Mathwords logoMathwords

Halley's Method — Definition, Formula & Example

Halley's method is a way to approximate a solution of f(x)=0f(x) = 0 by repeatedly improving a guess. It uses the first and second derivatives, so each step usually roughly triples the number of correct digits.

Let ff be three times continuously differentiable near a simple root rr, meaning f(r)=0f(r) = 0 and f′(r)≠0f'(r) \neq 0. Halley's method generates the sequence xn+1=xn−2f(xn)f′(xn)2[f′(xn)]2−f(xn)f′′(xn)x_{n+1} = x_n - \dfrac{2f(x_n)f'(x_n)}{2[f'(x_n)]^2 - f(x_n)f''(x_n)}. For x0x_0 sufficiently close to rr, the sequence converges to rr with order 3. That is, ∣xn+1−r∣≤C∣xn−r∣3|x_{n+1} - r| \le C|x_n - r|^3 for some constant CC. It is equivalent to applying Newton's method to g(x)=f(x)/∣f′(x)∣g(x) = f(x)/\sqrt{|f'(x)|}.

Key Formula

xn+1=xn−2 f(xn) f′(xn)2 [f′(xn)]2−f(xn) f′′(xn)x_{n+1} = x_n - \frac{2\,f(x_n)\,f'(x_n)}{2\,[f'(x_n)]^2 - f(x_n)\,f''(x_n)}
Where:
  • xnx_n = Current approximation of the root
  • xn+1x_{n+1} = Next, improved approximation
  • f(xn)f(x_n) = Function value at the current guess
  • f′(xn)f'(x_n) = First derivative at the current guess
  • f′′(xn)f''(x_n) = Second derivative at the current guess

Worked Example

Problem: Approximate 303\sqrt[3]{30} by applying one step of Halley's method to f(x)=x3−30f(x) = x^3 - 30 with starting guess x0=3x_0 = 3. Compare the result with one step of Newton's method.
Step 1: Find the derivatives, f′(x)=3x2f'(x) = 3x^2 and f′′(x)=6xf''(x) = 6x. Then evaluate all three at x0=3x_0 = 3.
f(3)=−3,f′(3)=27,f′′(3)=18f(3) = -3, \quad f'(3) = 27, \quad f''(3) = 18
Step 2: Substitute these values into Halley's formula and simplify the fraction.
x1=3−2(−3)(27)2(27)2−(−3)(18)=3+1621512=3+328=8728≈3.10714x_1 = 3 - \frac{2(-3)(27)}{2(27)^2 - (-3)(18)} = 3 + \frac{162}{1512} = 3 + \frac{3}{28} = \frac{87}{28} \approx 3.10714
Step 3: Compare with the true value 303≈3.10723\sqrt[3]{30} \approx 3.10723. One Newton step gives x1=3−−327=289≈3.11111x_1 = 3 - \frac{-3}{27} = \frac{28}{9} \approx 3.11111. Halley's single step is about 40 times more accurate.
Halley error≈9.0×10−5,Newton error≈3.9×10−3\text{Halley error} \approx 9.0 \times 10^{-5}, \qquad \text{Newton error} \approx 3.9 \times 10^{-3}
Answer: After one step of Halley's method, x1=8728≈3.10714x_1 = \frac{87}{28} \approx 3.10714. This is already within 10−410^{-4} of 303\sqrt[3]{30}.

Why It Matters

Halley's method is useful when f′′f'' costs little to compute, as with polynomials and many special functions. Some cube-root and reciprocal-square-root routines use Halley-type steps because fewer iterations mean faster code. In a numerical analysis course, it is the standard example of a third-order method. It shows how extra derivative information buys a higher order of convergence.

Common Mistakes

Mistake: Writing the denominator as 2[f′(xn)]2+f(xn)f′′(xn)2[f'(x_n)]^2 + f(x_n)f''(x_n), or dropping the factor of 2 in front of [f′(xn)]2[f'(x_n)]^2.
Correction: The denominator is 2[f′(xn)]2−f(xn)f′′(xn)2[f'(x_n)]^2 - f(x_n)f''(x_n), with a minus sign. A wrong sign or a missing 2 still gives numbers, but the cubic convergence is lost. A quick check for the missing 2 is to set f′′=0f''=0: the formula should reduce to Newton's step, xn−f(xn)/f′(xn)x_n - f(x_n)/f'(x_n). That check cannot catch a wrong sign, so compare the sign with the formula above.

Related Terms

  • Condition Number — Measures how sensitive the computed root is to errors
  • Euler's Method — Another iterative numerical scheme built from derivative information
  • Newton's Method — Simpler root-finding method that Halley's method extends
  • Second Derivative — Needed in each step of Halley's method
  • Iteration — Repeating the update step to improve the guess