First principles and rules
The derivative from first principles 1.07
- Derivative: f'(x) = \displaystyle\lim_{h\to0} \dfrac{f(x+h)-f(x)}{h}, the limit of the gradient of the chord joining to as the second point slides towards the first.
- Gradient function: is itself a function of , giving the gradient of the tangent at each point of the curve .
- Write down and expand it completely.
- Form the difference quotient and simplify the numerator; every term not containing must cancel.
- Divide the numerator through by .
- Let , discarding every remaining term that still contains a factor of .
- First principles is rarely the efficient way to differentiate, but it is the definition that justifies every shortcut rule below, and questions sometimes demand it explicitly to test that understanding.
- Keep the symbol in front of the quotient until the final line; setting before cancelling produces the meaningless .
- For , first principles gives — a concrete case worth rederiving from scratch to check the method is secure.
The power rule and index form 1.07
- , for any real power .
- , and the derivative of a constant is .
- Derivatives add term by term: \dfrac{d}{dx}\left(f(x)\pm g(x)\right) = f'(x)\pm g'(x).
- Negative powers: differentiates to .
- Fractional powers: differentiates to .
- The tangent to at has gradient m=f'(x_1) and equation ; the normal at the same point has gradient .
- Rewrite every term as a power of first: expand brackets, split fractions with a single term in the denominator, and convert roots to fractional indices.
- Differentiate term by term using the power rule.
- Convert back into root or fraction form if the question was set in that form.
- A quotient such as should be split into and differentiated directly; reaching for the quotient rule here wastes time and invites errors.
- Worked illustration of a tangent: for , the derivative is , which at equals ; since there, the tangent is , that is .
- The normal's gradient is only defined when — at a stationary point, the tangent is horizontal () and the normal is vertical, with no gradient to quote.
Worked example
Worked example
Differentiate:
from first principles, and hence state the gradient of the curve at .
Show worked solution
Subtracting:
leaves , so:
Letting gives:
which agrees with the power rule.
At the gradient is .
Chain, product and quotient rules
The chain rule 1.07
- If then \dfrac{dy}{dx} = f'(g(x))\cdot g'(x).
- Leibniz form: writing gives .
- Linear inner function: .
- Inverse form: , useful when is given as a function of .
- Identify the inner function and call it .
- Differentiate the outer function with respect to , keeping the inside unchanged.
- Differentiate the inner function with respect to .
- Multiply the two derivatives together and rewrite in terms of .
- The rule is often remembered as: differentiate the outside, keep the inside the same, then multiply by the derivative of the inside.
- Forgetting the final multiplication by g'(x) is the single most common differentiation slip; checking that the derivative of the inner function actually appears in the answer catches it every time.
- Nested chains (a function inside a function inside a function) apply the rule repeatedly, multiplying together one derivative per layer — work from the outermost layer inwards, one at a time.
The product rule 1.07
- If then .
- Split the expression into two factors and label them and .
- Differentiate each factor separately, writing and down before combining anything.
- Substitute into the formula.
- Factorise the result, since a product-rule answer almost always has a common factor and a factorised form is what later parts of a question need.
- Recognising when an expression is better rewritten before differentiating is part of the skill: expanding a product of two simple polynomials is quicker and safer than applying the product rule mechanically.
- The product rule is not 'differentiate each factor and multiply' — .
- Writing down , , and explicitly before substituting into the formula avoids the common error of using the wrong pair mid-substitution.
The quotient rule 1.07
- If then .
- The order of the numerator matters, because subtraction is not commutative: the term beginning with comes first, and reversing it changes the sign of the whole answer.
- The quotient rule is the product rule applied to together with the chain rule, so either route is valid; the quotient rule is simply the tidier bookkeeping.
- Choosing correctly between the chain, product and quotient rules — and spotting when none of them is needed — is often the real skill being tested rather than the differentiation itself.
- As with the product rule, differentiate and separately and write both out before substituting — combining derivatives mentally mid-formula is where sign errors creep in.
Worked examples
Worked example
Differentiate:
using the chain rule.
Show worked solution
Let the inside function be , so .
Then:
and:
By the chain rule:
Worked example
Differentiate:
giving your answer as a single fraction in its simplest form.
Show worked solution
Use the quotient rule with , so:
and , so:
Then:
Cancelling a factor of from numerator and denominator gives:
valid for .
Differentiating trig, exp and log functions
Trigonometric, exponential and logarithmic derivatives 1.07
- and , with in radians.
- , the defining property of from Exponentials and logarithms.
- , for .
- Composite forms follow from the chain rule: , \dfrac{d}{dx}e^{f(x)} = f'(x)e^{f(x)}, and \dfrac{d}{dx}\ln f(x) = \dfrac{f'(x)}{f(x)}.
- For example, , since the inner function differentiates to .
- These derivatives hold only in radians. In degrees the chain rule inserts an awkward factor, since , so calculus questions involving trigonometry are always set in radians.
- The negative sign on the derivative of cosine is a frequent source of error and is worth checking deliberately in any answer containing it.
- 's derivative \dfrac{f'(x)}{f(x)} is worth spotting directly in an integration context too — it's exactly the pattern that integrates back to .
Stationary points and curve sketching
Stationary points and their classification 1.07
- Stationary point: a point on the curve where , so the tangent there is horizontal.
- Point of inflection: a point where the curve changes its sense of curvature, that is, where changes sign.
- at a stationary point indicates a local minimum.
- at a stationary point indicates a local maximum.
- is inconclusive and requires checking the sign of on either side instead; this covers points of inflection as well as some maxima and minima the second-derivative test alone cannot distinguish.
- Away from stationary points, on an interval means the function is increasing there, and means it is decreasing.
- Differentiate and solve to find the -coordinates.
- Substitute each root into the original equation to obtain the corresponding -coordinates.
- Differentiate a second time and evaluate at each stationary point.
- Classify using the sign of the second derivative, falling back on a sign test of either side of the point if the second derivative is zero.
- The inconclusive case is genuinely ambiguous, not merely awkward: has at yet a clear minimum there, while has the same zero second derivative at and a stationary point of inflection instead.
- A stationary point is local, not global — the largest value of a function on a closed interval may occur at an endpoint rather than at any stationary point, so endpoints must be checked separately in optimisation problems.
- Always substitute the -coordinate back into the ORIGINAL function, not the derivative, to find the -coordinate of a stationary point — a surprisingly common slip under time pressure.
Curve sketching 1.07
- Find where the curve crosses the axes, by setting and then .
- Find the stationary points and classify them.
- Determine the behaviour as , using the dominant term of the expression.
- Identify any asymptotes, in particular values of excluded from the domain.
- Draw a smooth curve consistent with all of the above, labelling every feature found.
- A sketch does not need to be to scale, but every feature shown on it should be justified by a calculation rather than guessed from a calculator display.
- Sketching combines local information — stationary points and their nature — with global information about intercepts, end behaviour and asymptotes; a sketch missing either kind of information is incomplete.
- For a rational function, check the behaviour on BOTH sides of any vertical asymptote separately — the curve can approach from one side and from the other.
Worked examples
Worked example
Find and classify the stationary points of:
Show worked solution
Setting this to zero gives or .
At :
At :
The second derivative is:
At :
a local minimum at .
At :
a local maximum at .
Worked example
The curve has two stationary points.
Find their coordinates and determine the nature of each.
Show worked solution
Product rule with and :
Since always,
requires or:
The -values are at , and:
(3 s.f.) at:
Differentiating:
again gives:
At this is , a local minimum at .
At:
the bracket is , so the second derivative is , a local maximum at:
Rates of change and connected rates
Rates of change and connected rates 1.07
- Rate of change: in context, measures how fast changes per unit change in , and carries the units of divided by the units of .
- Connected rates follow from the chain rule: , and hence .
- Any chain of linked variables works the same way, for example for the surface area of an expanding sphere.
- Write down every rate given in the question in derivative notation, with its units.
- Write down the formula linking the variables involved, such as .
- Differentiate that formula to obtain the connecting derivative.
- Combine the derivatives by the chain rule, then substitute the values that apply at the required instant.
- Substitute the specific value, such as , only after differentiating; substituting first turns the formula into a constant whose derivative is zero.
- Interpreting the answer matters as much as calculating it: state the units, and read a negative rate as a decrease rather than reporting it as an error.
- Keep track of which variable each rate is 'per' — and look similar but mean very different things, and mixing them up is the most common error in connected-rates problems.
Worked example
Worked example
A spherical balloon is inflated so that its volume increases at a constant rate of cm s.
Find the rate at which its radius is increasing at the instant when the radius is cm.
Show worked solution
The volume of a sphere is:
so:
At this equals cm.
By the chain rule,
so:
(3 s.f.).
The value is positive, confirming the radius is increasing, and it is small because a large sphere needs a lot of extra volume for each extra centimetre of radius.
Per disputationem veritatem quaerimus