
Our gyrofluid model employs right-handed Cartesian coordinates (x, y, z) and a slab magnetic field with a radially varying magnitude . Here, R is the radial distance to the outboard mid-plane and B0 is the reference magnetic field magnitude. In this simple geometry a more accessible representation of the electron and ion gyrofluid moment equations (4) and (11) is obtained by expressing the set of equations with the help of Poisson brackets and curvature operators. The slab geometry approximation reduces the curvature operator to and the Poisson bracket to . We rearrange the terms in the gyrofluid moment equations (4) and (11) to highlight the highly nonlinear nature of our system:
Here, we introduced . In the isothermal limit the temperatures are constant and second order FLR effects vanish so that and . From this we obtain the set of full-F gyrofluid equations, which was studied in [28].
2.3. Energy theorem
The energy of our system can be determined by integrating the sum of the zeroth gyrofluid moment equation multiplied by the factor and the second gyrofluid moment equation over the domain. We obtain the energy theorem for electrons and ions by neglecting boundary terms:
The energy consists of the internal (thermal) energy densities for electrons and ions and the energy density. The dissipative terms enter the energy theorem via . The derived energy theorem resembles those in [38, 42].
2.4. Vorticity density equation
In order to understand the relation between gyro-centre and particle fields and to show the correspondence to drift-fluid models explicitly we derive the inherent vorticity density equation of the presented full-F gyrofluid model in the LWL [40]. Thus we first invert the polarisation equation (1) and apply the LWL. This yields an expression for Ni, which depends on the physical meaningful variables ne, ti and ϕ [42]:
Here, we defined the sum of the vorticity plus half of the ion diamagnetic vorticity as
The generalised vorticity is the sum of the magnetic field aligned component of the vorticity and the ion diamagnetic vorticity. We note here that the vorticity consists of the magnetic field aligned component of the common vorticity plus a cross term . Analogously the magnetic field aligned component of the ion diamagnetic vorticity consists of and a cross term The generalised vorticity density is given by . The LWL vorticity density equation is derived by taking the material derivative over the generalised vorticity density
where we have neglected FLR corrections in curvature operators and nonlinear terms that fall into the LWL. Dynamic FLR effects arise from the ion diamagnetic contribution of the generalised vorticity density and the second term on the right hand side of the vorticity density equation (21). The dissipation of the generalised vorticity density is represented by , which is discussed in detail in appendix. We note here that the Prandtl number and Schmidt number are fixed to unity. Hence, the mass diffusivity , the thermal diffusivity and the kinematic viscosity reduce to a single dissipative parameter ν. In the absence of dissipative terms we can rewrite the vorticity density equation (21) to a form, which equals the divergence of the ion polarisation flux. It reads
and allows a comparison to FLR-corrected drift-fluid models [43, 44]. Here, we defined .
2.5. Initialisation
Since gyro-centre fields match the particle fields only to zeroth order in and , an accurate initialisation procedure is required in order to avoid any residual vorticity. Hence, our initialisation routine relies on transformations from the gyro-centre fields Ni, Ti, to the particle fields , ti, te, without the generation of artificial initial vorticity. For the ion gyro-centre density and perpendicular temperature we mimic an initial blob by a Gaussian of the form
with initial amplitude A of the gyro-centre fields Ni and Ti. In order to initialise with zero vorticity we take the polarisation equation (1) and the equation for pi [42] in the limit . This yields the following initial particle fields
The initial electron pressure is related to ion pressure via so that the initial electron temperature is given by . We note here that the initial amplitudes and cross-field sizes of the particle fields ne, te and ti may differ from the initial parameters A and σ of the gyro-centre fields Ni and Ti due to the manifestation of FLR effects. Hence, the true initial amplitudes , and are taken into account for the derivation of quantities like the global interchange rate , which is defined in the next section 2.6. For the size of the blob we take the initial parameter σ of the gyro-centre fields.
2.6. Interchange dynamics
Now, we use the derived vorticity density equation (21) to obtain an estimate for the maximal perpendicular blob velocity . For this sake we first employ the thin-layer approximation on equation (21) and neglect all terms, which are related to FLR effects or dissipation. Our scaling analysis relies on the ideal interchange rate γ and the blob cross-field size σ as typical time and spatial scale, which results in the typical velocity scale . The characteristic scale of the generalised vorticity and perpendicular pressure is the ideal interchange rate γ and the total scalar pressure perturbation . Following [12] we obtain the explicit expression for the ideal interchange rate . We note that an initial electron and ion temperature perturbation enters the initial electron and ion pressure perturbation according to , which in the isothermal limit reduces to .
Now, we postulate the factor 'ex post' as a nonlinear correction factor to the blob size σ and the ideal interchange rate γ. As a result the typical spatial and time scale is the effective blob size and the global interchange rate [28]
This form of scaling fits our simulation results best. With the help of the global interchange rate we obtain the scaling law for the maximal perpendicular blob velocity
The derived velocity scaling estimate of equation (28) differs for high electron density amplitudes from the so called global scaling as reported in [28, 45]. However, in [22, 23] the amplitude scaling of equation (28) was verified numerically for high fluctuation amplitudes in the cold ion temperature and isothermal limit. In the isothermal limit our derived velocity scaling law agrees up to a factor with the inertial velocity scaling law of [12]. This is reasoned in the low beta approximation of the curvature operator. In this case the magnetic curvature is , whereas for a slab magnetic field the magnetic curvature is . Note that equation (28) also coincides with [17] up to a factor 2.
Now, we define a parameter, which is a measure of the FLR strength in the inertial regime. In order to do so we take the ratio of the magnetic field aligned component of the ion diamagnetic vorticity to the estimate for the magnetic field aligned component of the vorticity in the inertial regime. This defines the FLR strength parameter as oder
We note here that in the FLR strength parameter the ion diamagnetic vorticity stems from the LWL over the gyroaveraging operator .
3. Numerical experiments
The gyrofluid model, which consists of equations (1), (13)–(16), is numerically solved by the FELTOR library [46]. This library relies on discontinuous Galerkin (dG) methods to discretise perpendicular [47] and parallel spatial derivatives [48]. DG methods are very versatile in the choice of the desired order of accuracy and retain a high degree of parallelism in the resulting algorithm. FELTOR exploits this on shared as well as distributed memory systems and efficiently executes on CPUs and GPUs. This is required by the high computational cost of the underlying model.
We resolve our box of size by P = 3 polynomial coefficients and grid cells to ensure converged simulations. Boundary conditions of the fields are periodic in y and Dirichlet in x direction. We verified our code with the help of the energy theorem of equation (17) and by performing convergence tests in the L2-norm for all implemented numerical operators.
Our dimensionless set of equations employs the Bohm normalisation with drift scale , reference ion gyrofrequency and cold ion acoustic speed so that our variables are transformed to , , , , , , and .
The physical parameters match those of an exemplary SOL on the low field side of the ASDEX Upgrade (AUG) tokamak. Here, the electron temperature is and the magnetic field magnitude is B0 = 2T, which determines the drift scale to for a deuterium plasma. The remaining parameters are , , and enter the dimensionless equations via the curvature parameter . The parameter range of the ion background temperature is and the initial amplitude and width of the blob is A = {0.1, 0.5, 1, 2} and . The ratio between the effective gravity to the dissipative forces is known as the Rayleigh number . For unity Prandtl number we get in the isothermal case and if we include temperature dynamics (). We fix the Rayleigh number to in all our simulations. This determines the viscosity to be in the turbulent high Reynolds number regime, where no impact on the maximal velocity of the blob is expected [17, 49].
In the following we start with the description of the nonlinear evolution of thermal blobs in section 3.1. A detailed comparison of the propagation of isothermal and thermal blobs is given in section 3.2. The impact of the temperature dynamics on the persistence of the initial blob structure is given afterwards in section 3.3. In the end we focus on thermal effects on the particle transport of blobs (section 3.4).
3.1. Nonlinear evolution of thermal blobs
In the cold ion limit a finite electron pressure gradient induces an -vorticity dipole due to the drift. This accelerates the blob into the radial direction [18]. A steep and vertically extended pressure front develops finally into an up–down symmetric mushroom shape. This symmetry is broken in the late purely nonlinear phase due to the up–down asymmetric nature of the gyrofluid model. We depict this behaviour for electron density ne, electron temperature te and the vorticity of an exemplary cold blob in figure 2.
Download figure:
Standard image High-resolution imageFinite ion temperature effects change the behaviour of the blob fundamentally [27, 28]. We can see this by comparing the blob structures and positions of figures 2 and 3. FLR corrections arise, which contribute to the polarisation charge density in equation (1) and advect electrons and ions with different drifts and (see equations (10) and (5)). Moreover FLR effects manifest themselves by additional drifts (see equations (12) and (7)) in the gyrofluid moment equations and by the ion diamagnetic vorticity in equation (20). In the initial phase a tilted -vorticity dipole is generated by a finite ion diamagnetic vorticity and the term in the generalised vorticity density equation (21) (see figure 4 and [27]). Hence, the motion of the blobs is no more purely radial and an additional poloidal motion in the direction is observed. Subsequently, the tilted -vorticity dipole rolls itself up and FLR effects lead to the aligning of vorticity with ion pressure pi. The shear flow produces a differential rotation, which is strongest at the leading edge of the blob. This effect helps to retain the initial rotational symmetry of the blob and decreases the separation of small eddies. As a consequence the blob tends to retain its initial Gaussian structure. This is shown for electron density ne, ion temperature ti and the vorticity of an exemplary hot blob in figure 3. The electron temperature fields match to a great extent the electron density fields.
Download figure:
Standard image High-resolution imageDownload figure:
Standard image High-resolution image3.2. Thermal effects on the propagation of a blob
In this section we investigate the propagation of isothermal and thermal blobs for various ion temperatures, cross-field sizes and initial amplitudes. The position of the blob is tracked either by its centre of mass (COM) or by its maximal electron density amplitude. We define the COM of a blob by
The COM displacement is associated with the time integrated perpendicular particle transport via
Here, we introduced the perpendicular particle transport . The position of the maximal electron density amplitude is defined implicitly through and differs from the COM position after the initial acceleration phase. It is a good measure for the position of the blob front. The COM position and the position of the maximal electron density amplitude are visualised in figures 2 and 3. The displacement of the maximal electron density amplitude is defined by .
3.2.1. Cold ions.
As illustrated in section 3.1 the COM motion of cold ion temperature () blobs is purely radial before the up–down symmetry is broken. For differences in the propagation of isothermal and thermal blobs should be described by our scaling, which we used to derive the inertial velocity scaling law of equation (28). Figure 5(a) shows that the radial COM displacement for a given time t is increased by the initial amplitude A. In order to quantify the amplitude scaling of the radial COM displacement , we normalise the radial COM displacement by the effective blob size and the time by the ideal interchange rate . This must then result in nearly overlapping plots if the velocity scaling estimate of equation (28) is correct. This is shown in figure 5(b) and reveals that the radial COM velocities obey the velocity scaling estimate of equation (28) in the cold ion limit. Figure 5(b) also shows that in the initial phase the radial displacement experiences a constant acceleration phase with , which is followed by a constant velocity phase with . Hence, the time integrated radial particle transport depends in both time phases on the initial electron pressure perturbation . The initial electron pressure amplitude is proportional to for a thermal blob whereas it scales like for an isothermal blob. Consequently, thermal blobs travel faster radially than isothermal blobs as depicted by figure 5(a). We discuss the implications of the COM displacement for the (non time integrated) perpendicular particle transport in section 3.4.
Download figure:
Standard image High-resolution image3.2.2. Hot ions.
Finite ion temperature effects () redistribute the total COM momentum into the radial and poloidal directions. An ion temperature perturbation enhances the total (radial and poloidal) COM displacement compared to isothermal blobs. This behaviour is shown for the radial COM displacement in figure 6(a). The normalisation by the effective cross-field size and the ideal interchange rate is depicted in figure 6(b). This figure shows that the normalised radial COM displacement of isothermal and thermal blobs nearly coincide in the constant acceleration and velocity phases. Hence, the radial COM velocities Vc, x are captured by the inertial velocity scaling estimate of equation (28). Since this estimate is only suitable for the radial COM velocities, modifications for total and poloidal velocity estimates are required. In section 3.4 we show that the total and poloidal velocity scaling laws can be adjusted with the help of the FLR strength parameter .
Download figure:
Standard image High-resolution imageIn the following we track the radial and poloidal COM position and investigate the blob propagation for the complete parameter space.
In figure 7 we clarify the dependence of the COM displacement on the main parameters () for isothermal and thermal blobs. We observe that the radial and poloidal COM displacement and increases with amplitude A over the complete parameter space. This is in accordance with the scaling estimates of the radial COM displacement , presented in section 3.2.1, and holds similarly for the total COM displacement . The poloidal COM displacement is decreased by the cross-field size σ and is also enhanced by the ratio of ion to electron background temperature . In addition, a finite ion temperature perturbation enhances the poloidal COM displacement especially for high amplitudes A in comparison to isothermal blobs with constant ion temperature ti0. This is in line with the FLR effect strength parameter of equation (29) and is further discussed in section 3.4.
Download figure:
Standard image High-resolution imageWe ascribe the small increase of poloidal motion to dynamic FLR effects. We do so because a finite ion temperature perturbation together with dynamic FLR effects result in the FLR effect strength parameter . This exceeds its isothermal, constant FLR effect limit equivalent for all studied parameters. This difference stems from either constant or dynamic FLR effects in the polarisation equation (1). In case of ion temperature variations dynamic FLR effects appear in the polarisation equation (1) via the gyroaveraging operator whereas in the isothermal limit constant FLR effects enter the polarisation equation (1) via the gyroaveraging operator . In order to ensure that the poloidal motion depends on constant or dynamic FLR effects, we repeated the simulations with temperature dynamics but constant FLR effects . The related numerical results and the influence of FLR effects on the poloidal particle transport of a blob is presented in section 3.4.
3.3. Thermal effects on blob compactness
The ability of a blob to retain its initial shape is given by the blob compactness [27]
with the Heaviside function
and with defined as the position of the maximum electron density amplitude. The lowering of compactness is accompanied by the loss of mass via collisional dissipation and turbulent mixing and stretching whereas a finite Larmor radius contributes to long lived coherent compact structures. Consequently, from figures 2 and 3 we expect low compactness for cold blobs and high compactness for hot blobs.
The maximal radial particle transport of a blob is determined by the blob compactness and the maximal radial velocity. As a result we plot the blob compactness at the time , which approximately coincides with the first maximum of the radial velocity, versus the FLR effect strength (see equation (29)) in figure 8. Note that for high the maximal radial velocity is often reached at the second peak in time (see e.g. figure 11), which is accompanied by a small drop in compactness IC. In figure 8 a smooth transition to highly compact blobs is obtained at . Above this threshold typically only 10 percent of the initial particle density is lost whereas below the threshold up to 50 percent of the initial particle density is dissipated. For constant the transition to the ion diamagnetic vorticity dominated regime () is reached by decreasing the cross-field size σ or increasing the amplitude of the blob. The thermal and isothermal blobs approach similar values of compactness IC over the complete range. We note here that for the sake of clarity the zero ion temperature value was replaced by in order to compute a finite FLR strength .
Download figure:
Standard image High-resolution imageOn the other hand, the FLR strength parameter of thermal blobs is larger than of isothermal blobs for constant cross-field size σ, amplitude A and ion to electron background ratio . As an example for , A = 0.5 and the FLR strength parameter is for thermal blobs and for isothermal blobs. Consequently, figure 8 shows that thermal blobs remain more compact than isothermal blobs. This is underlined by figure 9, which reveals that thermal blobs with dynamic FLR effects lose less mass than thermal or isothermal blobs with constant FLR effects. This is due to differences in the polarisation charge density for constant and dynamic FLR effects, which is discussed in section 2.1. Since the polarisation charge density is associated to the vorticity, dynamic FLR effects lead to stronger shear flows in the blob edge in comparison to constant FLR effects. These shear flows help to retain the initial blob shape (see section 3.1). As a result the compactness of blobs with a dynamic gyro-radius is higher than of blobs with a constant gyro-radius.
Download figure:
Standard image High-resolution image3.4. Thermal effects on the particle transport of a blob
In the following we use the COM velocity as a measure for the perpendicular particle transport
The COM velocity is given by:
In the remainder of this section the influence of the ratio of ion to electron background temperature , cross-field size σ and initial amplitude A on the transport of isothermal and thermal blobs is studied.
3.4.1. Cold ions.
In case of zero ion temperature the radial and total transport coincide. In figure 10 the maximal COM velocity is reached after an initial constant acceleration phase and is captured within twenty percent by the inertial velocity scaling law of equation (28). The time period to reach the maximal COM velocity scales with the global ideal interchange rate of equation (27).
Download figure:
Standard image High-resolution image3.4.2. Hot ions.
Finite ion temperature induces a poloidal COM velocity in the direction. The poloidal COM velocity Vc, y is generated by FLR effects and is further decreased (see figure 7) if dynamic FLR effects arise as a consequence of gradients in ion temperature ti or magnetic field B. Consequently, in figure 11(b) thermal blobs with full FLR dynamics show increased absolute values for the normalised minimal poloidal COM velocities in comparison to blobs with constant FLR effects especially for high amplitudes.
Download figure:
Standard image High-resolution imageThe behaviour of the normalised radial COM velocities of thermal blobs is similar to blobs with constant FLR effects as depicted in figure 11(a). The maximal normalised radial COM velocities are well described by the velocity scaling estimates of equation (28). As in the cold ion case the maximal radial COM velocities of thermal blobs exceed those of isothermal blobs, which is shown in figure 12. However, the ion temperature scaling for the maximal radial and total COM velocity (see figure 13) is in line with the theoretical estimate of equation (28) for blobs with high cross-field size . For smaller blob sizes and we observe approximately a decreased scaling and , respectively.
Download figure:
Standard image High-resolution imageDownload figure:
Standard image High-resolution imageIn figure 14 the time dependence of the normalised total COM velocities are plotted. The maximal total COM velocities Vc agree with the velocity scaling estimates of equation (28) but are slightly underestimated by the velocity scaling estimates of equation (28) in case of dynamic FLR effects. The time period to reach this maximum is enhanced for high blob amplitudes A.
Download figure:
Standard image High-resolution imageTo verify the inertial velocity scaling law we plot the measured maximal radial COM velocities against the analytical estimates of equation (28). Before we do so we note that all the measured velocities are in the inertial regime. The ion pressure dominated RB (iRB) velocity estimate of [12] is not confirmed by our simulations even if some of the parameters fall into this regime. The iRB velocity underestimates the numerically obtained maximal COM velocities, which is in accordance with [50]. In figure 15 we prove numerically that the theoretical velocity estimate of the inertial regime (see equation (28)) is recovered over more than an order of magnitude for the thermal and isothermal blobs. All the derived maximal radial COM velocities lie within 20 percent of the estimate, which underlines the usefulness and quality of the inertial scaling law. We tested also the global amplitude velocity scaling of [28, 45]. However, we obtained distinct deviations from the numerically obtained maximal COM velocities by the global inertial scaling law . Hence, we emphasise that the inertial velocity scaling law of equation (28) captures the maximal radial COM velocities best.
Download figure:
Standard image High-resolution imageIn sections 3.2-3.4 we discussed the parameter defined in equation (29), which is a measure of the influence of FLR effects on blob convection. Specifically we argued that the poloidal and consequently the total transport is related to the FLR strength parameter . We will now derive an empirical scaling law for the poloidal and total particle transport based on the knowledge of our parameter study. The tracks of the COM position in figure 7 indicate that for high FLR strength (equation (29)) an angle of is taken by the blobs. Hence, we postulate that this angle is approached by
which is a shifted function with shift parameter . The angle α captures the blob compactness at up to the constants c1 and c2. We find that the constants , and the shift parameter fit the blob compactness with a threshold of (see equation (34)). With equation (37) we derive the total and poloidal velocity scaling law
In figures 16 and 17 we compare the derived total and poloidal inertial scaling law V of equation (38) and Vy of equation (39) with the numerically obtained maximal total COM velocities and the numerically obtained minimal poloidal COM velocities . For the total blob velocities we find again agreement up to percent. On the other hand, the minimal poloidal COM velocities of the thermal and isothermal blobs are not always within 20 percent of the theoretical estimate of equation (39). This occurs especially for very small poloidal velocities. Still, figure 17 underlines the quality of the empirical scaling law.
Download figure:
Standard image High-resolution imageDownload figure:
Standard image High-resolution image4. Discussion and conclusion
Our simulations reveal that variations in the temperature field increase the radial COM blob velocities. The maximal radial COM blob velocities obey the inertial velocity scaling law of equation (28) over the complete parameter scan, which encompasses a broad spectrum of typical blob amplitudes , ratios of ion to electron background temperatures and blob cross-field sizes . This coincidence occurs even in the limit of constant temperatures for the complete parameter space. On the other hand, the total and poloidal particle transport and the compactness of blobs is crucially determined by the FLR strength parameter of equation (29). This parameter describes the transition from the weak to the strong FLR effect regime. For weak FLR effects () we find plume-like blob structures in particle density, which primarily travel in the radial direction with maximal radial COM velocities captured by the inertial velocity scaling law of equation (28). The time to reach the maximal radial COM velocity is well described by the global interchange rate of equation (27). For strong FLR effects () compact radially and poloidally moving blob structures are observed in our numerical study. The inertial velocity scaling law of equation (28) estimates the maximal radial COM velocity over the complete range. In the limit the numerical results suggest that the absolute values of the radial and poloidal COM velocities are equal, leading to the empirical poloidal scaling estimate of equation (39). The overall perpendicular particle transport is affected by the inherent particle density of a blob, which in the strong FLR effect regime is roughly 50 percent higher than in the weak FLR effect regime. The radial particle transport scales with the square root of the total pressure perturbation (see equation (28)) and thus increases for finite ion temperature.
To gain insight into the transition between various blob regimes and the influence of poloidal particle transport on the divertor heat load fully three dimensional computations in X-point geometry are in process of planning. Future plans also involve the study of thermal effects in the sheath connected regime and the inclusion of FLR corrections to the polarisation density. According to [28] those corrections may slightly enhance the propagation of blobs with high ratios of ion to electron background temperature, small pressure amplitudes and small blob widths.
Acknowledgments
The authors want to thank R. Kube for contributing to this work with fruitful comments. This work was supported by the Austrian Science Fund (FWF) Y398. This work has been carried out within the framework of the EUROfusion Consortium and has received funding from the Euratom research and training programme 2014–2018 under grant agreement No 633053. The views and opinions expressed herein do not necessarily reflect those of the European Commission. The computational results presented have been achieved in part using the Vienna Scientific Cluster (VSC).
Appendix.: Dissipation of the generalised vorticity density
In the following we will show that the dissipative terms in the generalised vorticity density equation (21) reduce approximately to a hyperdiffusive term of the form . For the sake of convenience we show the derivation for common dissipative terms of first order and give the result for second order in the end of this section.
The dissipative terms for the gyro-centre densities and temperatures read:
Due to our choice of gyro-centre densities N and temperatures T as dependent variables the ion pressure pi is dissipated according to
In the generalised vorticity density equation (21) the full expression for the dissipation reads
We rewrite now the first two terms of equation (A.3) with the help of equation (19) to
and neglect variations in the magnetic field in the last term of equation (A.3)
Now it is clear that only for the complete dissipation of generalised vorticity density reduces to . This is only possible if the gyro-centre pressure P is evolved instead of the gyro-centre temperature T. With our choice the dissipative term of the generalised vorticity density reduces to:
The derivation with hyperdiffusive terms of second order is analogous and yields
We note here that the artificial dissipative term has no impact on the blob motion in the high Reynolds number regime.