P3 6.2 - Iterative solutions

P3 6.2 - Iterative solutions

Use a supplied rearrangement of an equation to build successive approximations to a root. You will generate iterates accurately, recognise when a sequence fails to settle, and check a claimed decimal answer using the original equation.

Rearranging to a supplied iterative form

Can two rearrangements of the same equation behave differently on a calculator? For x2−x−1=0x^2-x-1=0, both x=x+1x=\sqrt{x+1} for a positive root and x=x2−1x=x^2-1 are valid rearrangements. Repeatedly using their right-hand sides can produce very different sequences.

Iteration turns a rearrangement x=g(x)x=g(x) into an update rule: put the current approximation into gg to obtain the next approximation. First show that a fixed value of the rule satisfies the desired equation, with any required domain restrictions.

Iteration and fixed point

An iteration repeats a calculation using its latest output. In

xn+1=g(xn),x_{n+1}=g(x_n),

xnx_n is the current approximation and xn+1x_{n+1} is the next. A fixed point α\alpha satisfies α=g(α)\alpha=g(\alpha): the update leaves it unchanged.

A square-root rearrangement selects the nonnegative branch. State that restriction rather than silently discarding a negative solution.

Worked example 1: deriving a supplied square-root form

Show that a positive root of x2−x−1=0x^2-x-1=0 satisfies x=x+1x=\sqrt{x+1}, and state the corresponding iteration.

Plan. Isolate x2x^2, then take the positive square root because the required root is positive.

x2−x−1=0⟹x2=x+1⟹x=x+1.x^2-x-1=0 \quad\Longrightarrow\quad x^2=x+1 \quad\Longrightarrow\quad x=\sqrt{x+1}.

Answer. Use xn+1=xn+1x_{n+1}=\sqrt{x_n+1} from a supplied suitable starting value. The update is real when xn≥−1x_n\ge-1.

Check. A positive fixed point satisfies x2=x+1x^2=x+1, which rearranges back to x2−x−1=0x^2-x-1=0. Taking the positive square root is justified for this root.

Change only the coefficient of the linear term from −1-1 to −2-2. The same algebra gives a different update.

Worked example 2: a changed linear coefficient

Show that a positive root of x2−2x−1=0x^2-2x-1=0 satisfies x=2x+1x=\sqrt{2x+1}, and state the iteration.

Plan. Isolate the square and use the positive branch again.

x2−2x−1=0⟹x2=2x+1⟹x=2x+1.x^2-2x-1=0 \quad\Longrightarrow\quad x^2=2x+1 \quad\Longrightarrow\quad x=\sqrt{2x+1}.

Answer. xn+1=2xn+1x_{n+1}=\sqrt{2x_n+1}, real when xn≥−12x_n\ge-\frac12.

Check. Squaring a positive fixed point gives x2=2x+1x^2=2x+1, exactly the original equation after rearranging. The branch condition rules out introducing a negative fixed value.

A valid rearrangement identifies possible fixed points. It does not guarantee that the sequence from a particular starting value will reach one.

Watch out

Write xn+1=2xn+1x_{n+1}=\sqrt{2x_n+1}, not xn+1=2n+1x_{n+1}=\sqrt{2n+1}. The input is the current approximation's value, not its position in the sequence.

Change the linear coefficient once more and show the algebra.

Generating iterates

Once the rule and starting value are supplied, generate each new term from the latest one. Retain your calculator's full precision internally, rounding only the values you report.

For xn+1=xn+1x_{n+1}=\sqrt{x_n+1}, typing the expression with the calculator's previous-answer value lets you repeat the update. Check whether the initial value is labelled x0x_0 or x1x_1: this changes the labels on every subsequent term.

Initial approximation

An initial approximation, such as x0=1x_0=1, starts the iteration. One update gives x1x_1, two updates give x2x_2, and so on. The rule and starting index must both be retained.

Show at least one substitution to make the iterative method visible.

Worked example 3: starting below the root

Use xn+1=xn+1x_{n+1}=\sqrt{x_n+1}, with x0=1x_0=1, to find x1,x2,x3,x4x_1,x_2,x_3,x_4 to six decimal places.

Plan. Apply the square-root update four times, carrying unrounded values into each new calculation.

x1=1+1=1.414213562…x2=1.414213562…+1=1.553773974…x3=1.598053182…x4=1.611847754…\begin{aligned} x_1&=\sqrt{1+1}=1.414213562\ldots\\ x_2&=\sqrt{1.414213562\ldots+1}=1.553773974\ldots\\ x_3&=1.598053182\ldots\\ x_4&=1.611847754\ldots \end{aligned}

Answer. 1.414214, 1.553774, 1.598053, 1.6118481.414214,\ 1.553774,\ 1.598053,\ 1.611848, respectively.

Check. Reversing the final update gives x42−1=x3x_4^2-1=x_3, to the working precision. Each listed term is larger than its predecessor in this calculation.

Change only the initial value to 1.81.8. The same update now approaches from above.

Worked example 4: starting above the root

Use xn+1=xn+1x_{n+1}=\sqrt{x_n+1}, with x0=1.8x_0=1.8, to find x1,x2,x3,x4x_1,x_2,x_3,x_4 to six decimal places.

Plan. Keep the update unchanged and use the new starting approximation.

x1=1.8+1=1.673320053…x2=1.673320053…+1=1.635029068…x3=1.623277262…x4=1.619653439…\begin{aligned} x_1&=\sqrt{1.8+1}=1.673320053\ldots\\ x_2&=\sqrt{1.673320053\ldots+1}=1.635029068\ldots\\ x_3&=1.623277262\ldots\\ x_4&=1.619653439\ldots \end{aligned}

Answer. 1.673320, 1.635029, 1.623277, 1.6196531.673320,\ 1.635029,\ 1.623277,\ 1.619653, respectively.

Check. Again x42−1=x3x_4^2-1=x_3. The values decrease here, while Example 3's increase; both are approaching the same region near 1.6181.618.

Different starting values can give different early terms even when the eventual fixed point is the same.

Exam tip

Use the iteration specified in the question. A calculator's equation-solver answer does not show the requested successive approximations. Retain full precision in stored values and display the requested rounding.

Change only the starting value again.

Recognising convergence or failure

A useful iteration produces values that settle towards a fixed point. A correct algebraic rearrangement can also produce values that move away, cycle, or leave the real domain of the update.

For the same equation x2−x−1=0x^2-x-1=0, compare two updates starting at 1.71.7.

UpdateInitial value and next four values, rounded for display
xn+1=xn+1x_{n+1}=\sqrt{x_n+1}1.7, 1.643168, 1.625782, 1.620427, 1.6187731.7,\ 1.643168,\ 1.625782,\ 1.620427,\ 1.618773
xn+1=xn2−1x_{n+1}=x_n^2-11.7, 1.89, 2.5721, 5.615698, 30.5360691.7,\ 1.89,\ 2.5721,\ 5.615698,\ 30.536069

The first sequence moves closer to the positive fixed point. The second moves rapidly away from it. That second rule still has the correct fixed points; its failure is in the behaviour of its repeated updates from this start.

Continue this lesson.

You’re previewing NovaLearn without an account. Create a free account to read the rest.

Create free account