Non-thermal particle acceleration in multi-species kinetic plasmas: universal power-law distribution functions and temperature inversion in the solar corona Preprint: AIP/123-QED Uddipan Banik, Amitava Bhattacharjee Abstract The origin of non-thermal power-law distribution functions ubiquitously observed in astrophysical/space (e.g., the solar wind) and laboratory kinetic plasmas, is not well understood. Another puzzling phenomenon is temperature inversion in the solar corona. These two issues are deeply connected. We develop a self-consistent quasilinear theory (QLT) for electromagnetically driven kinetic plasmas, deriving a Fokker-Planck equation for the simultaneous relaxation of multiple species, with (i) a drive diffusion coefficient for the heating of dressed particles directly by the drive and indirectly by waves, and (ii) Balescu-Lenard diffusion and drag coefficients for internal turbulence and Coulomb collisions. Both electron and ion distributions relax towards a universal attractor with a v − 5 v^{-5} ( E − 2 ) (E^{-2}) tail, akin to a κ = 1.5 \kappa=1.5 distribution, under a super-Debye (but sub-Larmor) drive with a steep power-spectrum. This is an outcome of Debye screening: large-scale fields accelerate the unscreened, fast particles but not the screened, slow ones. The universality may be broken by shallow power-spectra and incomplete relaxation. Collisions cannot decelerate suprathermal particles, rendering a high v v tail immune to Maxwellianization. Such a tail may be generated in the solar corona by chromospheric convection despite collisional losses. The suprathermal particles escape sun’s gravity (velocity filtration), inverting the temperature profile and raising it to 10 6 10^{6} K. A proper analysis of velocity filtration with a κ ≈ 1.5 − 2 \kappa\approx 1.5-2 distribution inspired by QLT provides a reasonable fit to the spectroscopic data of heavy ions and explains the abrupt temperature rise, a consequence of the divergence of pressure in the κ → 1.5 \kappa\to 1.5 limit. I Introduction How collisionless plasmas relax to specific distribution functions (DFs) has been a persistent mystery. Collisions can drive a plasma towards a universal thermal/Maxwellian DF, a consequence of the Boltzmann H-theorem or the second law of thermodynamics. Collisionless (kinetic) plasmas, on the other hand, are governed by the Vlasov equation that admits infinitely many steady-state solutions. However, even in kinetic plasmas, while the fine-grained DF follows the Vlasov equation, the coarse-grained one can be shown to follow a (Balescu-Lenard type) kinetic equation with an effective collision operator whose steady state solution is a Maxwellian. In other words, an effective H-theorem arises from coarse-graining. Yet, kinetic plasmas in nature and numerical simulations commonly harbor non-thermal DFs with Maxwellian cores but extended power-law tails. A particular power-law DF, f_{0}(v)\sim v^{-5} , where v is the (non-relativistic) particle velocity, with a corresponding E^{-2} energy ( E ) distribution ( 18 ) , has been measured in the solar wind, the inner heliosphere by the Ulysses and ACE spacecraft ( 17 ) and the heliosheath by Voyager ( 24 ; 41 ) . The theory proposed to explain the universality of such DFs ( 18 ) has been questioned on both theoretical and observational grounds ( 23 ) , especially because the observed DFs show some variability, either due to collisions or a variety of plasma waves and instabilities not considered in the theory. Recent data by the Parker Solar Probe shows that the power-law exponent falls roughly in the range 5-6 ( \kappa=1.5-2 for a \kappa distribution fit) closer to the sun. To our knowledge, there is currently no rigorous understanding of Nature’s predilection for such power-laws, which calls for an explanation based on kinetic theory. This paper is one of a sequence of recent papers dedicated towards uncovering the rich physics of the relaxation of kinetic plasmas and explaining the universality of non-thermal power-law tails. Our recent self-consistent treatment of electrostatic kinetic plasmas demonstrates the presence an attractor DF, f_{0}(v)\sim v^{-5} , in a plasma relaxing under large-scale electric field fluctuations. Using quasilinear theory (QLT), we showed that this DF is an outcome of the self-consistent plasma response or Debye shielding that naturally yields a v dependent particle acceleration mechanism. Earlier test-particle treatments ( 18 ; 23 ) did not include the self-consistent Maxwell’s equations. Whether this asymptotic steady state is measured by in situ spacecrafts might depend on the collisional age of the plasma and electromagnetic (EM) effects, but the general agreement between our theoretical result and the spacecraft measurements suggests that our treatment captures essential aspects of kinetic plasma relaxation despite certain simplifying assumptions. This power-law tail also emerges in a kinetic plasma subject to electrostatic turbulence arising from the two-stream instability, as recently demonstrated by a particle-in-cell (PIC) simulation ( 15 ) . Simulations of magnetic reconnection and collisionless shocks show similar power-law tails ( 39 ; 21 ; 22 ; 20 ; 44 ) . The implications of non-thermal power-law DFs are potentially far reaching. Such plasma kinetic effects are believed to play a crucial role in heating the solar corona and inverting the coronal temperature profile. Regarding the coronal heating problem, the following questions have puzzled heliophysicists for a long time. Why is the corona significantly hotter than the chromosphere? How is an inverted temperature profile sustained despite the flow of energy from low to high temperatures in apparent violation of the second law of thermodynamics? Why is the temperature gradient so steep in the transition zone between the chromosphere and the corona? In other words, why is the transition region as thin as a few hundred kilometers? It is worth noting that this is even thinner than the solar tachocline that separates the radiative and convective zones. To explain the origin of temperature inversion in the corona, a physical mechanism called velocity filtration ( 37 ) was proposed, which suggests that the solar gravitational field can filter out the high velocity particles from a non-thermal \kappa type distribution at the coronal base and preferentially allow these suprathermal particles to escape outwards. While a Maxwellian distribution entails an isothermal plasma, a \kappa distribution with finite \kappa results in an inverted temperature profile through velocity filtration. Although the underlying assumption of a \kappa distribution is consistent with spacecraft measurements across the heliosphere ( 30 ) , sustaining such a distribution in the moderately collisional environment of the corona is apparently difficult because the collisional mean free path is around two orders of magnitude smaller than the pressure scale height, implying that \kappa distributions might Maxwellianize in this locale. Theory and idealized kinetic simulations ( 2 ; 25 ) demonstrate that, unless shallow power-law tails ( \kappa<4 ) are assumed at the coronal base, which is difficult to justify from the perspective of conventional wisdom, the temperature inversion cannot be explained. It is evident from the discussion above that the issue of steady-state distributions of kinetic plasmas and the coronal heating problem are deeply connected. As an attempt to resolve this problem, we develop a general quasilinear theory (QLT) ( 12 ; 3 ; 5 ) for the relaxation of electromagnetically driven kinetic plasmas. We investigate the simultaneous, self-consistent evolution of multiple species (electrons and ions) using a Balescu-Lenard (BL)-type framework for kinetic plasma turbulence and weak binary collisions. We study the relaxation of kinetic plasmas (evolving from a drive-free initial state), subject to sub-Larmor scale EM turbulence. It should be borne in mind though that only the quiescent region of the corona strictly falls within the regime of validity of QLT; violent instabilities and highly turbulent scenarios are not properly captured by this approach. Using this theory, we show that dielectric polarization or Debye shielding is the key phenomenon behind producing a universal v^{-5} tail in the DF of both electrons and ions and the E^{-2} tail in their energy distribution. The Debye screening of large-scale (super-Debye but sub-Larmor) fields suppresses particle heating at low velocities, which yields a v^{4} diffusion coefficient and a v^{-5} steady-state DF. Depending on the steepness of the power spectrum of EM fluctuations, the power-law exponent can deviate somewhat, albeit within a range that generally agrees with heliospheric observations. Sufficiently steep spectra lead to a universal v^{-5} scaling, which survives even in the presence of weak Coulomb collisions. This key result overcomes some of the fundamental challenges confronting the elegant velocity filtration model as a contender for resolving the coronal heating problem. This paper is organized as follows. In section II , we discuss the linear and quasilinear theories for the relaxation of kinetic plasmas. Using this theory, we derive a general transport equation for non-thermal particle acceleration (NTPA) in section III . We apply this theory and the velocity filtration model to the coronal heating problem in section IV , proposing a possible solution to the problem of temperature inversion in the solar corona. We summarize our findings in section V . II Relaxation theory for weakly collisional plasmas II.1 Governing equations A plasma is characterized by the DF or phase space ( {\bf x},{\bf v} ) density of particles of the {{\rm s}}^{\rm th} charged species, f_{\rm s}({\bf x},{\bf v},t) . The governing equations for a weakly collisional plasma are the Boltzmann-Maxwell equations. The Boltzmann equation, \displaystyle\frac{\partial f_{\rm s}}{\partial t}+{\bf v}\cdot{\bf\nabla}f_{\rm s}+\frac{q_{\rm s}}{m_{\rm s}}\,{\bf\nabla}_{\bf v}f_{\rm s}\cdot\left[\left({\bf E}^{({\rm P})}+{\bf E}\right)+\frac{{\bf v}}{c}\times\left({\bf B}^{({\rm P})}+{\bf B}\right)\right] \displaystyle=C\left[f_{\rm s}\right], (1) evolves the DF of each charged species (electrons and ions), with q_{\rm s} and m_{\rm s} the electric charge and mass of each species. Here, {\bf E} and {\bf B} are the self-generated electric and magnetic fields, sourced by the DF via Maxwell’s equations, \displaystyle\nabla\cdot{\bf E}={4\pi}\sum_{{\rm s}}q_{\rm s}\int{\rm d}^{3}v\,f_{\rm s}, \displaystyle\nabla\times{\bf B}=\frac{4\pi}{c}\sum_{{\rm s}}q_{\rm s}\int{\rm d}^{3}v\,{\bf v}f_{\rm s}+\frac{1}{c}\frac{\partial{\bf E}}{\partial t}, \displaystyle\nabla\cdot{\bf B}=0,\quad\nabla\times{\bf E}=-\frac{1}{c}\frac{\partial{\bf B}}{\partial t}. (2) We assume charge neutrality in equilibrium, i.e., the total equilibrium charge density is zero. {\bf E}^{({\rm P})} and {\bf B}^{({\rm P})} are perturbing electric and magnetic fields that are sourced by plasma “external" to our system, which we hereafter refer to as the drive. Since plasma is an open system, it is nearly impossible to switch off such fields. Background turbulence spontaneously arising from nonlinear wave activity or coherent structures (electrostatic, e.g., Bernstein-Greene-Kruskal (BGK) modes ( 8 ) or electromagnetic, e.g., plasmoids ( 28 ; 9 ; 11 ) ), copiously present within the solar plasma, can act as part of this drive. We assume the collision operator to be of the Balescu-Lenard form, \displaystyle C[f]=\frac{\partial}{\partial v_{i}}\left({\cal D}^{({\rm s})}_{ij}\left({\bf v}\right)\frac{\partial f_{\rm s}}{\partial v_{j}}+{\cal D}^{({\rm s})}_{i}\left({\bf v}\right)f_{\rm s}\right), (3) with the diffusion ( {\cal D}^{({\rm s})}_{ij} ) and drag ( {\cal D}^{({\rm s})}_{i} ) coefficients given by equation ( 15 ). We do not include an equilibrium guide magnetic field in our analysis. In other words, our current analysis strictly only applies to sub-Larmor scale EM fluctuations. The nonlinear Boltzmann-Maxwell equations are difficult to solve in their full generality, and demand the use of perturbation theory to make analytical progress. If the strength of the electrostatic potential, \Phi^{({\rm P})}=-\int{\bf E}^{({\rm P})}\cdot{\rm d}{\bf x} , is smaller than \sigma^{2}_{\rm s} , where \sigma_{\rm s} is the velocity dispersion or thermal velocity of the {\rm s}^{\rm th} species in the unperturbed near-equilibrium system, then the perturbation in f_{\rm s} can be expanded as a power series in the small perturbation parameter, \epsilon\sim\left|\Phi^{({\rm P})}\right|/\sigma^{2}_{\rm s} , i.e., f_{\rm s}=f_{{\rm s}0}+\epsilon f_{{\rm s}1}+\epsilon^{2}f_{{\rm s}2}+...\, ; {\bf E} and {\bf B} can also be expanded accordingly, assuming {\bf E}^{({\rm P})} and {\bf B}^{({\rm P})} to be {\cal O}(\epsilon) . We perform a Fourier transform with respect to {\bf x} and Laplace transform with respect to t of f_{{\rm s}i} , {\bf E}_{i} , {\bf E}^{({\rm P})} , {\bf B}_{i} and {\bf B}^{({\rm P})} in the ( i^{\rm th} order) perturbed Vlasov-Maxwell equations, to derive the response of the system order by order. II.2 Linear theory The Fourier-Laplace coefficients of the linear response for a weakly collisional plasma, in the large mean free path limit of \nu_{{\rm c}{\rm s}}\ll k\sigma_{\rm s} , where \nu_{{\rm c}{\rm s}}\sim\omega_{{\rm P}{\rm s}}\ln\Lambda/\Lambda is the collision frequency ( \omega_{{\rm P}s} is the plasma frequency and \Lambda is the plasma parameter or number of particles within the Debye sphere), can be expressed as \displaystyle\tilde{f}_{{\rm s}1{\bf k}}({\bf v},\omega)=-\frac{iq_{\rm s}}{m_{\rm s}}\,\frac{\left(\tilde{{\bf E}}^{({\rm P})}_{{\bf k}}(\omega)+\tilde{{\bf E}}_{1{\bf k}}(\omega)\right)\cdot{\partial f_{s0}}/{\partial{\bf v}}}{\omega-{\bf k}\cdot{\bf v}}+\frac{if_{{\rm s}1{\bf k}}\left({\bf v},0\right)}{\omega-{\bf k}\cdot{\bf v}}, \displaystyle\tilde{E}_{{\bf k}i}(\omega)=\tilde{E}^{({\rm P})}_{{\bf k}i}(\omega)+\tilde{E}_{1{\bf k}i}(\omega)=\varepsilon^{-1}_{{\bf k}\,ij}(\omega)\,\left[{\tilde{{\bf E}}^{({\rm P})}_{{\bf k}j}(\omega)}+g_{{\bf k}j}(\omega)\right], (4) with the dielectric tensor {\varepsilon}_{{\bf k}}(\omega)={\rm diag}\left(\varepsilon_{{\bf k}\perp},\varepsilon_{{\bf k}\perp},\varepsilon_{{\bf k}\parallel}\right) given by \displaystyle\varepsilon_{{\bf k}\parallel}(\omega)=1+\sum_{{\rm s}}\frac{\omega^{2}_{{\rm P}{\rm s}}}{k^{2}}\int{\rm d}^{3}v\,\frac{{\bf k}\cdot{\partial f_{{\rm s}0}}/{\partial{\bf v}}}{\omega-{\bf k}\cdot{\bf v}}, \displaystyle\varepsilon_{{\bf k}\perp}(\omega)=1-\frac{\omega}{c^{2}k^{2}-\omega^{2}}\sum_{\rm s}{\omega^{2}_{{\rm P}{\rm s}}}\int{\rm d}^{3}v\,\frac{v_{\perp}{\partial f_{{\rm s}0}}/{\partial v_{\perp}}}{\omega-{\bf k}\cdot{\bf v}}, (5) and the vector g_{{\bf k}}(\omega)=\left(g_{{\bf k}\perp},g_{{\bf k}\perp},g_{{\bf k}\parallel}\right) , corresponding to the initial perturbation, given by \displaystyle g_{{\bf k}\parallel}(\omega)=\frac{4\pi}{k}\sum_{{\rm s}}{n_{\rm s}q_{\rm s}}\int{\rm d}^{3}v\,\frac{f_{{\rm s}1{\bf k}}\left({\bf v},0\right)}{\omega-{\bf k}\cdot{\bf v}}, \displaystyle g_{{\bf k}\perp}(\omega)=-\frac{4\pi\omega}{c^{2}k^{2}-\omega^{2}}\sum_{{\rm s}}{n_{\rm s}q_{\rm s}}\int{\rm d}^{3}v\,\frac{v_{\perp}f_{{\rm s}1{\bf k}}\left({\bf v},0\right)}{\omega-{\bf k}\cdot{\bf v}}. (6) Here \omega_{{\rm P}{\rm s}}=\sqrt{4\pi n_{\rm s}q^{2}_{\rm s}/m_{\rm s}} is the plasma (electron Langmuir) frequency or that of the ion Langmuir waves, n_{\rm s} being the number density of the charged species. The subscript {\bf k} stands for the Fourier transform in {\bf x} , and the tilde represents the Laplace transform in t . Without loss of generality, we have assumed {\bf k}=k\,\hat{{\bf z}} with the perpendicular space spanned by \hat{{\bf x}} and \hat{{\bf y}} , and v_{\perp} denoting the component of {\bf v} perpendicular to {\bf k} , i.e., either v_{x} or v_{y} . We have assumed that the DF f_{{\rm s}0} is isotropic in {\bf v} , i.e., f_{{\rm s}0}({\bf v})=f_{{\rm s}0}(v) , such that \partial f_{{\rm s}0}/\partial{\bf v}=\partial f_{{\rm s}0}/\partial v\,\hat{{\bf v}} and therefore the magnetic Lorentz force is zero at linear order. Thus we ignore all plasma dynamics related to DF anisotropies, e.g., the Weibel instability, which we leave for future investigation. The dielectric tensor represents (i) the longitudinal polarization of the medium that manifests as Debye shielding/screening of the electric field as well as longitudinal plasma waves (electron Langmuir and ion acoustic and Langmuir waves) parallel to {\bf k} and (ii) the transverse EM or light waves, perpendicular to {\bf k} , that are modulated by plasma oscillations. The real part of the longitudinal component \varepsilon_{{\bf k}\parallel} universally scales as 1-\omega^{2}_{{\rm P}e}/\omega^{2} for \omega\gg k\sigma_{e} , independent of the detailed functional form of f_{{\rm s}0} , a property that plays a crucial role in NTPA, as we will demonstrate shortly. The zeros of the dielectric tensor denote the Landau modes or waves that oscillate and (Landau) damp due to collective wave-particle interactions. The electron Langmuir waves, with the dispersion relation \omega^{2}\approx\omega^{2}_{{\rm P}e}+3k^{2}\sigma^{2}_{e} , and the ion Langmuir and ion acoustic waves, with \omega^{2}\approx k^{2}c^{2}_{\rm s}/(1+k^{2}\lambda^{2}_{{\rm D}e}) , emerge from the zeros of \varepsilon_{{\bf k}\parallel}\left(\omega\right) . Here, \omega_{{\rm P}e}=\sqrt{4\pi n_{e}e^{2}/m_{e}} is the electron plasma frequency ( n_{e}= electron number density), c_{\rm s}=\sqrt{k_{\rm B}ZT_{e}/m_{i}} is the ion sound speed ( T_{e} = electron temperature, Z the atomic number of the dominant ionic species), and \lambda_{{\rm D}e}=\sigma_{e}/\omega_{{\rm P}e} is the electron Debye length. The electron Langmuir and ion acoustic and Langmuir waves are longitudinal waves, with electric field oscillations parallel to {\bf k} , that get Landau damped via wave-particle interactions. The transverse light waves, with EM oscillations perpendicular to {\bf k} , arise from the zeros of \varepsilon_{{\bf k}\perp}\left(\omega\right) , follow the dispersion relation \omega^{2}\approx c^{2}k^{2}+\sum_{\rm s}\omega^{2}_{{\rm P}{\rm s}}\left(1+\sigma^{2}_{\rm s}/c^{2}\right) , and do not undergo Landau damping due to the absence of superluminal particles and therefore wave-particle resonances ( \omega={\bf k}\cdot{\bf v} ) that might damp them. In the small mean free path limit (\nu_{{\rm c}e}\gg k\sigma_{\rm s}) , the response can still be described by equations ( 4 ) for v\gg\nu_{{\rm c}e}/k . The functional form of the dielectric tensor is, however, modified. Assuming a simplified Lenard-Bernstein form for the collision operator ( 26 ) , where D_{1}^{({\rm s})}(v)=\nu_{{\rm c}{\rm s}}v and D_{2}^{({\rm s})}(v)=\nu_{{\rm c}{\rm s}}\sigma^{2}_{\rm s} (the Balescu-Lenard coefficients scale the same way at v\lesssim\sigma_{\rm s} ), the longitudinal component, in the limit of \omega\gg k\sigma_{e} , can be written as follows ( 26 ; 4 ) : \displaystyle\varepsilon_{{\bf k}\parallel}(\omega)\approx 1-\frac{\omega^{2}_{{\rm P}e}}{\omega^{2}}\dfrac{1}{1+\dfrac{i\nu_{{\rm c}e}}{\omega}}. (7) Here we have neglected the sub-dominant ion term. For \omega\gg\nu_{{\rm c}e}\gg k\sigma_{e} , the real part scales as 1-\omega^{2}_{{\rm P}e}/\omega^{2} , just as in the large mean free path limit. Hence, the dielectric factor \varepsilon_{{\bf k}\parallel}({\bf k}\cdot{\bf v}) that multiplies the acceleration of a charged particle with velocity {\bf v} in a plasma, scales as 1-\omega^{2}_{{\rm P}e}/{\left({\bf k}\cdot{\bf v}\right)}^{2} for the high v particles, independent of whether the mean free path is small or large compared to the scale of the perturbation, k^{-1} . This is because, for these very fast particles, the diffusion length v/\nu_{{\rm c}e} always exceeds k^{-1} , even if the mean free path or the average free-streaming scale is small. In other words, the fastest particles are unaffected by collisions. II.3 Quasilinear theory Linear theory tells us how small perturbations propagate as undulations on a smooth background DF f_{{\rm s}0} . However, the non-linear coupling between these perturbations itself modifies the mean DF. The evolution of the mean DF of each species, f_{{\rm s}0}={\left(2\pi\right)}^{3}f_{{\rm s}2\,{\bf k}=0}/V , averaged over a volume V of the bulk plasma, can be studied by computing the second order response, f_{{\rm s}2\,{\bf k}} , taking the {\bf k}\to 0 limit and ensemble averaging the response over the random phases of the linear fluctuations (see Appendix A.2 of 3 ). This yields the following quasilinear equation for each species: \displaystyle\frac{\partial f_{{\rm s}0}}{\partial t} \displaystyle=-\frac{{\left(2\pi\right)}^{3}q_{\rm s}}{m_{\rm s}V}\int{\rm d}^{3}k\,\left<\left({\bf E}_{{\bf k}}^{\ast}+\frac{{\bf v}}{c}\times{\bf B}_{{\bf k}}^{\ast}\right)\cdot\nabla_{{\bf v}}f_{{\rm s}1{\bf k}}\right>, (8) where {\bf E}_{{\bf k}}={\bf E}_{1{\bf k}}+{\bf E}_{{\bf k}}^{({\rm P})} and {\bf B}_{{\bf k}}={\bf B}_{1{\bf k}}+{\bf B}_{{\bf k}}^{({\rm P})} , and the subscript 1 denotes linear perturbations. We have used the reality condition, {\bf E}_{1,-{\bf k}}={\bf E}_{1{\bf k}}^{\ast} , and similarly for the other quantities. Now, we need to make assumptions about the temporal correlation of the perturbing EM fields, E_{{\bf k}i}^{({\rm P})}(t) and B_{{\bf k}i}^{({\rm P})}(t) , where the subscript i denotes the i^{\rm th} component. Faraday’s law (fourth of Maxwell’s equations [ 2 ]) dictates that the electric and magnetic field perturbations are related by \tilde{{\bf B}}_{{\bf k}}^{({\rm P})}\left(\omega\right)=\frac{c}{\omega}\left({\bf k}\times\tilde{{\bf E}}_{{\bf k}}^{({\rm P})}\left(\omega\right)\right) . We assume that the perturbing electric field E_{{\bf k}i}^{({\rm P})}(t) is a generic red noise: \displaystyle\left \displaystyle={\cal E}_{ij}\left({\bf k}\right)\,{\cal C}_{t}\left(t-t^{\prime}\right), (9) where {\cal C}_{t} is the temporal correlation function. For white noise, this is simply \delta\left(t-t^{\prime}\right) , whose Fourier transform is {\cal C}_{\omega}\left(\omega\right)=1 . Substituting the expressions for the linear quantities, {\bf E}_{{\bf k}}(t) and f_{{\rm s}1{\bf k}}({\bf v},t) , obtained by performing the inverse Laplace transform of equations ( 4 ) in the quasilinear equation ( 8 ) above, and using the noise spectrum for the perturbing electric field given in equation ( 9 ), we obtain a simplified form for the quasilinear transport equation for each charged species, \displaystyle\frac{\partial f_{{\rm s}0}}{\partial t} \displaystyle=\frac{\partial}{\partial v_{i}}\left[\left(D_{ij}^{({\rm s})}({\bf v})+{\cal D}^{({\rm s})}_{ij}({\bf v})\right)\frac{\partial f_{{\rm s}0}}{\partial v_{j}}+{\cal D}^{({\rm s})}_{i}\left({\bf v}\right)\,f_{{\rm s}0}\right], (10) which is a Fokker-Planck equation with the diffusion tensor D_{ij}^{({\rm s})}+{\cal D}_{ij}^{({\rm s})} and the drag/friction tensor {\cal D}_{i}^{({\rm s})} . D_{ij}^{({\rm s})} is sourced by the EM drive, while {\cal D}_{ij}^{({\rm s})} and {\cal D}_{i}^{({\rm s})} are the BL coefficients sourced by the internal fields and collisions. We discuss the detailed velocity dependence of these coefficients as follows. II.3.1 Drive diffusion The drive diffusion tensor D_{ij}^{({\rm s})} is given by \displaystyle D_{ij}^{({\rm s})}({\bf v})\approx D_{p\,ij}^{({\rm s})}({\bf v})+D_{w\,ij}^{({\rm s})}({\bf v},t), (11) where D_{p\,ij}^{({\rm s})} denotes the direct diffusion of the dressed (Debye shielded) particles by the EM drive and is given by \displaystyle D_{p\,ij}^{({\rm s})}({\bf v})\approx\frac{8\pi^{4}q^{2}_{\rm s}}{m^{2}_{\rm s}V}\int{\rm d}^{3}k\,{\left[\varepsilon^{-1}_{{\bf k}}\left({\bf k}\cdot{\bf v}\right)\,\mathbb{P}_{{\bf k}}\left({\bf k}\cdot{\bf v}\right)\,\varepsilon^{-1\dagger}_{{\bf k}}\left({\bf k}\cdot{\bf v}\right)\right]}_{ij}, \displaystyle\mathbb{P}_{{\bf k}\,ij}\left({\bf k}\cdot{\bf v}\right)=\frac{k_{i}v_{l}\,{\cal E}_{lj}\left({\bf k}\right){\cal C}_{\omega}\left({\bf k}\cdot{\bf v}\right)}{{\bf k}\cdot{\bf v}}, (12) and D_{w\,ij}^{({\rm s})} denotes the diffusion mediated by the Landau modes or waves excited by the drive and is given by \displaystyle D_{w\,ij}^{({\rm s})}({\bf v},t)\approx\frac{8\pi^{4}q^{2}_{\rm s}}{m^{2}_{\rm s}V}\int{\rm d}^{3}k\,\frac{\left|\gamma_{\bf k}\right|}{{\left(\eta_{\bf k}-{\bf k}\cdot{\bf v}\right)}^{2}+\gamma^{2}_{\bf k}}\exp{\left[2\gamma_{\bf k}t\right]} \displaystyle\times{\varepsilon^{\prime-1}_{{\bf k}\,im}\left(\omega_{\bf k}\right)\,\mathbb{Q}_{{\bf k}\,mn}\,\varepsilon^{\prime-1\dagger}_{{\bf k}\,nj}\left(\omega_{\bf k}\right)}, \displaystyle\mathbb{Q}_{{\bf k}\,ij}=\left|{\cal C}^{\prime}_{\omega}(\omega_{\bf k})\right|\left[{\cal E}_{ij}({\bf k})\left(1-\frac{{\bf k}\cdot{\bf v}}{\omega_{\bf k}}\right)+\frac{k_{i}v_{l}}{\omega_{\bf k}}{\cal E}_{lj}({\bf k})\right]. (13) Note that both direct and wave-mediated diffusion are governed by the spatial power-spectrum {\cal E}_{ij}({\bf k}) and the temporal power-spectrum {\cal C}_{\omega} . Here, \omega_{\bf k}=\eta_{\bf k}+i\gamma_{\bf k} is the frequency of the least damped Landau mode, prime denotes the derivative of a quantity with respect to its argument, and the dielectric constant is given by \varepsilon_{{\bf k}}\left(\omega\right)={\rm diag}\left(\varepsilon_{{\bf k}\perp},\varepsilon_{{\bf k}\perp},\varepsilon_{{\bf k}\parallel}\right) with its components given by equations ( 5 ). The longitudinal component of \varepsilon_{{\bf k}} scales universally as 1-\omega^{2}_{{\rm P}e}/{\left({\bf k}\cdot{\bf v}\right)}^{2} ( \omega_{{\rm P}e}=\sqrt{4\pi n_{e}e^{2}/m_{e}} is the plasma frequency) over a large velocity range, \sigma_{e}{\bf k}\cdot{\bf v}) are less heated, since the polarizing field Debye screens the large-scale field. The faster ones (\omega_{{\rm P}e}<{\bf k}\cdot{\bf v}) , on the other hand, are unscreened and directly heated by the bare field. This velocity-dependent diffusion of dressed/Debye screened particles by the EM drive is described by the drive diffusion tensor D_{p\,ij}^{({\rm s})} given in equation ( 12 ). Contrary to equation ( 12 ) that describes the direct diffusion of particles by the EM drive, equation ( 13 ) denotes the wave diffusion tensor that describes the indirect heating of particles by the waves excited by the drive. If the plasma is stable to perturbations, then the waves Landau-damp, the wave diffusion tensor D_{w\,ij}^{({\rm s})} dies away and only the dressed particle diffusion tensor D_{p\,ij}^{({\rm s})} survives at long time. This is strictly only true if the Landau damping timescale is shorter than that of quasilinear relaxation, which typically holds, as we show in Appendix B . If, on the other hand, the plasma is unstable or marginally stable to perturbations, i.e., there exit plateaus and/or bumps in the DF, then \gamma_{\bf k}\approx 0 for the least damped mode (once an unstable bump has saturated to a plateau), for k\lambda_{\rm D}\lesssim 1 , and the waves keep contributing to diffusion along with direct heating. In this case, we have \displaystyle D_{w\,ij}^{({\rm s})}({\bf v},t)\to D_{w\,ij}^{({\rm s})}({\bf v}) \displaystyle\approx\frac{8\pi^{5}q^{2}_{\rm s}}{m^{2}_{\rm s}V}\int{\rm d}^{3}k\,\delta\left(\eta_{\bf k}-{\bf k}\cdot{\bf v}\right)\varepsilon^{\prime-1}_{{\bf k}\,im}\left(\eta_{\bf k}\right)\,\mathbb{Q}_{{\bf k}\,mn}\,\varepsilon^{\prime-1\dagger}_{{\bf k}\,nj}\left(\eta_{\bf k}\right), \displaystyle\mathbb{Q}_{{\bf k}\,ij}=\frac{k_{i}v_{l}\,{\cal E}_{lj}\left({\bf k}\right)\left|{\cal C}^{\prime}_{\omega}\left(\eta_{\bf k}\right)\right|}{\eta_{\bf k}}. (14) This is the classic expression for the diffusion tensor that describes the quasilinear saturation of instabilities or the ‘heating’ of particles due to the Landau damping of waves. Notably, only the resonant particles with \omega_{\bf k}\approx\eta_{\bf k}={\bf k}\cdot{\bf v} significantly exchange energy with the waves, as indicated by the \delta\left(\eta_{\bf k}-{\bf k}\cdot{\bf v}\right) factor above. The total energy of the waves is, however, distributed equally between the resonant and non-resonant particles. While the electromagnetic potential energy of the waves is exchanged with the resonant particles, the kinetic energy is exchanged with the non-resonant particles of the bulk. Although each resonant particle exchanges much more energy with the waves than a non-resonant one, the much greater abundance of non-resonant than resonant particles implies that an equipartition of energy exchanged is possible between the resonant and non-resonant species. The quasilinear evolution of the mean DF occurs hand in hand with the linear Landau damping (growth) of the waves in the marginally stable (unstable) regime. D^{({\rm s})}_{w\,ij} therefore denotes the heating of particles due to wave-particle interactions. II.3.2 Balescu-Lenard diffusion and drag The quasilinear equation ( 10 ) not only describes particle diffusion due to the drive as well as the waves excited by it, but also the diffusion and drag due to internal turbulence and collisions. The latter is described by the BL diffusion and drag coefficients, given by \displaystyle{\cal D}_{ij}^{({\rm s})}\left({\bf v}\right)=\frac{\pi}{m^{2}_{\rm s}}{\left(4\pi q_{\rm s}q_{{\rm s}^{\prime}}\right)}^{2}\sum_{{\rm s}^{\prime}}\int\frac{{\rm d}^{3}k}{{\left(2\pi\right)}^{3}}\frac{k_{i}k_{j}}{k^{4}} \displaystyle\times\frac{1}{{\left|\varepsilon_{{\bf k}\parallel}\left({\bf k}\cdot{\bf v}\right)\right|}^{2}}\int{\rm d}^{3}v^{\prime}\,\delta\left({\bf k}\cdot\left({\bf v}-{\bf v}^{\prime}\right)\right)f_{{\rm s}^{\prime}0}\left({\bf v}^{\prime}\right), \displaystyle{\cal D}_{i}^{({\rm s})}\left({\bf v}\right)=-\frac{\pi}{m_{\rm s}}{\left(4\pi q_{\rm s}q_{{\rm s}^{\prime}}\right)}^{2}\sum_{{\rm s}^{\prime}}\frac{1}{m_{{\rm s}^{\prime}}}\int\frac{{\rm d}^{3}k}{{\left(2\pi\right)}^{3}} \displaystyle\times\frac{1}{{\left|\varepsilon_{{\bf k}\parallel}\left({\bf k}\cdot{\bf v}\right)\right|}^{2}}\int{\rm d}^{3}v^{\prime}\,\delta\left({\bf k}\cdot\left({\bf v}-{\bf v}^{\prime}\right)\right)\frac{\partial f_{{\rm s}^{\prime}0}}{\partial v^{\prime}_{j}}. (15) Note that only the mutually resonant particles with {\bf k}\cdot{\bf v}={\bf k}\cdot{\bf v}^{\prime} exchange energy and momentum efficiently via collisions. Only the longitudinal modes contribute to BL diffusion and drag under isotropic conditions in a non-relativistic plasma. In the relativistic regime, the transverse modes also contribute ( 38 ) . While the diffusion tensor describes relaxation due to the exchange of energy between the particles via collisions and small-scale turbulence, the drag tensor describes the exchange of momentum. As we shall see, the BL coefficients generally dominate at small v , whereas the drive diffusion coefficient dominates at large v for a strong enough super-Debye scale drive. III A general transport equation for particle acceleration in unmagnetized plasmas We are interested in the quasilinear relaxation of an isotropic plasma \left(f_{{\rm s}0}({\bf v})=f_{{\rm s}0}(v)\right) in this paper. If we also assume spatially isotropic fluctuations 1 1 1 Note that the assumption of isotropic fluctuations, {\cal E}_{ij}({\bf k})={\cal E}(k)\,\hat{k}_{i}\hat{k}_{j} , is different from that in 3 . This, however, does not make any qualitative difference in the main results of the paper. , i.e., • Isotropic EM drive: {\cal E}_{ij}({\bf k})={\cal E}(k)\,\hat{k}_{i}\hat{k}_{j} • Isotropic DF: f_{{\rm s}0}({\bf v})=f_{{\rm s}0}(v) , with \hat{k}_{i}=k_{i}/k , then the dimensionality of the problem is greatly reduced. This assumption of isotropy is justified in the absence of a strong guide field. The multi-dimensional transport equation ( 10 ) then becomes the following one-dimensional equation in v : \displaystyle\frac{\partial f_{{\rm s}0}}{\partial t}=\frac{1}{v^{2}}\frac{\partial}{\partial v}\left[v^{2}\left(\left(D^{({\rm s})}(v)+{\cal D}^{({\rm s})}_{2}(v)\right)\frac{\partial f_{{\rm s}0}}{\partial v}+{\cal D}^{({\rm s})}_{1}(v)f_{{\rm s}0}\right)\right]. (16) The drive diffusion coefficient D^{({\rm s})} is equal to D^{({\rm s})}_{p}+D^{({\rm s})}_{w} with D^{({\rm s})}_{p} the dressed particle diffusion coefficient and D^{({\rm s})}_{w} the wave diffusion coefficient. {\cal D}^{({\rm s})}_{2} and {\cal D}^{({\rm s})}_{1} denote the BL diffusion and drag coefficients respectively. In what follows, we discuss and compare the velocity dependence of these different coefficients. III.1 Velocity dependence of the coefficients Figure 1: Electron diffusion coefficient due to direct dressed particle heating by the drive (solid lines) and indirect heating by the waves (dashed lines), in units of D^{(0)}_{e}=\left(32\pi^{5}e^{2}k^{2}{\cal E}_{0}/m_{e}^{2}V\right)\left[(3-\alpha)/\left(1-{\left(k_{\rm min}/k_{\rm c}\right)}^{3-\alpha}\right)\right] , vs v for different values of \alpha , with the electric field power spectrum of the drive given by {\cal E}(k)\sim k^{-\alpha} (see equation [ 25 ]). Ion diffusion coefficient scales similarly with v . The plasma is assumed to be marginally stable. We adopt T_{e}=T_{i} , k_{\rm min}\lambda_{\rm D}=10^{-3} , and \omega_{{\rm P}e}t_{\rm c}=10 . Note that the dressed particle coefficient D_{p}^{({\rm s})}(v) typically dominates over the wave coefficient D_{w}^{({\rm s})}(v) . For \alpha\geq 5 , D_{p}^{({\rm s})}(v) scales as v^{4} at \sigma_{e}1 , the v^{4} scaling persists in the range, \sigma_{e}T_{e} ), and similarly at \sigma_{i}=\sqrt{k_{\rm B}T_{i}/m_{i}}1 with t_{\rm c} the noise correlation time) that dominates on super-Debye scales ( k\lambda_{\rm D}\ll 1 ). To be fairly general, we assume that the spatial and temporal power spectra of the fluctuating drive are given by \displaystyle{\cal E}(k)={\cal E}_{0}\,\frac{3-\alpha}{1-{\left(k_{\rm min}/k_{\rm c}\right)}^{3-\alpha}}\,{\left(\frac{k}{k_{\rm c}}\right)}^{-\alpha}\exp{\left[-k/k_{\rm c}\right]}\,\Theta\left(k-k_{\rm min}\right), \displaystyle{\cal C}_{\omega}(\omega)=\frac{1}{1+\omega^{2}t^{2}_{\rm c}}, (25) with k_{\rm min} the minimum wavenumber of the drive and k_{\rm c} the maximum cutoff in k . The spatial power-spectrum can be as steep as {\cal E}(k)\sim k^{-(2d+2)} in d dimensions, i.e., k^{-8} in 3D, around the Debye scale ( 15 ; 32 ; 31 ; 19 ) . The Lorentzian form for {\cal C}_{\omega} arises from an exponentially decaying temporal correlation (c.f. equation [ 18 ]). With the power-law {\cal E}(k)\sim k^{-\alpha} and the Lorentzian {\cal C}_{\omega} , the dressed particle and wave diffusion coefficients can be obtained by integrating over k in equations ( 17 ) and ( 23 ). This yields power-law v scalings of D^{({\rm s})}_{p}(v) , which are listed for different values of \alpha and different v intervals in Table 1 . Power-spectrum Velocity Range D_{p}^{(\rm s)}(v) f_{\rm s0}(v) \alpha\geq 7 \sigma_{e}{\omega_{\rm Pe}}/{k_{\min}} v^{-3} \exp\!\left[-v^{2}/2\sigma^{\prime 2}_{\rm s}\right] 5\leq\alpha<7 \sigma_{e}{\omega_{\rm Pe}}/{k_{\min}} v^{-3} \exp\!\left[-v^{2}/2\sigma^{\prime 2}_{\rm s}\right] 0<\alpha<5 \sigma_{e}{\omega_{\rm Pe}}/{k_{\min}} v^{-3} \exp\!\left[-v^{2}/2\sigma^{\prime 2}_{\rm s}\right] Table 1: Velocity scalings of the dressed particle diffusion coefficient D^{({\rm s})}_{p} and the corresponding steady-state distribution f_{\rm s0}(v) for a drive with electric field power-spectrum \mathcal{E}(k)\sim k^{-\alpha} . We assume T_{e}=T_{i} and 1/\omega_{{\rm P}e}5 , the drive scales the same way with v (till \omega_{{\rm P}e}/k_{\rm min} instead of \omega_{{\rm P}e}/k_{0} ). The BL diffusion (drag) coefficient {\cal D}_{2}^{({\rm s})} ( {\cal D}_{1}^{({\rm s})} ) is constant (scales as \sim v ) at v\lesssim\sigma_{\rm s} and scales as \sim v^{-3} ( \sim v^{-2} ) at v\gtrsim\sigma_{\rm s} . For T_{e}\sim T_{i} , the drive diffusion coefficient D^{({\rm s})} is constant for v\lesssim\sigma_{e}=\sqrt{k_{\rm B}T_{e}/m_{e}} and scales as \sim v^{4} for \sigma_{e}\lesssim v\lesssim\omega_{{\rm P}e}/k=\sigma_{e}/k\lambda_{{\rm D}e} , as long as k\lambda_{{\rm D}e}<1 . For T_{e}\gg T_{i} , the v^{4} scaling also appears between the ion thermal speed \sigma_{i}=\sqrt{k_{\rm B}T_{i}/m_{i}} and the ion acoustic speed c_{\rm s}=\sqrt{k_{\rm B}T_{e}/m_{i}} . The wave particle resonance between the particles and the Langmuir waves causes D^{({\rm s})}(v) to spike at v=\omega_{{\rm P}e}/k for T_{e}=T_{i} . A similar resonance with the ion acoustic waves gives rise to a spike at v=c_{\rm s} when T_{e}\gg T_{i} , a necessary condition for the ion acoustic waves to be weakly damped. At T_{e}=T_{i} , these waves are strongly damped, which is why the large imaginary part of \varepsilon_{k\parallel} smooths out D^{({\rm s})}(v) at v=c_{\rm s} ; this effect is known as resonance broadening ( 12 ) . On the other hand, Langmuir waves are always weakly damped for k\lambda_{{\rm D}e}<1 , resulting in a pronounced spike at v=\omega_{{\rm P}e}/k irrespective of T_{i}/T_{e} . The drive diffusion coefficient always exceeds the BL coefficients at large enough v , independent of the strength of the drive, as long as it predominantly acts on super-Debye scales. This is because the drive coefficient is an increasing function of v in a large v range and falls off only beyond v=\omega_{{\rm P}e}/k , whereas the BL coefficients drop off beyond the thermal speed, which is much smaller than \omega_{{\rm P}e}/k . Hence, a large-scale drive always dominates over collisions or internal turbulence at high v . In other words, the fastest particles are unscreened and readily accelerated without slowing down due to collisions. III.2 Quasilinear relaxation of the distribution function Figure 3: Development of the non-thermal power-law tail in an initially Maxwellian ion distribution due to large-scale EM turbulence, obtained by solving equation ( 16 ). We adopt D^{(i)}(v=0)=10^{-2}{\cal D}^{(i)}_{2}(v=0) , \sigma_{i}=0.1\sigma_{e} , \alpha=8 , k_{\rm min}\lambda_{{\rm D}i}=10k_{\rm min}\lambda_{{\rm D}e}=10^{-3} , and \omega_{{\rm P}i}t_{\rm c}=0.1\omega_{{\rm P}e}t_{\rm c}=1 . Note the Maxwellianization of the bulk around \sigma_{i} and the emergence of the v^{-5} tail beyond, followed by a v^{1-\alpha}\sim v^{-7} tail and a Maxwellian fall-off. Figure 4: Same as Fig. 3 but for electrons. We adopt D^{(e)}(v=0)={\cal D}^{(e)}_{2}(v=0) , \alpha=8 , k_{\rm min}\lambda_{{\rm D}e}=10^{-3} , and \omega_{{\rm P}e}t_{\rm c}=10 . Just as in the ions, a v^{-5} tail develops, followed by a v^{1-\alpha}\sim v^{-7} tail and a Maxwellian fall-off. Figure 5: Maxwellianization of the electron DF, initialized as a \kappa=1.5 distribution ( v^{-5} ), after the EM drive has been switched off and the plasma relaxes due to turbulent and collisional relaxation through the BL equation (equation [ 16 ] without the drive). Note the continued presence of the v^{-5} tail of runaway electrons. It is not surprising that the power-law scaling of D^{({\rm s})}(v) would translate to a power-law f_{{\rm s}0}(v) . To see this explicitly, we evolve the electron and ion DFs by numerically integrating equation ( 16 ) for each species, using a flux conserving scheme detailed in Appendix C.1 of 3 . We assume an isotropic EM drive with a power spectrum of the form given in equation ( 25 ) with \alpha=8 , which is expected for Vlasov turbulence around the Debye length ( 31 ; 15 ) . We adopt \sigma_{i}=0.1\sigma_{e} , k_{\rm min}\lambda_{\rm D}=10^{-3} , and t_{\rm c}=10/\omega_{{\rm P}e}=1/\omega_{{\rm P}i} . In this case, as evident from Fig. 1 and Table 1 , D^{(s)}_{p}(v) dominates over D^{({\rm s})}_{w}(v) , and scales as v^{4} for \sigma_{e}\sigma_{e} . We plot the resulting ion DF f_{i0} in Fig. 3 . The electron DF has similar, albeit faster, evolution since D^{(e)}\sim D^{(i)}{\left(m_{i}/Zm_{e}\right)}^{2}\gg D^{(i)} , and is plotted in Fig. 4 . The maximum value of the diffusion coefficient is much larger for electrons than ions in the current setup, which significantly restricts the timestep (via the Courant condition) and total integration time in the former. Evidently, an initial Maxwellian-like DF develops an extended v^{-5} tail followed by a v^{-7} fall-off till \omega_{{\rm P}e}/k_{\rm min} , beyond which it forms a plateau and a Maxwellian fall-off. The ion DF, besides featuring these power-law tails, develops a break at the ion thermal speed as the bulk of the ions at v\lesssim\sigma_{e} tries to Maxwellianize, since D^{(i)}\ll{\cal D}_{2}^{(i)} at small velocities. The BL diffusion (drag) increases (decreases) the particle velocities, which is evident from the evolution around \sigma_{i} at earlier times. The drive diffusion coefficient gradually develops the v^{-5} and v^{-7} tails at higher v . Interestingly, the v^{-5} tail appears in the ions even at \sigma_{i}\sigma_{e} , D^{({\rm s})} exceeds both BL coefficients, which implies that f_{{\rm s}0} scales as \int_{v}^{\infty}{\rm d}v^{\prime}/v^{\prime 2}D^{({\rm s})}(v^{\prime}) at large v . Therefore, f_{{\rm s}0}(v)\sim\int_{v}^{\infty}{\rm d}v^{\prime}/v^{\prime 6}\sim v^{-5} at \sigma_{e}{\left(k_{\rm min}\lambda_{\rm D}\right)}^{7/2} , and \sigma_{e} otherwise (assuming T_{e}\sim T_{i} ). IV Application to the solar corona So far, in our treatment, we have assumed that the coarse-graining scale is smaller than macroscopic scales, e.g., the gravitational scale height h in a plasma under a gravitational field. To study the behavior of the plasma on scales \sim h , we must take into account the effect of the attractive gravitational potential. To describe the behavior of the macroscopic thermodynamic quantities such as density, pressure and temperature in the weakly collisional plasma of the solar corona, we can construct an effective fluid model from the kinetic description. Our goal here is to focus on the simplest, qualitative deviation from the original velocity filtration model ( 37 ) that serves as an Occam’s razor approach towards resolving the coronal heating problem. The Vlasov equation dictates that, although the DF is locally a function of v , under a global potential \Phi({\bf x}) , it is a function of the energy E=v^{2}/2+\Phi({\bf x}) . Therefore, under sun’s gravity, the electron and ion DFs, which locally scale as v^{-2\left(1+\kappa_{\rm s}\right)} , become f_{{\rm s}0}(v,r)\sim{\left[1+{(v^{2}+\Phi_{\rm G}(r)-\Phi_{\rm G}(r_{0}))}/\kappa_{\rm s}\sigma^{2}_{{\rm s}0}\right]}^{-\left(1+\kappa_{\rm s}\right)} ( 37 ) , where r is the radial distance from the center, r_{0} is some reference radius, \sigma_{{\rm s}0} is the thermal speed at r_{0} , \Phi_{\rm G}(r)\approx-GM_{\odot}/r is the gravitational potential ( M_{\odot} is the solar mass), and \kappa_{\rm s} is the value of \kappa for the {\rm s}^{\rm th} species. Here we have used the fact that the total potential \Phi(r) is a sum of the gravitational potential \Phi_{\rm G}(r)\approx-GM_{\odot}/r and the ambipolar electrostatic potential \Phi_{\rm E}(r)\approx-\Phi_{\rm G}(r)/2 that develops between the electrons and ions. As we showed earlier, kinetic turbulence implies a particular preference for 1.5\leq\kappa_{\rm s}\leq 2.5 (v^{-5}-v^{-7}) , which lies within a plausible range for protons and heavy ions in the fast solar wind (see 36 and references therein). There is a potential complication introduced to this picture of non-thermal tails by the presence of collisions in the plasma. The mean free path \lambda_{\rm mfp} of both ions and electrons in the coronal plasma is equal to \left(\Lambda/\ln\Lambda\right)\lambda_{\rm D}\approx 13\,{\left(T_{e}/10^{6}K\right)^{2}{\left(n_{e}/10^{8}{\rm cm}^{-3}\right)}^{-1}}{\rm km} , which is significantly smaller than the typical pressure scale height L\sim 10^{3} km. In other words, the Knudsen number \lambda_{\rm mfp}/L is approximately 100 . This may lead us to believe that, despite the weakly collisional nature of the plasma (\Lambda\gg 1) , the DF loses its kinetic power-law tails and Maxwellianizes on scales above \lambda_{\rm mfp} as the particles traverse a few mean free paths. This, however, would be an incorrect conclusion, since the drive diffusion coefficient, responsible for spawning the power-law tail, exceeds the Maxwellianizing BL coefficients over a large velocity range, \sigma_{\rm s}3/2 , the plasma has an effective equation of state (EOS) is polytropic, with p_{\rm s}\propto\rho^{\gamma_{\rm s}}_{\rm s} , with \gamma_{\rm s}=\left(\kappa_{\rm s}-3/2\right)/\left(\kappa_{\rm s}-1/2\right) . The polytropic coefficient tends to 1 at \kappa_{\rm s}\to\infty , i.e., the plasma becomes isothermal for a Maxwellian DF, as expected. For any non-thermal power law tail, however, we have \gamma_{\rm s}<1 , i.e., the EOS is always softer than an isothermal one for a kinetic plasma. This is because the collisionless nature of the plasma implies that the pressure does not sufficiently increase with increasing density. Figure 6: Temperature profile in the corona as a function of height above the photosphere, obtained from equation ( 35 ) using a single \kappa distribution with v_{\rm max}=3\sigma_{p0} . Left panel shows the predictions of the velocity filtration model for Fe and different values of \kappa as solid lines. Right panel shows the predictions for various elements as indicated, for \kappa=1.5 . The vertical dashed line indicates the boundary between the upper chromosphere and the lower corona. The symbols denote the temperature inferred from the UV spectroscopic observations of various ionic species ( 35 ; 43 ; 45 ) . The dashed curve denotes a smooth interpolation through the data. Note that the abrupt transition is reproduced by \kappa\approx 1.5-2 . The effective temperature T_{\rm s}=\left(m_{\rm s}/k_{\rm B}\right)\left(p_{\rm s}/\rho_{\rm s}\right) can be obtained from equations ( 28 ) and is given by \displaystyle T_{\rm s}(r) \displaystyle=\frac{2\kappa_{\rm s}}{3}\frac{\beta\left(\chi^{2}_{\rm s}(r),\frac{5}{2},\kappa_{\rm s}-\frac{3}{2}\right)}{\beta\left(\chi^{2}_{\rm s}(r),\frac{3}{2},\kappa_{\rm s}-\frac{1}{2}\right)}\left[T_{{\rm s}0}\,+\,\frac{m_{\rm s}}{\kappa_{\rm s}k_{\rm B}}\left[\Phi_{\rm G}(r)-\Phi_{\rm G}(r_{0})\right]\right], (30) where we have defined \sigma^{2}_{{\rm s}0}=k_{\rm B}T_{{\rm s}0}/m_{\rm s} , T_{{\rm s}0} being the temperature at the reference radius r_{0} . This, in the limit of v_{\rm max}\gg\sqrt{\Phi_{\rm G}(r)-\Phi_{\rm G}(r_{0})+3\sigma^{2}_{{\rm s}0}/2} or \chi_{\rm s}(r)\approx 1 , reduces to the following form: \displaystyle T_{\rm s}(r) \displaystyle\approx\frac{\kappa_{\rm s}}{\kappa_{\rm s}-3/2}\left[T_{{\rm s}0}\,+\,\frac{m_{\rm s}}{\kappa_{\rm s}k_{\rm B}}\left[\Phi_{\rm G}(r)-\Phi_{\rm G}(r_{0})\right]\right]. (31) Evidently, T_{\rm s} diverges at \kappa_{\rm s}\to 3/2 ( \gamma_{\rm s}\to 0 ), since the pressure diverges, provided that v_{\rm max}\gg\sqrt{\Phi_{\rm G}(r)-\Phi_{\rm G}(r_{0})+3\sigma^{2}_{{\rm s}0}/2} . This divergence is, however, logarithmic in v_{\rm max}/\sqrt{\Phi_{\rm G}(r)-\Phi_{\rm G}(r_{0})+3\sigma^{2}_{{\rm s}0}/2} , as we shall now see. The density, pressure and effective temperature for \kappa_{\rm s}=3/2 become \displaystyle\rho_{\rm s}(r) \displaystyle=4\pi{\cal N}_{\rm s}m_{\rm s}n_{\rm s}\,\frac{\chi^{3}_{\rm s}(r)}{3}\,{\left[1+\frac{2}{3}\frac{\Phi_{\rm s}(r)-\Phi_{\rm G}(r_{0})}{\sigma^{2}_{{\rm s}0}}\right]}^{-1}, \displaystyle p_{\rm s}(r) \displaystyle=4\pi{\cal N}_{\rm s}m_{\rm s}n_{\rm s}\sigma^{2}_{{\rm s}0}\,\left[\frac{1}{2}\ln{\left|\frac{1+\chi_{\rm s}(r)}{1-\chi_{\rm s}(r)}\right|}-\left(\,\chi_{\rm s}(r)+\frac{\chi^{3}_{\rm s}(r)}{3}\right)\right], \displaystyle T_{\rm s}(r) \displaystyle=\frac{3}{\chi^{3}_{\rm s}(r)}\left[\frac{1}{2}\ln{\left|\frac{1+\chi_{\rm s}(r)}{1-\chi_{\rm s}(r)}\right|}-\left(\,\chi_{\rm s}(r)+\frac{\chi^{3}_{\rm s}(r)}{3}\right)\right] \displaystyle\times\left[T_{{\rm s}0}\,+\,\frac{2m_{\rm s}}{3k_{\rm B}}\left[\Phi_{\rm G}(r)-\Phi_{\rm G}(r_{0})\right]\right]. (32) In the limit of v_{\rm max}\gg\sqrt{\Phi_{\rm G}(r)-\Phi_{\rm G}(r_{0})+3\sigma^{2}_{{\rm s}0}/2} , we have \chi_{\rm s}(r)\approx 1-\left(\Phi_{\rm G}(r)-\Phi_{\rm G}(r_{0})+3\sigma^{2}_{{\rm s}0}/2\right)/v^{2}_{\rm max} , and the corresponding density is \rho_{\rm s}(r)=(4\pi/3)\,{\left[1+2{\left(\Phi_{\rm G}(r)-\Phi_{\rm G}(r_{0})\right)}/{3\sigma^{2}_{{\rm s}0}}\right]}^{-1} , while the pressure is p_{\rm s}(r)=2\pi{\cal N}_{\rm s}m_{\rm s}n_{\rm s}\sigma^{2}_{{\rm s}0}\ln{\left(4v^{2}_{\rm max}\,\rho_{\rm s}(r)/3\sigma^{2}_{{\rm s}0}\,\rho_{\rm s}(r_{0})\right)} . This limit applies to \rho_{\rm s}(r)\gg(3\sigma^{2}_{{\rm s}0}/v^{2}_{\rm max})\,\rho_{\rm s}(r_{0}) , i.e., near the coronal base where the density is large. In the opposite limit of v_{\rm max}\ll\sqrt{\Phi_{\rm G}(r)-\Phi_{\rm G}(r_{0})+3\sigma^{2}_{{\rm s}0}/2} , \chi_{\rm s}(r)\approx{v_{\rm max}}/{\sqrt{\Phi_{\rm G}(r)-\Phi_{\rm G}(r_{0})+3\sigma^{2}_{{\rm s}0}/2}} , which implies that p_{\rm s}(r)\propto\rho_{\rm s}(r)\sim{\left[1+{2\left(\Phi_{\rm G}(r)-\Phi_{\rm G}(r_{0})\right)}/{3\sigma^{2}_{{\rm s}0}}\right]}^{-5/2} . This holds for \rho_{\rm s}(r)\ll(3\sigma^{2}_{{\rm s}0}/v^{2}_{\rm max})\,\rho_{\rm s}(r_{0}) , i.e., a low density environment. Thus, for \kappa_{\rm s}=3/2 , pressure increases very mildly (logarithmically) with density at high density, but linearly with density at low density. Often, the DF, especially that of the ions, is a broken power-law, as shown in Fig. 3 , in which case it can be expressed as a linear combination of different \kappa_{\rm s} distributions with different thermal speeds and cut-offs. If we assume \displaystyle f_{{\rm s}0}(v,r) \displaystyle=(1-x)f_{{\rm s}0}^{(1)}\left(v,r,\kappa_{{\rm s}1},\sigma_{{\rm s}0}^{(1)},v_{\rm max}^{(1)}\right) \displaystyle+xf_{{\rm s}0}^{(2)}\left(v,r,\kappa_{{\rm s}2},\sigma_{{\rm s}0}^{(2)},v_{\rm max}^{(2)}\right), (33) with 0\leq x\leq 1 , then the effective temperature can be expressed as: \displaystyle T_{\rm s}\left(r,\kappa_{{\rm s}1},\kappa_{{\rm s}2},\sigma_{{\rm s}0}^{(1)},v_{\rm max}^{(1)},\sigma_{{\rm s}0}^{(2)},v_{\rm max}^{(2)}\right) \displaystyle=\frac{m_{\rm s}}{k_{\rm B}}\frac{(1-x)\,p_{{\rm s}1}\left(r,\kappa_{{\rm s}1},\sigma_{{\rm s}0}^{(1)},v_{\rm max}^{(1)}\right)+x\,p_{{\rm s}2}\left(r,\kappa_{{\rm s}2},\sigma_{{\rm s}0}^{(2)},v_{\rm max}^{(2)}\right)}{(1-x)\,\rho_{{\rm s}1}\left(r,\kappa_{{\rm s}1},\sigma_{{\rm s}0}^{(1)},v_{\rm max}^{(1)}\right)+x\,\rho_{{\rm s}2}\left(r,\kappa_{{\rm s}2},\sigma_{{\rm s}0}^{(2)},v_{\rm max}^{(2)}\right)}, (34) where the density and pressure obtained from each individual \kappa distribution are \rho_{{\rm s}i}(r,\kappa_{{\rm s}i},\sigma_{{\rm s}0}^{(i)},v_{\rm max}^{(i)})=4\pi m_{\rm s}\int{\rm d}v\,v^{2}f_{{\rm s}0}^{(i)}(v,r,\kappa_{{\rm s}i},\sigma_{{\rm s}0}^{(i)},v_{\rm max}^{(i)}) and p_{{\rm s}i}(r,\kappa_{{\rm s}i},\sigma_{{\rm s}0}^{(i)},v_{\rm max}^{(i)})=(4\pi m_{\rm s}/3)\int{\rm d}v\,v^{4}f_{{\rm s}0}^{(i)}(v,r,\kappa_{{\rm s}i},\sigma_{{\rm s}0}^{(i)},v_{\rm max}^{(i)}) , with i=1,2 . Since both pressure and density fall off (radially) faster for larger \kappa_{\rm s} , the temperature is determined by the smaller value of \kappa_{\rm s} at large radii. This is the essence of velocity filtration ( 37 ; 30 ) , the fact that the gravitational potential only allows the harder/shallower power-law component to escape. IV.2 Velocity filtration, steep transition and temperature inversion Let us now discuss the behavior of the thermodynamic quantities in the solar atmosphere. As shown above, the profiles for the density and temperature for a single \kappa_{\rm s} distribution are given by \displaystyle\rho_{\rm s}(r) \displaystyle=4\pi{\cal N}_{\rm s}m_{\rm s}n_{\rm s}\;\beta\left(\chi^{2}_{\rm s}(r),\frac{3}{2},\kappa_{\rm s}-\frac{1}{2}\right) \displaystyle\times{\left[1+\frac{\Phi_{\rm G}(r)-\Phi_{\rm G}(r_{0})}{\kappa_{\rm s}\sigma^{2}_{{\rm s}0}}\right]}^{1/2-\kappa_{\rm s}}, \displaystyle T_{\rm s}(r) \displaystyle=\frac{2\kappa_{\rm s}}{3}\frac{\beta\left(\chi^{2}_{\rm s}(r),\frac{5}{2},\kappa_{\rm s}-\frac{3}{2}\right)}{\beta\left(\chi^{2}_{\rm s}(r),\frac{3}{2},\kappa_{\rm s}-\frac{1}{2}\right)}\left[T_{{\rm s}0}\,+\,\frac{m_{\rm s}}{\kappa_{\rm s}k_{\rm B}}\left[\Phi_{\rm G}(r)-\Phi_{\rm G}(r_{0})\right]\right], (35) for \kappa_{\rm s}>3/2 . The \kappa_{\rm s}=3/2 quantities are given by equations ( 32 ). Substituting \Phi_{\rm G}(r)=-GM_{\odot}/r in the above yields the following radial dependence: \displaystyle\rho_{\rm s}(r) \displaystyle=4\pi{\cal N}_{\rm s}m_{\rm s}n_{\rm s}\;\beta\left(\chi^{2}_{\rm s}(r),\frac{3}{2},\kappa_{\rm s}-\frac{1}{2}\right) \displaystyle\times{\left[1+\frac{m_{\rm s}}{m_{p}}\frac{T_{\rm vir}}{\kappa_{\rm s}T_{{\rm s}0}}\left(\frac{R_{\odot}}{r_{0}}-\frac{R_{\odot}}{r}\right)\right]}^{1/2-\kappa_{\rm s}}, \displaystyle T_{\rm s}(r) \displaystyle=\frac{2\kappa_{\rm s}}{3}\frac{\beta\left(\chi^{2}_{\rm s}(r),\frac{5}{2},\kappa_{\rm s}-\frac{3}{2}\right)}{\beta\left(\chi^{2}_{\rm s}(r),\frac{3}{2},\kappa_{\rm s}-\frac{1}{2}\right)}\left[T_{{\rm s}0}\,+\,\frac{m_{\rm s}}{m_{p}}\frac{T_{\rm vir}}{\kappa_{\rm s}}\left(\frac{R_{\odot}}{r_{0}}-\frac{R_{\odot}}{r}\right)\right], (36) with R_{\odot} the solar radius, the virial temperature T_{\rm vir} given by \displaystyle T_{\rm vir}=\frac{GM_{\odot}m_{p}}{k_{\rm B}R_{\odot}}, (37) and \displaystyle\chi_{\rm s}(r) \displaystyle=\sqrt{\dfrac{T_{\rm max}}{T_{\rm max}+\kappa_{\rm s}\left[T_{{\rm s}0}\,+\,\dfrac{m_{\rm s}}{m_{p}}\dfrac{T_{\rm vir}}{\kappa_{\rm s}}\left(\dfrac{R_{\odot}}{r_{0}}-\dfrac{R_{\odot}}{r}\right)\right]}}\,, \displaystyle T_{\rm max} \displaystyle=\frac{m_{p}v^{2}_{\rm max}}{k_{\rm B}}. (38) We have assumed that the temperature at the reference radius r_{0} of the upper chromosphere is T_{{\rm s}0}=m_{\rm s}\sigma^{2}_{{\rm s}0}/k_{\rm B} . In other words, we have assumed that the plasma is nearly Maxwellian (large \kappa ) in the collisional environment of the upper chromosphere. An important point to note is that T_{{\rm s}0} is not equal to T_{\rm s}(r\to r_{0}^{+}) , which is given by \displaystyle T_{\rm s}(r\to r_{0}^{+}) \displaystyle=\frac{2\kappa_{\rm s}}{3}\frac{\beta\left(\chi^{2}_{\rm s}(r\to r_{0}^{+}),\frac{5}{2},\kappa_{\rm s}-\frac{3}{2}\right)}{\beta\left(\chi^{2}_{\rm s}(r\to r_{0}^{+}),\frac{3}{2},\kappa_{\rm s}-\frac{1}{2}\right)}T_{{\rm s}0}, \displaystyle\xrightarrow{T_{\rm max}/\kappa_{\rm s}T_{{\rm s}0}\to\infty}\frac{\kappa_{\rm s}}{\kappa_{\rm s}-3/2}T_{{\rm s}0}, \displaystyle\chi_{\rm s}(r\to r_{0}^{+}) \displaystyle=\sqrt{\dfrac{T_{\rm max}}{T_{\rm max}+\kappa_{\rm s}T_{{\rm s}0}}}. (39) Note that, for \kappa_{\rm s}\to 3/2 , T_{\rm s}(r\to r^{+}_{0}) heavily exceeds T_{{\rm s}0} , since the pressure diverges (but the density remains finite) in this limit. For large \kappa_{\rm s} , T_{\rm s}\approx T_{{\rm s}0} , i.e., the plasma is nearly isothermal, as expected. The above scalings can also be obtained by solving the equation for hydrostatic equilibrium, \displaystyle\frac{{\rm d}p}{{\rm d}r}+\rho\frac{{\rm d}\Phi}{{\rm d}r}=0, (40) with a macroscopic equation of state (EOS), p=p\left(\rho\right) , obtained from equations ( 28 ) as shown above. For \kappa_{\rm s}>1.5 , the EOS is polytropic, i.e., p\propto\rho^{\gamma} with \gamma=\left(\kappa_{\rm s}-3/2\right)/\left(\kappa_{\rm s}-1/2\right) (as noted by 37 ), while, for \kappa_{\rm s}=1.5 , it is a much milder p\propto\ln{\left(4v^{2}_{\rm max}\,\rho_{\rm s}(r)/3\sigma^{2}_{{\rm s}0}\,\rho_{\rm s}(r_{0})\right)} . Figure 7: Temperature profile in the corona as a function of height above the photosphere, obtained from equation ( 34 ) using the double \kappa distribution given in equation ( 33 ). The symbols denote the temperature inferred from the UV spectroscopic observations of various ionic species ( 35 ; 43 ; 45 ) . The brown solid line denotes the best fit velocity filtration model (\kappa_{1}=1.5,\kappa_{2}=4,x=10^{-9}) to the Fe observations. It is evident from equations ( 36 ) that the temperature (density) is an increasing (decreasing) function of r in the corona. This is because the collisionless nature of the plasma yields a softer than isothermal EOS. As \rho decreases outwards due to gravity, p does so too but not quite enough, so that T\sim p/\rho rises outwards. As \kappa_{\rm s} approaches 1.5 , the EOS gets softer and the temperature rises dramatically. This temperature inversion due to a non-thermal distribution function is known as velocity filtration ( 37 ; 30 ) , which refers to the fact that the gravitational potential well of the sun filters out the fast suprathermal particles and lets them escape. How does the velocity filtration model with a single \kappa distribution compare to the observed temperature profile of the solar corona? Assuming the reference radius r_{0} to be the upper end of the chromosphere, i.e., r_{0}=R_{\odot}+h with R_{\odot}=7\times 10^{5} km and h=2\times 10^{3} km, T_{{\rm s}0}=10^{4} K, and M_{\odot}=2\times 10^{30} kg, the corresponding temperature profile is computed as a function of the height above the photosphere, and plotted for the case of Fe (m_{\rm s}=56) for different values of \kappa in the left panel Fig. 6 and for different species, assuming \kappa_{\rm s}=1.5 , in the right panel. The symbols denote the temperature inferred from the observed spectral (UV) linewidths of heavy ionic species in the upper chromosphere and lower corona ( 35 ; 43 ; 45 ) . We have adopted v_{\rm max}=3\,\sigma_{p0}=3\sqrt{m_{\rm s}/m_{p}}\,\sigma_{{\rm s}0} (the subcript p denotes protons), since this reproduces the coronal temperature of Fe reasonably well. While T_{\rm s} increases outwards in all cases, only for \kappa_{\rm s}\approx 1.5-2 it rises as drastically as the observed profile, with a pressure scale height less than 10^{3} km. For a Maxwellian-like (large \kappa_{\rm s} ) DF, the EOS is nearly isothermal and T_{\rm s} gradually rises, with a large scale height \gtrsim 10^{4} km. For \kappa_{\rm s} closer to 3/2 , the base temperature exceeds the chromospheric temperature T_{{\rm s}0}=10^{4} K by a greater degree. T_{{\rm s}} rises rapidly to a steady value that roughly scales linearly with the atomic mass m_{\rm s}/m_{p} and is as large as 10^{6} K for heavier elements. Leaving aside the deviations due to the details of ionization physics not incorporated in the current model, the predictions of velocity filtration are in reasonable agreement with the data. Although velocity filtration with a single \kappa distribution with \kappa\approx 1.5-2 explains the observed drastic transition in the temperature fairly well, the agreement with the Fe observations farther out in the corona is not as good. The more gradual rise in the corona calls for a large value of \kappa than 1.5 beyond the transition zone. So far, we have assumed the ion DF to be a single \kappa distribution that mimics the first break and power-law fall-off beyond the ion thermal speed. As we showed in Fig. 3 , though, the ion DF is typically a broken power-law, with one power-law between \sigma_{i0} and \sigma_{e0} and another beyond \sigma_{e0} . Such a DF can be expressed as a linear combination of two \kappa distributions as f_{{\rm s}0}(v,r)=(1-x)f_{{\rm s}0}^{(1)}\left(v,r,\kappa_{{\rm s}1},\sigma_{{\rm s}0}^{(1)},v_{\rm max}^{(1)}\right)+xf_{{\rm s}0}^{(2)}\left(v,r,\kappa_{{\rm s}2},\sigma_{{\rm s}0}^{(2)},v_{\rm max}^{(2)}\right) , with \sigma_{{\rm s}0}^{(1)} and \sigma_{{\rm s}0}^{(2)} respectively equal to the ion thermal speed \sigma_{i0} and the electron thermal speed \sigma_{e0} at the upper end of the chromosphere. The corresponding temperature profile is given by equation ( 34 ). We compute this for Fe, adopting \kappa_{1}=1.5 , \kappa_{2}=4 , x=10^{-9} ( x is small since the second \kappa distribution mimics the power-law fall-off beyond \sigma_{e0} and is therefore heavily suppressed relative to the first), v^{(1)}_{\rm max}=2.5\sigma_{p0} and v^{(2)}_{\rm max}=1.2\sigma_{e0} , and plot this as a function of the height above the photosphere as a brown solid line in Fig. 7 . We have denoted the spectroscopic observations ( 35 ; 43 ; 45 ) using the symbols as before. Needless to say, the double \kappa velocity filtration model is a much better fit to the data than a single \kappa one. While the \kappa_{1}=1.5 distribution dictates the steep temperature rise, the \kappa_{2}=4 one governs the gradual increase farther up. The preference for larger values of \kappa farther away from the transition zone could be a consequence of the tendency of the plasma to Maxwellianize through collisions on large scales. The proportionality of T_{\rm s} to the mass of the species m_{\rm s} implies that velocity filtration predicts a much more gradual rise of the temperature of electrons than that of ions. The theory also predicts that the ion to electron temperature ratio T_{i}/T_{e} in the corona is always larger than 1 and can be as high as 10^{2} . This is because the velocities of non-thermal ions and electrons in the weakly collisional corona are determined by the solar gravitational potential, which is the same for all masses, and are therefore more comparable than in the strongly collisional plasma of the solar interior. However, a proper analysis of T_{i}/T_{e} likely requires finite Larmor radius effects in a magnetized plasma, a topic that is beyond the scope of this paper and is left for future work. The basic physics behind coronal heating within the scope of the filtration model is as follows. Gravity acts as a filter and allows the high energy suprathermal particles to escape outwards. The virial temperature of the sun, GM_{\odot}m_{\rm P}/k_{\rm B}R_{\odot} , is \sim 10^{7} K, which indicates that the filtering effect of gravity can invert the temperature profile and raise it to a million degrees K in the corona. However, this occurs only in a non-Maxwellian kinetic plasma with a shallow enough power-law DF, i.e., small enough \kappa . The limiting value of \kappa at which the pressure and temperature diverge (for v_{\rm max}\gg\sigma_{{\rm s}0} ) is 1.5 , which corresponds to a v^{-5} DF and an E^{-2} energy distribution. Such a shallow non-thermal tail works together with gravitational filtering to dramatically raise the temperature within a few 100 km. Interestingly, this is exactly the power-law that is predicted by QLT for a kinetic plasma driven by super-Debye (but sub-Larmor) EM turbulence. Photospheric or chromospheric convection, which manifests as a large number of small-scale reconnection events or nano-flares (footprint motion of the field lines) ( 34 ) , can itself serve as a turbulent drive that heats the plasma to a non-thermal distribution, which can then enable temperature inversion via velocity filtration. As we showed earlier, Maxwellianization through collisions cannot suppress the high v suprathermal tail as long as super-Debye turbulence keeps stirring the plasma. This alleviates the concerns raised ( 2 ; 25 ) about the interference of collisional effects with velocity filtration. Even if collisions Maxwellianize the plasma, it happens over scales much larger than the width of the transition zone between the chromosphere and the corona. What gives rise to the remarkably narrow transition zone after all? This can be reasoned as follows. As we go from the collisional chromosphere into the kinetic corona, the turbulent electric field exerted by chromospheric convection exceeds the collisional one and accelerates particles into a non-thermal tail. QLT predicts that this heating occurs as a diffusive process. The particles accelerate until they reach speeds comparable to the gravitational escape speed, at which point the ambipolar electric field kicks in and keeps the electrons and ions from readily escaping the sun. The time taken by the protons to diffuse from the thermal speed to the (gravitational) escape speed is t_{\rm diff}=v^{2}_{\rm esc}/D^{(p)}(\sigma_{p0})\approx v^{2}_{\rm esc}/{\cal D}_{2}^{(p)}(\sigma_{p0}) , assuming that the drive and collisional diffusion coefficients are comparable. The distance traveled by the protons is therefore given by \displaystyle l_{\rm diff}\approx v_{\rm esc}t_{\rm diff} \displaystyle\approx\lambda_{\rm D}\frac{\Lambda}{\ln\Lambda}\,{\left(\frac{v_{\rm esc}}{\sigma_{p0}}\right)}^{3} \displaystyle\approx 100\,\frac{10^{9}{\rm cm}^{-3}}{n_{e}}\sqrt{\frac{T_{p0}}{10^{4}\,{\rm K}}}\,{\rm km}, (41) which is roughly the observed width of the transition zone. Hence, within the scope of our quasilinear framework, the width is simply determined by the diffusion length of the non-thermal particles escaping from the chromosphere to the corona. Note that this estimate, as well as the predictions of velocity filtration (see Figs. 6 and 7 ), is smaller than that predicted by idealized 1D simulations of 6 ; 7 , which hints at the need for 3D simulations with realistic plasma turbulence to resolve the coronal heating problem. V Conclusion We have formulated a novel theory for NTPA in electromagnetically driven, multi-species, kinetic (weakly collisional) plasmas. The EM drive can either be of external origin or describe the forces exerted by non-linear coherent structures (e.g., BGK holes and plasmoids) on the bulk plasma. A crucial ingredient of the theory is self-consistency through the Maxwell equations, manifested through the Debye screening of large-scale EM fluctuations, something that is ignored in earlier test particle treatments ( 33 ; 23 ) . Using the quasilinear framework for the Vlasov-Maxwell equations, we have derived a general quasilinear transport equation for the relaxation of the mean coarse-grained DF of each charged species under the action of the EM drive, modulated by self-consistent fields, as well as internal turbulence and weak collisions. The transport equation is a Fokker-Planck equation with diffusion and friction/drag coefficients that describe the energy and momentum exchange between the particles and fields. While the Balescu-Lenard (BL) coefficients describe the relaxation due to internal turbulence and Coulomb collisions, the drive diffusion coefficient represents the heating of particles either directly by the EM drive or by the waves excited by it or both. We derive the expressions for the transport coefficients in stable, unstable and marginally stable plasmas, assuming that the instability is not violent enough to negate the quasilinear approximation (e.g., a classic bump-on-tail instability). We find that the direct heating of particles generally dominates over wave-mediated heating in stable plasmas. This is also the case in marginally stable plasmas once the instability has saturated. The drive diffusion coefficient for direct heating typically scales as a power-law in the particle velocity v over an extended range of v . If the plasma is forced by a large-scale (super-Debye) EM drive, i.e., the spatial power-spectrum of the driving electric field scales as k^{-\alpha} with \alpha\geq 5 and k_{\rm min}\lambda_{\rm D}<1 , then D^{({\rm s})}(v) scales as \sim v^{4} between the electron thermal speed \sigma_{e} and the phase-velocity of the Langmuir waves, \omega_{{\rm P}e}/k_{\rm min} . This arises from the universal 1-\omega^{2}_{{\rm P}e}/{({\bf k}\cdot{\bf v})}^{2} scaling of the longitudinal component of the dielectric tensor in the range \sigma_{e}\omega_{{\rm P}e}/k_{\rm min} , D^{({\rm s})}(v) scales as v^{-3} , similar to the BL diffusion coefficient, which, together with the v^{-2} scaling of the BL drag coefficient, yields a Maxwellian high v cut-off for the DF. Therefore, a generic prediction of our theory is that kinetic plasmas driven by super-Debye scale EM turbulence harbor power-law tails in their DFs, with a particular preference for the v^{-5} tail, even in the presence of collisions and instabilities. Our transport equation is more general than the conventional Parker transport equation ( 33 ) , since, unlike the latter, we incorporate the effects of self-consistent EM turbulence as well as collisions on particle acceleration. Our theory provides a general description of NTPA in hot and tenuous kinetic plasmas, be that in fusion plasmas, e.g., inertial confinement fusion devices, or in astrophysical and space plasmas, e.g., the solar corona and solar wind, circumstellar coronae, collsionless shocks from supernova explosions, the hot ionized interstellar medium, and so on. In this paper, we discuss the implications of the theory for the specific problem of solar coronal heating. The velocity filtration model ( 37 ) predicts that the presence of a non-thermal power-law tail in the DF of a plasma confined in an attractive potential raises its temperature (but decreases its density) towards the region of shallower potential. An attractive gravitation potential would therefore filter out the high energy suprathermal particles. This explains the temperature inversion in the solar corona as an outcome of the filtering action of the solar gravitational potential. However, the drastic temperature rise, especially of the heavy ionic species, inferred from the observed linewidths ( 35 ) in the transition region between the upper chromosphere and the lower corona requires very shallow power-law tails with \kappa\approx 1.5-2 for velocity filtration to explain the data. This was deemed implausible ( 2 ; 25 ) since the smallness of the collisional mean free path relative to the pressure scale height was believed to Maxwellianize the plasma on large scales even if kinetic power-law tails exist on small scales. We demonstrate though that this assumption is incorrect. Even in the presence of collisions, a plasma driven by large-scale (super-Debye but sub-Larmor) EM turbulence invariably develops a power-law tail with a particular preference for \kappa=1.5 . Although the mean free path is short, the diffusion length of the fast non-thermal population significantly exceeds the scale height, indicating that these particles accelerate and diffuse away without collisional losses. Turbulent reconnection in the chromosphere in the form of nano-flares ( 34 ) can itself serve as the EM drive that engenders the power-law tail, inverting the coronal temperature profile via velocity filtration, in reasonable agreement with the spectroscopic data. Only hard power-law tails with \kappa\approx 1.5-2 can reproduce the steep transition of temperature and density from the chromosphere to the corona with a scale height \approx 100 km. This abrupt rise in the temperature is an outcome of two facts: (i) the pressure diverges in the \kappa_{\rm s}\to 1.5 limit and (ii) the width of the transition zone is set by the diffusion length of the non-thermal particles (see equation [ 41 ]). Interestingly, Parker Solar Probe measurements of the solar wind DF in the inner heliosphere reveal a similar range of \kappa values as what is required to explain the abrupt transition from the chromosphere to the corona. These measurements, along with the general agreement of the velocity filtration theory with the observed coronal temperature profile, suggest that the non-thermal v^{-5} or a similar power-law DF is very plausibly present in the solar corona. The simple requirements for the generation of this non-thermal tail and its remarkable resilience against collisions imply that it is a universal feature of all weakly collisional plasmas. The general success of filtration in explaining the observed temperature inversion in the solar corona indicates that this mechanism is operant in the weakly collisional environment of all circumstellar and circumplanetary coronae as well as those around black holes and active galactic nuclei. It is important to note that our theory for NTPA only applies to non-relativistic unmagnetized plasmas with super-Debye but sub-Larmor scale turbulence. In a follow up paper, we will present a self-consistent theory for the relaxation of magnetized plasmas with super-Larmor scale (e.g., Alfvénic) turbulence, incorporate relativistic effects, and derive a generalized version of the Parker transport equation for NTPA in the context of the solar wind as well as cosmic ray transport. To this end we will also explore the energy partition between electrons and ions. While our current framework for unmagnetized plasmas predicts similar non-thermal power-laws in the electron and ion DFs, the scenario might change in magnetized plasmas due to their significantly different Larmor radii. In future work we intend to perform a detailed investigation of the various factors responsible for determining the ion to electron temperature ratio, such as the relative strength of sub-Larmor and super-Larmor scale turbulence, the helicity barrier in imbalanced turbulence ( 40 ; 46 ; 1 ) , and cyclotron heating. Acknowledgements. The authors are thankful to Srijan Das, Mihir Desai, Robert Ewart, Barry Ginat, Matthew Kunz, Nuno Loureiro, Michael Nastac, Alex Schekochihin, Anatoly Spitkovsky, Dmitri Uzdensky and Vladimir Zhdankin for stimulating discussions and valuable suggestions. The first author (UB) is especially thankful to Nuno Loureiro for illuminating exchanges and the Plasma Science Fusion Center at MIT for providing an intellectually stimulating environment. This research is supported by the National Science Foundation Award 2206607 at the Multi-Messenger Plasma Physics Center (MPPC), Princeton University, and the Bezos Membership grant at the IAS. Data Availability Statement AVAILABILITY OF DATA STATEMENT OF DATA AVAILABILITY Data available on request from the authors The data that support the findings of this study are available from the corresponding author upon reasonable request. Appendix A Computation of D_{p}^{(s)}(v) The dressed particle diffusion coefficient D_{p}^{(s)}(v) is given by \displaystyle D^{({\rm s})}_{p}(v) \displaystyle=\frac{32\pi^{5}q^{2}_{\rm s}}{m^{2}_{\rm s}V} \displaystyle\times\int_{0}^{\infty}{\rm d}k\,k^{2}{\cal E}(k)\int_{0}^{1}{\rm d}\cos{\theta}\,\cos^{2}\theta\,\frac{{\cal C}_{\omega}\left(kv\cos{\theta}\right)}{{\left|\varepsilon_{k\parallel}\left(kv\cos\theta\right)\right|}^{2}}. (42) In the range \sigma_{e}1 ), I\approx 1/5{\left(kvt_{\rm c}\right)}^{2} . Therefore, the asymptotic scalings of D^{({\rm s})}_{p}(v) (for \omega_{{\rm P}e}t_{\rm c}>1 ) are: \displaystyle D^{({\rm s})}_{p}(v)\approx\dfrac{32\pi^{5}q^{2}_{\rm s}}{m^{2}_{\rm s}V}\times\begin{cases}\dfrac{v^{4}}{7\omega^{4}_{{\rm P}e}}\mathop{\text{\Large$\int_{\text{\normalsize$\scriptstyle\kern-1.63885pt\;0$}}^{\text{\normalsize$\scriptstyle\infty$}}$}}\nolimits{\rm d}k\,k^{6}\,{\cal E}(k),\qquad\;\sigma\ll v\ll\dfrac{1}{k\tau_{\rm c}},\\ \\ \dfrac{v^{2}}{5\omega^{4}_{{\rm P}e}t^{2}_{\rm c}}\mathop{\text{\Large$\int_{\text{\normalsize$\scriptstyle\kern-1.63885pt\;0$}}^{\text{\normalsize$\scriptstyle\infty$}}$}}\nolimits{\rm d}k\,k^{4}\,{\cal E}(k),\quad\dfrac{1}{kt_{\rm c}}\ll v\ll\dfrac{\omega_{{\rm P}e}}{k},\\ \\ \dfrac{1}{t^{2}_{\rm c}v^{2}}\mathop{\text{\Large$\int_{\text{\normalsize$\scriptstyle\kern-1.63885pt\;0$}}^{\text{\normalsize$\scriptstyle\infty$}}$}}\nolimits{\rm d}k\,{\cal E}(k),\qquad\quad\;\;v\gg\dfrac{\omega_{{\rm P}e}}{k}.\end{cases} (47) In deriving the last scaling, we have used the fact that, at v\gg\omega_{{\rm P}e}/k , \varepsilon_{k\parallel}\approx 1 . Appendix B Landau damping vs quasilinear relaxation The wave-mediated diffusion can be neglected relative to direct diffusion by the drive in the stable regime if the Landau damping timescale is shorter than that of quasilinear relaxation. In the large mean free path case, \nu_{{\rm c}{\rm s}}\ll k\sigma_{\rm s} , and on super-Debye scales ( k\lambda_{{\rm D}}\ll 1 ), the electron and ion Langmuir waves damp away at a rate faster than the quasilinear relaxation rate \sim{\left(k\lambda_{\rm D}\right)}^{4}\sigma^{2}_{{\rm s}}/2D_{0}^{({\rm s})} ( D_{0}^{({\rm s})} is the diffusion coefficient at v\gtrsim\omega_{{\rm P}e}/k ), at which the power-law tail develops at intermediate v , by a factor of \sim{\left(k\lambda_{{\rm D}}\right)}^{-2} . The ion-acoustic waves damp away even faster, at a rate \sim{\left(k\lambda_{{\rm D}}\right)}^{-3} higher than the quasilinear relaxation rate. On sub-Debye ( k\lambda_{{\rm D}}\gtrsim 1 ) scales, Landau damping is efficient and rapidly damps away the waves on a timescale \sim{\left(k\sigma_{\rm s}\right)}^{-1}\sim{\left(k\lambda_{{\rm D}}\right)}^{-1}{\omega}^{-1}_{{\rm P}{\rm s}} . Therefore, the longitudinal modes do not contribute to quasilinear diffusion at long time. The transverse light waves do not Landau damp away, though, and can potentially contribute. However, for an isotropic plasma subject to isotropic perturbations, the transverse components do not show up in the quasilinear transport equation. Under anisotropic conditions, they do, and the undamped light waves contribute, albeit with a strength suppressed by a factor \sim{\left(\sum_{\rm s}\omega^{2}_{{\rm P}{\rm s}}/c^{2}k^{2}\right)}^{2} relative to direct diffusion for sub-skin depth ( k\gtrsim\sqrt{\sum_{\rm s}\omega^{2}_{{\rm P}{\rm s}}}/c ) perturbations. In the small mean free path limit (\nu_{{\rm c}{\rm s}}\gg k\sigma_{\rm s}) , the least damped Landau mode (the Lenard-Bernstein mode ( 26 ; 4 ) ), damps at the rate, {\left(k\sigma_{\rm s}\right)}^{2}/\nu_{{\rm c}{\rm s}}\sim{\left(k\lambda_{\rm D}\right)}^{2}\omega_{{\rm P}{\rm s}}\,\Lambda/\ln\Lambda . The ratio of this damping rate to the quasilinear diffusion rate is equal to \sim{\left(k\lambda_{\rm D}\right)}^{-2}\Lambda/\ln\Lambda . Hence, the Landau modes damp faster than the power-law tail develops for super-Debye (k\lambda_{\rm D}\ll 1) perturbations, even more so in a more collisionless environment (larger \Lambda ). Based on the above considerations, the waves/Landau term can be neglected in the diffusion coefficient for stable plasmas (see also equation [A19] of 3 ). Appendix C Entropy considerations Can we construct an entropy functional for the effective collision operator in equation ( 16 ), that satisfies the H-theorem (never decreases with time) and yields the above DF as an extremal solution? It can be shown that S=-\int{\rm d}^{3}x\,{\rm d}^{3}v\,G(f) with any convex function G(f) follows the H-theorem. Since the DF approaches a \kappa distribution at long times in the presence of a large-scale EM drive, G(f)\sim\left(f^{q}-1\right)/\left(q-1\right) with q=\kappa/\left(1+\kappa\right) is the function for which the entropy S , when maximized with the constraints of total energy, momentum and particle number conservation, yields this DF as the extremal solution. The corresponding entropy S is nothing but the Tsallis entropy ( 42 ; 27 ; 47 ; 48 ) . And, since our DF typically scales as v^{-5} at large v , \kappa=1.5 and q=3/5 are the preferred values. Interestingly, 13 and 14 recast the entropy into a generalized Boltzmann-Shannon form, S=\int{\rm d}{\bf v}\int{\rm d}\eta\,P({\bf v},\eta)\ln{P({\bf v},\eta)} , with P(\eta) the probability that f (treated as a random variable) takes the value \eta . Maximizing this with the constraints of the conservation of probability, total energy and phase-volume (waterbag content), \rho(\eta)=\int{\rm d}^{3}v\,P({\bf v},\eta)=\frac{1}{V}\int{\rm d}{\bf x}\int{\rm d}{\bf v}\,\delta\left(f({\bf x},{\bf v})-\eta\right) , they obtain a Fermi-Dirac distribution for P({\bf v},\eta) , along the lines of 29 . Noting that \rho(\eta)=\int{\rm d}{\bf v}\,\delta\left(f_{\rm G}-\eta\right) ( f_{\rm G} is the Gardener distribution that is a function of E and not {\bf x} and has the same \rho(\eta) as f ) scales as \eta^{-1} for exponentially truncated (or steeper) f_{\rm G} , they find that the Fermi-Dirac P({\bf v},\eta) yields N(E)\sim E^{-2} in the non-degenerate limit. It is, however, a priori, not clear why the Gardener distribution would be of such a form. It is easy to see that, for f_{\rm G}(v)\sim v^{-n} , \rho(\eta)\sim\eta^{-\left(1+1/n\right)} (as also pointed out by 13 ), which would yield a shallower tail than E^{-2} . The derivation of the E^{-2} tail from this generalized Lynden-Bell approach hinges on (i) a sufficient deviation of f_{{\rm s}0} from f_{\rm G} and (ii) the \eta^{-1} scaling of \rho(\eta) . As shown by 14 using \left(1x,1v\right) PIC simulations of two-stream instability in electrostatic plasmas, the former depends on the precise nature of Vlasov turbulence (e.g., presence or absence of BGK holes). Interestingly, the E^{-2} tail appears in the spatially averaged energy distribution after the saturation of the two-stream instability, only when BGK holes appear in the phase-space, such as in an electron-ion plasma. The electron-positron case neither features large-scale holes nor the E^{-2} tail. Therefore, it appears that the appearance of the tail is intimately related to the occurrence of a large-scale drive, which, in this case, consists of the electric fields exerted by the BGK holes themselves on the bulk plasma. And the BGK holes are an outcome of the self-consistent plasma response. In other words, self-consistency through the Poisson equation, something that the entropy approach does not explicitly take into account but our kinetic formalism does, is crucial for the emergence of the power-law tail. As we show, it is the Debye-shielding of large-scale EM fields (e.g., those exerted by coherent structures as in 14 ’s simulation) and the consequent suppression of particle heating at low v , that fundamentally gives rise to the tail.