What is the Gauss-Legendre Quadrature Calculator?
This tool computes the nodes (abscissas) and weights of the n-point Gauss-Legendre quadrature rule on the reference interval [-1, 1]. Gauss-Legendre quadrature is a numerical integration method that approximates a definite integral as a weighted sum of function values: the integral of \(f(x)\) from \(-1\) to \(1\) is approximately the sum over \(i\) of \(w_i\) times \(f(x_i)\). With just \(n\) points it integrates any polynomial up to degree \(2n-1\) exactly, making it far more accurate than equally-spaced rules like the trapezoidal or Simpson methods.
How to Use It
Pick the order \(n\) (the number of points, 2 to 100) and optionally the number of significant digits to display. The calculator returns a table of \(n\) rows, each with a node \(x_i\) and its weight \(w_i\). The nodes are symmetric about 0 and all lie strictly inside \((-1, 1)\); the weights are all positive and sum to exactly 2, the length of the interval. To integrate over an arbitrary interval \([a, b]\), map each node with \(t_i = \frac{b-a}{2} x_i + \frac{a+b}{2}\) and multiply each weight by \(\frac{b-a}{2}\).
The Formula Explained
The nodes are the \(n\) roots of the Legendre polynomial \(P_n(x)\), built with Bonnet's recurrence: \(P_0=1\), \(P_1=x\), and \(P_k = \frac{(2k-1) x P_{k-1} - (k-1) P_{k-2}}{k}\). Each weight is $$w_i = \frac{2}{\left(1 - x_i^{2}\right)\left[P_n^{\prime}(x_i)\right]^{2}},$$ where the derivative is \(P_n^{\prime}(x) = \frac{n\left(x P_n(x) - P_{n-1}(x)\right)}{x^2 - 1}\). Roots are found by Newton's method seeded at \(x = \cos\left(\frac{\pi (i - 0.25)}{n + 0.5}\right)\), which converges in a few iterations.
Worked Example (n = 3)
The roots of \(P_3\) are \(x = 0\) and \(x = \pm\sqrt{3/5} = \pm 0.7745966692\). The weight at \(x = 0\) is \(\frac{8}{9} = 0.888888889\), and the weight at each \(x = \pm 0.7745966692\) is \(\frac{5}{9} = 0.555555556\). The weights sum to $$\frac{5}{9} + \frac{8}{9} + \frac{5}{9} = 2,$$ and this 3-point rule integrates polynomials up to degree 5 exactly.
Applying the Rule to a Definite Integral Step by Step
The calculator returns nodes \(x_i\) and weights \(w_i\) on the standard interval \([-1,1]\). To integrate over a general interval \([a,b]\), follow these steps.
- Choose n. Pick the order (2 to 100) in the calculator. An n-point rule integrates every polynomial of degree up to \(2n-1\) exactly. Smooth integrands usually need only a small n.
- Get \(x_i\) and \(w_i\). Read the nodes and weights for that n. The nodes are the roots of the Legendre polynomial \(P_n\), and the weights satisfy \(w_i = \dfrac{2}{(1-x_i^2)\,[P_n'(x_i)]^2}\).
- Map the nodes to \([a,b]\). $$t_i = \frac{b-a}{2}\,x_i + \frac{a+b}{2}$$
- Scale the weights. $$w_i' = \frac{b-a}{2}\,w_i$$ The factor \(\frac{b-a}{2}\) is the Jacobian of the change of variable.
- Evaluate the function at each mapped node: \(f(t_i)\).
- Sum. $$\int_a^b f(t)\,dt \approx \sum_{i=1}^{n} w_i'\, f(t_i)$$
Check 1: \(x^2\) on \([-1,1]\) with n = 2
For n = 2 the nodes are \(x_{1,2} = \mp 1/\sqrt{3} \approx \mp 0.5773503\) and both weights equal 1. Then $$\sum w_i x_i^2 = 1\cdot\tfrac13 + 1\cdot\tfrac13 = \tfrac23,$$ which equals the exact value \(\int_{-1}^{1} x^2\,dx = 2/3\). This is expected, because degree 2 is at most \(2n-1 = 3\).
Check 2: \(e^x\) on \([0,1]\) with n = 2
Here \(a=0\) and \(b=1\), so \(t_i = 0.5\,x_i + 0.5\), which gives \(t_1 \approx 0.2113249\) and \(t_2 \approx 0.7886751\). The scaled weights are \(w_i' = 0.5\cdot 1 = 0.5\).
$$\int_0^1 e^x\,dx \approx 0.5\,(e^{0.2113249} + e^{0.7886751}) \approx 0.5\,(1.235313 + 2.200479) \approx 1.71790$$
The exact value is \(e - 1 \approx 1.71828\), so the error is about \(4\times 10^{-4}\) with just two function evaluations. Increasing n shrinks the error very quickly.
Interpreting the Nodes and Weights
- Symmetry. The nodes are symmetric about 0: \(x_i = -x_{n+1-i}\). The matching weights are equal: \(w_i = w_{n+1-i}\). For odd n, the middle node is exactly 0.
- Positive weights that sum to 2. Every weight is strictly positive, and \(\sum w_i = 2\), the length of \([-1,1]\), since the rule integrates the constant 1 exactly. Positive weights make the rule numerically stable, with no cancellation between large terms of opposite sign. After scaling to \([a,b]\), the weights sum to \(b-a\).
- Node locations. All nodes lie strictly inside \((-1,1)\), and the endpoints are never used. As n grows, the nodes cluster toward the endpoints. Their density follows roughly \(1/\sqrt{1-x^2}\), and the weights near the ends become small.
- Degree of exactness. An n-point rule is exact for all polynomials of degree \(\le 2n-1\). This is the maximum possible for n nodes, twice what equally spaced Newton-Cotes rules with n points typically achieve.
- Error term. On \([-1,1]\) the error is $$E_n = \frac{2^{2n+1}\,(n!)^4}{(2n+1)\,[(2n)!]^3}\; f^{(2n)}(\xi), \qquad \xi \in (-1,1).$$ For n = 2 the coefficient is \(1/135\). On \([a,b]\) the error is multiplied by \(\left(\frac{b-a}{2}\right)^{2n+1}\). The error is therefore governed by the size of the \(2n\)-th derivative.
- Smooth integrands. For analytic functions such as \(e^x\), \(\sin x\) or rational functions with poles far from the interval, convergence is geometric (exponentially fast in n). A modest n often reaches near machine precision.
- Singular or non-smooth integrands. If f has a kink, a jump, an endpoint singularity (such as \(\sqrt{x}\) or \(\ln x\) at 0), or a nearby pole, the high derivatives are large or undefined. Convergence then drops to algebraic and may be slow. Practical remedies are to split the interval at the trouble point, apply a variable substitution that removes the singularity, use a rule with a matching weight function (such as Gauss-Jacobi or Gauss-Chebyshev), or use composite or adaptive Gauss rules.
- High orders. Very large n gives no benefit if f is not smooth. Compare results for two different n to estimate the error.
FAQ
Why do the weights sum to 2? Integrating the constant function \(f(x) = 1\) over \([-1, 1]\) gives 2, and quadrature must reproduce constants exactly, so the weights must total the interval length.
How accurate are the values? This is a double-precision rebuild giving about 15 significant digits. Display options requesting more digits are rounded to what double precision can represent.
What is the maximum degree integrated exactly? An \(n\)-point rule is exact for all polynomials of degree up to \(2n-1\).