Why does the secant method require two initial points rather than one, unlike Newton's method?
ABecause the secant method applies only to polynomials, which need two roots specified to initialize
BBecause it approximates the derivative using a finite difference slope between two known function values, requiring two previous points at every step
CBecause it checks convergence by comparing consecutive iterates, and convergence checking needs two points
DBecause two starting points allow the method to bracket the root, guaranteeing convergence
Newton's method uses the tangent line at xₙ, requiring only f(xₙ) and f'(xₙ). The secant method replaces f'(xₙ) with the finite difference [f(xₙ) − f(xₙ₋₁)] / [xₙ − xₙ₋₁]. This approximation of the slope requires two points on the curve. At startup, there is no previous iterate, so you must supply two starting points x₀ and x₁. Option D is wrong: unlike bisection, the secant method does not maintain a bracket and does not guarantee convergence.
Question 2 Multiple Choice
A function f(x) has an expensive-to-compute but analytically available derivative. Which root-finding method is most appropriate, and why?
ASecant method — because it always uses fewer function evaluations regardless of derivative cost
BBisection — because its guaranteed convergence outweighs any speed advantage
CNewton's method — the cheap derivative makes quadratic convergence dominant, minimizing total steps to full precision
DSecant method — because avoiding the derivative is always safer than using it
The secant method's advantage is specifically when computing f' is expensive relative to f. If the derivative is cheap, Newton's method is superior: its quadratic convergence (order 2) doubles the digits of accuracy per step, reaching 16-digit precision in roughly 5 iterations from a good start. The secant method would need 7–8 iterations for the same accuracy. When f and f' cost equally, Newton's wins on total computational work. Only when f' significantly exceeds f in cost does the secant method's per-step saving compensate for its lower convergence order.
Question 3 True / False
The secant method converges faster than the bisection method, though it does not guarantee convergence from arbitrary starting points.
TTrue
FFalse
Answer: True
The secant method has superlinear convergence with order φ ≈ 1.618 (the golden ratio), while bisection has linear convergence (order 1) — adding only about one bit of precision per step. The secant method is much faster when it converges. However, unlike bisection, the secant method does not maintain a bracket around the root and can diverge if starting points are poorly chosen or if the function has pathological behavior near the root.
Question 4 True / False
The secant method's convergence order is exactly 2, the same as Newton's method, because the finite difference approximation becomes exact near the root.
TTrue
FFalse
Answer: False
The secant method's convergence order is φ = (1+√5)/2 ≈ 1.618, not 2. The golden ratio arises from the two-point error recurrence |eₙ₊₁| ≈ C|eₙ||eₙ₋₁|, which leads to the equation α² = α + 1 when you assume order α. Newton's quadratic convergence (order 2) comes from a one-point recurrence |eₙ₊₁| ≈ C|eₙ|², which is possible only because it uses the exact derivative. The finite difference is never exact — even near the root, it introduces an approximation error that costs convergence order.
Question 5 Short Answer
Explain why the secant method's convergence order is the golden ratio φ ≈ 1.618 rather than a simpler value like 1.5 or 2.0.
Think about your answer, then reveal below.
Model answer: The golden ratio emerges from the error recurrence. Near the root, |eₙ₊₁| ≈ C|eₙ||eₙ₋₁| for some constant C. To find the order α such that |eₙ| ~ |eₙ₋₁|^α, substitute |eₙ| ≈ |eₙ₋₁|^α into the recurrence: |eₙ₋₁|^α ≈ C|eₙ₋₁|^α · |eₙ₋₁|, which gives α = 1 + 1/α, or equivalently α² − α − 1 = 0. The positive solution is (1+√5)/2 = φ. The golden ratio is not arbitrary — it is the fixed point of this two-step recurrence.
The golden ratio appears because the secant method couples two consecutive errors (it uses the two most recent iterates), while Newton's method couples only one. The recurrence α = 1 + 1/α defines the golden ratio uniquely among positive reals, making φ the natural and exact convergence order for this two-point structure. Any method using only one previous iterate in its error recurrence will generically have order 2 (Newton's) or order 1 (simple iterations).