In this paper, we propose algorithms using a branch and bound method for finding all real solutions to an equation f(x)=0 of one variable x on an interval [a,b]. We denote by P([c,d]) the subproblem in which we find all approximate solutions to f(x)=0 on [c,d]⊂ [a,b]. The algorithms repeat the following procedure for each subproblem P([c,d]) (the first subproblem is P([a,b])) until we terminate all the subproblems: If we solve the subproblem P([c,d]) then we terminate it, otherwise we split it into two new subproblems P([c,e]) and P([e,d]) for e=(c+d)/2. Here we can solve the subproblem P([c,d]) when there is no solution of f(x)=0 on [c,d] or when there exists a solution of f(x)=0 on [c,d] and the width d-c is sufficiently small. We assume that there are two functions g(x) and h(x) such that f(x) = g(x)-h(x) and one of the following conditions holds: (1) The functions g(x) and h(x) are continuous and monotone increasing. (2) The first derivatives g'(x) and h'(x) are continuous and monotone increasing. (3) The second derivatives g"(x) and h"(x) are continuous and monotone increasing . Then we present new sufficient conditions that the equation f(x)=0 has no solution on [c,d]. Those conditions are used in the algorithms for solving subproblems. We also show that if the function f(x) is a sum of elementary functions then there exist the functions g(x) and h(x) which satisfy the above as sumption. Furthermore we present sufficient conditions that the equation f(x)=0 has exactly one solution on [c,d]. We propose two algorithms which solve the subproblem P([c,d]) efficiently when the conditions hold.
View full abstract