---
title: 'Rotational Transform: Definition and Applications'
url: https://www.emergentmind.com/topics/rotational-transform
type: topic
---

# Rotational Transform: Definition and Applications

Rotational transform is a rotation number that quantifies the winding of trajectories, field lines, or transformed coordinates around a periodic structure. In magnetic confinement, it is the number of poloidal turns made by a magnetic field line per toroidal circuit, conventionally denoted $\iota$; in circle dynamics, it characterizes the action of a map through an irrational rotation factor; in geometric and computational settings, related terminology describes rotation-adapted transformations, symmetry reductions, or representations. The term is therefore domain-dependent: its most established physical meaning concerns magnetic field-line topology in stellarators and toroidal plasmas, while mathematically related notions occur in circle maps, interval exchanges, and Floquet theory.

## 1. Magnetic definition and geometric interpretation

In a toroidal magnetic configuration, a field line winds simultaneously around the poloidal and toroidal directions. The rotational transform is

$$
\iota=\frac{\text{poloidal turns}}{\text{toroidal turns}}.
$$

For a field line intersecting a fixed toroidal section, if $\theta_n$ is the poloidal angular advance between successive intersections, then

$$
\iota=\lim_{N\rightarrow\infty}\frac{1}{2\pi N}\sum_{n=0}^{N-1}\theta_n.
$$

For nested magnetic surfaces with toroidal flux $\psi$ and poloidal flux $\chi$, it can also be written

$$
\iota=\frac{d\chi}{d\psi}.
$$

Thus $\iota$ is simultaneously a field-line winding number, a ratio of differential magnetic fluxes, and the rotation number of a Poincaré return map. Its reciprocal is commonly identified with the tokamak safety factor,

$$
q\simeq\frac{1}{\iota}.
$$

The sign depends on the orientations selected for the toroidal and poloidal angles and on the direction in which the magnetic field line is followed. In an exact nonsymmetric vacuum field constructed in closed form, every off-axis field line closes after one toroidal transit and one poloidal turn, with $\iota=-1$ under the paper’s orientation conventions [2610.29921]. The same geometric winding would be reported with a different sign if either the poloidal orientation or field-line direction were reversed.

Rotational transform is distinguished from magnetic-field magnitude and from the continuity of individual field components. In the Hahm–Kulsrud–Taylor problem, the tangential field $B_\phi$ can possess a finite jump and a delta-function current sheet while $\iota$ remains continuous. The relevant field-line expression is governed by a harmonic, rather than arithmetic, average:

$$
\iota=\frac{R}{sB_\zeta\langle B_\phi^{-1}\rangle^{-1}}.
$$

Flux preservation gives

$$
B_\phi=\frac{B_{0\phi}}{\partial_x r},
\qquad
\langle\partial_xr\rangle=1,
$$

and consequently

$$
\iota=\frac{RB_{0\phi}}{sB_{0\zeta}}
      =\frac{Rj_0x}{sB_{0\zeta}}.
$$

The transform therefore passes continuously through the resonant surface even though the tangential magnetic field is discontinuous [1810.08268].

## 2. Sources of rotational transform in toroidal equilibria

Rotational transform may be generated by plasma current, by external non-axisymmetric coils, or by the geometry of the magnetic axis and surrounding flux surfaces. In a quasi-axisymmetric stellarator, it is useful to decompose the total transform schematically as

$$
\iota(s)=\iota_{\mathrm{plasma}}(s)+\iota_{\mathrm{ext}}(s),
$$

where $\iota_{\mathrm{plasma}}$ is produced by toroidal plasma current and $\iota_{\mathrm{ext}}$ by external coils.

External transform reduces the plasma current required to obtain a prescribed field-line pitch. In simulations of external kinks in a quasi-axisymmetric stellarator, the original configuration had $\iota_{\mathrm{ext}}\approx0.15$ and a plasma current of $8.25\ \mathrm{MA}$, compared with $11.5\ \mathrm{MA}$ for an equivalent tokamak. The external contribution also modifies magnetic shear, including vacuum shear outside the plasma. An axisymmetric “virtual current” was used to represent this geometry in nonlinear simulations; it reproduces the external transform but is not physical plasma current and does not provide plasma heating or free magnetic energy [2102.03080].

Increasing external transform reduced the $n=1$ external-kink growth rate and delayed saturation. The reported growth rates were $4.367\times10^4\ \mathrm{s}^{-1}$ at $11.5\ \mathrm{MA}$, $1.840\times10^4\ \mathrm{s}^{-1}$ at $8.25\ \mathrm{MA}$, $8.760\times10^3\ \mathrm{s}^{-1}$ at $6.75\ \mathrm{MA}$, $3.427\times10^3\ \mathrm{s}^{-1}$ at $5.25\ \mathrm{MA}$, and $3.413\times10^2\ \mathrm{s}^{-1}$ at $4.00\ \mathrm{MA}$; the $2.75\ \mathrm{MA}$ case was linearly stable. The reduction of the primary external kink did not imply monotonic improvement in confinement, because nonlinear current redistribution could trigger internal $(7,4)$ and $(9,5)$ modes. Consequently, confinement loss and thermal-energy loss depended non-monotonically on the external-transform fraction.

The on-axis transform also has a geometric contribution. For an elliptic magnetic axis, Mercier’s formula is

$$
\iota_0=N+\frac{1}{2\pi}\oint
\frac{d\ell}{\cosh\eta}
\left(
\frac{J_0}{2B_0}-\tau+\delta'
\right),
$$

where $J_0/(2B_0)$ is the current-driven contribution, $\tau$ is the axis torsion, $\delta'$ describes the rotation of elliptical cross-sections, $\eta$ measures ellipticity, and $N$ is an integer determined by the phase convention. The formula separates current-driven and geometric transform [2404.17531].

A variational derivation identifies this on-axis transform with the Floquet rotation number of neighboring field lines. The second variation of the magnetic field-line action produces the linearized transverse dynamics around the periodic magnetic axis. After rescaling, these reduce to

$$
X'+\Omega_0Y=0,\qquad
Y'-\Omega_0X=0,
$$

with

$$
\Omega_0=
\frac{J_0/(2B_0)-\tau+\delta'}{\cosh\eta}.
$$

The accumulated phase is

$$
\nu=\oint\Omega_0\,d\ell\pmod{2\pi},
$$

and

$$
\nu=2\pi\iota_0\pmod{2\pi}.
$$

A discrete action formulation reduces the numerical calculation of the Floquet multipliers to the eigenvalues of a $6\times6$ transfer matrix [2404.17531].

## 3. Rotational transform in stellarator design and stability

The magnitude and radial profile of $\iota$ are independent design variables. A family of quasi-axisymmetric equilibria was generated by continuation in the mean transform

$$
\bar{\iota}=\langle\iota\rangle_\psi,
$$

primarily over

$$
0.12\leq\bar{\iota}\leq0.75.
$$

The configurations retained low two-term quasisymmetry error over approximately $0.2\lesssim\bar{\iota}\lesssim0.65$. At larger transform, increasing elongation and limited boundary Fourier resolution degraded quasisymmetry, with equilibrium calculations eventually failing near $\bar{\iota}\simeq0.81$ [2401.09021].

Higher transform reduces neoclassical orbit width. For quasi-axisymmetry, the small-orbit-width scaling is approximately

$$
\Delta\psi\propto\frac{1}{\iota}.
$$

Collisionless guiding-center simulations of $3.52\ \mathrm{MeV}$ alpha particles found that loss fractions decreased as $\bar{\iota}$ increased up to approximately $\bar{\iota}=0.73$. Beyond this value, degradation of quasisymmetry offset the orbit-width benefit. The dependence was not described completely by $1/\iota$ because geometry, orbit topology, quasisymmetry error, and flux-surface compression also contributed.

The radial transform profile is characterized through magnetic shear. If

$$
\iota(s)\simeq c_0+c_1s,
$$

then the mean global shear is defined as

$$
\bar{\hat{s}}=-\frac{c_1}{\bar{\iota}}.
$$

At fixed $\bar{\iota}=0.42$, sufficiently positive shear, approximately $\bar{\hat{s}}\gtrsim0.085$, increased the maximum linear ITG growth rate and the nonlinear ion heat flux by roughly $25\%$. The destabilizing mode occurred at high $k_y$. These results show that transform magnitude alone does not determine stability or transport.

Linear gyrokinetic calculations also exhibit sensitivity near rational $\iota$. On rational surfaces, field lines close after a finite number of toroidal and poloidal turns, and individual flux tubes may sample curvature and finite-Larmor-radius damping differently. A growth rate computed on a single flux tube can therefore be unrepresentative. Taking the maximum over several flux tubes mitigates this sensitivity. Nonlinear ion heat flux showed a non-monotonic dependence on $\bar{\iota}$ and greater flux-tube variation than the linear growth rate [2401.09021].

## 4. Rotational transform in exact and numerical magnetic configurations

An exact vacuum magnetic field can possess nested toroidal flux surfaces and nonzero rotational transform without continuous Euclidean symmetry. In the closed-form construction of [2610.29921], a complex root $\zeta$ of

$$
2i\zeta^3+(\bar w+2)\zeta^2-2z\zeta-w=0
$$

defines the scalar potential

$$
\Phi=\Im\log\zeta
$$

and the magnetic field

$$
\mathbf B=\nabla\Im\log\zeta.
$$

The field is both curl-free and divergence-free, nowhere vanishing, and globally represented by coordinates $(\lambda,\beta,\varphi)$, where $\lambda$ and $\beta$ label field lines and $\varphi\in S^1$ parametrizes their toroidal progression. The flux function is

$$
\psi=Q(\lambda,\beta),
$$

with nested toroidal level sets. Since $\varphi$ advances by $2\pi$ on every field line, each line closes after one toroidal transit. A geometric analysis of the normal-plane displacement shows a poloidal change of $-2\pi$, yielding $\iota=-1$ on every regular off-axis surface.

The construction is significant as an exact benchmark for three-dimensional equilibrium codes. It provides analytically specified values for $\nabla\cdot\mathbf B$, $\nabla\times\mathbf B$, flux surfaces, field-line closure, and rotational transform. It also demonstrates that constant transform does not require axisymmetry, helical symmetry, or another continuous Euclidean symmetry.

In numerical equilibrium and stability calculations, rotational transform is commonly represented through field-line tracing, Poincaré maps, flux coordinates, or equilibrium profiles. The on-axis action-based method provides an alternative local calculation based on Floquet multipliers. In external-kink simulations, an axisymmetric approximation can retain the rotational-transform and safety-factor profiles of a three-dimensional stellarator while omitting magnetic-well effects, full three-dimensional shaping, and toroidal harmonic coupling. Such models are therefore useful for qualitative or first-order studies but are not exact substitutes for fully three-dimensional equilibria [2102.03080].

## 5. Circle dynamics, interval exchanges, and rotation factors

In one-dimensional dynamics, a rotational subset is a compact, invariant, minimal subset of the circle on which the map preserves cyclic order. For a continuous circle map with finite fibers, every infinite proper rotational subset is an extension of an irrational circle rotation. More precisely, there is a factor map

$$
\phi:X\to\mathbb T
$$

such that

$$
\phi\circ T=\tau\circ\phi,
\qquad
\tau(z)=e^{2\pi i\theta_0}z,
$$

where $\theta_0$ is irrational. The factor map is one-to-one except over countably many points corresponding to complementary gaps of $X$; each exceptional fiber contains the two endpoints of one gap. The system is uniquely ergodic, and the invariant measure projects to normalized Lebesgue measure [1705.03851].

The same circle-rotation structure appears in rotational interval exchange transformations. An interval exchange is rotational when its interval-exchange scheme and its dual scheme are both interval-exchange schemes. For irreducible systems, this is equivalent to being the first-return map of an irrational circle rotation to a finite union of arcs. The first-return map has finitely many return times and is piecewise translational. Its dual scheme records the return-time structure. The associated natural extension is a genus-one translation surface, namely a torus without singularities [2404.11211].

The equivalence can be summarized as

$$
\text{first-return map of a circle rotation to finitely many arcs}
\Longleftrightarrow
\text{interval exchange with an interval-exchange dual scheme}.
$$

This framework is extended to circle maps with breaks through linear-fractional rotational interval exchanges. For $N$ breaks, the commuting-pair description used in the one-break case is replaced by a commuting collection of $2N$ increasing Möbius maps,

$$
F_\alpha(x)=\frac{a_\alpha x+b_\alpha}{c_\alpha x+d_\alpha},
\qquad
a_\alpha d_\alpha-b_\alpha c_\alpha>0.
$$

The maps satisfy commutation relations determined by the dual rotational scheme, with one break factor for each stationary symbol. A toroidal patchwork construction yields an involutive duality, and allowed induction steps satisfy the time-reversal relation

$$
\Pi^{-1}=\Pi^*.
$$

The construction supplies an algebraic and combinatorial basis for renormalization studies of circle diffeomorphisms with multiple breaks, but it does not establish a multiple-break rigidity theorem, universal geometric realizability, or hyperbolicity of the associated renormalization operator [2609.40261].

## 6. Related uses of “rotational transform”

The term is also used for transformations that encode or compensate rotation in optics, computer vision, relativistic kinematics, and differentiable computation. These applications are conceptually distinct from magnetic rotational transform and from rotation numbers in dynamical systems.

In image processing, a Laguerre–Gaussian transform decomposes an image into modes with azimuthal index $l$ and radial index $p$. Under rigid rotation through $\Delta\phi$, the coefficients transform as

$$
A_{p,l}\longrightarrow A_{p,l}e^{il\Delta\phi}.
$$

The modal magnitudes remain invariant while their phases evolve predictably. Physical rotation therefore supplies a temporal encoding at frequencies $l\Omega$, enabling reconstruction and processing of rapidly rotating objects using temporal detection and radial decomposition [2005.07428].

In computer vision, a Rotational Rectification Network estimates a global in-plane angle and uses a spatial transformer to warp feature maps toward an upright configuration. Its Global Polar Pooling operator converts rotation in Cartesian coordinates into approximate translation along a polar angular coordinate. The estimated angle $\theta$ determines a planar transformation matrix,

$$
M=
\begin{bmatrix}
\cos\theta&-\sin\theta&0\\
\sin\theta&\cos\theta&0
\end{bmatrix}.
$$

The resulting rectified features can be supplied to an ordinary pedestrian detector. On rotated Caltech data, GP-Pooling reduced the reported angle error from $18.33^\circ$ for VGG-S to $15.82^\circ$ for GP-Pooling plus VGG-S; inserting the rectification network reduced the RPN-BF average miss rate from $83.3\%$ to $57.6\%$ [1706.08917].

For dense oriented-object detection, RotaTR incorporates rotation directly into transformer attention. A rotated reference box is represented as

$$
A_q=(x_q,y_q,w_q,h_q,\theta_q),
$$

and learned deformable-attention offsets are scaled by width and height and rotated by the reference angle:

$$
\Delta\widetilde{\mathbf p}_{qk}
=
\alpha
\begin{pmatrix}
w_q&0\\
0&h_q
\end{pmatrix}
\left[
\sigma(\Delta\mathbf p_{qk})-
\begin{pmatrix}
1/2\\1/2
\end{pmatrix}
\right]
R^T(\theta_q).
$$

This Rotation Sensitive deformable attention makes orientation an active component of feature sampling rather than merely an output regression variable. The model retains DETR’s Hungarian matching and NMS-free set prediction [2312.02821].

In relativistic kinematics, “rotational transformations” denotes alternative prescriptions for rotating frames. The compared Langevin, Post, Franklin, and rotational absolute Lorentz transformations differ in simultaneity, time dilation, length contraction, one-way light speed, two-way light speed, and Sagnac predictions. The cited analysis favors the rotational absolute Lorentz transformation because it combines anisotropic one-way propagation, isotropic two-way propagation, a nonzero conventional Sagnac effect, time dilation, and length contraction [2106.08885]. These claims concern a particular comparison of kinematic prescriptions and should not be conflated with the experimentally standard use of rotating coordinates in relativity.

For differentiable computation, modern spatial-transformation software represents three-dimensional rotations using unit quaternions and rotation-vector exponentials. A finite angular-velocity update is implemented as

$$
\operatorname{rot}_{k+1}
=
\operatorname{rot}_k
*
\operatorname{from\_rotvec}(\boldsymbol\omega\,dt).
$$

The framework described in [2511.18157] dispatches operations through the Python Array API, supporting NumPy, JAX, PyTorch, CuPy, batching, JIT compilation, GPU and TPU execution, and native autodiff. Here “rotational transform” refers to a computational transformation on $SO(3)$ or $SE(3)$, not to a field-line winding invariant.

Across these domains, the common structural idea is the extraction, preservation, or manipulation of rotational phase. In magnetic confinement it is a topological winding rate of field lines; in circle dynamics it is a rotation number or rotation factor; in Floquet theory it is a phase accumulated by neighboring trajectories; and in computational or optical systems it is a representation designed to make rotation explicit, invariant, or differentiable.

Source: https://www.emergentmind.com/topics/rotational-transform