Piecewise Omnigenity in Stellarators

• Piecewise omnigenity is defined as a stellarator magnetic-field concept where the trapped-particle second invariant remains constant only within distinct orbit classes on a flux surface.
• It enables reactor-relevant benefits by suppressing the deleterious 1/ν transport, achieving tokamak-like banana transport and potentially zero bootstrap current.
• The approach employs explicit parameterizations and hybrid constructions, merging quasi-isodynamic or omnigenous low-field regions with high-field piecewise structures to optimize confinement.

Searching arXiv for papers on piecewise omnigenity and related stellarator optimization. Piecewise omnigenity is a stellarator magnetic-field concept in which the trapped-particle second adiabatic invariant is constant only within distinct orbit classes or regions on a flux surface, rather than being a single flux-surface constant as in standard omnigenity. In this framework, trapped-particle phase space is partitioned into well types or orbit classes, with discontinuous jumps of the invariant at junctures where particles transition between classes. The principal motivation is reactor design: piecewise omnigenous fields can suppress the stellarator-specific low-collisionality $1/\nu$ transport and yield tokamak-like banana-regime transport while relaxing the global contour-topology constraints that standard omnigenity imposes on $B$-contours (Velasco et al., 2024). Subsequent work has developed explicit field parameterizations, optimization methods, hybrid constructions that combine quasi-isodynamic or poloidally omnigenous structure with piecewise omnigenity, and fully magnetohydrodynamic equilibria exhibiting near-piecewise-omnigenous behavior (Velasco et al., 2024, Liu et al., 13 Feb 2025, Calvo et al., 5 May 2025, Fernández-Pacheco et al., 21 Jan 2026, Liu et al., 12 Mar 2026, Velasco et al., 12 Mar 2026).

1. Definition and conceptual scope

In standard stellarator theory, omnigenity means that for all trapped particles the orbit-averaged radial drift vanishes. A sufficient condition is that the second adiabatic invariant satisfy

$\partial_\alpha J = 0,$

with $\alpha$ the field-line label on a flux surface. In the low-collisionality $1/\nu$ regime, the source term of the drift-kinetic equation is proportional to $\partial_\alpha J$, so global omnigenity removes the $1/\nu$ regime and leaves tokamak-like banana transport (Velasco et al., 2024).

Piecewise omnigenity relaxes this condition. Its formal definition is that, for orbit classes $\mathrm w=\mathrm{I,II,III},\ldots$,

$J = J^{(\mathrm w)}(s,\mathcal E,\mu),$

while at junctures

$$\lim_{\Delta\alpha\to 0}\left[J(s,\alpha+\Delta\alpha,\mathcal E,\mu)-J(s,\alpha,\mathcal E,\mu)\right]\neq 0.$$

Thus $B$0 is constant only piecewise, not globally, and trapped particles may transition between different well types as they precess on a surface (Velasco et al., 2024). Later parameterization work adopts the same structure, emphasizing that pwO fields are qualitatively different from omnigenous, quasi-isodynamic, or quasisymmetric fields because $B$1 is constant only within distinct orbit classes, with jumps at transitions (Velasco et al., 2024).

This broader concept is distinct from a naive geometric patching of the surface. In the direct optimization and OOPS-based literature, the “piecewise” character is tied to branches or regions of trapped-particle motion associated with different parts of the magnetic-well structure, especially low-field versus high-field regions, rather than a simple subdivision of the surface into independent geometric patches (Liu et al., 13 Feb 2025, Liu et al., 12 Mar 2026). A plausible implication is that pwO is best interpreted as a phase-space organization principle rather than solely a contour-geometry prescription.

2. Relation to omnigenity, quasi-isodynamicity, and quasisymmetry

Classical omnigenity imposes strong global conditions on each flux surface. The standard geometric characterization used in the recent pwO literature requires that constant-$B$2 contours close toroidally, poloidally, or helically, and that the distance along a field line between equal-$B$3 bounce points be independent of field line (Velasco et al., 12 Mar 2026). More general omnigenous fields can be classified by the global topology of these contour closures, with quasi-isodynamicity corresponding to poloidally closed contours and quasisymmetry corresponding to $B$4 depending on a single helical angle in Boozer coordinates (Landreman et al., 2011, Dudt et al., 2023).

Piecewise omnigenity relaxes precisely these global contour-topology requirements. In pwO fields, constant-$B$5 contours need not close toroidally, poloidally, or helically, and transitioning particles are permitted (Velasco et al., 2024). This is the principal geometric novelty. The gain is that one can retain low radial neoclassical transport without requiring the full Cary–Shasharina contour structure of omnigenous fields.

The relation to quasi-isodynamicity is especially important. Quasi-isodynamic fields have been attractive because, in addition to low radial neoclassical transport, they give zero bootstrap current at low collisionality, which is useful for island-divertor reactor concepts (Velasco et al., 12 Mar 2026). Piecewise omnigenity initially broadened the space of fields with tokamak-like radial transport (Velasco et al., 2024), and later work proved that pwO can also support zero bootstrap current for arbitrary density and temperature profiles in the low-collisionality regime (Calvo et al., 5 May 2025). This establishes pwO as an alternative route to the two reactor-relevant properties previously associated primarily with quasi-isodynamicity: small radial neoclassical transport and vanishing bootstrap current.

At the same time, omnigenity remains a fundamentally global property of a flux surface in the classical theory. Earlier work treating omnigenity as generalized quasisymmetry emphasizes that contour topology, helicity, and branch relations are whole-surface constraints, not merely local ones (Landreman et al., 2011). The pwO development therefore represents not a reinterpretation of omnigenity itself, but a relaxation of it.

3. Trapped-particle dynamics and the second invariant

The trapped-particle interpretation is central. In the pwO framework, a flux surface contains several trapped-orbit classes, each with field-line-independent bounce geometry inside its class. Within each region,

$B$6

so trapped particles behave omnigenously there; discontinuities occur only at class transitions (Velasco et al., 2024). The key result is that these discontinuities do not automatically generate $B$7 transport. In the prototypical construction, the branches satisfy a compatibility condition

$B$8

so the source term in the drift-kinetic equation vanishes even at the juncture (Velasco et al., 2024).

This branchwise structure can also be written directly in terms of the orbit-averaged drifts. Using

$B$9

piecewise constancy of $\partial_\alpha J = 0,$0 suppresses radial drift inside each orbit class, while the branch-matching condition prevents the junctures from reintroducing the $\partial_\alpha J = 0,$1 source term (Velasco et al., 2024). This is the precise sense in which pwO preserves the main neoclassical benefit of omnigenity while tolerating transitioning particles.

A related line of work frames piecewise omnigenity through kinetic solvability. In the low-collisionality drift-kinetic problem, the existence of functions $\partial_\alpha J = 0,$2 satisfying

$\partial_\alpha J = 0,$3

encodes vanishing orbit-averaged radial drift, and for a prototypical pwO field this structure can be made explicit (Calvo et al., 5 May 2025). This kinetic formulation is what later enables an analytic proof of zero bootstrap current.

The notion of piecewise maximum-$\partial_\alpha J = 0,$4 extends the idea further. In a nearly pwO configuration approximated as

$\partial_\alpha J = 0,$5

with $\partial_\alpha J = 0,$6, one obtains

$\partial_\alpha J = 0,$7

a piecewise version of maximum-$\partial_\alpha J = 0,$8. This suggests favorable trapped-electron precession and links pwO to magnetic-well formation and MHD stability, though those implications are explored more fully only in the equilibrium work (Fernández-Pacheco et al., 21 Jan 2026).

4. Geometric constructions and parameterizations

The earliest explicit pwO construction is a high-$\partial_\alpha J = 0,$9 parallelogram in Boozer coordinates. A prototypical field is

$\alpha$0

with $\alpha$1 giving a piecewise-constant limit in which $\alpha$2 inside a parallelogram and $\alpha$3 outside (Velasco et al., 2024). Two sufficient conditions for piecewise omnigenity in that limit are that all contours $\alpha$4 collapse into a single parallelogram and that the rotational transform be such that only two field lines connect the four corners (Velasco et al., 2024).

A later systematic treatment develops this into a broader parameter space. It introduces a piecewise-quasisymmetric prototype

$\alpha$5

with rotational transform

$\alpha$6

This construction yields three trapped-orbit classes, regions I, II, and III, associated with different side pairings of the high-$\alpha$7 domain (Velasco et al., 2024). Their bounce-point separations $\alpha$8 are class-dependent but field-line-independent, and satisfy

$\alpha$9

which reproduces the branch structure of $1/\nu$0 (Velasco et al., 2024).

The same work generalizes the straight-segment prototype to deformed non-straight contours by introducing side deformations $1/\nu$1 expanded in Fourier-like series, subject to corner-fixing and tangency-avoidance constraints (Velasco et al., 2024). The tangency conditions,

$1/\nu$2

identify when unwanted local trapping structures appear, so avoiding these tangencies is essential for remaining in the pwO topological class (Velasco et al., 2024).

The same parameterization work also constructs hybrids in which deeply trapped particles see a conventional omnigenous field while barely trapped particles see pwO structure. This already foreshadows the later QI-pwO and PO-pwO constructions (Velasco et al., 2024).

5. Hybrid constructions: low-field omnigenity and high-field pwO

A major development has been the deliberate combination of omnigenity or quasi-isodynamicity in the low-field region with piecewise omnigenity in the high-field region. In the OOPS framework this appears as PO-pwO, where the low-field side remains poloidally omnigenous and the high-field side is “squeezed” into an approximately pwO region (Liu et al., 12 Mar 2026). The Landreman–Catto mapping is modified by restricting the accessible $1/\nu$3-range and altering the bounce-distance function $1/\nu$4 so that the branch separation $1/\nu$5 reaches $1/\nu$6 early, causing the outermost high-$1/\nu$7 contour to close locally and form an approximate pwO domain (Liu et al., 12 Mar 2026).

The same conceptual structure appears in direct optimization work, where restricting the contour-label domain $1/\nu$8 to a narrower interval such as $1/\nu$9 makes neighboring $\partial_\alpha J$0 contours intersect, leaving the $\partial_\alpha J$1 region omnigenous and the high-field region piecewise omnigenous (Liu et al., 13 Feb 2025). In that context, the resulting field is described as combining omnigenity in the low-field region and pwO in the high-field region.

The most explicit hybrid is the QI-pwO field. There the magnetic field is modeled as

$\partial_\alpha J$2

with $\partial_\alpha J$3 generating an approximately parallelogram-shaped high-$\partial_\alpha J$4 region and $\partial_\alpha J$5 imposing a quasi-isodynamic modulation only in a band around the minimum-$\partial_\alpha J$6 contour (Velasco et al., 12 Mar 2026). The field is quasi-isodynamic in the low-field region and departs strongly from quasi-isodynamicity in the high-field region.

The particle interpretation is explicit. For

$\partial_\alpha J$7

bounce points lie on poloidally closed contours as in a QI field; for

$\partial_\alpha J$8

bounce points lie on a parallelogram-shaped contour as in a pwO field (Velasco et al., 12 Mar 2026). Thus “piecewise” refers both to regions of the surface and to different trapped-particle classes in velocity space.

An important result of the QI-pwO analysis is that the high-field departure cannot be arbitrary. The pwO part must satisfy inherited transport-oriented constraints such as

$\partial_\alpha J$9

which guarantee tokamak-like radial transport in the limit $1/\nu$0 and zero bootstrap current at low collisionality in that same limit (Velasco et al., 12 Mar 2026). Parameter scans show that $1/\nu$1 remains close to zero only around $1/\nu$2, demonstrating that not every high-$1/\nu$3 deviation from QI is benign (Velasco et al., 12 Mar 2026).

6. Transport, bootstrap current, and reactor-relevant consequences

The original transport claim of pwO is that it removes the deleterious $1/\nu$4 regime despite permitting transitioning particles. Numerical scans of the monoenergetic radial transport coefficient $1/\nu$5 show strong reduction as the sharpness parameter $1/\nu$6 is increased, with low-collisionality transport about two orders of magnitude smaller than in W7-X standard and LHD for a representative nearly pwO case, though still about one order of magnitude larger than in a very precisely optimized QHS configuration (Velasco et al., 2024). The qualitative conclusion is that pwO can reproduce tokamak-like banana-regime collisional transport while being much farther from conventional omnigenous topology.

A later analytical milestone is the proof that pwO can also give zero bootstrap current. For a prototypical piecewise omnigenous field with parallelogram-shaped $1/\nu$7 patches, the bootstrap current takes the form

$1/\nu$8

with

$1/\nu$9

The zero-bootstrap criterion is simply

$\mathrm w=\mathrm{I,II,III},\ldots$0

When this holds, MHD consistency and the axis condition force $\mathrm w=\mathrm{I,II,III},\ldots$1, hence $\mathrm w=\mathrm{I,II,III},\ldots$2 (Calvo et al., 5 May 2025). MONKES simulations confirm that the bootstrap-related coefficient $\mathrm w=\mathrm{I,II,III},\ldots$3 vanishes at $\mathrm w=\mathrm{I,II,III},\ldots$4 with remarkable accuracy (Calvo et al., 5 May 2025).

The reactor relevance of pwO was strengthened further by the construction of CIEMAT-pw1, a fixed-boundary ideal-MHD equilibrium optimized toward a smoothed pwO target (Fernández-Pacheco et al., 21 Jan 2026). On the optimized surface $\mathrm w=\mathrm{I,II,III},\ldots$5, the relative difference from the smoothed target with $\mathrm w=\mathrm{I,II,III},\ldots$6 is below $\mathrm w=\mathrm{I,II,III},\ldots$7 and of order $\mathrm w=\mathrm{I,II,III},\ldots$8 over most of the surface (Fernández-Pacheco et al., 21 Jan 2026). The configuration has lower effective ripple than W7-X standard for all studied $\mathrm w=\mathrm{I,II,III},\ldots$9, a bootstrap coefficient comparable to W7-X high-mirror, Mercier and ballooning stability, and alpha heating efficiency of $J = J^{(\mathrm w)}(s,\mathcal E,\mu),$0 at $J = J^{(\mathrm w)}(s,\mathcal E,\mu),$1 (Fernández-Pacheco et al., 21 Jan 2026).

The PO-pwO optimization study extends these reactor-oriented findings. It reports a family of optimized configurations spanning field periods $J = J^{(\mathrm w)}(s,\mathcal E,\mu),$2 and several aspect ratios, with low $J = J^{(\mathrm w)}(s,\mathcal E,\mu),$3, favorable $J = J^{(\mathrm w)}(s,\mathcal E,\mu),$4, and maximum alpha loss below $J = J^{(\mathrm w)}(s,\mathcal E,\mu),$5 for all nine scanned cases, with below $J = J^{(\mathrm w)}(s,\mathcal E,\mu),$6 for the magnetic-well example (Liu et al., 12 Mar 2026). A vacuum magnetic well of depth $J = J^{(\mathrm w)}(s,\mathcal E,\mu),$7 is obtained in a representative case without adding a separate magnetic-well cost term (Liu et al., 12 Mar 2026).

The QI-pwO paper adds a comparative perspective. Using a W7-X-like surface with

$J = J^{(\mathrm w)}(s,\mathcal E,\mu),$8

and computing $J = J^{(\mathrm w)}(s,\mathcal E,\mu),$9 and $$\lim_{\Delta\alpha\to 0}\left[J(s,\alpha+\Delta\alpha,\mathcal E,\mu)-J(s,\alpha,\mathcal E,\mu)\right]\neq 0.$$0 with MONKES, it finds both coefficients much smaller at low collisionality than in the W7-X high-mirror comparison case (Velasco et al., 12 Mar 2026). However, $$\lim_{\Delta\alpha\to 0}\left[J(s,\alpha+\Delta\alpha,\mathcal E,\mu)-J(s,\alpha,\mathcal E,\mu)\right]\neq 0.$$1 is expected to deviate from zero at sufficiently low $$\lim_{\Delta\alpha\to 0}\left[J(s,\alpha+\Delta\alpha,\mathcal E,\mu)-J(s,\alpha,\mathcal E,\mu)\right]\neq 0.$$2, coinciding with onset of a $$\lim_{\Delta\alpha\to 0}\left[J(s,\alpha+\Delta\alpha,\mathcal E,\mu)-J(s,\alpha,\mathcal E,\mu)\right]\neq 0.$$3 regime in $$\lim_{\Delta\alpha\to 0}\left[J(s,\alpha+\Delta\alpha,\mathcal E,\mu)-J(s,\alpha,\mathcal E,\mu)\right]\neq 0.$$4, because the analytic proof of zero bootstrap current assumes tokamak-like banana transport (Velasco et al., 12 Mar 2026). This is an important caveat: the low-bootstrap behavior of finite-$$\lim_{\Delta\alpha\to 0}\left[J(s,\alpha+\Delta\alpha,\mathcal E,\mu)-J(s,\alpha,\mathcal E,\mu)\right]\neq 0.$$5, finite-$$\lim_{\Delta\alpha\to 0}\left[J(s,\alpha+\Delta\alpha,\mathcal E,\mu)-J(s,\alpha,\mathcal E,\mu)\right]\neq 0.$$6 models is approximate rather than exact at arbitrarily low collisionality.

7. Interpretation, historical context, and limitations

One of the important interpretive claims of the pwO literature is that it explains why some successful stellarator configurations have good transport despite being visibly far from classical omnigenous contour topology. W7-X standard, inward-shifted LHD, and configuration A have been discussed as realistic cases in which $$\lim_{\Delta\alpha\to 0}\left[J(s,\alpha+\Delta\alpha,\mathcal E,\mu)-J(s,\alpha,\mathcal E,\mu)\right]\neq 0.$$7 clusters around several discrete values rather than a single branch, with approximate relations such as

$$\lim_{\Delta\alpha\to 0}\left[J(s,\alpha+\Delta\alpha,\mathcal E,\mu)-J(s,\alpha,\mathcal E,\mu)\right]\neq 0.$$8

suggesting near-piecewise omnigenity (Velasco et al., 2024). The QI-pwO work further argues that W7-X bootstrap reduction, achieved historically by tuning a few low-order Fourier harmonics,

$$\lim_{\Delta\alpha\to 0}\left[J(s,\alpha+\Delta\alpha,\mathcal E,\mu)-J(s,\alpha,\mathcal E,\mu)\right]\neq 0.$$9

can retrospectively be interpreted as producing a field approximately QI near $B$00 with a high-field deviation resembling the pwO bootstrap-reducing structure (Velasco et al., 12 Mar 2026).

This has led to a more rigorous reinterpretation of what had often been called “quasi-omnigenous.” The hybrid literature argues that many configurations have good confinement because they are close to omnigenous or QI only in some parts of the surface, especially around $B$01, while differing substantially elsewhere; QI-pwO then identifies a more precise and reactor-relevant subset of such fields (Velasco et al., 12 Mar 2026).

Several limitations are common across the literature. Exact pwO constructions are typically idealized and become discontinuous in the sharp limit $B$02 or $B$03 (Velasco et al., 2024, Calvo et al., 5 May 2025, Fernández-Pacheco et al., 21 Jan 2026). Practical equilibria therefore target smoothed approximations, not exact pwO. Existing optimization tools based on smooth variance reduction of $B$04 are not naturally aligned with piecewise constancy plus jumps, which is why new coordinate-based methods such as OOPS or range-restricted mappings were introduced (Velasco et al., 2024, Liu et al., 13 Feb 2025, Liu et al., 12 Mar 2026). Coil realization remains an open question: several papers motivate pwO by the possibility of reduced shaping and coil complexity, but none provides a definitive stage-2 coil solution (Fernández-Pacheco et al., 21 Jan 2026, Liu et al., 12 Mar 2026, Velasco et al., 12 Mar 2026).

A further limitation is scope. Some studies emphasize neoclassical transport and bootstrap current while leaving turbulence, energetic particles, and engineering metrics only partially explored (Velasco et al., 2024, Calvo et al., 5 May 2025). Even where gyrokinetic or orbit-following results are provided, they remain configuration-specific rather than constituting a general theorem (Fernández-Pacheco et al., 21 Jan 2026). Likewise, hybrid methods have so far been demonstrated mainly for PO-pwO and QI-pwO, with QS-pwO, toroidal omnigenity plus pwO, and helical omnigenity plus pwO identified as future directions rather than established design classes (Liu et al., 12 Mar 2026).

Taken together, the recent literature defines piecewise omnigenity as a broadened confinement principle in which the trapped-particle second invariant is piecewise constant across orbit classes, permitting transitioning particles and nonstandard $B$05-contour topology while preserving the low-collisionality confinement advantages usually associated with omnigenity. The concept now spans formal branchwise definitions, explicit Boozer-space parameterizations, optimization strategies, hybrid low-field-omnigenous/high-field-pwO constructions, and reactor-candidate equilibria, making pwO a distinct and increasingly systematic branch of stellarator optimization research (Velasco et al., 2024, Velasco et al., 2024, Liu et al., 13 Feb 2025, Calvo et al., 5 May 2025, Fernández-Pacheco et al., 21 Jan 2026, Liu et al., 12 Mar 2026, Velasco et al., 12 Mar 2026).