x² − 2 from 1.5
x₁ = 1.5 − 0.25/3 = 1.416667
xₙ₊₁ = xₙ − f/f′.
Near a simple root , Newton's method satisfies : the error is roughly squared at each step, so the number of correct digits about doubles.
This page is easier if you are comfortable with:
For , : . Starting at (error ): , error . The ratio , close to the predicted constant 0.3536. The next error is about , then .
The formula comes from the Taylor expansion of about the root (Taylor's theorem with the Lagrange remainder): . Compare linear convergence of bisection (bisection error bound) and the secant method (the secant method) of order 1.618; Newton is the case of fixed-point convergence (fixed-point convergence).
The convergence is quadratic only for simple roots and sufficiently good starting values. For a repeated root it falls back to linear convergence. The method itself is in Newton–Raphson, and in optimisation it appears as Newton's method for optimisation.
Newton's method roughly doubles the number of correct digits at each step near a simple root.
The errors fall far faster than a fixed ratio.
These examples are worked by hand, step by step, so you can compare your own working.
Find for at .
Find from for .
Find .
A few situations come up again and again. Knowing them saves time and helps you spot answers that cannot be right.
x₁ = 1.5 − 0.25/3 = 1.416667
xₙ₊₁ = xₙ − f/f′.
x₂ = 1.414216
Error ≈ 2 × 10⁻⁶.
e₁ = 0.00245, e₀ = 0.0858: e₁/e₀² ≈ 0.33
Constant ≈ f″/(2f′).
f″/(2f′) = 2/(2 × 2.828) ≈ 0.3536
At the root.
Newton’s method xₙ₊₁ = xₙ − f(xₙ)/f′(xₙ) converges quadratically near a simple root: eₙ₊₁ ≈ (f″/2f′) eₙ², so the number of correct digits roughly doubles each step.
Apply Newton to x² − 2 from x₀ = 1.5.
Find |x₁ − √2| to 5 decimal places.
Find e₁/e₀² to 2 decimal places.
If e = 0.0025, estimate the next error using 0.3536 e².
x₀ = 0 for x² − 2 ✗
f′(0) = 0 ✓
Division by zero.
error halves ✗
error squares ✓
Quadratic convergence.
f′ = x for x² − 2 ✗
f′ = 2x ✓
Differentiate carefully.
quadratic at a double root ✗
only linear ✓
Convergence slows.
Try each one on paper, then open the answer. Use the solver above to see the full working.
3.125
1.5
1.41429
2.2963
3
0.0000354
Unlimited fresh questions with instant marking. Choose a level, work it out on paper, then type your answer.
Solve: …
The error is roughly squared each step.
xₙ₊₁ = xₙ − f(xₙ)/f′(xₙ).
When f′ = 0 or the start is poor.
About f″/(2f′).
fixed-point iteration.
secant method.
More in Numerical methods and optimisation and the formula sheets.