Approximation · Topic 19 of 23
Linear Approximation and Differentials
Near a point where is differentiable, replace the curve by its tangent: . The differential estimates the change in , the sign of says whether the estimate is too big or too small, and bounds the error.
Key ideas
5 things to remember- 1
The tangent line is the estimate
is exactly the tangent at , and for near . Two hypotheses carry everything: exists, and is close to .
- 2
Choose the nearest exact base point
Pick the closest at which and are computable exactly — a perfect square or cube, , , . For take , not .
- 3
Differentials measure the tangent rise
is an independent variable you may set to anything; is then defined. With , is the rise along the tangent and the rise along the curve.
- 4
Concavity decides over or under
If between and the tangent lies below the curve, so under-estimates; if it over-estimates. Which side of the point sits on is irrelevant.
- 5
Errors propagate by the exponent
For , : a error in a radius gives about in an area and in a volume, but only in a square root.
Formulas
What to have memorisedLinearisation at
The derivative is frozen at , which is what makes linear.
Tangent-line estimate
Needs to exist and to be small — nothing is promised far from .
Differential
rides the tangent line, rides the curve.
Standard approximations near
Also and ; the trig ones need radians.
Relative error propagation
Percentage error in is about times the percentage error in .
Direction from concavity
Concave up: under-estimates. Concave down: over-estimates.
Error bound (first-order Taylor)
must bound on the whole interval between and .
Estimate a value by linear approximation
The steps, in order- 1
Name the function and pick the nearest base point where and are exact.
- 2
Compute and , then write .
- 3
Set — converting degrees to radians first — and evaluate .
- 4
Check the sign of between and : concave up means under-estimate, concave down means over-estimate.
- 5
If a bound is asked, find with on that interval and quote .
- 6
For measurement error instead, differentiate the model, then divide by it: .
Watch out
The mistakes that cost marks✗ Wrong
✓ Right
— the slope is the number .
Why: With left in it, is not a linear function at all.
✗ Wrong
Estimating with
✓ Right
Convert first: and .
Why: The derivative formulas for sine and cosine assume radians.
✗ Wrong
", so the estimate is too small."
✓ Right
The side of is irrelevant — only the sign of decides the direction.
✗ Wrong
✓ Right
, and is an approximation, never an equality.
Why: Without the dx the equation is dimensionally meaningless.
✗ Wrong
Taking in the error bound
✓ Right
must dominate on the entire interval from to .
Why: Any valid overestimate of M is fine; a value only at a is not.
✗ Wrong
Reporting a relative error of as the percentage error
✓ Right
Percentage error is the relative error: .
Quick check
Commit to an answer before you reveal one- Q1easy
Find the linearisation of at and use it to approximate . Is the estimate too large or too small? Justify the direction.
Hint
is a perfect square, which is why it was chosen; the sign of settles the direction.
Show answer
Answer
, so — an over-estimate.
Steps
and , so and
on , so is concave down and the tangent lies above the graph: the estimate is too large.
Check by squaring: . (True value .)
- Q2easy
Let .
(a) Find the differential . (b) Evaluate and the actual change as goes from to . (c) What happens to the gap if the step is instead?
Hint
uses the tangent; must be computed from itself.
Show answer
Answer
(a) . (b) and . (c) The gap falls from to .
Steps
(a) , so .
(b) At with : . Along the curve, and , so .
(c) For a general step , while , so
At that is ; at it is . A tenth of the step gives about a hundredth of the gap — the error is quadratic.
- Q3medium
Use a linear approximation to estimate . Give an exact expression and a decimal to six places, state whether it over- or under-estimates, and bound the error.
Hint
Base at and work in radians, so .
Show answer
Answer
: an over-estimate, with error at most .
Steps
Take , , . Then and , so
across , so is concave down and this over-estimates.
Since , take : . (True value .)
- Q4medium
The radius of a sphere is measured as cm with a possible error of at most cm. Use differentials to estimate the maximum error in the calculated volume, and give the relative and percentage errors.
Hint
Divide by symbolically before substituting any numbers.
Show answer
Answer
cm; relative error ; percentage error about .
Steps
, so . With and :
For the relative error divide first: , hence
i.e. about — three times the error in , as the exponent predicts.
- Q5medium
The area of a disk is computed from its radius, nominally cm, and must be correct to within a percentage error of . Using differentials, how accurately must the radius be measured — as a percentage, and in centimetres?
Hint
Run the propagation rule backwards: , then solve for .
Show answer
Answer
, i.e. ; with that is cm.
Steps
gives , so — the power rule with .
Impose the tolerance :
So the radius needs accuracy — half the tolerance on the area, because the exponent is . In centimetres, cm.
- Q6hard
is twice differentiable on with , and for every .
(a) Use the linearisation at to estimate and . (b) Is each estimate too large or too small, and why is the answer the same for both? (c) If also on , give an interval that certainly contains .
Hint
The over/under question is settled by the sign of alone — not by which side of the point lies on.
Show answer
Answer
(a) , . (b) Both too small. (c) .
Steps
(a) , so and .
(b) everywhere means is concave up, so its graph lies above every tangent line on both sides of : for . Both are under-estimates. That is decreasing at changes nothing.
(c) Taylor's inequality with gives
so . Combining with (b): .
On the exam
How this topic is markedNearly every question carries three separate marks: the linearisation, the number, and a sentence naming the direction with the sign of as the reason. Writing only the number loses two of them.
For measurement error, divide symbolically first — — then substitute. It removes the arithmetic and hands you the percentage directly.
Work in radians, quote the final decimal with units, and remember the bound is quadratic: halving quarters it.
Keep going
Read next
Stuck on this topic?
One session with a tutor who teaches it every term usually settles it.
Book a session