Solve equations using the Newton-Raphson method and other recurrence relations of the form xₙ₊₁ = g(xₙ); understand how such methods can fail

Overview

Numerical methods are what mathematicians use when an equation refuses to give a neat algebraic answer. In Pearson Edexcel A-Level Mathematics, you must turn an equation into a root problem, use Newton-Raphson or a recurrence relation to generate increasingly accurate approximations, and explain when the process is unreliable. This is a high-value topic because it joins algebra, differentiation, graphs and calculator fluency in one place.

The question wording often tells you the method to use. Expect instructions such as “use the Newton-Raphson method”, “show that the iteration converges”, “use the recurrence relation” or “explain why this method may fail”. Marks are awarded for the mathematical set-up as well as the decimal answer. A calculator can produce the numbers, but candidates earn credit by defining the right function, showing enough substitution, and interpreting the graph correctly.

Key Concepts

1. Roots, fixed points and a safe starting point

$A root is a value, often written (\alpha), for which (f(\alpha)=0). If the original equation is (x^3+x=4), do not start Newton-Raphson with that form. Move everything to one side first:$

[
f(x)=x^3+x-4.
]

Now the root is exactly the value where the graph of (y=f(x)) crosses the x-axis. This first line is frequently worth a method mark. A sign change in a short interval can help you choose a sensible initial estimate. For example, (f(1)=-2) and (f(2)=6), so a root lies between 1 and 2 because the polynomial is continuous.

A recurrence relation has a different-looking target. If (x_{n+1}=g(x_n)), the limiting value must satisfy (\alpha=g(\alpha$)). Such a value is called a fixed point. Rearranging an equation into (x=g(x)) creates an iteration, but not every rearrangement is useful. The algebra may be correct while the sequence fails to settle.$

Exam habit: write down the supplied starting value as (x_0), retain full calculator precision, and round only the final answer. Repeated rounding can create a misleading final digit or even make a slowly converging iteration appear stationary too early.

2. Newton-Raphson: tangent, intercept, repeat

Newton-Raphson turns local gradient information into a better estimate of a root. For a differentiable function (f), the iteration is

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

Here (x_n) is the current estimate, (f(x_n)) is its vertical distance from the x-axis, and (f'(x_n)) is the gradient of the curve at that point. Geometrically, draw the tangent to (y=f(x)) at ((x_n,f(x_n))). Where that tangent meets the x-axis is the next estimate, (x_{n+1}). The method can be extremely fast once the estimate is close to a well-behaved root.

Reliable method for a calculation question:

1. Rearrange to (f(x)=0).
2. Differentiate to find (f'(x)).
3. State the Newton-Raphson formula before using it.
4. Substitute the supplied (x_0), then continue until the required accuracy is secure.
5. State the root clearly, for example (x=1.379) correct to 3 decimal places.

The diagram shows why a tangent is useful. It is a straight-line local approximation to the curve. Its x-intercept is usually closer to the true intercept than the starting point. It is not magic: if the tangent is nearly horizontal, its x-intercept can be very far away.

3. General recurrence relations and the cobweb picture

For a recurrence of the form (x_{n+1}=g(x_n)), repeatedly put the previous answer into the same expression. For instance, (x_{n+1}=e^{-x_n}) with (x_0=0.5) gives (0.60653, 0.54524, 0.57970,\ldots). The values alternate around a fixed point near (0.567), with the gap becoming smaller.

A cobweb or staircase diagram uses the two curves (y=g(x)) and (y=x). Start at (x_n) on the x-axis, move vertically to (y=g(x)), then horizontally to (y=x). Repeating these vertical-horizontal moves traces the sequence. If the steps head into the intersection, the iteration converges. If they move away, cycle, or wander, it fails from that start or with that rearrangement.

A helpful local test is the size of the gradient of (g) near the fixed point. If (|g'(\alpha)|<1), nearby iterations normally converge; a negative gradient produces an alternating approach. If (|g'(\alpha)|>1), nearby errors tend to grow. Treat this as an explanation tool rather than a replacement for the recurrence the question has specified.

4. How and why methods fail

Pearson explicitly expects geometric understanding of failure near small gradients. In Newton-Raphson, the denominator is (f'(x_n)). If (f'(x_n)=0), the tangent is horizontal and never meets the x-axis, so the formula is undefined. If (f'(x_n)) is very small, dividing by it creates a large correction. The next estimate can jump away from the intended root, possibly towards another root or a region where the function is unsuitable.

For example, with (f(x)=x^3-3x+1) and (x_0=1), (f'(x)=3x^2-3), so (f'(1)=0). The correct explanation is not merely “the calculator does not work”. Write: “At (x=1), the gradient is zero, so the tangent is parallel to the x-axis and has no x-intercept. Therefore the Newton-Raphson iteration is undefined.” That links algebra to geometry and earns the explanatory mark.

General iterations can fail because an unsuitable rearrangement has a steep gradient near its fixed point, because the starting value is too far away, or because the values enter a cycle. Do not claim that every rearrangement of the same equation behaves identically. Choosing (g(x)) is part of the mathematics.

5. Exam-ready presentation and calculator control

Approximately one minute per mark is a sensible planning rule. A 4-mark numerical-methods task should normally have a visible function, derivative or recurrence, a substitution, numerical evidence, and a correctly rounded conclusion. If the instruction says “show that”, your lines must make the stated result unavoidable. If it says “explain”, a bare decimal receives no credit.

Use the calculator’s Ans key or stored variables to repeat an iteration, but copy at least the first one or two substitutions into the answer. State enough digits to justify the requested accuracy. For 3 decimal places, show estimates that agree to 3 decimal places, preferably one further iteration after the agreement is first seen. Check the calculator is in radians for any trigonometric function, and never use a rounded previous answer unless the question explicitly tells you to.

6. Choosing between Newton-Raphson and simple iteration

The method named in the question is the method you use. When a question asks for Newton-Raphson, do not replace it with a simple iteration just because a calculator sequence is easier to type. The examiner is testing the tangent formula and your derivative. Similarly, when a recurrence is supplied, use that recurrence even if you can see a different rearrangement that would converge more quickly. You may comment on its behaviour only if the question asks you to explain, compare or evaluate.

When the method is not prescribed, begin with the structure of the equation. Newton-Raphson is usually effective when you can differentiate accurately and have a sensible starting value away from a stationary point. A general recurrence is convenient if a natural rearrangement is supplied or if a fixed-point graph is the focus. Change of sign is useful before either method because it brackets a root and gives a defensible region for a starting estimate. It is not, by itself, a final decimal solution unless the question asks only for an interval.

Candidates sometimes write a recurrence that looks almost right but uses the new value on both sides, for example (x_{n+1}=g(x_{n+1})). Check the subscripts: the old value goes into (g), and the new value comes out. For Newton-Raphson, put brackets around the full numerator and denominator before pressing =. This prevents a calculator precedence error such as subtracting only part of the fraction.

7. Graph and data skills that secure explanation marks

A sketch need not be perfectly scaled, but it must communicate the mathematical idea. For Newton-Raphson, show the curve, the x-axis, a point on the curve and a tangent that meets the axis at the next estimate. For a cobweb diagram, plot both (y=g(x)) and (y=x), then use vertical and horizontal segments. A diagonal line alone is not a cobweb. Label the intersection as the fixed point and make the direction of iteration unambiguous.

Read an iteration table as evidence, not decoration. Values such as (1.442, 1.368, 1.381, 1.378) suggest convergence because the estimates tighten around one value. Values growing in magnitude, jumping between distant values, or repeating two distinct values suggest failure. If the values alternate but the interval between them shrinks, that is still convergence. State exactly what your figures show rather than using “converges” as a guess.

Mathematical Relationships

The differentiation rules required to form (f'(x)) must be known and applied accurately. A formula sheet cannot repair an incorrectly differentiated function.

Practical Applications

Numerical methods are used whenever an exact rearrangement is impractical: engineers locate where a design curve meets a safety limit, programmers solve equations inside simulations, and scientists calibrate models against observations. In the examination, the context may be a volume, a time, or an optimisation boundary, but the mathematics is unchanged: define the equation, approximate the required root, then interpret the answer in context. There is no required practical or tier split for this A-Level topic. Unit conversion is only needed when the surrounding context supplies units.

Listen: 10-minute tutor podcast