Gauss-Lobatto Quadrature Nodes and Weights Calculator

Gauss-Lobatto Quadrature Nodes and Weights Calculator

Emerging AI Skill PRO

Enter Calculation

Formula

∫−11f(x)dx≈∑i=1nwif(xi)where{x1=−1,xn=1xi:Pn−1′(xi)=0(interior)w1=wn=2n(n−1)wi=2n(n−1)[Pn−1(xi)]2\begin{gathered} \int_{-1}^{1} f(x)\,dx \approx \sum_{i=1}^{\text{n}} w_i\, f(x_i) \\[1.5em] \text{where}\quad \left\{ \begin{aligned} x_1 &= -1,\quad x_{\text{n}} = 1 \\ x_i &: P_{\text{n}-1}^{\prime}(x_i) = 0 \quad(\text{interior}) \\ w_{1} &= w_{\text{n}} = \frac{2}{\text{n}\,(\text{n}-1)} \\ w_i &= \frac{2}{\text{n}\,(\text{n}-1)\,\left[P_{\text{n}-1}(x_i)\right]^{2}} \end{aligned} \right. \end{gathered}
Usage stats

Advertisement

Results

Gauss-Lobatto rule, n = 20
20
points on the interval [-1, 1] · sum of weights = 2
i Node x_i Weight w_i
1 -1 0.005263157894737
2 -0.980743704893914 0.032237123188489
3 -0.935934498812665 0.057181802127567
4 -0.86687797808995 0.08063176399612
5 -0.775368260952056 0.101991499699451
6 -0.663776402290311 0.120709227628675
7 -0.534992864031886 0.136300482358724
8 -0.392353183713909 0.148361554070917
9 -0.239551705922986 0.156580102647475
10 -0.080545937238822 0.160743286387846
11 0.080545937238822 0.160743286387846
12 0.239551705922986 0.156580102647475
13 0.392353183713909 0.148361554070917
14 0.534992864031886 0.136300482358724
15 0.663776402290311 0.120709227628675
16 0.775368260952056 0.101991499699451
17 0.86687797808995 0.08063176399612
18 0.935934498812665 0.057181802127567
19 0.980743704893914 0.032237123188489
20 1 0.005263157894737

Report an error on this page

What this calculator does

This tool computes the nodes (abscissas) \(x_i\) and weights \(w_i\) of the n-point Gauss-Lobatto quadrature rule on the reference interval \([-1, 1]\) with weight function \(w(x) = 1\). Unlike standard Gauss-Legendre quadrature, the Gauss-Lobatto rule always forces the two endpoints \(x = -1\) and \(x = +1\) to be quadrature nodes, which is valuable when boundary values matter (for example in spectral element methods). This is pure numerical analysis and applies identically everywhere; it is not region-specific.

Number line from -1 to 1 with quadrature nodes including both endpoints, each marked by a vertical weight bar
Gauss-Lobatto nodes on [-1, 1] include both endpoints; bar heights suggest the associated weights.

How to use it

Pick the number of points \(n\) (between 2 and 100) and optionally a display precision. The calculator returns a table of \(n\) rows, each giving the node \(x_i\) and its weight \(w_i\). The nodes are symmetric about 0 and the weights are symmetric too, so \(x_i\) and \(-x_i\) share the same weight. As a built-in sanity check, the sum of all weights equals 2, the length of the interval.

The formula explained

The rule approximates the integral as the sum \(w_1 f(x_1) + \ldots + w_n f(x_n)\) and is exact for polynomials up to degree \(2n-3\). The interior nodes \(x_2, \ldots, x_{n-1}\) are the \(n-2\) zeros of \(P'_{n-1}(x)\), the derivative of the Legendre polynomial of degree \(n-1\). $$\int_{-1}^{1} f(x)\,dx \approx \sum_{i=1}^{n} w_i\, f(x_i)$$ $$\text{where}\quad \left\{ \begin{aligned} x_1 &= -1,\quad x_{n} = 1 \\ x_i &: P_{n-1}^{\prime}(x_i) = 0 \quad(\text{interior}) \\ w_{1} &= w_{n} = \frac{2}{n\,(n-1)} \\ w_i &= \frac{2}{n\,(n-1)\,\left[P_{n-1}(x_i)\right]^{2}} \end{aligned} \right.$$ The endpoints take weight \(2 / (n(n-1))\), while each interior node \(x_i\) takes weight \(2 / (n(n-1)[P_{n-1}(x_i)]^2)\). The calculator finds the interior roots by Newton iteration starting from Chebyshev-Gauss-Lobatto guesses \(\cos(\pi j / (n-1))\), giving full double precision (about 15-16 significant digits).

Advertisement
Curve f(x) over [-1,1] approximated by weighted samples at Gauss-Lobatto nodes including the endpoints
The integral is approximated by a weighted sum of function values at the nodes, with the two endpoints always included.

Worked example (n = 4)

The interior nodes solve \(P'_3(x) = (15x^2 - 3)/2 = 0\), so \(x = \pm 1/\sqrt{5} = \pm 0.4472135955\). The endpoint weight is \(2/(4\cdot 3) = 1/6 = 0.1666666667\). For the interior nodes \(P_3(1/\sqrt{5}) = -0.4472135955\), whose square is \(0.2\), giving weight \(2/(4\cdot 3\cdot 0.2) = 5/6 = 0.8333333333\). The sum \(1/6 + 5/6 + 5/6 + 1/6 = 2\), confirming the rule.

Gauss-Lobatto Nodes and Weights for n = 2 to 6

The n-point Gauss-Lobatto rule on [-1, 1] fixes the endpoints \(x=\pm 1\) as nodes. The remaining n-2 interior nodes are the roots of \(P_{n-1}'(x)\), and the rule is exact for polynomials up to degree \(2n-3\). Nodes are symmetric about 0, and each node shares its weight with its mirror image. Decimals are given to about 10 digits.

n Node \(x_i\) (exact) Node (decimal) Weight \(w_i\) (exact) Weight (decimal) Exact up to degree
2 ±1 ±1.0000000000 1 1.0000000000 1
3 ±1 ±1.0000000000 1/3 0.3333333333 3
0 0.0000000000 4/3 1.3333333333
4 ±1 ±1.0000000000 1/6 0.1666666667 5
±1/√5 ±0.4472135955 5/6 0.8333333333
5 ±1 ±1.0000000000 1/10 0.1000000000 7
±√(3/7) ±0.6546536707 49/90 0.5444444444
0 0.0000000000 32/45 0.7111111111
6 ±1 ±1.0000000000 1/15 0.0666666667 9
±√(1/3 + 2√7/21) ±0.7650553239 (14 − √7)/30 0.3784749563
±√(1/3 − 2√7/21) ±0.2852315165 (14 + √7)/30 0.5548583770

Check for n = 6: \(2\left(\tfrac{1}{15}+\tfrac{14-\sqrt7}{30}+\tfrac{14+\sqrt7}{30}\right)=2\left(\tfrac{2}{30}+\tfrac{28}{30}\right)=2\). In every row the weights sum to 2, the length of [-1, 1]. The endpoint weights follow \(w_1=w_n=\dfrac{2}{n(n-1)}\): 1, 1/3, 1/6, 1/10 and 1/15 for n = 2 to 6.

Gauss-Lobatto vs Gauss-Legendre Comparison

Rule (n points) Degree of exactness Endpoints are nodes? Free nodes Typical use
Gauss-Legendre \(2n-1\) No n (all nodes and weights are free, 2n parameters) General-purpose smooth integrals; highest accuracy per function evaluation
Gauss-Lobatto \(2n-3\) Yes (both) n − 2 interior nodes Spectral element methods (GLL points), problems needing values at the boundaries, nodes shared between adjacent elements
Clenshaw-Curtis at least \(n-1\) Yes (both) 0 (nodes are fixed Chebyshev extrema) Nested rules, FFT-based weights, adaptive integration

Gauss-Lobatto gives up two degrees of freedom (the endpoint positions), which costs two degrees of exactness compared with Gauss-Legendre at the same n.

Same integrand, n = 4

Integrate \(f(x)=x^6\) over [-1, 1]. The exact value is \(2/7\approx 0.285714\).

  • 4-point Gauss-Legendre is exact to degree 7, so it returns \(2/7\) exactly.
  • 4-point Gauss-Lobatto is exact only to degree 5. With nodes ±1 (weight 1/6) and ±1/√5 (weight 5/6): $$2\left[\tfrac{1}{6}(1)^6+\tfrac{5}{6}\left(\tfrac{1}{\sqrt5}\right)^6\right]=2\left[\tfrac{1}{6}+\tfrac{5}{6}\cdot\tfrac{1}{125}\right]=2\cdot\tfrac{26}{150}=\tfrac{26}{75}\approx 0.346667$$ The error is about 0.0610, because degree 6 exceeds \(2n-3=5\).

For \(f(x)=x^4\) (degree 4 ≤ 5), Lobatto is exact: \(2\left[\tfrac16+\tfrac56\cdot\tfrac1{25}\right]=2\cdot\tfrac{6}{30}=\tfrac25\).

Advertisement

Interpreting the Nodes and Weights

Mapping to [a, b]. For an interval [a, b], substitute \(x=\tfrac{b-a}{2}\,t+\tfrac{a+b}{2}\), where t is a node on [-1, 1]. Then \(dx=\tfrac{b-a}{2}\,dt\), so $$\int_a^b f(x)\,dx\approx \frac{b-a}{2}\sum_{i=1}^{n} w_i\, f\!\left(\tfrac{b-a}{2}\,t_i+\tfrac{a+b}{2}\right).$$ The scaled weights \(\tfrac{b-a}{2}w_i\) sum to \(b-a\).

Example. Integrate \(x^3\) over [0, 2] with n = 3. Here \((b-a)/2=1\), and the nodes \(t=-1,0,1\) map to \(x=0,1,2\). Then \(1\cdot\left[\tfrac13(0)^3+\tfrac43(1)^3+\tfrac13(2)^3\right]=\tfrac43+\tfrac83=4\), which equals the exact value \(2^4/4=4\). This is expected because degree 3 equals \(2n-3\).

  • Clustering: interior nodes bunch toward ±1, as with Chebyshev-type points. This is why the points suit high-order polynomial interpolation and avoid the Runge phenomenon.
  • Positive weights: all Gauss-Lobatto weights are positive, so the rule is numerically stable with no cancellation from mixed-sign weights.
  • Sum of weights: the weights sum to 2, the integral of the constant 1 over [-1, 1]. This is a quick sanity check on any table.
  • Symmetry: the nodes come in pairs ±x, and mirrored nodes have equal weights. Odd integrands over [-1, 1] therefore integrate to exactly 0. For odd n, the center node is 0.
  • Degree of exactness: the rule is exact up to degree \(2n-3\). A polynomial of degree \(2n-2\) or higher generally produces an error, as the \(x^6\) example with n = 4 shows.
  • Non-polynomial integrands: for functions such as \(e^x\) or \(\sin x\), accuracy depends on how well a polynomial approximates them. Smooth, analytic functions converge quickly as n grows. Functions with singularities, kinks, discontinuities or rapid oscillation can converge slowly. Split the interval at such features or use a rule designed for them.

FAQ

How does this differ from Gauss-Legendre? Gauss-Legendre places all nodes strictly inside \((-1, 1)\) and is exact to degree \(2n-1\). Gauss-Lobatto fixes both endpoints as nodes and is exact to degree \(2n-3\), trading two degrees of precision for boundary inclusion.

How do I use these on a general interval [a, b]? Map each node with \(x \to (b-a)/2 \cdot x + (a+b)/2\) and multiply every weight by \((b-a)/2\). This page only outputs the \([-1, 1]\) values.

Why must the weights sum to 2? Integrating \(f(x) = 1\) over \([-1, 1]\) gives 2, and the rule is exact for constants, so the weights must add up to the interval length.

Last updated:

Most popular in Math and Statistics

View all Math and Statistics calculators