🔢 mathematics

Analytic First Derivatives of SIDER Interpolation

This paper derives analytic first-derivative formulas for arbitrary-order SIDER spherical interpolation by applying chain-rule propagation through its recursive SLERP structure, thereby providing tangent-consistent differential information essential for high-order numerical methods involving sphere-valued data.

Original authors: Shingyu Leung

Published 2026-06-15

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Shingyu Leung

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Picture: Drawing on a Ball

Imagine you have a basketball and you want to draw a smooth, curved line on its surface that passes through a specific set of dots you've marked.

In the normal world (flat paper), if you want to connect dots with a smooth curve, you might use a flexible ruler or a polynomial formula. But if you try to do that on a ball, a standard ruler will cut straight through the air, leaving the surface of the ball. If you force the line to stay on the ball by squishing it back down, the math gets messy, and the "speed" or "direction" of the line becomes hard to calculate accurately.

This paper introduces a better way to draw these lines called SIDER. Instead of using a straight ruler, SIDER uses the ball's own geometry. It connects dots by following the shortest path along the surface (like a plane flying the great circle route between two cities). This ensures the line never leaves the ball.

The Problem: Knowing the Direction

The paper solves a specific problem: How do you know which way the line is pointing at any given moment?

In math and computer simulations, knowing the "first derivative" is like knowing the exact direction and speed a car is traveling at a specific second.

The old way: You could calculate the position of the line, but figuring out the direction required complex, messy math that often broke the rule of staying on the ball.
The new way (this paper): The author, Shingyu Leung, figured out a clean, step-by-step recipe to calculate that direction (the derivative) without ever leaving the surface of the ball.

The Secret Ingredient: The "Tree" of Connections

The paper explains that SIDER works like a family tree or a Russian nesting doll.

The Building Block: The basic tool is called SLERP (Spherical Linear Interpolation). Think of SLERP as a single, perfect arc connecting two dots.
The Construction: To connect many dots, SIDER doesn't draw one big line. It builds a tree. It takes two smaller arcs, finds the middle of those, and connects them with another arc. It keeps doing this recursively until it connects all the dots.

The paper's main breakthrough is realizing that because this structure is a tree, you can calculate the direction (derivative) by working your way up the tree, just like a computer program. You calculate the direction of the small branches, and then use those to figure out the direction of the big branches.

The "Magic" Result: Always Tangent

The most important claim in the paper is about Tangency.
Imagine a bug walking along the line you drew on the ball. The "derivative" is the direction the bug is facing.

If you were on flat paper, the bug faces forward.
On a ball, the bug must face along the surface. It cannot point up into the air or down into the ball.

The paper proves that their new math formulas automatically make the bug face exactly along the surface. You don't need to add a "correction step" to force it to stay on the ball; the math naturally does it. The direction is always "tangent" (touching) the sphere, never piercing it.

Why This Matters (According to the Paper)

The author mentions this is useful for:

Computer Graphics: Making smooth animations of objects that rotate or move on spheres (like 3D models or camera angles).
Scientific Computing: Helping computers solve complex physics problems (like fluid dynamics or weather patterns) where the data represents directions (like wind direction or magnetic fields) rather than just numbers.
High-Order Methods: These are fancy math techniques used to get very precise answers. This paper gives those techniques the ability to handle "directions" on a sphere just as well as they handle numbers on a flat line.

A Note on How the Paper Was Made

The paper includes a unique disclosure at the end. The author states that they used an AI (specifically GPT-5.5) to do the heavy lifting of the mathematical modeling and derivation. The author acted as the "director," setting the goals and checking the logic, but the actual complex math was generated by the AI. The author presents this as a preprint (a draft) and notes that the results are "AI-assisted hypotheses" that need future verification by human experts.

Summary

In short, this paper provides a new, clean set of instructions for calculating the direction of a smooth line drawn on a sphere. It uses a recursive "tree" method to ensure the direction always stays perfectly on the surface, making it a powerful tool for advanced computer simulations and 3D geometry.

Technical Summary: Analytic First Derivatives of SIDER Interpolation

Problem Statement
In numerical analysis and geometric modeling, there is a growing need to interpolate data constrained to nonlinear manifolds, specifically the unit sphere $S^2 \subset \mathbb{R}^3$ . Standard Euclidean polynomial interpolation fails to preserve the spherical constraint between data points, often requiring pointwise normalization which complicates accuracy and derivative analysis. While Spherical Interpolation of orDER $n$ (SIDER- $n$ ) was previously introduced as a high-order reconstruction tool for Spherical Essentially Non-Oscillatory (SENO) methods, it lacked an analytic theory for first derivatives. The absence of derivative information limits the application of SIDER in numerical schemes requiring local slopes, directional changes, or differential quantities (e.g., generalized Riemann solvers, inverse Lax–Wendroff procedures, and high-order quadrature).

Methodology
The paper develops analytic first-derivative formulas for SIDER curves of arbitrary order by exploiting the recursive structure of the interpolation. The core methodology relies on the following principles:

Recursive Composition: SIDER- $n$ is constructed entirely from Spherical Linear Interpolation (SLERP) operations, forming a binary tree of SLERP maps.
Chain-Rule Propagation: Instead of differentiating a fully expanded high-order expression, the authors apply the chain rule directly through the recursive definition. The derivative of a SIDER- $n$ curve is expressed as the total derivative of a SLERP operation applied to two lower-order SIDER- $(n-1)$ branches.
Total SLERP Derivative: A fundamental building block is derived for the total derivative of SLERP with moving endpoints and a moving interpolation parameter. This formula accounts for the variation of the geodesic angle and the endpoints.
Specialization: The general recursive formula is specialized for:
- Interior Interpolation Nodes: Where the two lower-order branches meet at the same data point, allowing for a simplified weighted combination of derivatives.
- Endpoints: Where one branch is evaluated outside the standard interpolation interval (extrapolation), introducing logarithmic map terms.
- Middle Points: Locations halfway between sampling data, which are critical for high-order reconstruction-based algorithms.

Key Contributions

Recursive Derivative Formulas: The paper provides a compact, implementation-ready recursive algorithm for computing $D_n(t) = \frac{d}{dt}S_n(t)$ for any order $n$ . This avoids unwieldy expanded expressions for high-order curves.
Explicit Base Cases: The base case SIDER2 is treated explicitly, and SIDER3 and SIDER4 are used to illustrate the recursive mechanism and the specific handling of endpoints and interior nodes.
Tangency Preservation: A rigorous proof is provided demonstrating that the analytically derived derivative vector is inherently tangent to the sphere at every reconstructed point (both sampling nodes and middle points). This property ( $S_n(t) \cdot D_n(t) = 0$ ) arises from the algebraic structure of the SLERP derivative and does not require post-processing projection.
Middle-Point Derivatives: The paper derives specific formulas for derivatives at half-integer index coordinates ( $\xi = j + 1/2$ ), which are essential for evaluating reconstructions between data samples in finite-volume and WENO-type schemes.

Results

General Recursion: The derivative $D_n$ is computed by evaluating the total SLERP derivative on the two lower-order branches $A(t)$ and $B(t)$ , scaled by the reparameterization factor $n/(n-1)$ .
Node Simplification: At interior nodes $t_j = j/n$ , the derivative simplifies to a weighted sum of the derivatives of the two lower-order branches meeting at $p_j$ . The weights sum to $n/(n-1)$ , reflecting the scaling of the parameterization.
Endpoint Behavior: At endpoints, the derivative includes terms involving the logarithmic map ( $\text{Log}_p$ ) to account for the direction of the extrapolated branch values.
Geometric Consistency: Numerical visualizations confirm that the computed derivative vectors lie exactly in the tangent space of the sphere at the interpolated points, validating the theoretical proof of tangency.

Significance and Scope
The paper claims that these formulas extend the original SIDER/SENO framework by supplying necessary differential information for sphere-valued reconstructions. The significance lies in enabling the use of SIDER in high-order finite-volume, ENO/WENO, and SENO-type methods for conservation laws and evolution problems where the reconstructed field must be differentiated while strictly preserving the manifold constraint. The work is presented as an analytic theory that interprets the derivative computation as a forward-mode differentiation of the SIDER evaluation algorithm.

The authors note that the implementation remains compact and follows the recursive definition directly. The paper concludes by acknowledging that while the mathematical depth of the results is significant, the work is intended as a preprint for dissemination and future formal peer verification, with potential future work involving non-uniform parameter spacing and extensions to other manifolds like rotation groups.