Lag Polyhedral Derivative Ensemble (LPDE)

• Lag Polyhedral Derivative Ensemble (LPDE) is a technique that uses multi-spacecraft lag vectors assembled into tetrahedra to estimate the full 3D divergence of the Yaglom flux in anisotropic MHD turbulence.
• The method employs precise tetrahedral ensemble sampling, avoiding assumptions like isotropy and Taylor’s hypothesis, to compute the inertial-range energy cascade rate.
• LPDE leverages both statistical and geometric strategies to convert sparse in-situ measurements into robust estimates of turbulent energy transfer, highlighting sensitivity to spacecraft separation and lag-space quality.

Lag Polyhedral Derivative Ensemble (LPDE) is a multi-spacecraft measurement technique for estimating the three-dimensional turbulent energy cascade rate and the associated Yaglom flux in anisotropic incompressible magnetohydrodynamic turbulence. It is formulated in lag space, uses polyhedral—effectively tetrahedral—sets of lag vectors, and evaluates the differential form of the anisotropic third-order law directly from simultaneous measurements, without imposing isotropy and, in its core implementation, without relying on Taylor’s frozen-in-flow hypothesis. In the Earth’s magnetosheath, LPDE was introduced as a practical strategy for making the incompressible MHD von Kármán–Howarth/Yaglom exact law operational with four-point Magnetospheric Multiscale (MMS) observations; a later simulation study placed it alongside direction-averaging (DA) as one of two non-isotropic estimators for inertial-range energy transfer (Pecora, 2024, Gao et al., 5 May 2026).

1. Definition and scope

LPDE is named for its three constitutive ideas. “Lag” indicates that the governing exact law is written in terms of derivatives with respect to the lag vector $\boldsymbol{\ell}$, so the natural domain of the method is three-dimensional lag space rather than physical space or a one-dimensional sampling direction. “Polyhedra” refers to the use of collections of lag-space points assembled into 3D simplices, specifically tetrahedra, over which a divergence can be estimated by curlometer-like multi-point methods. “Derivative” refers to the estimation of $\nabla_{\boldsymbol{\ell}}\!\cdot \mathbf{Y}$, the lag-space divergence of the Yaglom flux vector. “Ensemble” has a dual meaning: the statistical average in the definition of the third-order moment, and the combinatorial ensemble of many lag-space tetrahedra produced from a single four-spacecraft snapshot (Pecora, 2024).

The method does not introduce a new turbulence law. Its role is to provide a practical measurement strategy for evaluating the full anisotropic third-order law in three dimensions from sparse in-situ data. The central scientific motivation is that standard single-spacecraft approaches commonly invoke Taylor’s hypothesis and then often assume isotropy, reducing the cascade estimate to a one-dimensional radial relation. LPDE instead targets the quantity that the exact anisotropic law directly governs, namely the full three-dimensional divergence in lag space, $$\nabla_{\boldsymbol\ell}\!\cdot \mathbf{Y}^\pm(\boldsymbol\ell)$$, using simultaneous multi-spacecraft measurements (Pecora, 2024, Gao et al., 5 May 2026).

A basic geometric requirement follows immediately. Because a three-dimensional divergence estimate in lag space requires simultaneous sampling at multiple lag vectors, LPDE requires at least four spacecraft. This four-point minimum is fundamental to the method’s measurement logic rather than an incidental implementation detail (Pecora, 2024).

2. Exact-law foundation in incompressible MHD

The theoretical basis of LPDE is the incompressible MHD extension of the von Kármán–Howarth equation due to Politano and Pouquet, written in terms of the Elsässer fields

$\mathbf{z}^{\pm} = \mathbf{v} \pm \mathbf{b},$

where $\mathbf{v}$ is the plasma velocity and $\mathbf{b}$ is the magnetic field in Alfvén units, normalized by $\sqrt{4\pi n_p m_p}$. The field increments over lag $\boldsymbol{\ell}$ are

$$\delta \mathbf{z}^{\pm}(\boldsymbol{\ell}) = \mathbf{z}^{\pm}(\mathbf{x}) - \mathbf{z}^{\pm}(\mathbf{x}+\boldsymbol{\ell}).$$

The full exact law is

$\frac{\partial}{\partial t}\langle |\delta \mathbf{z}^\pm|^2 \rangle + \boldsymbol{\nabla}_{\boldsymbol{\ell} \cdot \langle \delta \mathbf{z}^\mp |\delta \mathbf{z}^\pm|^2 \rangle - 2 \nu \nabla_{\boldsymbol{\ell}^2 \langle |\delta \mathbf{z}^\pm|^2 \rangle = -4 \epsilon^\pm,$

with temporal, nonlinear-transfer, and dissipation contributions conventionally labeled $\nabla_{\boldsymbol{\ell}}\!\cdot \mathbf{Y}$0, $\nabla_{\boldsymbol{\ell}}\!\cdot \mathbf{Y}$1, and $\nabla_{\boldsymbol{\ell}}\!\cdot \mathbf{Y}$2 (Pecora, 2024).

The mixed third-order moment

$\nabla_{\boldsymbol{\ell}}\!\cdot \mathbf{Y}$3

is the Yaglom flux vector. In an inertial-range regime where the temporal and dissipation terms may be neglected, the exact law reduces to

$\nabla_{\boldsymbol{\ell}}\!\cdot \mathbf{Y}$4

The mean transfer rate is

$\nabla_{\boldsymbol{\ell}}\!\cdot \mathbf{Y}$5

The cascade rate is therefore obtained from the divergence of the Yaglom flux in lag space (Pecora, 2024).

A crucial point is that this divergence form holds under homogeneity whether or not the turbulence is isotropic. Isotropy becomes necessary only when one integrates the divergence law into the standard one-dimensional radial formula

$\nabla_{\boldsymbol{\ell}}\!\cdot \mathbf{Y}$6

with

$\nabla_{\boldsymbol{\ell}}\!\cdot \mathbf{Y}$7

LPDE is defined precisely by retaining the full vector/divergence form rather than collapsing the problem to a radial projection (Gao et al., 5 May 2026).

3. Lag-space geometry and derivative ensemble

Operationally, LPDE converts the inter-spacecraft baselines of a four-spacecraft constellation into lag vectors. For spacecraft $\nabla_{\boldsymbol{\ell}}\!\cdot \mathbf{Y}$8 and $\nabla_{\boldsymbol{\ell}}\!\cdot \mathbf{Y}$9, with positions $$\nabla_{\boldsymbol\ell}\!\cdot \mathbf{Y}^\pm(\boldsymbol\ell)$$0 and $$\nabla_{\boldsymbol\ell}\!\cdot \mathbf{Y}^\pm(\boldsymbol\ell)$$1,

$$\nabla_{\boldsymbol\ell}\!\cdot \mathbf{Y}^\pm(\boldsymbol\ell)$$2

is treated as a lag vector, and the corresponding increment is

$$\nabla_{\boldsymbol\ell}\!\cdot \mathbf{Y}^\pm(\boldsymbol\ell)$$3

The MMS application emphasizes this direct spatial construction as a core advantage, since the derivative calculation does not require Taylor’s hypothesis (Pecora, 2024).

With four spacecraft there are six distinct baselines. Because the Yaglom flux is an odd function of lag,

$$\nabla_{\boldsymbol\ell}\!\cdot \mathbf{Y}^\pm(\boldsymbol\ell)$$4

the six baselines generate twelve lag-space points once reflections are included. Choosing subsets of four lag vectors produces

$$\nabla_{\boldsymbol\ell}\!\cdot \mathbf{Y}^\pm(\boldsymbol\ell)$$5

candidate tetrahedra. In the MMS formulation, tetrahedra whose barycenter lies at the origin are excluded, and reflection duplicates are removed, leaving

$$\nabla_{\boldsymbol\ell}\!\cdot \mathbf{Y}^\pm(\boldsymbol\ell)$$6

independent, physically acceptable estimates. This combinatorial amplification is the distinctive “ensemble” feature of LPDE: one four-spacecraft snapshot yields hundreds of local derivative estimates rather than one (Pecora, 2024).

For each lag-tetrahedron, LPDE estimates the lag-space divergence using curlometer-like linear reconstruction. In the simulation study this is written explicitly in reciprocal-vector form: $$\nabla_{\boldsymbol\ell}\!\cdot \mathbf{Y}^\pm(\boldsymbol\ell)$$7 where

$$\nabla_{\boldsymbol\ell}\!\cdot \mathbf{Y}^\pm(\boldsymbol\ell)$$8

with $$\nabla_{\boldsymbol\ell}\!\cdot \mathbf{Y}^\pm(\boldsymbol\ell)$$9 a cyclic permutation of $\mathbf{z}^{\pm} = \mathbf{v} \pm \mathbf{b},$0. Under LPDE,

$\mathbf{z}^{\pm} = \mathbf{v} \pm \mathbf{b},$1

so that

$\mathbf{z}^{\pm} = \mathbf{v} \pm \mathbf{b},$2

and hence

$\mathbf{z}^{\pm} = \mathbf{v} \pm \mathbf{b},$3

Each tetrahedron therefore yields one local estimate of the cascade rate at the scale set by its mesocenter (Gao et al., 5 May 2026).

4. Measurement workflow and quality control

In the MMS magnetosheath application, magnetic field measurements are taken from the Fluxgate Magnetometers (FGM) at 128 Hz, while proton velocity and electron density are taken from the Fast Plasma Investigation (FPI), with ion velocity at 150 ms cadence and electron density at 30 ms cadence. Electron density is used instead of ion density because it is generally more accurate, with quasi-neutrality assumed. Velocities are despun using the standard MMS product, and all quantities are resampled to a common cadence of 150 ms. Density values above $\mathbf{z}^{\pm} = \mathbf{v} \pm \mathbf{b},$4 are discarded as potentially instrumentally contaminated (Pecora, 2024).

The practical LPDE workflow is explicit. First, the magnetic field is converted to Alfvén units and the Elsässer variables $\mathbf{z}^{\pm} = \mathbf{v} \pm \mathbf{b},$5 are formed. Second, for each spacecraft pair $\mathbf{z}^{\pm} = \mathbf{v} \pm \mathbf{b},$6, the separation vector $\mathbf{z}^{\pm} = \mathbf{v} \pm \mathbf{b},$7 and the increments $\mathbf{z}^{\pm} = \mathbf{v} \pm \mathbf{b},$8 are computed. Third, the mixed third-order quantity

$\mathbf{z}^{\pm} = \mathbf{v} \pm \mathbf{b},$9

is averaged over the interval to estimate $\mathbf{v}$0 at each lag vector. Fourth, all admissible sets of four lag-space vertices are assembled into tetrahedra. Fifth, a curlometer-like derivative estimate gives $\mathbf{v}$1. Sixth,

$\mathbf{v}$2

is used to obtain the partial cascade rate, and the resulting cloud of estimates is summarized statistically by its mean and standard deviation (Pecora, 2024).

Tetrahedron shape quality is a central numerical issue. The MMS study evaluates elongation and planarity using the volumetric tensor

$\mathbf{v}$3

with $\mathbf{v}$4 and barycenter

$\mathbf{v}$5

If $\mathbf{v}$6 are the eigenvalues of $\mathbf{v}$7, then

$\mathbf{v}$8

and

$\mathbf{v}$9

Based on Paschmann and Daly, tetrahedra with

$\mathbf{b}$0

should yield errors potentially below $\mathbf{b}$1. The authors also use relaxed thresholds and examine convergence of the interval-averaged cascade rate as the cutoff $\mathbf{b}$2 is increased; depending on interval and threshold, roughly $\mathbf{b}$3–$\mathbf{b}$4 acceptable tetrahedra remain (Pecora, 2024).

In the controlled simulation study, LPDE quality control is supplemented by an inertial-range filter: the main analysis retains lag-tetrahedra satisfying

$\mathbf{b}$5

and

$\mathbf{b}$6

to remove highly irregular tetrahedra and lags too small to plausibly lie in the inertial range (Gao et al., 5 May 2026).

5. MMS magnetosheath application and observed Yaglom-flux structure

The MMS study examines five magnetosheath intervals in burst mode, although LPDE results are reported for intervals I, II, III, and V, with interval IV discussed previously. The intervals are:

• I: 2017 Sep 28 06:31:33–07:01:43
• II: 2017 Nov 10 22:35:43–22:52:03
• III: 2017 Dec 21 07:21:54–07:48:01
• IV: 2017 Dec 26 06:12:43–06:52:23
• V: 2018 Apr 19 05:10:23–05:41:53

The paper also tabulates $\mathbf{b}$7, $\mathbf{b}$8, plasma $\mathbf{b}$9, correlation time $\sqrt{4\pi n_p m_p}$0, correlation length $\sqrt{4\pi n_p m_p}$1, and ion inertial length $\sqrt{4\pi n_p m_p}$2, using the normalized magnetic autocorrelation

$\sqrt{4\pi n_p m_p}$3

with $\sqrt{4\pi n_p m_p}$4 defined by the $\sqrt{4\pi n_p m_p}$5-folding time and $\sqrt{4\pi n_p m_p}$6 obtained by multiplying $\sqrt{4\pi n_p m_p}$7 by the average flow speed. The listed $\sqrt{4\pi n_p m_p}$8 are much larger than $\sqrt{4\pi n_p m_p}$9, while MMS separations are small; this is the basis for the statement that the measurements are not centered in the inertial range (Pecora, 2024).

Using the relaxed threshold $\boldsymbol{\ell}$0, the reported mean transfer rates are:

Interval	$\boldsymbol{\ell}$1 in $\boldsymbol{\ell}$2	$\boldsymbol{\ell}$3
I	$\boldsymbol{\ell}$4	0.95
II	$\boldsymbol{\ell}$5	0.95
III	$\boldsymbol{\ell}$6	0.95
V	$\boldsymbol{\ell}$7	0.95

At the conservative threshold $\boldsymbol{\ell}$8, the values remain similar:

Interval	$\boldsymbol{\ell}$9 in $$\delta \mathbf{z}^{\pm}(\boldsymbol{\ell}) = \mathbf{z}^{\pm}(\mathbf{x}) - \mathbf{z}^{\pm}(\mathbf{x}+\boldsymbol{\ell}).$$0	$$\delta \mathbf{z}^{\pm}(\boldsymbol{\ell}) = \mathbf{z}^{\pm}(\mathbf{x}) - \mathbf{z}^{\pm}(\mathbf{x}+\boldsymbol{\ell}).$$1
I	$$\delta \mathbf{z}^{\pm}(\boldsymbol{\ell}) = \mathbf{z}^{\pm}(\mathbf{x}) - \mathbf{z}^{\pm}(\mathbf{x}+\boldsymbol{\ell}).$$2	0.6
II	$$\delta \mathbf{z}^{\pm}(\boldsymbol{\ell}) = \mathbf{z}^{\pm}(\mathbf{x}) - \mathbf{z}^{\pm}(\mathbf{x}+\boldsymbol{\ell}).$$3	0.6
III	$$\delta \mathbf{z}^{\pm}(\boldsymbol{\ell}) = \mathbf{z}^{\pm}(\mathbf{x}) - \mathbf{z}^{\pm}(\mathbf{x}+\boldsymbol{\ell}).$$4	0.6
V	$$\delta \mathbf{z}^{\pm}(\boldsymbol{\ell}) = \mathbf{z}^{\pm}(\mathbf{x}) - \mathbf{z}^{\pm}(\mathbf{x}+\boldsymbol{\ell}).$$5	0.6

The central methodological result is the stability of the interval mean despite substantial scatter in individual tetrahedral estimates and occasional opposite-sign outliers. The central physical result is strong interval-to-interval variability, including interval I, which exhibits a negative average transfer rate and is interpreted as opposite in sign to the usual forward cascade (Pecora, 2024).

LPDE also enables direct visualization of the vector Yaglom flux in lag space. In isotropic turbulence one expects $$\delta \mathbf{z}^{\pm}(\boldsymbol{\ell}) = \mathbf{z}^{\pm}(\mathbf{x}) - \mathbf{z}^{\pm}(\mathbf{x}+\boldsymbol{\ell}).$$6 to point radially inward toward the origin. In the reported MMS intervals, however, the flux is distinctly non-radial. In interval I the arrows point away from the origin, consistent with positive divergence of $$\delta \mathbf{z}^{\pm}(\boldsymbol{\ell}) = \mathbf{z}^{\pm}(\mathbf{x}) - \mathbf{z}^{\pm}(\mathbf{x}+\boldsymbol{\ell}).$$7 and negative $$\delta \mathbf{z}^{\pm}(\boldsymbol{\ell}) = \mathbf{z}^{\pm}(\mathbf{x}) - \mathbf{z}^{\pm}(\mathbf{x}+\boldsymbol{\ell}).$$8. The authors describe the fluxes as “swirling” rather than radial. This suggests strong anisotropy and possibly local inhomogeneity in magnetosheath turbulence, and constitutes the paper’s main observational evidence that the lag-space transfer structure is genuinely three-dimensional rather than well approximated by a radial scalar proxy (Pecora, 2024).

6. Comparative assessment, limitations, and outlook

A later simulation study performed a systematic comparison between LPDE and direction-averaging (DA), which uses the integral form of the same exact law rather than the differential form. DA estimates the solid-angle average of the longitudinal third-order structure function over a sphere in lag space, whereas LPDE estimates a local lag-space divergence from tetrahedral samples. The comparison identifies a sharp division of sensitivities: DA exhibits clear polar-angle and azimuthal-angle dependence but is weakly sensitive to spacecraft configuration, while LPDE is strongly affected by spacecraft separation and tetrahedral shape but comparatively insensitive to the sampling trajectory (Gao et al., 5 May 2026).

The strongest LPDE sensitivity is to separation scale. In the simulation, baseline families

$$\delta \mathbf{z}^{\pm}(\boldsymbol{\ell}) = \mathbf{z}^{\pm}(\mathbf{x}) - \mathbf{z}^{\pm}(\mathbf{x}+\boldsymbol{\ell}).$$9

were tested, with $\frac{\partial}{\partial t}\langle |\delta \mathbf{z}^\pm|^2 \rangle + \boldsymbol{\nabla}_{\boldsymbol{\ell} \cdot \langle \delta \mathbf{z}^\mp |\delta \mathbf{z}^\pm|^2 \rangle - 2 \nu \nabla_{\boldsymbol{\ell}^2 \langle |\delta \mathbf{z}^\pm|^2 \rangle = -4 \epsilon^\pm,$0 and $\frac{\partial}{\partial t}\langle |\delta \mathbf{z}^\pm|^2 \rangle + \boldsymbol{\nabla}_{\boldsymbol{\ell} \cdot \langle \delta \mathbf{z}^\mp |\delta \mathbf{z}^\pm|^2 \rangle - 2 \nu \nabla_{\boldsymbol{\ell}^2 \langle |\delta \mathbf{z}^\pm|^2 \rangle = -4 \epsilon^\pm,$1 in the inertial range and smaller separations roughly in the dissipation range. LPDE improves as the baseline approaches inertial-range scales; for

$\frac{\partial}{\partial t}\langle |\delta \mathbf{z}^\pm|^2 \rangle + \boldsymbol{\nabla}_{\boldsymbol{\ell} \cdot \langle \delta \mathbf{z}^\mp |\delta \mathbf{z}^\pm|^2 \rangle - 2 \nu \nabla_{\boldsymbol{\ell}^2 \langle |\delta \mathbf{z}^\pm|^2 \rangle = -4 \epsilon^\pm,$2

the estimated normalized mean is about

$\frac{\partial}{\partial t}\langle |\delta \mathbf{z}^\pm|^2 \rangle + \boldsymbol{\nabla}_{\boldsymbol{\ell} \cdot \langle \delta \mathbf{z}^\mp |\delta \mathbf{z}^\pm|^2 \rangle - 2 \nu \nabla_{\boldsymbol{\ell}^2 \langle |\delta \mathbf{z}^\pm|^2 \rangle = -4 \epsilon^\pm,$3

The same study shows that increasing tetrahedral irregularity tends to move the mean estimate below the true dissipation rate, while trajectory-direction dependence remains weak. This makes LPDE a physically direct but geometrically demanding estimator (Gao et al., 5 May 2026).

These sensitivities clarify the principal limitations already visible in the MMS application. Without Taylor’s hypothesis, LPDE is limited to the separations actually spanned by the spacecraft constellation. For MMS, those separations are too small to span the inertial range broadly, so the measured value is only the nonlinear-transfer contribution at the sampled scales, not the full asymptotic inertial-range cascade rate. Because the temporal and dissipation terms of the full exact law are not independently measured, the inferred $\frac{\partial}{\partial t}\langle |\delta \mathbf{z}^\pm|^2 \rangle + \boldsymbol{\nabla}_{\boldsymbol{\ell} \cdot \langle \delta \mathbf{z}^\mp |\delta \mathbf{z}^\pm|^2 \rangle - 2 \nu \nabla_{\boldsymbol{\ell}^2 \langle |\delta \mathbf{z}^\pm|^2 \rangle = -4 \epsilon^\pm,$4 from the reduced law is described as a partial cascade rate and should be regarded as a lower threshold on the total cascade; elsewhere the interpretation is stated as effectively a lower bound on the total cascade (Pecora, 2024).

The broader implication is mission-design specific. For MMS, Cluster, HelioSwarm, and Plasma Observatory, LPDE is promising only when constellation planning provides baselines in or near the inertial range and well-conditioned tetrahedral geometries. The 2026 study explicitly presents HelioSwarm and Plasma Observatory as the natural next platforms for fully scale-dependent LPDE analyses. The MMS paper goes further in suggesting that the observed non-radial, possibly swirling Yaglom flux may motivate an additional equation involving the curl of $\frac{\partial}{\partial t}\langle |\delta \mathbf{z}^\pm|^2 \rangle + \boldsymbol{\nabla}_{\boldsymbol{\ell} \cdot \langle \delta \mathbf{z}^\mp |\delta \mathbf{z}^\pm|^2 \rangle - 2 \nu \nabla_{\boldsymbol{\ell}^2 \langle |\delta \mathbf{z}^\pm|^2 \rangle = -4 \epsilon^\pm,$5, paired with a measure of anisotropy. This suggests that LPDE is not only a measurement procedure for the anisotropic third-order law but also a way of exposing three-dimensional transfer structure that is invisible in isotropic reductions (Gao et al., 5 May 2026, Pecora, 2024).