Chebyshev–Gauss–Lobatto Nodes

• Chebyshev–Gauss–Lobatto nodes are a set of collocation points defined as the extrema of Chebyshev polynomials, including endpoints and clustering near boundaries for spectral accuracy.
• They enable efficient quadrature, interpolation, and spectral collocation in solving differential equations with closed-form expressions and O(N) computational complexity.
• Their design minimizes interpolation error and suppresses the Runge phenomenon, making them pivotal in high-precision numerical methods in applied mathematics and computational physics.

Chebyshev–Gauss–Lobatto nodes are a canonical set of collocation and quadrature points derived from the extrema of Chebyshev polynomials of the first kind. Their defining property is that they include domain endpoints, cluster algebraically near boundaries, and enable spectrally accurate approximation, collocation, and integration of functions on finite intervals. Chebyshev–Gauss–Lobatto nodes form the foundation of a variety of spectral algorithms, including collocation methods in differential equations and near-optimal polynomial interpolation and quadrature schemes. The nodes admit closed-form expressions, enable efficient $\mathcal{O}(N)$ construction, and underlie high-accuracy machine-precision algorithms for polynomial-based computations (Gil et al., 20 Sep 2025).

1. Definition and Construction

Chebyshev–Gauss–Lobatto (CGL) nodes are defined on the canonical interval $x \in [-1,1]$ as the extrema of the $N$th-degree Chebyshev polynomial of the first kind $T_N(x)$. The explicit formula is

$$x_j = \cos\left(\frac{\pi j}{N}\right), \qquad j=0,1,\dots,N,$$

where $N$ is the total polynomial degree, so there are $N+1$ nodes in total. These include both endpoints $x_0=1$ and $x_N=-1$, distinguishing them from Gauss or Gauss–Radau nodes which do not necessarily include both interval boundaries. The CGL nodes cluster quadratically near the endpoints $x=\pm1$.

For Jacobi weights $x \in [-1,1]$0 with $x \in [-1,1]$1, the CGL rule specializes as a Gauss–Lobatto–Jacobi quadrature, with interior nodes at the zeros of $x \in [-1,1]$2, the Chebyshev polynomial of the second kind of degree $x \in [-1,1]$3 (Gil et al., 20 Sep 2025).

2. Quadrature Weights and Barycentric Structure

CGL quadrature rules provide closed-form expressions for both nodes and weights. For $x \in [-1,1]$4 CGL nodes,

$x \in [-1,1]$5

where $x \in [-1,1]$6 are the quadrature weights for integrals of the form

$x \in [-1,1]$7

This quadrature rule integrates all polynomials of degree up to $x \in [-1,1]$8 exactly (Gil et al., 20 Sep 2025).

For polynomial interpolation, the associated barycentric weights are

$x \in [-1,1]$9

These barycentric weights enable stable and efficient evaluation of the interpolating polynomial through the CGL nodes.

3. Mapping and Applications in Spectral Collocation

CGL nodes can be linearly mapped to an arbitrary finite interval $N$0, or, as in cosmological applications, to $N$1 for the variable $N$2: $N$3 This affine transformation preserves endpoint clustering, now near $N$4 and $N$5, and allows the Chebyshev spectral method to be applied directly to physical domains (Rana, 17 Oct 2025).

In the spectral collocation approach, a target function $N$6 (e.g., the cosmological Hubble parameter, normalized) is expanded as a truncated Chebyshev series: $N$7 where $N$8 and the coefficients $N$9 are determined by enforcing the governing equations at each $T_N(x)$0. This transforms differential or boundary-value problems into algebraic systems, with derivatives calculated via Chebyshev differentiation matrices. The method is spectrally accurate and numerically robust, particularly for problems exhibiting stiffness or boundary layers (Rana, 17 Oct 2025).

4. Theoretical Properties and Spectral Approximation

CGL nodes possess multiple optimality and convergence attributes:

• Endpoint Clustering: The algebraic clustering ($T_N(x)$1 accumulate near $T_N(x)$2) enables resolution of boundary layers and is essential for high accuracy in problems with sharp endpoint features.
• Minimax Interpolation: Among all possible $T_N(x)$3-point node distributions, CGL nodes nearly minimize the uniform error of polynomial interpolation—closely realizing the Chebyshev equioscillation (minimax) property.
• Spectral Convergence: For analytic target functions, the Chebyshev expansion coefficients $T_N(x)$4 decay exponentially with degree $T_N(x)$5, yielding global approximation errors that decrease faster than any polynomial in $T_N(x)$6.
• Runge Phenomenon Suppression: The clustering suppresses large oscillations that afflict equispaced interpolants, preserving stability and accuracy for high-degree interpolations (Rana, 17 Oct 2025).
• Exact Integration: CGL quadrature rules, for the Chebyshev weight, integrate polynomials of degree up to $T_N(x)$7 exactly and achieve machine-precision for smooth integrands (Gil et al., 20 Sep 2025).

5. Computational Aspects and Accuracy

All CGL node, weight, and barycentric evaluations admit $T_N(x)$8 complexity: a single cosine evaluation per node, and a fixed number of floating-point operations per weight. In IEEE 64-bit arithmetic, CGL quadrature and interpolation maintain full precision (15–16 digits) for $T_N(x)$9 up to $$x_j = \cos\left(\frac{\pi j}{N}\right), \qquad j=0,1,\dots,N,$$0 or more, constrained primarily by memory resources (Gil et al., 20 Sep 2025).

For the Chebyshev–Gauss–Lobatto special case ($$x_j = \cos\left(\frac{\pi j}{N}\right), \qquad j=0,1,\dots,N,$$1), no iterative or nonlinear solves are needed to compute the nodes—even for large $$x_j = \cos\left(\frac{\pi j}{N}\right), \qquad j=0,1,\dots,N,$$2. In more general Jacobi–Lobatto rules, globally convergent fourth-order fixed point iterations are available, but for CGL the nodes are exact by construction, with barycentric and quadrature weights in immediately computable closed form.

Empirically, CGL quadrature rules match theoretical accuracy: polynomials of degree up to $$x_j = \cos\left(\frac{\pi j}{N}\right), \qquad j=0,1,\dots,N,$$3 are integrated to machine precision. Interpolation stability at these nodes is optimal due to the minimal Lebesgue constant, further enhanced by the barycentric form (Gil et al., 20 Sep 2025).

6. Applications in Collocation and Quadrature

CGL nodes are a standard discretization in spectral methods for solving ODEs and PDEs, underpinning the spectral collocation (“pseudospectral”) schemes. In numerical cosmology, they enable globally accurate resolution of stiff or nonlinear equations, such as the modified Friedmann equations in $$x_j = \cos\left(\frac{\pi j}{N}\right), \qquad j=0,1,\dots,N,$$4 gravity, by transforming continuous boundary value problems to nonlinear algebraic systems for spectral coefficients. The algebraic endpoint clustering is crucial when physical solutions develop rapid variation near boundaries or require imposition of precise endpoint conditions (Rana, 17 Oct 2025).

CGL quadrature and interpolation with barycentric weights remain a core routine in numerical analysis, fast transforms, and data-driven polynomial approximation, due to their optimal convergence and stability. The efficiency of CGL construction ensures scalability to very high polynomial degrees, a requirement for high-precision computational physics and scientific computing (Gil et al., 20 Sep 2025).

7. Summary of Algorithmic Construction

The algorithmic generation of the CGL rule is streamlined:

Step	Operation	Formula/Action
Node set	Compute $$x_j = \cos\left(\frac{\pi j}{N}\right), \qquad j=0,1,\dots,N,$$5	$$x_j = \cos\left(\frac{\pi j}{N}\right), \qquad j=0,1,\dots,N,$$6
Weight set	$$x_j = \cos\left(\frac{\pi j}{N}\right), \qquad j=0,1,\dots,N,$$7, $$x_j = \cos\left(\frac{\pi j}{N}\right), \qquad j=0,1,\dots,N,$$8	$$x_j = \cos\left(\frac{\pi j}{N}\right), \qquad j=0,1,\dots,N,$$9
Barycentric	$N$0 for $N$1; else $N$2

All steps achieve $N$3 cost, use only elementary functions, and are numerically stable for large-scale computation. For interpolation, the barycentric formula enables rapid evaluation and minimal roundoff error, while, in collocation contexts, the nodes' properties yield well-conditioned systems and robust spectral convergence (Gil et al., 20 Sep 2025).