Why Bisection Converges When Newton's Method Fails

In numerical mathematics, finding the roots of a function often comes down to a choice between absolute reliability and computational speed. This article provides an overview of why the bisection method is mathematically guaranteed to converge to a root, whereas the much faster Newton's method can easily fail. By examining the core mechanics of both algorithms—specifically the interval-trapping nature of bisection versus the derivative-based extrapolations of Newton's method—we will break down the exact mathematical conditions that cause Newton's method to diverge and why bisection remains fail-safe.

Why the Bisection Method Guarantees Convergence

The bisection method is a "bracketing" algorithm, meaning it traps the root within a defined boundary. To start, it requires only two things: a continuous function and two initial guesses, \(a\) and \(b\), where the function outputs have opposite signs (one positive, one negative).

Because the function is continuous and changes signs between \(a\) and \(b\), the Intermediate Value Theorem guarantees that the function must cross the x-axis at least once within that interval. That crossing point is the root.

The bisection method works by calculating the midpoint of the interval, checking the sign of the function at that midpoint, and replacing whichever endpoint shares that same sign. This cuts the search interval in half while ensuring the root remains trapped between the new endpoints. Because the interval shrinks by 50% with every single step, and the root cannot escape the interval, the method is mathematically forced to converge on the root. It does not care about the shape, slope, or complexity of the function inside the bracket.

How Newton's Method Sacrifices Safety for Speed

Newton's method (also known as the Newton-Raphson method) is an "open" method. Instead of trapping the root between two points, it starts with a single guess and uses calculus to project where the root might be.

At the initial guess, the algorithm calculates the derivative (slope) of the function and draws a tangent line straight to the x-axis. Where that tangent line hits the x-axis becomes the next guess. The formula is:

\[x_{n+1} = x_n - \frac{f(x_n)}{f'(x_n)}\]

While riding the tangent line usually finds the root much faster than halving an interval, relying on local slopes leaves the algorithm blind to the broader behavior of the function.

Specific Reasons Why Newton's Method Fails

Because Newton's method relies on derivatives and a single starting point rather than a closed bracket, it can fail in several ways:

  • Division by Zero (Flat Tangents): If the algorithm lands on a point where the curve is completely flat (a local maximum, minimum, or inflection point), the derivative \(f'(x_n)\) becomes zero. You cannot divide by zero, so the formula breaks down entirely. Geometrically, a horizontal tangent line runs parallel to the x-axis and will never intersect it.
  • Near-Zero Derivatives: If the slope is not quite zero, but very close to it, the tangent line will shoot off at an extreme angle. This flings the next guess miles away from the actual root, often into a region where the function behaves unpredictably.
  • Infinite Oscillations: For certain shapes of functions, Newton's method can get caught in a perpetual loop. The tangent line at point \(A\) might send the next guess to point \(B\), and the tangent line at point \(B\) sends the guess right back to point \(A\). The algorithm will bounce back and forth forever without ever getting closer to the root.
  • Poor Initial Guesses: Newton's method only guarantees convergence if your first guess is already sufficiently close to the actual root. If you start too far away, the local slope at your starting point may have no relevance to where the root actually lies, causing the guesses to diverge toward infinity.
  • Non-Differentiable Points: The bisection method only requires a function to be continuous. Newton's method requires the function to be differentiable. If the function has a sharp corner, a cusp, or a vertical tangent line, the derivative does not exist at that point, and Newton's method fails.

The Fundamental Difference

Ultimately, the bisection method succeeds because it relies on global containment. Once a root is bracketed, the algorithm systematically destroys the space where the root isn't, leaving nowhere for the root to hide. Newton's method relies on local extrapolation, assuming that the slope at a single point accurately points toward the root. When a function is curvy, flat, or complex, that assumption fails, and the method diverges.