Field-Particle Correlation Technique

• Field–particle correlation is a diagnostic that correlates electric-field fluctuations with velocity-space gradients to quantify net energy exchange in plasmas.
• It uses single-point measurements and a carefully chosen correlation interval to cancel oscillatory effects and expose resonant interactions like Landau damping.
• Various implementations—from 1D–1V to gyrokinetic and Fourier-space methods—provide detailed insights into energy transfer mechanisms in turbulent plasmas.

Field–particle correlation is a plasma diagnostic that uses single-point measurements of electromagnetic fields and particle velocity distribution functions to isolate the secular transfer of energy between fields and particles and to reveal its structure in velocity space. In the plasma-kinetic literature, the technique is formulated by correlating electric-field fluctuations with velocity-space gradients of the distribution function, over a correlation interval long enough to suppress oscillatory “sloshing” and retain the net energy exchange. Its defining output is not only a local rate of energization, but also a velocity-space signature that can be compared with resonant conditions such as Landau, cyclotron, and related wave–particle interactions (Klein et al., 2016, Klein et al., 2017, Klein et al., 2020).

1. Vlasov basis and the definition of the correlation

The technique is derived from the collisionless Vlasov equation. For species $s$,

$$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$

Multiplication by the single-particle kinetic energy $m_s v^2/2$ defines the phase-space energy density,

$$\Theta_s(\mathbf r,\mathbf v,t)\equiv \frac{m_s v^2}{2}f_s(\mathbf r,\mathbf v,t),$$

or equivalently $w_s=\tfrac12 m_s v^2 f_s$ in the 1D–1V formulation. Its evolution contains a ballistic term, a field–particle term, and a Lorentz-force term. Under periodic or open boundary conditions the ballistic contribution vanishes after the appropriate spatial integration, and the $(\mathbf v\times \mathbf B)$ term integrates to zero over velocity, so the electric-field term is isolated as the part responsible for net energization (Klein et al., 2016, Klein et al., 2020).

This leads to the central correlation

$$C_E(\mathbf r,\mathbf v,t,\tau)\equiv \left\langle -\,q_s\,\frac{v^2}{2}\, \left[\frac{\partial \delta f_s(\mathbf r,\mathbf v,\cdot)}{\partial \mathbf v}\right]\cdot \mathbf E(\mathbf r,\cdot) \right\rangle_{t-\tau/2}^{t+\tau/2},$$

where $\delta f_s\equiv f_s-F_{0,s}$, $F_{0,s}$ is the time-average over the interval $T$, and $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$0 denotes a uniform average over the time window $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$1. In the original 1D–1V electrostatic formulation this appears as

$$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$2

Provided $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$3 exceeds the characteristic wave period of the dominant fluctuation, the fast oscillatory exchange cancels out in the sum, leaving a nonzero mean only where secular damping or growth occurs (Klein et al., 2016).

A central conceptual point is that the method retains the full $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$4 dependence of the energy-transfer term at a fixed point. The velocity integral of the correlation recovers the local field–particle work, but the unintegrated structure shows which particles gain or lose energy. This is the basis for mechanism identification from single-point data (Klein et al., 2017).

2. Operational forms of the technique

In magnetized plasmas the electric field is split into components parallel and perpendicular to the local magnetic field. In field-aligned coordinates, the hybrid Vlasov–Maxwell implementation defines

$$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$5

and

$$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$6

The discrete correlation is

$$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$7

Velocity integration gives the local spatial energy-density transfer rates,

$$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$8

with $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$9. In the Alfvén–Ion Cyclotron study, all $m_s v^2/2$0 are normalized by the time-averaged local proton energy density

$m_s v^2/2$1

and the accumulated transfer is

$m_s v^2/2$2

These constructions make the technique quantitative rather than merely qualitative (Klein et al., 2020).

A practically important variant avoids the velocity derivative. For 1D–1V electrostatic data one may define

$m_s v^2/2$3

and in gyrokinetic form

$m_s v^2/2$4

By integration by parts in velocity, $m_s v^2/2$5, and the same holds for the gyrokinetic proxy. This derivative-free form is explicitly motivated by the difficulty of computing $m_s v^2/2$6 from noisy, discretized spacecraft measurements (Klein et al., 2016, Klein et al., 2017).

3. Velocity-space signatures and mechanism identification

The distinctive feature of the technique is that different energization mechanisms imprint different velocity-space signatures. In resonant interactions such as Landau damping, the correlation is localized near the resonant velocity $m_s v^2/2$7 or $m_s v^2/2$8. In the electrostatic tutorial, if the wave is damped, $m_s v^2/2$9 is positive for $$\Theta_s(\mathbf r,\mathbf v,t)\equiv \frac{m_s v^2}{2}f_s(\mathbf r,\mathbf v,t),$$0 and negative for $$\Theta_s(\mathbf r,\mathbf v,t)\equiv \frac{m_s v^2}{2}f_s(\mathbf r,\mathbf v,t),$$1, corresponding to flattening of the VDF around $$\Theta_s(\mathbf r,\mathbf v,t)\equiv \frac{m_s v^2}{2}f_s(\mathbf r,\mathbf v,t),$$2 (Klein et al., 2016).

In gyrokinetic turbulence, where the dominant resonance is $$\Theta_s(\mathbf r,\mathbf v,t)\equiv \frac{m_s v^2}{2}f_s(\mathbf r,\mathbf v,t),$$3, the reduced parallel correlation $$\Theta_s(\mathbf r,\mathbf v,t)\equiv \frac{m_s v^2}{2}f_s(\mathbf r,\mathbf v,t),$$4 shows a bipolar signature in $$\Theta_s(\mathbf r,\mathbf v,t)\equiv \frac{m_s v^2}{2}f_s(\mathbf r,\mathbf v,t),$$5 about $$\Theta_s(\mathbf r,\mathbf v,t)\equiv \frac{m_s v^2}{2}f_s(\mathbf r,\mathbf v,t),$$6. For fully developed 3D turbulence at $$\Theta_s(\mathbf r,\mathbf v,t)\equiv \frac{m_s v^2}{2}f_s(\mathbf r,\mathbf v,t),$$7, the net $$\Theta_s(\mathbf r,\mathbf v,t)\equiv \frac{m_s v^2}{2}f_s(\mathbf r,\mathbf v,t),$$8 is strongly localized in $$\Theta_s(\mathbf r,\mathbf v,t)\equiv \frac{m_s v^2}{2}f_s(\mathbf r,\mathbf v,t),$$9 around the range of $w_s=\tfrac12 m_s v^2 f_s$0 corresponding to peak proton damping, and a resonant fraction

$w_s=\tfrac12 m_s v^2 f_s$1

is found to satisfy $w_s=\tfrac12 m_s v^2 f_s$2–$w_s=\tfrac12 m_s v^2 f_s$3, well above the $w_s=\tfrac12 m_s v^2 f_s$4–$w_s=\tfrac12 m_s v^2 f_s$5 expected for a uniform-in-velocity transfer (Klein et al., 2017).

The Alfvén–Ion Cyclotron turbulence study extends the method beyond purely Landau channels. There, $w_s=\tfrac12 m_s v^2 f_s$6 shows the canonical bipolar signature of Landau damping around $w_s=\tfrac12 m_s v^2 f_s$7, with negative correlation for $w_s=\tfrac12 m_s v^2 f_s$8 and positive for $w_s=\tfrac12 m_s v^2 f_s$9, extended out to $(\mathbf v\times \mathbf B)$0–$(\mathbf v\times \mathbf B)$1, and with weak $(\mathbf v\times \mathbf B)$2 dependence for $(\mathbf v\times \mathbf B)$3. By contrast, $(\mathbf v\times \mathbf B)$4 peaks in a broad band $(\mathbf v\times \mathbf B)$5, $(\mathbf v\times \mathbf B)$6, forming a “quasilinear plateau” consistent with contours

$(\mathbf v\times \mathbf B)$7

The two channels are therefore separable even when they act simultaneously (Klein et al., 2020).

The method also distinguishes unstable mechanisms. In counter-streaming electrostatic instability, each electron beam’s $(\mathbf v\times \mathbf B)$8 shows a sign-flip at its drift speed $(\mathbf v\times \mathbf B)$9, with no sharp structure at $$C_E(\mathbf r,\mathbf v,t,\tau)\equiv \left\langle -\,q_s\,\frac{v^2}{2}\, \left[\frac{\partial \delta f_s(\mathbf r,\mathbf v,\cdot)}{\partial \mathbf v}\right]\cdot \mathbf E(\mathbf r,\cdot) \right\rangle_{t-\tau/2}^{t+\tau/2},$$0, and the ion $$C_E(\mathbf r,\mathbf v,t,\tau)\equiv \left\langle -\,q_s\,\frac{v^2}{2}\, \left[\frac{\partial \delta f_s(\mathbf r,\mathbf v,\cdot)}{\partial \mathbf v}\right]\cdot \mathbf E(\mathbf r,\cdot) \right\rangle_{t-\tau/2}^{t+\tau/2},$$1 is even in $$C_E(\mathbf r,\mathbf v,t,\tau)\equiv \left\langle -\,q_s\,\frac{v^2}{2}\, \left[\frac{\partial \delta f_s(\mathbf r,\mathbf v,\cdot)}{\partial \mathbf v}\right]\cdot \mathbf E(\mathbf r,\cdot) \right\rangle_{t-\tau/2}^{t+\tau/2},$$2 and positive overall, implying net ion heating. In bump-on-tail instability, the bump population shows $$C_E(\mathbf r,\mathbf v,t,\tau)\equiv \left\langle -\,q_s\,\frac{v^2}{2}\, \left[\frac{\partial \delta f_s(\mathbf r,\mathbf v,\cdot)}{\partial \mathbf v}\right]\cdot \mathbf E(\mathbf r,\cdot) \right\rangle_{t-\tau/2}^{t+\tau/2},$$3 around the acoustic resonant velocity, while the core population shows $$C_E(\mathbf r,\mathbf v,t,\tau)\equiv \left\langle -\,q_s\,\frac{v^2}{2}\, \left[\frac{\partial \delta f_s(\mathbf r,\mathbf v,\cdot)}{\partial \mathbf v}\right]\cdot \mathbf E(\mathbf r,\cdot) \right\rangle_{t-\tau/2}^{t+\tau/2},$$4 peaked at the same resonance (Klein, 2017).

In a perpendicular collisionless shock, the ion correlation $$C_E(\mathbf r,\mathbf v,t,\tau)\equiv \left\langle -\,q_s\,\frac{v^2}{2}\, \left[\frac{\partial \delta f_s(\mathbf r,\mathbf v,\cdot)}{\partial \mathbf v}\right]\cdot \mathbf E(\mathbf r,\cdot) \right\rangle_{t-\tau/2}^{t+\tau/2},$$5 exhibits nested blue/red crescents around the reflected population, identifying shock-drift acceleration, while in a transverse-drift frame the electron $$C_E(\mathbf r,\mathbf v,t,\tau)\equiv \left\langle -\,q_s\,\frac{v^2}{2}\, \left[\frac{\partial \delta f_s(\mathbf r,\mathbf v,\cdot)}{\partial \mathbf v}\right]\cdot \mathbf E(\mathbf r,\cdot) \right\rangle_{t-\tau/2}^{t+\tau/2},$$6 yields a two-lobed positive signature of adiabatic heating. These signatures were reproduced with simplified “step-shock” and “linear-ramp” models, giving a direct interpretive bridge between Eulerian correlations and Lagrangian particle dynamics (Juno et al., 2020).

4. Numerical, spectral, and analytic realizations

The first explicit development of the technique was in the electrostatic 1D–1V Vlasov–Poisson system, where it was introduced as a practical, single-point means to diagnose how collisionless plasma fluctuations transfer energy to particles and to discuss how the method could be implemented on spacecraft data in the solar wind (Klein et al., 2016).

It was then extended to strongly driven electromagnetic gyrokinetic turbulence. In that setting, gyrokinetics provides the rigorous low-frequency, anisotropic limit of Vlasov–Maxwell, retaining Landau damping while filtering out high-frequency cyclotron resonances. Single-point time series of the complementary gyrocenter distribution $$C_E(\mathbf r,\mathbf v,t,\tau)\equiv \left\langle -\,q_s\,\frac{v^2}{2}\, \left[\frac{\partial \delta f_s(\mathbf r,\mathbf v,\cdot)}{\partial \mathbf v}\right]\cdot \mathbf E(\mathbf r,\cdot) \right\rangle_{t-\tau/2}^{t+\tau/2},$$7 and $$C_E(\mathbf r,\mathbf v,t,\tau)\equiv \left\langle -\,q_s\,\frac{v^2}{2}\, \left[\frac{\partial \delta f_s(\mathbf r,\mathbf v,\cdot)}{\partial \mathbf v}\right]\cdot \mathbf E(\mathbf r,\cdot) \right\rangle_{t-\tau/2}^{t+\tau/2},$$8 are recorded at fixed points, and the gyrotropic correlation $$C_E(\mathbf r,\mathbf v,t,\tau)\equiv \left\langle -\,q_s\,\frac{v^2}{2}\, \left[\frac{\partial \delta f_s(\mathbf r,\mathbf v,\cdot)}{\partial \mathbf v}\right]\cdot \mathbf E(\mathbf r,\cdot) \right\rangle_{t-\tau/2}^{t+\tau/2},$$9 or its reduced form is constructed directly from the turbulent simulation output (Klein et al., 2017).

A spectral generalization decomposes the transfer rate mode by mode in a periodic domain. Writing

$\delta f_s\equiv f_s-F_{0,s}$0

one obtains

$\delta f_s\equiv f_s-F_{0,s}$1

and

$\delta f_s\equiv f_s-F_{0,s}$2

This supplies scale-by-scale accounting of energy flow in kinetic turbulence, but the same work notes that it requires full 3D spatial information and is therefore impractical for single-spacecraft data (Li et al., 2019).

The hybrid Vlasov–Maxwell implementation resolves both Landau and cyclotron channels. In the reported simulation, the ion Vlasov equation is integrated on a $\delta f_s\equiv f_s-F_{0,s}$3 grid in physical space and $\delta f_s\equiv f_s-F_{0,s}$4 in velocity space, with isothermal fluid electrons, box size $\delta f_s\equiv f_s-F_{0,s}$5, $\delta f_s\equiv f_s-F_{0,s}$6, $\delta f_s\equiv f_s-F_{0,s}$7, $\delta f_s\equiv f_s-F_{0,s}$8, and a velocity grid $\delta f_s\equiv f_s-F_{0,s}$9. Sixty-four fixed points are selected, and at each point the full 3-dimensional ion distribution and the simulation-frame fields are recorded. The analysis transforms the fields and $F_{0,s}$0 to the local mean-flow frame and rotates into an instantaneous field-aligned basis before computing $F_{0,s}$1 and $F_{0,s}$2 (Klein et al., 2020).

The shock application uses a 1D–2V continuum Vlasov–Maxwell simulation in Gkeyll. There, because the self-consistent shock is quasi-stationary, the analysis uses $F_{0,s}$3 and directly evaluates componentwise instantaneous correlations $F_{0,s}$4 and $F_{0,s}$5 in the shock-rest frame, after the Galilean transformation $F_{0,s}$6 and $F_{0,s}$7 (Juno et al., 2020).

A recent analytic realization is JET-PLUME, an extension of the PLUME hot-plasma dispersion solver. In Fourier space and for a single $F_{0,s}$8 mode in the weak-damping limit,

$F_{0,s}$9

or equivalently

$T$0

Because $T$1 contains the resonant denominator $T$2, the analytic correlation isolates individual resonances directly and can separate degenerate entropy-mode components (Brown et al., 18 Jun 2026).

5. Quantitative findings and observational reach

The Alfvén–Ion Cyclotron turbulence study supplies one of the clearest quantitative demonstrations of the technique. Averaging over 64 points and over one nonlinear turnover time yields

$T$3

so $T$4 mediates $T$5 of the total transfer and $T$6 $T$7. At every point, the velocity-space pattern is reproducible: bipolar in $T$8 for the parallel channel and plateau-like in $T$9 for the perpendicular channel. Over the chosen correlation window $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$00, approximately $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$01 of the largest-scale Alfvén period, the oscillatory component is largely removed and the net secular patterns persist with small scatter across points (Klein et al., 2020).

The Fourier-space treatment shows that, in a $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$02, solar-wind-like, low-frequency Alfvénic turbulence simulation, all diagnosed Fourier modes covering the dissipation range display resonant energy transfer localized at the Landau resonances for each mode. In the reported scale survey, the width in $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$03 remains narrow, $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$04, the accumulated $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$05 peaks at the lowest $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$06 mode and drops by $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$07 across the dissipation range, electrons and ions receive comparable transfer at $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$08, and electron transfer remains significant out to $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$09–$$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$10 whereas ions are effectively damped only around $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$11–$$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$12 (Li et al., 2019).

The technique has also been tested under severe cadence limitations. In a high-resolution gyrokinetic simulation, 38 distinct Landau signatures were tracked through systematic downsampling; 26, approximately $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$13, remained detectable at super-Nyquist frequencies, in some cases with $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$14. A rule of thumb was extracted from the undersampling study: to recover the true time-average of the product within $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$15, one needs at least $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$16 samples spanning the correlation interval, or $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$17. Signatures appearing with fewer than twelve samples in $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$18 may not be statistically significant (Horvath et al., 2022).

The observational motivation is explicit throughout the literature. The original proposal discussed application to spacecraft missions such as the Magnetospheric Multiscale and Solar Probe Plus missions, and the later turbulence studies state that the method is readily applicable to in situ spacecraft data because it requires only local $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$19 and $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$20 (Klein et al., 2016, Klein et al., 2020). Electron Landau damping has been found in both simulations and observations from Earth’s magnetosheath using this technique, and the undersampling analysis concludes that Parker Solar Probe should be able to identify Landau damping signatures by correlating over intervals of a few seconds (Horvath et al., 2022).

6. Assumptions, interpretive issues, and recent extensions

The method depends critically on the choice of correlation interval $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$21. The interval must be long enough to cancel oscillatory exchange and short enough not to smear secular evolution. In the gyrokinetic and electrostatic treatments, $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$22 is typically chosen to exceed one or two wave periods of the target mode or broadband fluctuation. The Fourier-space study notes that overlapping mode spectra can complicate isolation if mode lifetimes are short (Klein et al., 2016, Li et al., 2019).

An important interpretive caution is that the technique isolates secular transfer, but collisionless interactions are not simply equivalent to monotonic heating. In the Alfvén–Ion Cyclotron simulation, the instantaneous sign of local transfer can reverse because collisionless interactions are reversible, even though the time-averaged velocity-space signatures persist. A plausible implication is that the correlation is best understood as a diagnostic of phase-space energization channels, with collisional thermalization remaining a separate step (Klein et al., 2020, Li et al., 2019).

Noise and discretization in velocity space are a recurrent practical limitation. Both the original and gyrokinetic papers therefore introduce alternative proxy correlations that avoid direct computation of $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$23, while retaining the same integrated $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$24. This is particularly relevant for spacecraft VDFs with finite cadence and finite velocity resolution (Klein et al., 2016, Klein et al., 2017).

Model assumptions also delimit which mechanisms can be accessed. Gyrokinetics filters out high-frequency cyclotron resonances and excludes stochastic ion heating in large-amplitude turbulence, so perpendicular energization requires extensions beyond gyrokinetics. The full-Vlasov and hybrid Vlasov–Maxwell studies were introduced precisely to diagnose cyclotron, shock, and related perpendicular channels (Klein et al., 2017, Klein et al., 2020, Juno et al., 2020).

Recent analytic work extends the technique from diagnostics to controlled resonance isolation. In JET-PLUME, the susceptibility tensor is used to decompose parallel energy transfer into a direct Landau term and an off-diagonal term,

$$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$25

For kinetic Alfvén waves at $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$26, the off-diagonal channel can drive a parallel ion current at large $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$27 with phase relative to $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$28 that returns field energy, partially counteracting the direct Landau damping from $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$29. The resulting velocity-space signature “twists” sign as $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$30 increases. Because the analytic solution is built from a sum over cyclotron harmonics $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$31, Landau ($$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$32) and cyclotron ($$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$33) resonances can be isolated by retaining the corresponding term in the Bessel sums, and degenerate entropy modes can be combined or separated by keeping the mixed cross terms that survive the time average (Brown et al., 18 Jun 2026).

Taken together, these developments define the field–particle correlation technique as a family of closely related diagnostics that transform local measurements of $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$34 and fields into a phase-space map of collisionless energy transfer. Its distinctive contribution is that it does not stop at the scalar work rate $$\frac{\partial f_s}{\partial t} + \mathbf v\cdot \nabla f_s + \frac{q_s}{m_s}\left[\mathbf E + \frac{\mathbf v\times \mathbf B}{c}\right]\cdot \frac{\partial f_s}{\partial \mathbf v}=0.$$35: it identifies which particles, at which velocities, are involved in the exchange, and thereby ties measured energization directly to the underlying physical resonance.