Numerical Analysis. Bisection method.

What will happen if the bisection method is used with the function f(x) = tan(x) and

a) [3,4]

b) [1,3]

Attempt:

Check the signs of the function:

f(x) = tan(x)

a) f(3)*f(4) = -0.165 <0 => the root is between 3 and

b) f(1)*f(3)= -0.222<0 => the root is between 1 and 3 ,

but this is obviously wrong.

Please, help.

Re: Numerical Analysis. Bisection method.

Hey vercammen.

Are you trying to find the root to f(x) = tan(x) = 0 with the bisection method?

Re: Numerical Analysis. Bisection method.

Remember, the bisection root finding method works for __continuous__ functions on the interval. So the bisection method applied to tan(x) on [3,4] does fine. But tan(x) has an infinite discontinuity on [1,3] at x=pi/2. So there's no reason to suppose bisection will work; in fact tan(x) is never 0 on [1,3].

Re: Numerical Analysis. Bisection method.

The question asks "What will happen if the bisection method is used with the function..." . I guess I need to answer that. I'm confused...

Re: Numerical Analysis. Bisection method.

If you apply the bisection method and get the sequence {xn}, then xn approaches pi/2 as n approaches infinity. I really don't want to give an analytic proof of this, but if you try it by hand, you'll see what is happening. Alternatively, write a little program to compute say the first 1000 terms to see what you get:

#include <stdio.h>

#include <math.h>

double bisect(double (*f)(double),double x0, double x1, int n);

int main(void) {

double x0=1,x1=3,result;

result=bisect(tan,x0,x1,1000);

printf("After 1000 iterations, xn = %g with tan(xn) = %g\n",result,tan(result));

return(0);

}

/* bare bones bisection method; not really a workable version since no

tolerance is supplied, just do the method n times.

*/

double bisect(double (*f)(double x), double x0, double x1, int n) {

double m;

int i;

for (i=1;i<=n;i++) {

m=(x0+x1)/2;

if ((*f)(m)*(*f)(x0)<0) {

x1=m;

}

else {

x0=m;

}

}

return(m);

}

The above program produces result 1.5708 (approximately pi/2) with

tan(result)=1.63e+016, as expected a large value in magnitude.

Re: Numerical Analysis. Bisection method.

Quote:

Originally Posted by

**vercammen** The question asks "What will happen if the bisection method is used with the function..." . I guess I need to answer that. I'm confused...

The bisection method finds points at which the function changes sign, which yours does do at the singular points, so if you start with an interval [1,3] it will converge to the singularity at $\displaystyle \pi/2$.

.