1.04c, 1.04d - Rational binomial expansions and approximation

1.04c, 1.04d - Rational binomial expansions and approximation

A short polynomial can approximate a square root or reciprocal surprisingly well near a suitable input. You will build rational binomial expansions, factor the constant correctly, state where the expansion is valid and distinguish that condition from numerical accuracy.

Expanding a rational power

Can four easy terms approximate 4.08\sqrt{4.08} to better than one millionth? Since 4.08=4(1+0.02)4.08=4(1+0.02), the real question is how 1+u\sqrt{1+u} behaves when uu is small. We will build the polynomial first and return to the number at the end.

You already know

For positive integer powers, the binomial coefficients terminate: (1+u)2=1+2u+u2(1+u)^2=1+2u+u^2. Fractional and negative powers obey the index laws on suitable real domains; for example, (1+u)−1=1/(1+u)(1+u)^{-1}=1/(1+u).

The descending-factor pattern still works for rational powers, but it need not stop. A half-power gives coefficient factors 1/2,−1/2,−3/2,…1/2,-1/2,-3/2,\ldots; none becomes zero.

Rational binomial expansion

For rational nn and ∣u∣<1|u|<1,

(1+u)n=1+nu+n(n−1)2!u2+n(n−1)(n−2)3!u3+⋯ .(1+u)^n=1+nu+\frac{n(n-1)}{2!}u^2+\frac{n(n-1)(n-2)}{3!}u^3+\cdots.

For positive integer nn, the expansion terminates and is exact for every real input. For n=0n=0, the value is 11 when the base is non-zero; u=−1u=-1 would give the undefined expression 000^0. For other rational powers the displayed finite part is a truncation of an infinite expansion. The sheet gives this descending-factor pattern.

In the formula sheet.

You need to calculate specified terms, rather than memorise a general term or prove convergence. Keep fractions exact until the numerical approximation stage.

Worked example 1: a square-root expansion

Question. Expand (1+x)1/2(1+x)^{1/2} through the term in x3x^3.

Plan. Substitute n=1/2n=1/2 into the first four coefficient factors.

(1+x)1/2=1+x2+(1/2)(−1/2)2x2+(1/2)(−1/2)(−3/2)6x3+⋯descending factors=1+x2−x28+x316+⋯simplify fractions\begin{aligned} (1+x)^{1/2}&=1+\frac{x}{2}+\frac{(1/2)(-1/2)}{2}x^2+\frac{(1/2)(-1/2)(-3/2)}{6}x^3+\cdots &&\text{descending factors}\\ &=1+\frac{x}{2}-\frac{x^2}{8}+\frac{x^3}{16}+\cdots &&\text{simplify fractions} \end{aligned}

Why is the cubic coefficient positive? Two of its three coefficient factors are negative.

Answer. 1+x/2−x2/8+x3/16+⋯1+x/2-x^2/8+x^3/16+\cdots, valid for ∣x∣<1|x|<1.

Check. Squaring this polynomial gives constant and first three power terms 1+x+0x2+0x31+x+0x^2+0x^3, as a square-root expansion should.

Change only the exponent to −1/2-1/2. Do not assume that every rational expansion has the same signs.

Worked example 2: the reciprocal square root

Question. Expand (1+x)−1/2(1+x)^{-1/2} through the term in x3x^3.

Plan. Use the same input, changing the descending coefficient factors.

(1+x)−1/2=1−x2+(−1/2)(−3/2)2x2+(−1/2)(−3/2)(−5/2)6x3+⋯use n=−1/2=1−x2+3x28−5x316+⋯simplify\begin{aligned} (1+x)^{-1/2}&=1-\frac{x}{2}+\frac{(-1/2)(-3/2)}2x^2+\frac{(-1/2)(-3/2)(-5/2)}6x^3+\cdots &&\text{use }n=-1/2\\ &=1-\frac{x}{2}+\frac{3x^2}{8}-\frac{5x^3}{16}+\cdots &&\text{simplify} \end{aligned}

Why retain every sign? Here the quadratic has two negative factors and the cubic has three.

Answer. 1−x/2+3x2/8−5x3/16+⋯1-x/2+3x^2/8-5x^3/16+\cdots, for ∣x∣<1|x|<1.

Check. Multiplying the two worked polynomials gives 1+0x+0x2+0x31+0x+0x^2+0x^3, matching (1+x)1/2(1+x)−1/2=1(1+x)^{1/2}(1+x)^{-1/2}=1 to this degree.

A scaled input must also be powered: (1+2x)−1=1−2x+4x2−8x3+⋯(1+2x)^{-1}=1-2x+4x^2-8x^3+\cdots, for ∣x∣<1/2|x|<1/2. It is the whole 2x2x that is squared and cubed.

Watch out

Wrong: (1+x)1/2=1+x/2−x2/8+x3/16(1+x)^{1/2}=1+x/2-x^2/8+x^3/16 for general xx. Right: retain the ellipsis for the infinite expansion, or use ≈\approx for the finite polynomial. Equality to a truncation would claim that the omitted terms are zero.

Keep the input and degree, changing only the exponent again.

Factor out the constant before expanding

The formula begins with 11 inside the bracket. If the constant is 44, factor it out before expanding, then apply the exponent to that constant too.

Normalising the bracket

For a>0a>0, a real input in the required domain and rational nn,

(a+bx)n=an(1+bxa)n.(a+bx)^n=a^n\left(1+\frac{bx}{a}\right)^n.

This is an index-law rearrangement used with the supplied expansion; it is not separately printed as a formula.

Not in the formula sheet: learn it.

The outside factor scales every term, not just the constant term.

Worked example 3: factor out a positive constant

Question. Expand 4+2x\sqrt{4+2x} through x3x^3.

Plan. Write the expression as 2(1+x/2)1/22(1+x/2)^{1/2}, then substitute into the half-power expansion.

4+2x=2(1+x2)1/2factor 4 and take its root=2[1+x4−x232+x3128+⋯ ]use u=x/2=2+x2−x216+x364+⋯multiply every term by 2\begin{aligned} \sqrt{4+2x}&=2\left(1+\frac{x}{2}\right)^{1/2} &&\text{factor }4\text{ and take its root}\\ &=2\left[1+\frac{x}{4}-\frac{x^2}{32}+\frac{x^3}{128}+\cdots\right] &&\text{use }u=x/2\\ &=2+\frac{x}{2}-\frac{x^2}{16}+\frac{x^3}{64}+\cdots &&\text{multiply every term by }2 \end{aligned}

Why is the outside factor 22, not 44? It is 41/24^{1/2}.

Answer. 2+x/2−x2/16+x3/64+⋯2+x/2-x^2/16+x^3/64+\cdots, for ∣x∣<2|x|<2.

Check. At x=0x=0, the constant is 2=42=\sqrt4. Squaring the polynomial gives 4+2x4+2x through cubic degree.

Change only the sign of the variable term. The outside scale remains 22.

Worked example 4: a negative inner input

Question. Expand 4−2x\sqrt{4-2x} through x3x^3.

Plan. Use u=−x/2u=-x/2, keeping its sign inside each power.

4−2x=2(1−x2)1/2normalise the bracket=2[1−x4−x232−x3128+⋯ ]substitute u=−x/2=2−x2−x216−x364+⋯restore the outside scale\begin{aligned} \sqrt{4-2x}&=2\left(1-\frac{x}{2}\right)^{1/2} &&\text{normalise the bracket}\\ &=2\left[1-\frac{x}{4}-\frac{x^2}{32}-\frac{x^3}{128}+\cdots\right] &&\text{substitute }u=-x/2\\ &=2-\frac{x}{2}-\frac{x^2}{16}-\frac{x^3}{64}+\cdots &&\text{restore the outside scale} \end{aligned}

Why is the cubic negative now? Its positive half-power coefficient multiplies a negative cube.

Answer. 2−x/2−x2/16−x3/64+⋯2-x/2-x^2/16-x^3/64+\cdots, for ∣x∣<2|x|<2.

Check. Replacing xx by −x-x in example 3 gives these same signs.

Negative exponents need the outside power too: (3−x)−2=1/9+2x/27+x2/27+4x3/243+⋯(3-x)^{-2}=1/9+2x/27+x^2/27+4x^3/243+\cdots, for ∣x∣<3|x|<3. The outside factor is 3−2=1/93^{-2}=1/9.

For a negative constant, check the real domain first. Near zero, −8+2x3=−2(1−x/4)1/3\sqrt[3]{-8+2x}=-2(1-x/4)^{1/3} uses the real cube root. The square root of −8+2x-8+2x is not real there. Factoring cannot repair an invalid real input.

Watch out

Wrong: (3−x)−2=3(1−x/3)−2(3-x)^{-2}=3(1-x/3)^{-2}. Right: (3−x)−2=3−2(1−x/3)−2(3-x)^{-2}=3^{-2}(1-x/3)^{-2}. The exponent applies to the factor taken out.

Change the positive constant and its associated scale in another square-root bracket.

Validity and combining truncated expansions

Validity applies to the entire inner input: ∣bx/a∣<1|bx/a|<1. For example, (9−6x)−1/2=13(1−2x/3)−1/2(9-6x)^{-1/2}=\tfrac13(1-2x/3)^{-1/2} has the condition −3/2<x<3/2-3/2<x<3/2, not merely x<3/2x<3/2.

Continue this lesson.

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

Create free account