|
|
(One intermediate revision by one other user not shown) |
Line 1: |
Line 1: |
− | A formula for calculating an integral over a finite interval <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/r/r080/r080120/r0801201.png" />: | + | {{TEX|done}} |
| + | A formula for calculating an integral over a finite interval $[a,b]$: |
| | | |
− | <table class="eq" style="width:100%;"> <tr><td valign="top" style="width:94%;text-align:center;"><img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/r/r080/r080120/r0801202.png" /></td> <td valign="top" style="width:5%;text-align:right;">(*)</td></tr></table>
| + | \begin{equation}\label{eq:1} |
| + | \int\limits_a^bf(x)dx\cong h\sum_{k=1}^Nf(\alpha+(k-1)h), |
| + | \end{equation} |
| | | |
− | where <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/r/r080/r080120/r0801203.png" /> and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/r/r080/r080120/r0801204.png" />. Its algebraic degree of accuracy is 1 if <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/r/r080/r080120/r0801205.png" /> and 0 otherwise. | + | where $h=(b-a)/N$ and $\alpha\in[a,a+h]$. Its algebraic degree of accuracy is 1 if $\alpha=a+h/2$ and 0 otherwise. |
| | | |
− | The [[Quadrature formula|quadrature formula]] (*) is exact for the trigonometric functions | + | The [[quadrature formula]] \eqref{eq:1} is exact for the trigonometric functions |
| | | |
− | <table class="eq" style="width:100%;"> <tr><td valign="top" style="width:94%;text-align:center;"><img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/r/r080/r080120/r0801206.png" /></td> </tr></table>
| + | $$\cos\frac{2\pi}{b-a}kx,\quad\sin\frac{2\pi}{b-a}kx,\quad k=0,\dots,N-1.$$ |
| | | |
− | If <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/r/r080/r080120/r0801207.png" />, then (*) is exact for all trigonometric polynomials of order at most <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/r/r080/r080120/r0801208.png" />; moreover, its trigonometric degree of accuracy is <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/r/r080/r080120/r0801209.png" />. No other quadrature formula with <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/r/r080/r080120/r08012010.png" /> real nodes can have trigonometric degree of accuracy larger than <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/r/r080/r080120/r08012011.png" />, so that the rectangle rule with <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/r/r080/r080120/r08012012.png" /> has the highest trigonometric degree of accuracy. | + | If $b-a=2\pi$, then \eqref{eq:1} is exact for all trigonometric polynomials of order at most $N-1$; moreover, its trigonometric degree of accuracy is $N-1$. No other quadrature formula with $N$ real nodes can have trigonometric degree of accuracy larger than $N-1$, so that the rectangle rule with $b-a=2\pi$ has the highest trigonometric degree of accuracy. |
| | | |
− | Let <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/r/r080/r080120/r08012013.png" /> be the error of the rectangle rule, i.e. the difference between the left- and right-hand sides of (*). If the integrand <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/r/r080/r080120/r08012014.png" /> is twice continuously differentiable on <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/r/r080/r080120/r08012015.png" />, then for <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/r/r080/r080120/r08012016.png" /> one has | + | Let $R(f,\alpha)$ be the error of the rectangle rule, i.e. the difference between the left- and right-hand sides of \eqref{eq:1}. If the integrand $f$ is twice continuously differentiable on $[a,b]$, then for $\alpha=a+h/2$ one has |
| | | |
− | <table class="eq" style="width:100%;"> <tr><td valign="top" style="width:94%;text-align:center;"><img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/r/r080/r080120/r08012017.png" /></td> </tr></table>
| + | $$R\left(f,a+\frac h2\right)=\frac{b-a}{24}h^2f''(\xi),$$ |
| | | |
− | for some <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/r/r080/r080120/r08012018.png" />. If <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/r/r080/r080120/r08012019.png" /> is a periodic function with period <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/r/r080/r080120/r08012020.png" /> and has a continuous derivative of order <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/r/r080/r080120/r08012021.png" /> (where <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/r/r080/r080120/r08012022.png" /> is a natural number) on the entire real axis, then for any <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/r/r080/r080120/r08012023.png" />, | + | for some $\xi\in[a,b]$. If $f$ is a periodic function with period $b-a$ and has a continuous derivative of order $2k$ (where $k$ is a natural number) on the entire real axis, then for any $\alpha\in[a,a+h]$, |
| | | |
− | <table class="eq" style="width:100%;"> <tr><td valign="top" style="width:94%;text-align:center;"><img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/r/r080/r080120/r08012024.png" /></td> </tr></table>
| + | $$R(f,\alpha)=-(b-a)B_{2k}\frac{h^{2k}}{(2k)!}f^{(2k)}(\eta),$$ |
| | | |
− | for some <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/r/r080/r080120/r08012025.png" />, where <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/r/r080/r080120/r08012026.png" /> is the Bernoulli number (cf. [[Bernoulli numbers|Bernoulli numbers]]). | + | for some $\eta\in[a,b]$, where $B_{2k}$ is the Bernoulli number (cf. [[Bernoulli numbers]]). |
| | | |
| | | |
Latest revision as of 17:35, 24 March 2018
A formula for calculating an integral over a finite interval $[a,b]$:
\begin{equation}\label{eq:1}
\int\limits_a^bf(x)dx\cong h\sum_{k=1}^Nf(\alpha+(k-1)h),
\end{equation}
where $h=(b-a)/N$ and $\alpha\in[a,a+h]$. Its algebraic degree of accuracy is 1 if $\alpha=a+h/2$ and 0 otherwise.
The quadrature formula \eqref{eq:1} is exact for the trigonometric functions
$$\cos\frac{2\pi}{b-a}kx,\quad\sin\frac{2\pi}{b-a}kx,\quad k=0,\dots,N-1.$$
If $b-a=2\pi$, then \eqref{eq:1} is exact for all trigonometric polynomials of order at most $N-1$; moreover, its trigonometric degree of accuracy is $N-1$. No other quadrature formula with $N$ real nodes can have trigonometric degree of accuracy larger than $N-1$, so that the rectangle rule with $b-a=2\pi$ has the highest trigonometric degree of accuracy.
Let $R(f,\alpha)$ be the error of the rectangle rule, i.e. the difference between the left- and right-hand sides of \eqref{eq:1}. If the integrand $f$ is twice continuously differentiable on $[a,b]$, then for $\alpha=a+h/2$ one has
$$R\left(f,a+\frac h2\right)=\frac{b-a}{24}h^2f''(\xi),$$
for some $\xi\in[a,b]$. If $f$ is a periodic function with period $b-a$ and has a continuous derivative of order $2k$ (where $k$ is a natural number) on the entire real axis, then for any $\alpha\in[a,a+h]$,
$$R(f,\alpha)=-(b-a)B_{2k}\frac{h^{2k}}{(2k)!}f^{(2k)}(\eta),$$
for some $\eta\in[a,b]$, where $B_{2k}$ is the Bernoulli number (cf. Bernoulli numbers).
References
[a1] | D.M. Young, R.T. Gregory, "A survey of numerical mathematics" , Dover, reprint (1988) pp. 362ff |
How to Cite This Entry:
Rectangle rule. Encyclopedia of Mathematics. URL: http://encyclopediaofmath.org/index.php?title=Rectangle_rule&oldid=13753
This article was adapted from an original article by I.P. Mysovskikh (originator), which appeared in Encyclopedia of Mathematics - ISBN 1402006098.
See original article