A bracketing root-finding technique that repeatedly halves an interval known to contain a sign change of a continuous function, thereby converging to a root guaranteed by the intermediate value theorem.

Principle

If f is continuous on [a,b] and f(a) and f(b) have opposite signs, then there exists at least one root in (a,b); bisect the interval, select the subinterval that preserves the sign change, and iterate to reduce interval width geometrically.

Demonstration

Given f continuous with f(0)<0 and f(1)>0, compute midpoint m=0.5, evaluate f(m); choose [0,m] or [m,1] according to sign change and repeat; interval length halves each step, guaranteeing convergence to a root with linear (geometric) rate.

Misapplication

Applying bisection when the function is discontinuous on the interval, when there is no sign change, or attempting to use it for roots of multiplicity without sign change will fail or be inapplicable.

Consequence

A robust, simple algorithm with guaranteed global convergence provided the initial bracket contains a sign change; it converges linearly and is often used to supply starting intervals for faster local methods.

Reversal

Reverse trade-off: sacrifice the bisection's global guarantee for faster local methods (Newton, secant) that can converge superlinearly but may diverge without good initialization.

Boundary

Applicable to continuous scalar functions on an interval with known sign change; does not exploit derivative information, is limited to one-dimensional real problems, and is inefficient relative to superlinear methods near the root.

Semantic Tension

Often compared with regula falsi (false position): both are bracketing methods, but regula falsi uses secant-based interpolation to choose subintervals and can be faster yet may stall; bisection is slower but simpler and monotone.

Synthesis

The bisection method repeatedly halves a bracketing interval that contains a sign change, providing a guaranteed and simple route to a real root under continuity; it trades speed for reliability and is a standard fallback or initialization step for faster algorithms.