Numerical integration#
Numerical integration approximates analytic integration, and is particularly useful because:
Analytic integrals may be hard to find if they exist at all!
Integration tends to damp experimental noise (in contrast with differentiation which tends to amplify it)
In 1D, integration is simply finding the area under the curve \(s = \int_a^b f(x) \ dx\) in the range [a,b]:
![Integral_as_region_under_curve.svg](data:image/svg+xml;base64,PD94bWwgdmVyc2lvbj0iMS4wIiBlbmNvZGluZz0iVVRGLTgiIHN0YW5kYWxvbmU9Im5vIj8+CjxzdmcgeG1sbnM6c3ZnPSJodHRwOi8vd3d3LnczLm9yZy8yMDAwL3N2ZyIgIHhtbG5zPSJodHRwOi8vd3d3LnczLm9yZy8yMDAwL3N2ZyIKCXZlcnNpb249IjEuMCIgd2lkdGg9Ijc1MHB4IiBoZWlnaHQ9IjcwMHB4IiB2aWV3Qm94PSIwIDAgNzUwIDcwMCI+Cgk8ZGVmcz4KCTwvZGVmcz4KCTxnIHN0eWxlPSJzdHJva2U6IzAwMDAwMDtzdHJva2Utd2lkdGg6NDtzdHJva2UtbGluZWNhcDpidXR0O3N0cm9rZS1saW5lam9pbjptaXRlcjtmaWxsLXJ1bGU6ZXZlbm9kZDsiPgoJCTxwYXRoIHN0eWxlPSJmaWxsOiNkZGRkZGQ7IiBkPSJNIDIzNy42LDE4OC44IEMgMjc5LjUsMjAxLjMgMzM1LjQsMjM1IDM4NiwyNjMuMSA0NjEuNSwzMDUgNTM3LjYsMzM1LjIgNjM2LjksMzAzLjMgTCA2MzgsNTcxLjMgMjM3LjYsNTcwLjIgMjM3LjYsMTg4LjggeiIgLz4KCQk8cGF0aCBzdHlsZT0iZmlsbDojMDAwMDAwOyIgZD0iTSA0Ni44OCw1NzEuMyA2ODcuMyw1NzIuNSIgLz4KCQk8cGF0aCBzdHlsZT0iZmlsbDojMDAwMDAwOyIgZD0iTSAxMDcuMiw2MzYuNyAxMDYsMzcuMzMgMTA2LDU5LjM3IiAvPgoJCTxwYXRoIHN0eWxlPSJmaWxsOm5vbmU7ICAgIiBkPSJNIDYzNi4zLDMwMy4zIEMgNjY4LjEsMjkzLjggNjg2LDI4NyA3MTMuNywyNzQuNiIgLz4KCQk8cGF0aCBzdHlsZT0iZmlsbDpub25lOyAgICIgIGQ9Ik0gMTEuMjIsMjQxLjQgQyA3My4wNywxOTguOCAxMzcuNywxNjEuMyAyMzgsMTg4LjciIC8+CgkJPHBhdGggc3R5bGU9ImZpbGw6IzAwMDAwMDtzdHJva2Utd2lkdGg6MTsiIGQ9Ik0gMTA0LjEsNjMuODEgODYuNzQsODAuNTQgMTA1LjksMTIuMzkgMTI1LjEsNzkuOTIgMTA4LjQsNjMuODEgMTA0LjEsNjMuODEgeiIgLz4KCQk8cGF0aCBzdHlsZT0iZmlsbDojMDAwMDAwO3N0cm9rZS13aWR0aDoxOyIgZD0iTSA2NzguNCw1NzAuNSA2NjEuOCw1NTMgNzI5LjgsNTcyLjcgNjYyLjEsNTkxLjQgNjc4LjQsNTc0LjggNjc4LjQsNTcwLjUgeiIgLz4KCTwvZz4KCTxnIHN0eWxlPSJmb250LXNpemU6NzBweDtmb250LXN0eWxlOml0YWxpYztmb250LXdlaWdodDpib2xkO3RleHQtYW5jaG9yOnN0YXJ0O2ZpbGw6IzAwMDAwMDtzdHJva2U6bm9uZTtmb250LWZhbWlseTpUaW1lcyBOZXcgUm9tYW47Ij4KCQk8dGV4dCB4PSIyMTMiIHk9IjYzMyI+YTwvdGV4dD4KCQk8dGV4dCB4PSI2MDkiIHk9IjYzOCI+YjwvdGV4dD4KCQk8dGV4dCB4PSI2ODIiIHk9IjYzMyI+eDwvdGV4dD4KCQk8dGV4dCB4PSIxNDYiIHk9IjUzIiA+eTwvdGV4dD4KCQk8dGV4dCB4PSI0MzkiIHk9IjI0NyI+Zih4KTwvdGV4dD4KCQk8dGV4dCB4PSI0MTEiIHk9IjQ2MCI+UzwvdGV4dD4KCTwvZz4KPC9zdmc+Cg==)
For definite integrals (i.e.: with finite limits), numerical integration is called numerical quadrature.
Aside: The integral sign \(\int\) looks like an elongated ‘S’ because that ‘summa’ (latin for summation) is exactly what we are doing!
The methods discussed in this section consider 2 cases:
The function \(f(x)\) is available
The data \(f(x), x\) is known at a set of points.