Analytic Coordinate Discretization

• Analytic Coordinate Discretization is a cross-disciplinary methodology that converts continuous analytic structures into discrete frameworks while preserving key geometric invariants such as separability and orthogonality.
• It is applied across fields like MRI trajectory design, discrete differential geometry, PDE discretization, and mesh adaptation, ensuring structure-preserving outcomes through analytic templates.
• The method enhances numerical efficiency and convergence by embedding discretization within analytic models, thereby reducing bias and improving computational performance in various applications.

Analytic Coordinate Discretization denotes, in its most explicit recent usage, a framework in which a practical non-Cartesian MRI trajectory is interpreted as a discretized version of an analytic coordinate defined by a set of template trajectories (Jang et al., 21 Mar 2025). Across neighboring literatures, closely related constructions discretize continuous coordinates, coordinate transformations, coordinate-adapted operators, or coordinate-based dynamical laws while preserving separability, orthogonality, monotonicity, conservation laws, or analyticity (Bobenko et al., 2015, Benamou et al., 2014, Suh et al., 2022). This suggests a broader methodological theme: discrete objects are derived from an analytic coordinate description rather than introduced as purely local grid rules.

1. Terminology and scope

The current literature uses the expression in more than one sense. In MRI, it is a named trajectory-design framework; in discrete differential geometry, mesh adaptation, PDE discretization, optimization, and functional analysis, it appears as a family of structure-preserving constructions built from analytic coordinates, analytic coordinate transformations, or analytic coordinate functionals. A plausible implication is that the term is best understood as a cross-disciplinary methodology rather than as a single closed formalism.

Setting	Analytic structure	Discretized outcome
Fast volumetric MRI	analytic coordinate defined by template trajectories	spokes/interleaves sampled from a continuous spiral path (Jang et al., 21 Mar 2025)
Confocal coordinates	separable confocal quadrics and EPD structure	discrete Koenigs nets and discrete confocal coordinates (Bobenko et al., 2015)
Curvature-line geometry	orthogonal coordinate systems and Dupin cyclides	cyclidic nets as discrete analogs (Bobenko et al., 2011)
Mesh adaptation	continuous coordinate transformation Jacobian	simplicial functional via affine element mappings (Huang et al., 2014)
Monge-Ampère PDE	continuous operator $\det(D^2u)$	MA-LBR monotone and consistent stencil scheme (Benamou et al., 2014)
Geometric modeling	analytic shape field from radial kernels	grid-free discretization by countable unions of balls (Behandish et al., 2017)

A recurrent misconception is that coordinate discretization is synonymous with uniform Cartesian gridding. The literature contains grid-free discretization of arbitrary shapes as countable unions of balls, adaptive stencil generation along the Stern-Brocot tree, and spiral sampling on spheres, ellipsoids, or cylinders, all of which are explicitly nonuniform or non-Cartesian (Behandish et al., 2017, Benamou et al., 2014, Jang et al., 21 Mar 2025).

2. Discrete geometry and coordinate systems

A major branch of analytic coordinate discretization concerns coordinate systems whose geometric invariants are retained in discrete form. For confocal quadrics, an integrable discretization of the Euler-Poisson-Darboux equation is used to preserve two crucial properties of confocal coordinates: separability and all two-dimensional coordinate subnets being isothermic surfaces, equivalently supporting orthogonal Koenigs nets. The discrete EPD equation is

$$\Delta_i \Delta_j x = \frac{\gamma}{n_i+\epsilon_i-n_j-\epsilon_j}\left(\Delta_j x-\Delta_i x\right),$$

and for $\gamma=\tfrac12$ the corresponding coordinate functions are given explicitly in terms of gamma functions, with $(u)_{1/2}=\Gamma(u+\tfrac12)/\Gamma(u)$ acting as a discrete square root (Bobenko et al., 2015). The companion geometric development characterizes classical confocal coordinates as factorizable orthogonal coordinate systems, constructs discrete nets via polarity with respect to a sequence of classical confocal quadrics, and computes the coordinate functions explicitly for parametrizations that include gamma functions, trigonometric and hyperbolic functions, and Jacobi elliptic functions (Bobenko et al., 2017).

Cyclidic nets furnish a different, but closely aligned, discretization paradigm. They are introduced as discrete analogs of curvature line parametrized surfaces and orthogonal coordinate systems. A 2-dimensional cyclidic net is a piecewise smooth $C^1$-surface built from surface patches of Dupin cyclides, each patch bounded by curvature lines of the supporting cyclide, while 3-dimensional cyclidic nets are discrete analogs of triply-orthogonal coordinate systems. The construction is based on the Lie geometric description of Dupin cyclides, and explicit formulas are derived and implemented in a computer program (Bobenko et al., 2011).

Derivative-coordinate formulations for analytic tree fractals show the same pattern in a curve-and-branch setting. In two dimensions,

$$p(s)=\int \nabla r \cdot \exp\!\left(i\int \nabla \varphi\,ds\right)ds,$$

and branching is introduced by a multivalued periodic unit function $u(s)$, yielding a root-to-canopy analytic formulation. Computational realization is obtained by discretizing the integral as a Riemann sum, and the paper shows that iterative, discrete tree fractals correspond to analytic fractals whose derivative coordinate functions are piecewise constant; under matching integrated values, the canopies of analytic and discrete fractals coincide up to scaling, rotation, and translation (Mulder, 2015). This suggests that, in geometric settings, analytic coordinate discretization often means replacing pointwise coordinates by differential, separable, or curvature-adapted primitives and only then sampling.

3. Coordinate mappings, meshing, and grid-free solids

A second branch starts from coordinate transformations themselves. For curvilinear coordinates, one analytical method uses the solution of the Dirichlet problem of the Poisson equation to determine $\xi(x,y)$ and $\eta(x,y)$, while the algebraic method constructs continuous mapping functions by polynomial interpolation fitted at discrete points with the least square method: $$\xi(x,y)=\sum_{m=0}^M\sum_{n=0}^{M-m}A_{mn}x^m y^n,\qquad \eta(x,y)=\sum_{m=0}^M\sum_{n=0}^{M-m}B_{mn}x^m y^n.$$ The same work applies the generated mappings to boundary value problems after transforming the PDE to the mapped coordinates (Isshiki et al., 2017).

Variational mesh generation and adaptation adopts a direct geometric discretization of meshing functionals on simplicial meshes. Instead of discretizing the Jacobian matrix of the continuous coordinate transformation directly, the Jacobian is replaced, on each element, by the Jacobian matrix of the affine mapping between corresponding physical and computational elements. With edge matrices

$$\Delta_i \Delta_j x = \frac{\gamma}{n_i+\epsilon_i-n_j-\epsilon_j}\left(\Delta_j x-\Delta_i x\right),$$0

one has $$\Delta_i \Delta_j x = \frac{\gamma}{n_i+\epsilon_i-n_j-\epsilon_j}\left(\Delta_j x-\Delta_i x\right),$$1 and $$\Delta_i \Delta_j x = \frac{\gamma}{n_i+\epsilon_i-n_j-\epsilon_j}\left(\Delta_j x-\Delta_i x\right),$$2, so the discretized functional preserves the basic geometric structure of the continuous functional and admits simple analytical derivatives with respect to vertex coordinates (Huang et al., 2014).

Analytic geometric modeling via spherical decomposition replaces uniform grid-based sampling by a grid-free discretization of arbitrary shapes as countable unions of balls,

$$\Delta_i \Delta_j x = \frac{\gamma}{n_i+\epsilon_i-n_j-\epsilon_j}\left(\Delta_j x-\Delta_i x\right),$$3

with the shape encoded as a sublevel set of a sum of compactly-supported smooth radial kernels. Using a geometric lifting trick, the representation becomes a convolution of an impulsive skeletal density in $$\Delta_i \Delta_j x = \frac{\gamma}{n_i+\epsilon_i-n_j-\epsilon_j}\left(\Delta_j x-\Delta_i x\right),$$4 and primitive kernels with conical support, and the method leverages nonequispaced FFTs rather than standard FFTs on uniform grids (Behandish et al., 2017). A plausible implication is that analytic coordinate discretization need not privilege lattices at all; it can instead privilege an analytic embedding in a higher-dimensional coordinate space.

4. PDE operators, consistency, and singular structure

In nonlinear elliptic PDEs, the Monge-Ampère using Lattice Basis Reduction scheme provides a novel discretization of the Monge-Ampère operator that is simultaneously consistent and degenerate elliptic on a two dimensional cartesian grid. With second differences

$$\Delta_i \Delta_j x = \frac{\gamma}{n_i+\epsilon_i-n_j-\epsilon_j}\left(\Delta_j x-\Delta_i x\right),$$5

and a symmetric stencil $$\Delta_i \Delta_j x = \frac{\gamma}{n_i+\epsilon_i-n_j-\epsilon_j}\left(\Delta_j x-\Delta_i x\right),$$6, the discrete operator is defined as a minimum over superbases,

$$\Delta_i \Delta_j x = \frac{\gamma}{n_i+\epsilon_i-n_j-\epsilon_j}\left(\Delta_j x-\Delta_i x\right),$$7

Its analysis uses the geometry of two dimensional lattices, $$\Delta_i \Delta_j x = \frac{\gamma}{n_i+\epsilon_i-n_j-\epsilon_j}\left(\Delta_j x-\Delta_i x\right),$$8-obtuse superbases, Selling’s algorithm, and the Stern-Brocot tree for systematic, parameter-free, and adaptive stencil generation. The adaptive search is proved equivalent to a brute-force search over exponentially large stencils, but at far less cost (Benamou et al., 2014).

For separable radial equations with rotational symmetry, standard central differences fail to reproduce the singular behavior at the origin when the exact local behavior is $$\Delta_i \Delta_j x = \frac{\gamma}{n_i+\epsilon_i-n_j-\epsilon_j}\left(\Delta_j x-\Delta_i x\right),$$9. The improved discretization replaces the centrifugal coefficient by the discrete analog

$\gamma=\tfrac12$0

so that the discrete equation preserves the known local analytic behavior near $\gamma=\tfrac12$1. The paper reports dramatic improvements in convergence for non-integer $\gamma=\tfrac12$2, including a change from $\gamma=\tfrac12$3 to $\gamma=\tfrac12$4 for the 2D hydrogen ground state and from $\gamma=\tfrac12$5 to $\gamma=\tfrac12$6 for a skyrmion zero mode (Laliena et al., 2018).

In 3D coordinate-space Hartree-Fock-Bogoliubov calculations, finite-difference discretization on a cubic box is evaluated against harmonic oscillator reference solutions. The implementation uses a seven-point formula for derivatives, a nine-point formula for Laplacians, and trapezoidal-rule integration. For grid spacing $\gamma=\tfrac12$7, the off-diagonal elements of the resulted HFB matrix elements are extremely small $\gamma=\tfrac12$8, the quasi-particle spectra differ from those of HO calculations by a few keV, and self-consistent HF and HFB calculations give results similar to those of existing HO basis and coordinate-space methods (Shi, 2018).

Hyperbolic discretization via Riemann invariants transforms conservative variables into characteristic coordinates and then discretizes the invariant transport equations with upwinding aligned to the signs of the characteristic speeds. The stated novelty is the possibility to allow for an efficient discretization of the boundary and coupling conditions at nodal points of the network; the original discretization is analyzed for correctly recovering steady states and resolving possible analytic solutions (Grundel et al., 2020). Across these PDE examples, analytic coordinate discretization is less about point sampling per se than about embedding the discrete scheme in the right invariant, characteristic, or arithmetic structure.

5. Trajectories, dynamical coordinates, and alternatives to parameter-grid discretization

In fast volumetric MRI, Analytic Coordinate Discretization is developed as a unified framework for designing 3D Radial, 3D Cones, and Stack-of-Spirals trajectories. A non-Cartesian trajectory is interpreted as a discretized version of an analytic coordinate defined by a set of template trajectories; equivalently, the analytic coordinate is conceptualized as a non-Cartesian trajectory composed of an infinite number of copies of a set of template trajectories. The practical trajectory is obtained by constructing a continuous spiral path on a surface and sampling points along this path at unit intervals, leaving only the essential spokes or interleaves. The framework derives analytic density compensation factors via Jacobian determinants, analytic formulae for the number of readouts based on prescribed parameters, and variable-density variants with smoothly distributed spokes or interleaves in $\gamma=\tfrac12$9-space even for a small number of readouts. In a preliminary phantom study, the proposed method demonstrated improved sampling efficiency and image quality compared to the conventional approach (Jang et al., 21 Mar 2025).

Optimization theory contains adjacent developments that replace parameter-space discretization by global coordinate parameterizations. For a family of optimization problems indexed by $(u)_{1/2}=\Gamma(u+\tfrac12)/\Gamma(u)$0, one recent method parameterizes the entire solution path as $(u)_{1/2}=\Gamma(u+\tfrac12)/\Gamma(u)$1 and solves a single stochastic optimization problem instead of discretizing $(u)_{1/2}=\Gamma(u+\tfrac12)/\Gamma(u)$2. When the solution path is analytic on $(u)_{1/2}=\Gamma(u+\tfrac12)/\Gamma(u)$3, the method requires $(u)_{1/2}=\Gamma(u+\tfrac12)/\Gamma(u)$4 gradient calls to obtain $(u)_{1/2}=\Gamma(u+\tfrac12)/\Gamma(u)$5 uniform error, whereas the best-known discretization schemes in these settings require at least $(u)_{1/2}=\Gamma(u+\tfrac12)/\Gamma(u)$6 discretization points (Dong et al., 2024). This is not presented as Analytic Coordinate Discretization in name, but it is a closely related anti-discretization result: an analytic parameter coordinate can be learned globally rather than sampled locally.

Continuous-time analysis of accelerated methods via dilated coordinate systems provides a further coordinate-transformation perspective. Instead of analyzing $(u)_{1/2}=\Gamma(u+\tfrac12)/\Gamma(u)$7, the work analyzes $(u)_{1/2}=\Gamma(u+\tfrac12)/\Gamma(u)$8, derives a conservation law in the dilated coordinates, and shows that a semi-second-order symplectic Euler discretization in the dilated coordinate system leads to an $(u)_{1/2}=\Gamma(u+\tfrac12)/\Gamma(u)$9 rate on smooth convex minimization without any further assumptions such as infinite differentiability (Suh et al., 2022). For adaptive sparse-grid regression, a preprocessing rotation on the Stiefel manifold is used to determine an optimized, problem-dependent coordinate system that reduces effective dimensionality in the ANOVA sense before sparse-grid discretization; the paper reports substantial gains in regression accuracy and computational efficiency on synthetic and real-world data (Bohn et al., 2018). Taken together, these works suggest that coordinate design and coordinate discretization are increasingly co-optimized.

6. Analyticity, bias, and open questions

Theoretical work on adapted coordinates clarifies when analytic coordinates exist prior to discretization. For a finite collection of $C^1$0 vector fields on a $C^1$1 manifold spanning the tangent space at every point, there are necessary and sufficient, coordinate-free conditions for the existence of local coordinate charts in which the pulled-back vector fields are real analytic. The chart is constructed as a joint flow

$C^1$2

and the resulting charts can be viewed as scaling maps for sub-Riemannian geometry (Street, 2018). A plausible implication is that some coordinate discretizations inherit their effectiveness from an antecedent real-analytic normal form.

In Banach spaces, continuity of coordinate functionals associated with filter bases holds whenever the underlying filter is analytic. This gives a ZFC solution to Kadets’ problem for the filter of statistical convergence. The same paper proves that if a basis with respect to an arbitrary filter has continuous coordinate functionals, then it is also a basis with respect to a filter that is analytic; at the same time, automatic continuity beyond the analytic class remains a mystery, and the existence of an $C^1$3-basis with discontinuous coordinate functionals is posed as an open problem (Rancourt et al., 2022). Here the “coordinate” is functional rather than geometric, but the emphasis on analyticity as the threshold for stable coordinate extraction is closely parallel.

A different caution appears in analytic continuation. In the Average Spectrum Method, discretization of the real-frequency grid biases the result: the distribution of the grid points plays the role of a default model, and the number of grid points acts as a regularization parameter. The paper gives a quantitative explanation for this behavior and emphasizes the practical necessity of choosing the grid density so that the corresponding default model is compatible with the data (Ghanem et al., 2019). This directly contradicts any naive view that discretization is a neutral implementation detail. Within the broader landscape of analytic coordinate discretization, the warning is general: analytic structure can improve a discretization, but discretization choices also feed back into the effective prior, admissible invariants, and asymptotic bias.