Numerical solution of generalized Fresnel integral

563 Views Asked by At

We need to find an approximate solution for the generalized Fresnel integral:

$\int_0^S \cos(as+\frac{bs^2}{2}+\frac{cs^3}{3}+\frac{ds^4}{4})ds$

Our approach is to use the Simpsons rule: $\int_a^bf(x)dx\approx \frac{h}{3}(f(x_0)+2\sum_{j=1}^{n/2-1}f(x_{2j})+4\sum_{j=1}^{n/2}f(x_{2j-1})+f(x_n))$

where $h=\frac{b-a}{n}$ and $x_j=a+jh$ and $j\in{[0,n]}$.

We don't know if we can apply the Simpsons rule directly to the function as the argument of the cosine is a polynomial. Can we solve the argument for $s=x_0$ and then evaluate the cosine or do we have to treat the argument somehow to be able to apply the Simpsons rule?

Thanks in advance!

1

There are 1 best solutions below

0
On

The application of Simpson's rule to any definite integral $I(f) = \int_a^b f(x)dx$ requires only that the integrand $f$ is continuous on the interval $[a,b]$. In particular, the application of Simpson's rule to the integral $$I(g) = \int_0^S \cos\left(a \cdot s+\frac{b \cdot s^2}{2}+\frac{c\cdot s^3}{3}+\frac{d\cdot s^4}{4}\right)ds$$ is perfectly valid because the integrand $g : [0,S] \rightarrow \mathbb{R}$ given by $$ g(s) = \cos\left(a \cdot s+\frac{b \cdot s^2}{2}+\frac{c\cdot s^3}{3}+\frac{d\cdot s^4}{4}\right)$$ is continuous for all $s \in [0,S]$. In fact, $g$ is infinitely often differentiable, i.e., $f \in C^{\infty}([0,S],\mathbb{R})$. In particular, $g \in C^4([0,S],\mathbb{R})$ and the classical error formula holds true. There exists at least one $\xi \in [0,S]$ such that $$ I(g) - S_h(g) = - \frac{S}{180} g^{(4)}(\xi) h^4$$ where $S_h(g)$ denotes the composite trapezoidal rule for $g$ corresponding to the uniform step size $h >0$. Now, given a positive integer $k$ you divide $[0,S]$ into $k$ subintervals of length $2h$, i.e., $h = S/(2k)$. Define $x_j = jh$ for $j=0,1,2,\dots,2k$. Then $S_h(g)$ is given by $$ S_h(g) = \frac{h}{3} \sum_{j=0}^{k-1} \Big[ g(x_{2j}) + 4 g(x_{2j+1}) + g(x_{2j+2})\Big].$$ Many textbooks will rewrite this sum to avoid repetitions. This is a dubious practice which invites index errors when implementing the rule.