Everything, on one sheet

The card

Cancel factors only, never terms: (Ak)/(Bk) = A/B needs k multiplying the WHOLE top and bottom.
Improper (deg num ≥ deg denom): divide first. f(x) ≡ D(x)Q(x) + R(x), deg R < deg D.
Remainder theorem: divide by (x−a), remainder is f(a). Factor theorem: (x−a) a factor ⟺ f(a)=0.
"Show that" [fraction simplifies]: finish the division, THEN factorise and cancel the remainder against the divisor.
"Show that" [a given constant = value, e.g. D=0]: no factor to cancel — set up the equation it must satisfy and substitute to show it holds.
Domain: exclusions come from the ORIGINAL denominator, before cancelling — cancelled factors still count.

Why it works — Why the remainder theorem is true — and why a quadratic divisor is harder than a linear one

Start from the division algorithm itself, which is a definition, not a discovery: dividing any polynomial $f(x)$ by any nonzero divisor $D(x)$ produces a UNIQUE quotient $Q(x)$ and remainder $R(x)$ satisfying $f(x) \equiv D(x)Q(x) + R(x)$, with $\deg R < \deg D$. The $\equiv$ is doing real work here: it means the equation holds for every value of $x$, not just some — the two sides are literally the same expression, differently arranged. That licenses substituting ANY value of $x$ into it and getting a true statement, which is the entire mechanism behind everything else in this lesson. Take $D(x) = x-a$, linear, degree 1. Then $\deg R < 1$ forces $R$ to be a constant — call it $R$, with no $x$ in it at all — so $f(x) \equiv (x-a)Q(x) + R$ for every $x$. Substitute the one value that kills the first term: at $x=a$, $f(a) = (a-a)Q(a) + R = 0 + R = R$. So $R = f(a)$. That derivation IS the remainder theorem — not a separate fact to memorise alongside the division algorithm, but that identity evaluated at the single point where it collapses to something trivial. The factor theorem falls out immediately: $(x-a)$ divides $f(x)$ exactly precisely when $R=0$, which by the line above happens exactly when $f(a)=0$. Now take $D(x)$ quadratic instead, which is what spec 1.1's own guidance says this paper actually tests. $\deg R < 2$ this time, so $R(x)$ can be linear, $R(x) = mx+n$ — TWO unknowns, not zero. Substituting a root of $D(x)$ still gives one true equation, exactly as before, but one equation can't pin down two unknowns. This is the precise reason a quadratic divisor genuinely needs the division carried out in full (or, equivalently, the coefficients of $Q(x)$ and $R(x)$ found by matching coefficients on both sides of the identity) rather than being shortcut by a single clever substitution the way a linear divisor sometimes can be. It is also why the ONLY thing that makes the leftover fraction reducible afterwards is luck of a specific kind: $R(x)$ happening to share an actual factor with $D(x)$, which is exactly the "show that" pattern the worked chain below walks through in full.

Traps — 6

cancel-and-simplify-step-abandoned

Confirmed directly, verbatim, against both the examiner report AND the real mark scheme for the same question. Oct 2020 Q9(a) — divide $\frac{x^4-x^3-10x^2+3x-9}{x^2-x-12}$ — awards the last two of its four marks as "M1: Writes the given expression in the required form using $x^2-x-12=(x-4)(x+3)$... A1: Correct answer... Note that Q = 5 is given so it must be shown from correct work, not just stated." Candidates who divided correctly still lost both marks: "candidates did not continue to factorise the denominator and cancel (x+3) and hence not prov[e]... that Q is 5, they just stated it instead" — over 70% scored full marks, but the report names this specific abandonment as the reason the rest did not.

given-value-not-verified-by-substitution

A genuinely different failure from the one above — worth telling apart precisely, because the two are marked differently and fixed differently. Not every division question ends in a fraction to simplify: some GIVE you one of the constants outright and mark it as an independent, "shown" accuracy mark for proving that stated value, not for reaching it — the mark-scheme convention from §4 of the facts bank for "the answer is printed on the paper." Verified verbatim, Jan 2025 Q4(a)(ii) — divide $\frac{4x^3+2x^2+3x+8}{x^2+4} \equiv Ax+B+\dfrac{Cx+D}{x^2+4}$ and "show that $D=0$" — the real mark scheme awards this as "B1*: Fully shows that D = 0 from clear and correct work... they would need to set up (at least) two correct equations and solve, with appropriate substitutions seen, to show that D = 0." The examiner report confirms candidates who found $A$, $B$ and $C$ correctly still lost this mark: "did not subsequently establish that D = 0 as they did not show the substitution." There is no factor to cancel here — $x^2+4$ has no real linear factor, so "factorise and cancel" is not the fix. The fix is the same "show that" discipline applied to a different target: write down the actual equation the given value must satisfy, substitute into it, and show it holds — not assert the printed value because it is, in fact, printed.

combined-fraction-not-fully-justified

A separate, independently confirmed pattern across at least three series (Oct 2020, Jan 2024, Jan 2025): candidates combine rational expressions over a common denominator and reach the correct final simplified form, but without showing the intermediate working that justifies it — and lose the mark attached to the justification even though the answer on the page is right. This is the general "show that" rule from the paper's own general marking guidance applied to this specific topic: an answer that happens to be correct isn't the same thing as an answer that's been shown to be correct, and only the second earns a 'show that' mark.

method-not-set-up-before-the-arithmetic

The general marking guidance, verified verbatim from the January 2023 mark scheme and cross-checked against October 2023 and June 2022: "Where a method involves using a formula that has been learnt, the advice given in recent examiners' reports is that the formula should be quoted first... Where the formula is not quoted, the method mark can be gained by implication from correct working with values but may be lost if there is any mistake in the working." Algebraic division has no single formula to quote, but the identity it's built on does: writing $f(x) \equiv D(x)Q(x)+R(x)$ before diving into the subtraction protects the method mark the same way quoting the quadratic formula does — implied method from unlabelled working is real credit, but it is credit that a single early slip can destroy entirely.

exact-form-abandoned-for-a-decimal

Verified verbatim, same general marking guidance: "Examiners' reports have emphasised that where, for example, an exact answer is asked for, or working with surds is clearly required, marks will normally be lost if the candidate resorts to using rounded decimals." A simplified rational expression or a constant found by cancelling factors is an exact algebraic answer by nature — decimalising it (turning $\frac{5}{x+3}$ into something like "$\approx 1.67$ when $x=0$", or rounding a found constant) answers a question that wasn't asked and drops marks a correct exact form would have kept.

calculator-technology-cited-as-the-method

The paper-wide rubric, verified verbatim from a real WMA13 question paper: "In this question you must show all stages of your working. Solutions relying entirely on calculator technology are not acceptable." Simplification and division questions are exactly where this lands hardest, because a calculator can often produce the simplified fraction directly. Examiner reports across the paper document candidates losing marks for a correct final answer with no algebraic method shown on a question carrying this instruction — the number being right is not the thing being marked.

The card

1 hour 30 minutes, 75 marks. Students answer all questions — no choice of question.
9 or 10 compulsory questions — not fixed by the spec, varies by series. No sections: one flat sequence, each with its own mark tariff.
≈1.2 minutes (72 seconds) per mark on average — budget against each question's own printed marks, not a flat average.
Ordinary scientific calculator only — no CAS or symbolic algebra. Some questions explicitly ban calculator-only solutions; working must be shown there regardless.
Marked in M/A/B lines (method/accuracy/standalone), not AO bands. A marks depend on M — M0 blocks every chained A.

Why it works — What the Examiner Is Actually Reading For, and Why a Right Answer Isn't Always Enough

Three questions decide whether a line of working earns its mark: what does the examiner actually read, what are they looking for in it, and what would change the decision. What they read is the working itself, not the final boxed value in isolation — a mark scheme built from M/A/B lines is, by construction, reading process, not just output. What they're looking for on an M line is one specific thing: a correct method, or a recognisable attempt at one, applied to this question — not correctness of the final number, and not effort in general. This is why the paper's own general marking principles, repeated across schemes covering entirely different topics, tell examiners — and, by extension, tell a candidate exactly what to write — the same thing every time: "Where a method involves using a formula that has been learnt, the advice given in recent examiners' reports is that the formula should be quoted first... Where the formula is not quoted, the method mark can be gained by implication from correct working with values but may be lost if there is any mistake in the working" (verified verbatim, the Pure-Mathematics-specific general marking principles printed in every P3 scheme reviewed). Quoting the formula first is not a stylistic nicety — it is the single cheapest way to make the M mark independently visible, so a slip two lines later does not retroactively cast doubt on whether the method itself was ever really there. What would change the decision, then, is exactly what a calculator-only answer withholds: a visible, checkable method line. A correct final value produced with no method shown gives the examiner nothing to read on the M line — and because every A mark on this paper is explicitly dependent on the M mark that precedes it ("M0 A1 is impossible"), the absence of a checkable method does not cost one mark, it caps the whole question's method-dependent marks at zero, regardless of how the final number was actually reached.

The card

Cancel factors only, never terms: (Ak)/(Bk) = A/B needs k multiplying the WHOLE top and bottom.
Improper (deg num ≥ deg denom): divide first. f(x) ≡ D(x)Q(x) + R(x), deg R < deg D.
Remainder theorem: divide by (x−a), remainder is f(a). Factor theorem: (x−a) a factor ⟺ f(a)=0.
"Show that" [fraction simplifies]: finish the division, THEN factorise and cancel the remainder against the divisor.
"Show that" [a given constant = value, e.g. D=0]: no factor to cancel — set up the equation it must satisfy and substitute to show it holds.
Domain: exclusions come from the ORIGINAL denominator, before cancelling — cancelled factors still count.

Why it works — Why the remainder theorem is true — and why a quadratic divisor is harder than a linear one

Start from the division algorithm itself, which is a definition, not a discovery: dividing any polynomial $f(x)$ by any nonzero divisor $D(x)$ produces a UNIQUE quotient $Q(x)$ and remainder $R(x)$ satisfying $f(x) \equiv D(x)Q(x) + R(x)$, with $\deg R < \deg D$. The $\equiv$ is doing real work here: it means the equation holds for every value of $x$, not just some — the two sides are literally the same expression, differently arranged. That licenses substituting ANY value of $x$ into it and getting a true statement, which is the entire mechanism behind everything else in this lesson. Take $D(x) = x-a$, linear, degree 1. Then $\deg R < 1$ forces $R$ to be a constant — call it $R$, with no $x$ in it at all — so $f(x) \equiv (x-a)Q(x) + R$ for every $x$. Substitute the one value that kills the first term: at $x=a$, $f(a) = (a-a)Q(a) + R = 0 + R = R$. So $R = f(a)$. That derivation IS the remainder theorem — not a separate fact to memorise alongside the division algorithm, but that identity evaluated at the single point where it collapses to something trivial. The factor theorem falls out immediately: $(x-a)$ divides $f(x)$ exactly precisely when $R=0$, which by the line above happens exactly when $f(a)=0$. Now take $D(x)$ quadratic instead, which is what spec 1.1's own guidance says this paper actually tests. $\deg R < 2$ this time, so $R(x)$ can be linear, $R(x) = mx+n$ — TWO unknowns, not zero. Substituting a root of $D(x)$ still gives one true equation, exactly as before, but one equation can't pin down two unknowns. This is the precise reason a quadratic divisor genuinely needs the division carried out in full (or, equivalently, the coefficients of $Q(x)$ and $R(x)$ found by matching coefficients on both sides of the identity) rather than being shortcut by a single clever substitution the way a linear divisor sometimes can be. It is also why the ONLY thing that makes the leftover fraction reducible afterwards is luck of a specific kind: $R(x)$ happening to share an actual factor with $D(x)$, which is exactly the "show that" pattern the worked chain below walks through in full.

Traps — 6

cancel-and-simplify-step-abandoned

Confirmed directly, verbatim, against both the examiner report AND the real mark scheme for the same question. Oct 2020 Q9(a) — divide $\frac{x^4-x^3-10x^2+3x-9}{x^2-x-12}$ — awards the last two of its four marks as "M1: Writes the given expression in the required form using $x^2-x-12=(x-4)(x+3)$... A1: Correct answer... Note that Q = 5 is given so it must be shown from correct work, not just stated." Candidates who divided correctly still lost both marks: "candidates did not continue to factorise the denominator and cancel (x+3) and hence not prov[e]... that Q is 5, they just stated it instead" — over 70% scored full marks, but the report names this specific abandonment as the reason the rest did not.

given-value-not-verified-by-substitution

A genuinely different failure from the one above — worth telling apart precisely, because the two are marked differently and fixed differently. Not every division question ends in a fraction to simplify: some GIVE you one of the constants outright and mark it as an independent, "shown" accuracy mark for proving that stated value, not for reaching it — the mark-scheme convention from §4 of the facts bank for "the answer is printed on the paper." Verified verbatim, Jan 2025 Q4(a)(ii) — divide $\frac{4x^3+2x^2+3x+8}{x^2+4} \equiv Ax+B+\dfrac{Cx+D}{x^2+4}$ and "show that $D=0$" — the real mark scheme awards this as "B1*: Fully shows that D = 0 from clear and correct work... they would need to set up (at least) two correct equations and solve, with appropriate substitutions seen, to show that D = 0." The examiner report confirms candidates who found $A$, $B$ and $C$ correctly still lost this mark: "did not subsequently establish that D = 0 as they did not show the substitution." There is no factor to cancel here — $x^2+4$ has no real linear factor, so "factorise and cancel" is not the fix. The fix is the same "show that" discipline applied to a different target: write down the actual equation the given value must satisfy, substitute into it, and show it holds — not assert the printed value because it is, in fact, printed.

combined-fraction-not-fully-justified

A separate, independently confirmed pattern across at least three series (Oct 2020, Jan 2024, Jan 2025): candidates combine rational expressions over a common denominator and reach the correct final simplified form, but without showing the intermediate working that justifies it — and lose the mark attached to the justification even though the answer on the page is right. This is the general "show that" rule from the paper's own general marking guidance applied to this specific topic: an answer that happens to be correct isn't the same thing as an answer that's been shown to be correct, and only the second earns a 'show that' mark.

method-not-set-up-before-the-arithmetic

The general marking guidance, verified verbatim from the January 2023 mark scheme and cross-checked against October 2023 and June 2022: "Where a method involves using a formula that has been learnt, the advice given in recent examiners' reports is that the formula should be quoted first... Where the formula is not quoted, the method mark can be gained by implication from correct working with values but may be lost if there is any mistake in the working." Algebraic division has no single formula to quote, but the identity it's built on does: writing $f(x) \equiv D(x)Q(x)+R(x)$ before diving into the subtraction protects the method mark the same way quoting the quadratic formula does — implied method from unlabelled working is real credit, but it is credit that a single early slip can destroy entirely.

exact-form-abandoned-for-a-decimal

Verified verbatim, same general marking guidance: "Examiners' reports have emphasised that where, for example, an exact answer is asked for, or working with surds is clearly required, marks will normally be lost if the candidate resorts to using rounded decimals." A simplified rational expression or a constant found by cancelling factors is an exact algebraic answer by nature — decimalising it (turning $\frac{5}{x+3}$ into something like "$\approx 1.67$ when $x=0$", or rounding a found constant) answers a question that wasn't asked and drops marks a correct exact form would have kept.

calculator-technology-cited-as-the-method

The paper-wide rubric, verified verbatim from a real WMA13 question paper: "In this question you must show all stages of your working. Solutions relying entirely on calculator technology are not acceptable." Simplification and division questions are exactly where this lands hardest, because a calculator can often produce the simplified fraction directly. Examiner reports across the paper document candidates losing marks for a correct final answer with no algebraic method shown on a question carrying this instruction — the number being right is not the thing being marked.

The card

sec θ = 1/cos θ; cosec θ = 1/sin θ; cot θ = cos θ/sin θ = 1/tan θ. Not the other pairing.
A coefficient in front stays outside the reciprocal: 3 cosec θ = 3/sin θ, never 1/(3 sin θ) — a real, mark-scheme-penalised WMA13 slip.
sec²θ ≡ 1 + tan²θ; cosec²θ ≡ 1 + cot²θ — both from dividing cos²θ+sin²θ≡1 by cos²θ or sin²θ. Memorise; not in the booklet.
sec θ, cosec θ: |value| ≥ 1 always; undefined where cos θ, sin θ = 0. cot θ: any real value; undefined where sin θ = 0.
arcsin: domain [−1,1], range −90°–90°. arccos: domain [−1,1], range 0°–180°. arctan: domain ℝ, range −90°<y<90°, open.
Inverse exists only where the original is one-one — that's why each range above is a restriction, not the full curve.

Why it works — Where sec²θ = 1 + tan²θ actually comes from

Start from the one identity that is genuinely foundational — $\cos^2\theta + \sin^2\theta \equiv 1$ — and divide every term by $\cos^2\theta$ (valid wherever $\cos\theta \neq 0$, which is exactly where $\sec\theta$ is defined in the first place, so nothing is lost). $\frac{\cos^2\theta}{\cos^2\theta} + \frac{\sin^2\theta}{\cos^2\theta} \equiv \frac{1}{\cos^2\theta}$. The first term is $1$. The second term is $\left(\frac{\sin\theta}{\cos\theta}\right)^2 = \tan^2\theta$. The right-hand side is $\left(\frac{1}{\cos\theta}\right)^2 = \sec^2\theta$. So the division produces, without any further assumption, $1 + \tan^2\theta \equiv \sec^2\theta$ — the identity was already sitting inside $\cos^2\theta + \sin^2\theta \equiv 1$, waiting to be uncovered by dividing through by the right thing. Divide the SAME starting identity by $\sin^2\theta$ instead, and the companion identity falls out the same way: $\frac{\cos^2\theta}{\sin^2\theta} + \frac{\sin^2\theta}{\sin^2\theta} \equiv \frac{1}{\sin^2\theta}$, giving $\cot^2\theta + 1 \equiv \text{cosec}^2\theta$. Both identities are the same starting fact, viewed through two different divisions — which is also why they are not independent things to memorise separately so much as one thing to know how to re-derive twice. This matters for exam technique specifically: general marking guidance for this paper records that where a method uses a learned formula or identity, examiners advise quoting it BEFORE substituting values, because the method mark can otherwise only be inferred from correct working and is then vulnerable to any slip inside that working. Writing "$\sec^2\theta \equiv 1 + \tan^2\theta$" as its own line, before using it, is the safer habit — and it is genuinely the same advice this course's own quadratic-functions lesson gives for quoting the quadratic formula before substituting into it.

Traps — 7

cosec-written-as-reciprocal-of-cosine

Confirmed directly on a real WMA13 identity question: candidates "struggled writing cosec θ = 1/cos θ" (Jan 2022, Q2) — reaching for cosine because of the shared first syllable, when the correct pairing is cosec θ = 1/sin θ. The same report records the flip side too: "even some of the weakest candidates were able to earn a mark for stating the identity cosec θ = 1/sin θ" (Jan 2022, Q2) — this is a genuinely accessible mark, and losing it to a mispairing rather than a harder step is the most avoidable loss in this whole topic. Check any pairing by confirming it multiplies to 1: cosec θ · sin θ = 1, not cosec θ · cos θ.

coefficient-folded-into-the-reciprocal

A second, distinct error on the exact same real WMA13 question named above, re-verified directly against the primary mark scheme (not just the examiner-report prose): the same series' report records candidates "sometimes" writing "3 cosecθ = 1/ 3sinθ" (Jan 2022, Q2) — folding the coefficient into the denominator alongside sin θ, rather than leaving it multiplying the finished reciprocal. The real mark scheme is explicit that this scores nothing for the identity mark: "Note that 3cosecθ = 1/(3sinθ) is B0 unless there is an aside that does state cosecθ = 1/sinθ." The fix is mechanical, not a new idea: reciprocate the trig function alone first — cosec θ = 1/sin θ — then multiply by whatever coefficient sits in front, e.g. 3 cosec θ = 3 × (1/sin θ) = 3/sin θ. The coefficient never moves inside the denominator next to sin θ.

quadratic-in-tan-or-cot-missing-a-second-solution

Documented on a different WMA13 trig equation, not one reducing via sec²θ or cosec²θ specifically, but the exact same shape of failure: "it was disappointing that many candidates failed to identify the second solution here to gain the full marks" (Jun 2022, Q7). Every equation in this lesson that reduces to a quadratic in tan θ or cot θ produces TWO values of the trig ratio, and each of those typically produces TWO angles in a full 0°–360° range — up to four solutions from one equation. Stopping after the first branch, or after one solution within a branch, is this trap in miniature.

solved-ratio-mistaken-for-the-angle-itself

Confirmed on a WMA13 question that solved for sin x rather than x directly: "a few interpreted the solution to their equation as being the value for x rather than for sin x, thereby losing both marks" (Jan 2024, Q6(c)). The identical failure is available here with tan θ or cot θ: solving cot²θ − cotθ − 6 = 0 gives values of cot θ (3 and −2), not values of θ. The step from "cot θ = 3" to "θ = 18.4° or 198.4°" is not optional bookkeeping — it is a required, markable step that a genuine number of candidates skip on the analogous sin-based question.

identity-not-quoted-before-use

General marking guidance for this paper states the advice explicitly: "where a method involves using a formula that has been learnt, the advice given in recent examiners' reports is that the formula should be quoted first... where the formula is not quoted, the method mark can be gained by implication from correct working with values but may be lost if there is any mistake in the working" (Jan 2023 general marking guidance, cross-checked against Oct 2023 and Jun 2022). Applied here: writing "sec²θ ≡ 1 + tan²θ" as its own line before substituting is what makes the method mark secure even if the next line contains a slip — skipping straight to the substituted equation makes the method mark depend on everything downstream staying correct.

restricted-domain-of-the-inverse-left-unstated

Confirmed as the single most consistently dropped mark on real inverse-function questions on this paper: "despite being a standard question, the majority of candidates still failed to state the domain for their inverse function and did not achieve the B mark" (Jun 2022, Q2), and independently, "a majority of candidates were not aware that the domain was required, and therefore by far the most common score seen was 2/3" (Jan 2022, Q6). Those quotes are about inverse functions generally (spec 1.2), not about arcsin/arccos/arctan specifically — but the failure is identical in shape: a question asking for the domain OR range of arcsin, arccos or arctan is asking for exactly the kind of stated restriction these reports record candidates omitting, even when the rest of the working is correct.

missing-intermediate-line-in-a-prove-that

A general pattern confirmed across a WMA13 series' full set of reports: "it was noticeable in this series that many candidates omitted important lines when proceeding to the given solution resulting in the loss of some vital marks" (Jan 2022, general summary). Applied to this topic specifically: a "show that sec²θ ≡ 1 + tan²θ" question is asking for the division-by-cos²θ derivation shown above written out, not just asserted — jumping from cos²θ + sin²θ ≡ 1 straight to the answer, with the dividing-through step invisible, is exactly the omission these reports repeatedly penalise.

The card

Function: one input → exactly one output. One-one: no two inputs share an output. Many-one: some do.
fg(x) = f(g(x)) — do g first, then f. fg ≠ gf in general.
f⁻¹ exists as a function only if f is one-one (restrict the domain if it isn't).
Swap x and y to find f⁻¹ ⇒ domain of f⁻¹ = range of f; range of f⁻¹ = domain of f. Always state both.
f⁻¹f(x) = ff⁻¹(x) = x. Graph of f⁻¹ is the reflection of f's graph in y = x.

Why it works — Why an inverse needs one-one — and why domain and range swap places

An inverse function $f^{-1}$ is meant to undo $f$: feed it an output of $f$, and it should hand back the input that produced it. Follow that requirement to its logical end and one-one-ness stops being a rule to memorise and becomes unavoidable. Suppose $f$ is many-one: some output $y_0$ is shared by two different inputs, $a$ and $b$, with $f(a) = f(b) = y_0$ and $a \neq b$. Feed $y_0$ into $f^{-1}$ and ask what comes out. It has to be $a$, because $f^{-1}$ is meant to undo $f(a)$. It also has to be $b$, because $f^{-1}$ is meant to undo $f(b)$. One input, $y_0$, with two required outputs — but that is precisely the property that disqualifies something from being a function at all, the same one-input-one-output requirement the very first teach block opened with. So $f^{-1}$ exists as a function if and only if $f$ is one-one: not a separate condition bolted onto the definition, but the definition of "function" applied to the reversed rule. This is exactly why domain restriction matters so much in this topic — $f(x) = x^2$ on all of $\mathbb{R}$ is many-one and has no inverse function, but $f(x) = x^2$ restricted to $x \geq 0$ is one-one (every non-negative output now comes from exactly one non-negative input), and that restricted version does have an inverse: $f^{-1}(x) = \sqrt{x}$. Nothing about the formula changed; what changed is which inputs are on offer. Now follow the mechanics of actually finding $f^{-1}$, and a second fact falls out for free. The standard method is: write $y = f(x)$, swap $x$ and $y$ (this is the "walk it backwards" step — the new equation asks "what $x$ produces this $y$?" using the old equation's own rule), then rearrange to make $y$ the subject again. Because the swap literally exchanges the two letters' roles, whatever set of values $x$ was allowed to range over becomes the set $y$ ranges over in the new equation, and vice versa. Translated back into function language: the domain of $f^{-1}$ is the range of $f$, and the range of $f^{-1}$ is the domain of $f$. This is not a separate fact to memorise alongside "swap $x$ and $y$" — it is what swapping $x$ and $y$ *means*. It is also why the spec's own guidance, transcribed verbatim during this course's own research pass, states $f^{-1}f(x) = ff^{-1}(x) = x$: applying $f$ and then $f^{-1}$ (or the other way round) walks forward along the rule and then immediately back along it, landing exactly where it started, for every $x$ in the appropriate domain.

Traps — 5

domain-of-inverse-omitted

The single most consistently reported error anywhere in the WMA13 archive for this sub-topic, confirmed across at least three separate series. Verified verbatim, Jun 2022 Q2: "Despite being a standard question, the majority of candidates still failed to state the domain for their inverse function and did not achieve the B mark." Verified verbatim, Jan 2022 Q6: "a majority of candidates were not aware that the domain was required, and therefore by far the most common score seen was 2/3" (out of 3) — independently confirmed again in Oct 2021 Q1(b). The mechanism is structural, not carelessness alone: the domain of f⁻¹ is an independent B mark (see the next trap), so getting every line of algebra correct still leaves it unclaimed unless it is written down as its own separate statement.

domain-of-inverse-derived-from-scratch-not-recognised-as-range-of-f

A distinct pattern from simply omitting the domain: some candidates DO attempt to state a domain for f⁻¹, but re-derive it from first principles on the new expression rather than recognising it is already sitting in the answer to an earlier part of the question, as the range of f. Verified, Oct 2022 Q2(b) — the report notes candidates deriving the inverse's domain "from scratch" instead of using the range of f they had, or could have had, already. This costs time even where it does not cost the mark outright, and it is the exact opposite of the efficient route the worked chain above demonstrates: find the range of f early, and the domain of f⁻¹ is already answered before the algebra for f⁻¹(x) itself has even begun.

inverse-algebra-and-its-domain-are-independently-marked

Not an error a candidate makes so much as a fact about the mark scheme that, misunderstood, produces one: the domain of an inverse function is not a bonus tacked onto the M1 A1 for finding its formula, and it is not lost automatically if the algebra goes wrong, or gained automatically if it goes right. This course's own research verifies this directly from a real mark scheme, Jan 2023 Q1(c) (there, finding g⁻¹(x) for a fraction g rather than f⁻¹(x) for a quadratic, but the structure is identical to every example in this lesson): the scheme awards M1 A1 for finding the inverse and a SEPARATE, independent B1 for the domain. Treating the domain as "the last part of finding the inverse" rather than its own distinct, separately-earned statement is the root cause behind both traps above.

inverse-rearrangement-shown-as-calculator-output-with-no-algebra

WMA13 carries an explicit, paper-wide, verbatim rubric on exactly this kind of algebra: certain questions state "In this question you must show all stages of your working. Solutions relying entirely on calculator technology are not acceptable" (verified verbatim from the January 2023 question paper, and confirmed recurring across multiple series). Finding f⁻¹(x) is a rearrangement — precisely the kind of step a calculator or an equation solver can shortcut silently. Where a question carries this instruction, arriving at a correct f⁻¹(x) with no visible swap-and-rearrange working risks the method mark outright, whatever the final expression says.

inverse-answer-attached-to-the-wrong-function-letter

A distinct, purely notational failure mode, confirmed directly in a real mark scheme rather than an examiner report. When a question defines more than one function — exactly the situation in every worked example, chain drill and marked solution in this lesson, since composition and inverses are normally taught and tested together — a correctly-derived inverse can still lose its accuracy mark if it is labelled with the wrong function's letter. Verified verbatim, Jan 2023 Q1(c), which defines both f(x) and g(x) and asks specifically for g⁻¹(x): the scheme's note on the A1 mark reads "Condone y = (3−x)/(2x) o.e and even g⁻¹ = (3−x)/(2x) but NOT f⁻¹ = (3−x)/(2x) o.e." Right rearrangement, right final expression, zero marks — because f is the letter more often inverted in practice, and writing "f⁻¹(x) = ..." out of habit when the question asked for g⁻¹(x) is an easy, silent slip that the algebra itself never reveals.

The card

eˣ: domain ℝ, range >0, asymptote y=0 as x→−∞. ln x: domain >0, range ℝ, asymptote x=0. ln(eˣ)=x, e^(ln x)=x.
Solve e^(ax+b)=p: take ln both sides. Solve ln(ax+b)=q: apply e^(...) both sides.
y=axⁿ → log y = n log x + log a: plot log y vs log x, gradient=n, intercept=log a.
y=kbˣ → log y = x log b + log k: plot log y vs x, gradient=log b, intercept=log k.
Un-log every intercept (10^intercept) before it's a genuine constant. Gradient never needs un-logging.

Why it works — Why ln x exists at all — built on the one-one/domain-range mechanism from the functions lesson

The functions lesson established a chain of reasoning that applies to $e^x$ exactly as it applies to any other function: an inverse exists as a function if and only if the original is one-one, because a many-one function would need its inverse to hand back two different outputs for one input, which is precisely what disqualifies something from being a function at all. $e^x$ is strictly increasing on all of $\mathbb{R}$ — established two blocks above — so no two different $x$-values ever give the same $e^x$: it is one-one on its whole domain, no restriction needed (unlike the quadratics in the earlier WMA11 lesson, which needed a domain restriction before an inverse could exist at all). So $e^x$ has an inverse function, defined on all of $\mathbb{R}$, and it is given its own name rather than left as "$e^x$, walked backwards": $\ln x$. The functions lesson's domain/range-swap fact applies immediately: the domain of $e^x$ is $\mathbb{R}$ and its range is $\{y : y > 0\}$, so the domain of $\ln x$ is $\{x : x > 0\}$ and its range is all of $\mathbb{R}$ — exactly what the first prequestion question above asked for, before this mechanism had been written out. The defining cancellation follows the same pattern the spec's own guidance states for inverses in general, applied here: $\ln(e^x) = x$ for every real $x$, and $e^{\ln x} = x$ for every $x > 0$. This is not a separate rule to memorise — it is what "inverse function" means, applied to this particular pair. And it is the whole method for solving the two equation types spec 3.2 names. To solve $e^{ax+b} = p$: apply $\ln$ to both sides, and the left side collapses by the cancellation above to exactly $ax+b$, leaving $ax+b = \ln p$ to rearrange for $x$ — an ordinary linear equation, once the exponential has been undone. To solve $\ln(ax+b) = q$: apply $e^{(\ldots)}$ to both sides instead, and the left side collapses to $ax+b$, leaving $ax+b = e^q$. Concretely: $e^{5-2x} = 7 \Rightarrow 5-2x = \ln 7 \Rightarrow x = \dfrac{5-\ln 7}{2} \approx 1.53$ (3 s.f.); and $\ln(4-x) = 2 \Rightarrow 4-x = e^2 \Rightarrow x = 4-e^2 \approx -3.39$ (3 s.f.). Both moves are the same one move, in opposite directions — apply whichever function undoes the one already sitting in the equation.

Traps — 6

modelling-answer-missing-units-or-context-word

The single most consistently reported error anywhere in the WMA13 exponentials/logs archive, confirmed near-identically across at least four separate series: "the units (tonnes) were often omitted" (Oct 2021, Q3(b)); "only about half of them remembered that... they needed to also state the units" (Jan 2022, Q4(c)); "many lost the accuracy mark by either omitting the units '£' or not referring to a 'loss'" (Oct 2022, Q5(a)); "many did not gain the mark as they did not include the units (m2)" (Jan 2025, Q2(a)). A correct number with no unit or no context word attached is treated as an incomplete answer, not a minor presentation slip — write the unit or the contextual word every time a modelling question's answer is a quantity, not just an $x$.

calculator-equation-solver-used-with-no-algebra-shown

WMA13 papers carry an explicit, verbatim, paper-wide rubric on specific parts — "In this question you must show all stages of your working. Solutions relying entirely on calculator technology are not acceptable" (verified verbatim from the question paper itself) — and equations of the form $e^{ax+b}=p$ or $\ln(ax+b)=q$ are exactly where it lands. An examiner report confirms the consequence directly: "candidates should make sure that they do not resort to using a general equation solver... as this may not necessarily score full marks... it may be that full marks will not be awarded in the future, despite a correct answer" (Jun 2022, §5.3). Show the $\ln$ or $e^{(\ldots)}$ line explicitly, every time, on a question carrying this instruction.

log-base-confused-on-a-log-log-graph

Verified real context, Jun 2023 Q2: a log-log graph set explicitly in base $6$ — the examiner report records "being confused by log base 6, with some attempts to use log base 10 or 'e' and 'ln'" as the main documented error. The mechanism block above proves the gradient of a log-log graph is the same whatever base is chosen (true because the exponent $n$ in $y=ax^n$ doesn't depend on which base the logs are taken in — this is NOT true for a log-linear graph's gradient, $\log_c b$ for $y=kb^x$, which does change with the base $c$) — but the INTERCEPT is not, because it equals $\log_c(\text{constant})$ for whichever base $c$ the graph was actually plotted in. Un-logging an intercept with the wrong base — reading a base-6 intercept as if it were base 10 — produces a wrong constant even when every other step was correct.

exact-value-required-but-answer-rounded-early

Verified verbatim, from the general marking guidance itself, applying across every WMA13 topic but landing especially hard here because solving $e^{ax+b}=p$ so often produces an answer that is only exact as a logarithm: "Examiners' reports have emphasised that where, for example, an exact answer is asked for... marks will normally be lost if the candidate resorts to using rounded decimals". Where a question asks for an exact value of $k$ or $x$, stop at the $\ln(\ldots)$ or $\dfrac{\ln(\ldots)}{\ldots}$ form — a rounded decimal answers a different question than the one asked, even if a later part of the same question does then want 3 significant figures.

domain-restriction-from-a-non-positive-argument-not-explained

Confirmed real context: a temperature-decay model (Jan 2024, Q5) carries a hard lower-bound domain restriction that has to be explained using log-of-a-non-positive-number reasoning — the facts bank describes this context without quoting the examiner report verbatim on this specific point, so it is presented here as a described real question type, not a direct quote. The underlying mechanism is exact, though: $\ln(ax+b)$ is only defined where $ax+b>0$, so a model built around a $\ln(\ldots)$ term is only valid on the part of its domain where that argument stays positive — and a question that asks WHY a model breaks down, or for how long it remains valid, is asking for exactly this restriction to be identified and explained, not for more algebra on the equation itself.

log-notation-dropped-or-misplaced-when-reading-a-log-log-graph

Verified verbatim, Jun 2023 Q2(a)(i) examiner report — a distinct notation failure from the base-confusion trap above, on the very same real question: writing down the linear equation from a given log-log graph "proved to be straight forward and was generally done correctly for one mark. If this was not achieved, for example, a few students simply wrote $y = 4 - 2x$" — dropping every $\log_6$ entirely and treating the two labelled axes as if they were plain, unlogged $x$ and $y$. A second, separate slip the same report names: "the argument was often written as a superscript within the log" — writing something that reads as $\log 6^x$ rather than $\log_6 x$, putting the base in the wrong position. Neither is a calculation error — the axis labels on the graph already state the base and which variable is logged; the failure is not reading them onto the equation correctly, not anything that happens afterward.

The card

Function: one input → exactly one output. One-one: no two inputs share an output. Many-one: some do.
fg(x) = f(g(x)) — do g first, then f. fg ≠ gf in general.
f⁻¹ exists as a function only if f is one-one (restrict the domain if it isn't).
Swap x and y to find f⁻¹ ⇒ domain of f⁻¹ = range of f; range of f⁻¹ = domain of f. Always state both.
f⁻¹f(x) = ff⁻¹(x) = x. Graph of f⁻¹ is the reflection of f's graph in y = x.

Why it works — Why an inverse needs one-one — and why domain and range swap places

An inverse function $f^{-1}$ is meant to undo $f$: feed it an output of $f$, and it should hand back the input that produced it. Follow that requirement to its logical end and one-one-ness stops being a rule to memorise and becomes unavoidable. Suppose $f$ is many-one: some output $y_0$ is shared by two different inputs, $a$ and $b$, with $f(a) = f(b) = y_0$ and $a \neq b$. Feed $y_0$ into $f^{-1}$ and ask what comes out. It has to be $a$, because $f^{-1}$ is meant to undo $f(a)$. It also has to be $b$, because $f^{-1}$ is meant to undo $f(b)$. One input, $y_0$, with two required outputs — but that is precisely the property that disqualifies something from being a function at all, the same one-input-one-output requirement the very first teach block opened with. So $f^{-1}$ exists as a function if and only if $f$ is one-one: not a separate condition bolted onto the definition, but the definition of "function" applied to the reversed rule. This is exactly why domain restriction matters so much in this topic — $f(x) = x^2$ on all of $\mathbb{R}$ is many-one and has no inverse function, but $f(x) = x^2$ restricted to $x \geq 0$ is one-one (every non-negative output now comes from exactly one non-negative input), and that restricted version does have an inverse: $f^{-1}(x) = \sqrt{x}$. Nothing about the formula changed; what changed is which inputs are on offer. Now follow the mechanics of actually finding $f^{-1}$, and a second fact falls out for free. The standard method is: write $y = f(x)$, swap $x$ and $y$ (this is the "walk it backwards" step — the new equation asks "what $x$ produces this $y$?" using the old equation's own rule), then rearrange to make $y$ the subject again. Because the swap literally exchanges the two letters' roles, whatever set of values $x$ was allowed to range over becomes the set $y$ ranges over in the new equation, and vice versa. Translated back into function language: the domain of $f^{-1}$ is the range of $f$, and the range of $f^{-1}$ is the domain of $f$. This is not a separate fact to memorise alongside "swap $x$ and $y$" — it is what swapping $x$ and $y$ *means*. It is also why the spec's own guidance, transcribed verbatim during this course's own research pass, states $f^{-1}f(x) = ff^{-1}(x) = x$: applying $f$ and then $f^{-1}$ (or the other way round) walks forward along the rule and then immediately back along it, landing exactly where it started, for every $x$ in the appropriate domain.

Traps — 5

domain-of-inverse-omitted

The single most consistently reported error anywhere in the WMA13 archive for this sub-topic, confirmed across at least three separate series. Verified verbatim, Jun 2022 Q2: "Despite being a standard question, the majority of candidates still failed to state the domain for their inverse function and did not achieve the B mark." Verified verbatim, Jan 2022 Q6: "a majority of candidates were not aware that the domain was required, and therefore by far the most common score seen was 2/3" (out of 3) — independently confirmed again in Oct 2021 Q1(b). The mechanism is structural, not carelessness alone: the domain of f⁻¹ is an independent B mark (see the next trap), so getting every line of algebra correct still leaves it unclaimed unless it is written down as its own separate statement.

domain-of-inverse-derived-from-scratch-not-recognised-as-range-of-f

A distinct pattern from simply omitting the domain: some candidates DO attempt to state a domain for f⁻¹, but re-derive it from first principles on the new expression rather than recognising it is already sitting in the answer to an earlier part of the question, as the range of f. Verified, Oct 2022 Q2(b) — the report notes candidates deriving the inverse's domain "from scratch" instead of using the range of f they had, or could have had, already. This costs time even where it does not cost the mark outright, and it is the exact opposite of the efficient route the worked chain above demonstrates: find the range of f early, and the domain of f⁻¹ is already answered before the algebra for f⁻¹(x) itself has even begun.

inverse-algebra-and-its-domain-are-independently-marked

Not an error a candidate makes so much as a fact about the mark scheme that, misunderstood, produces one: the domain of an inverse function is not a bonus tacked onto the M1 A1 for finding its formula, and it is not lost automatically if the algebra goes wrong, or gained automatically if it goes right. This course's own research verifies this directly from a real mark scheme, Jan 2023 Q1(c) (there, finding g⁻¹(x) for a fraction g rather than f⁻¹(x) for a quadratic, but the structure is identical to every example in this lesson): the scheme awards M1 A1 for finding the inverse and a SEPARATE, independent B1 for the domain. Treating the domain as "the last part of finding the inverse" rather than its own distinct, separately-earned statement is the root cause behind both traps above.

inverse-rearrangement-shown-as-calculator-output-with-no-algebra

WMA13 carries an explicit, paper-wide, verbatim rubric on exactly this kind of algebra: certain questions state "In this question you must show all stages of your working. Solutions relying entirely on calculator technology are not acceptable" (verified verbatim from the January 2023 question paper, and confirmed recurring across multiple series). Finding f⁻¹(x) is a rearrangement — precisely the kind of step a calculator or an equation solver can shortcut silently. Where a question carries this instruction, arriving at a correct f⁻¹(x) with no visible swap-and-rearrange working risks the method mark outright, whatever the final expression says.

inverse-answer-attached-to-the-wrong-function-letter

A distinct, purely notational failure mode, confirmed directly in a real mark scheme rather than an examiner report. When a question defines more than one function — exactly the situation in every worked example, chain drill and marked solution in this lesson, since composition and inverses are normally taught and tested together — a correctly-derived inverse can still lose its accuracy mark if it is labelled with the wrong function's letter. Verified verbatim, Jan 2023 Q1(c), which defines both f(x) and g(x) and asks specifically for g⁻¹(x): the scheme's note on the A1 mark reads "Condone y = (3−x)/(2x) o.e and even g⁻¹ = (3−x)/(2x) but NOT f⁻¹ = (3−x)/(2x) o.e." Right rearrangement, right final expression, zero marks — because f is the letter more often inverted in practice, and writing "f⁻¹(x) = ..." out of habit when the question asked for g⁻¹(x) is an easy, silent slip that the algebra itself never reveals.

The card

|t| = t if t ≥ 0, −t if t < 0. Never negative. Two branches, always both checked.
y=af(x): (p,q)→(p,aq). y=f(x)+a: →(p,q+a). y=f(x+a): →(p−a,q). y=f(ax): →(p/a,q).
y = |f(x)|: reflects parts below the x-axis upward; parts already ≥0 unchanged.
y = f(|x|): keeps x ≥ 0; discards x < 0 and rebuilds it as a mirror of x ≥ 0.
Combined: one input change + one/two output changes. f(ax+b) is NOT required.
|A| = B: solve A=B and A=−B, check each root against the case it came from.

Why it works — Why the input change and the output change can be done in either order

Take a concrete combination — $y = 3f(x+2) - 1$ — and track a single point $(p, q)$ on $y = f(x)$ through it two different ways, to see whether the order in which the two moves are made actually matters. Route one, input first: apply the translation $x \to x+2$ first, which by the rule above moves $(p, q)$ to $(p - 2, q)$; then apply the two output changes in the order they are written, $\times 3$ then $-1$: $(p-2, q) \to (p-2, 3q) \to (p-2, 3q-1)$. Route two, output first: apply $\times 3$ then $-1$ to the ORIGINAL point first, giving $(p, 3q-1)$ on $y = 3f(x)-1$; then apply the same translation to that: $(p-2, 3q-1)$. Both routes land on exactly the same point, $(p-2, 3q-1)$ — and substituting directly into the formula confirms it independently: $3f((p-2)+2) - 1 = 3f(p) - 1 = 3q - 1$. The reason this works is not a coincidence of this particular example. The input change touches only the x-coordinate; the two output changes touch only the y-coordinate. Operations that act on different coordinates cannot interfere with each other, so it makes no difference which is imagined as happening "first" — the combined picture is the same either way. Contrast this with the TWO output changes done together, $\times 3$ then $-1$: swap THEIR order and the result changes. $3(f(x) - 1) = 3f(x) - 3$, not $3f(x) - 1$ — a different function entirely. Two operations on the same coordinate generally do not commute; an input change and an output change, touching different coordinates, always do. This is also the real reason $y = f(ax+b)$ is excluded from this spec: it is two input changes sharing one coordinate, which — like the two output changes above — do not commute, and getting the order and the exact translation amount right together is a genuinely harder skill than anything this paper's guidance actually lists.

Traps — 5

positive-x-branch-not-mirrored-for-f-of-modulus-x

Confirmed directly on a real f(|x|) reflective-symmetry question: "very few students understood that... the negative x part of the graph is a reflection of the positive x part... so [the second solution] is also a solution" (Jun 2023, Q6(d)). The mechanism is the one this lesson derives rather than states: for x < 0, f(|x|) equals f evaluated at the corresponding positive value, so the entire left-hand branch is a rebuild, not a survival of the curve's original left half.

only-one-branch-of-modulus-equation-solved

A confirmed, recurring pattern across at least three series (Jan 2022 Q7, Oct 2020 Q4(c), Jun 2023 Q6(c)): candidates solve only one of the two cases a modulus equation or inequality actually splits into, or find both critical values correctly but then fail to select the correct combined region at the end. Jan 2022's own examiner report calls the specific question this pattern showed up on "very demanding by many." The fix is structural, not a matter of care: always write out both cases before solving either.

critical-values-correct-but-wrong-region-selected

Confirmed by the mark scheme's own design, not just examiner commentary: Pearson attaches a dedicated, separately-earned mark to the single step of choosing which region satisfies a modulus inequality — distinct from, and dependent on, the earlier mark(s) that found the critical values in the first place. Verbatim from a plain linear modulus-inequality question of exactly the shape this lesson teaches: "dM1: Selects outside region for their critical values... It is dependent upon having attempted to solve one correct equation" (Jan 2022, Q7(b)). On a harder WMA13 question that combines the modulus with a second technique — so that TWO separate method marks, not one, precede this same step — the dependency escalates and the mark scheme uses a doubly-dependent ddM1 instead, verified verbatim elsewhere in this paper's mark schemes: "ddM1: Chooses the outside region for their values... It is dependent on both previous method marks" (Oct 2023, Q9(c)). Either way, the region-selection step is never folded into the mark that found the values — it is always its own separate checkpoint, exactly because getting the two boundary values right and then picking the wrong side of them is common enough to need one.

combined-transformation-only-partially-applied

Confirmed on a real combined-transformation coordinate question: common errors included giving "an error of (−8, −3) or (−8, −6)" and, separately, that "a common error was to subtract 1 from the x coordinate to give (−5, −9)" (Jan 2024, Q1). Both are the same underlying failure: applying only one of the two combined moves correctly, or applying the right move to the wrong coordinate. The check from the mechanism above is the direct fix — track one point through both moves independently and confirm each one only touched the coordinate it was supposed to.

calculator-used-with-no-algebraic-method-shown

WMA13 carries an explicit no-calculator-methods rubric on specific questions, verified verbatim from the January 2023 question paper: "In this question you must show all stages of your working. Solutions relying entirely on calculator technology are not acceptable." The facts bank confirms this same requirement recurred specifically on the January 2024 modulus simultaneous-equation question (Q8) — the exact topic this lesson covers. Treat the rubric as a cue that the algebra itself, not just the final numbers, is what earns the marks on that question.

The card

|t| = t if t ≥ 0, −t if t < 0. Never negative. Two branches, always both checked.
y=af(x): (p,q)→(p,aq). y=f(x)+a: →(p,q+a). y=f(x+a): →(p−a,q). y=f(ax): →(p/a,q).
y = |f(x)|: reflects parts below the x-axis upward; parts already ≥0 unchanged.
y = f(|x|): keeps x ≥ 0; discards x < 0 and rebuilds it as a mirror of x ≥ 0.
Combined: one input change + one/two output changes. f(ax+b) is NOT required.
|A| = B: solve A=B and A=−B, check each root against the case it came from.

Why it works — Why the input change and the output change can be done in either order

Take a concrete combination — $y = 3f(x+2) - 1$ — and track a single point $(p, q)$ on $y = f(x)$ through it two different ways, to see whether the order in which the two moves are made actually matters. Route one, input first: apply the translation $x \to x+2$ first, which by the rule above moves $(p, q)$ to $(p - 2, q)$; then apply the two output changes in the order they are written, $\times 3$ then $-1$: $(p-2, q) \to (p-2, 3q) \to (p-2, 3q-1)$. Route two, output first: apply $\times 3$ then $-1$ to the ORIGINAL point first, giving $(p, 3q-1)$ on $y = 3f(x)-1$; then apply the same translation to that: $(p-2, 3q-1)$. Both routes land on exactly the same point, $(p-2, 3q-1)$ — and substituting directly into the formula confirms it independently: $3f((p-2)+2) - 1 = 3f(p) - 1 = 3q - 1$. The reason this works is not a coincidence of this particular example. The input change touches only the x-coordinate; the two output changes touch only the y-coordinate. Operations that act on different coordinates cannot interfere with each other, so it makes no difference which is imagined as happening "first" — the combined picture is the same either way. Contrast this with the TWO output changes done together, $\times 3$ then $-1$: swap THEIR order and the result changes. $3(f(x) - 1) = 3f(x) - 3$, not $3f(x) - 1$ — a different function entirely. Two operations on the same coordinate generally do not commute; an input change and an output change, touching different coordinates, always do. This is also the real reason $y = f(ax+b)$ is excluded from this spec: it is two input changes sharing one coordinate, which — like the two output changes above — do not commute, and getting the order and the exact translation amount right together is a genuinely harder skill than anything this paper's guidance actually lists.

Traps — 5

positive-x-branch-not-mirrored-for-f-of-modulus-x

Confirmed directly on a real f(|x|) reflective-symmetry question: "very few students understood that... the negative x part of the graph is a reflection of the positive x part... so [the second solution] is also a solution" (Jun 2023, Q6(d)). The mechanism is the one this lesson derives rather than states: for x < 0, f(|x|) equals f evaluated at the corresponding positive value, so the entire left-hand branch is a rebuild, not a survival of the curve's original left half.

only-one-branch-of-modulus-equation-solved

A confirmed, recurring pattern across at least three series (Jan 2022 Q7, Oct 2020 Q4(c), Jun 2023 Q6(c)): candidates solve only one of the two cases a modulus equation or inequality actually splits into, or find both critical values correctly but then fail to select the correct combined region at the end. Jan 2022's own examiner report calls the specific question this pattern showed up on "very demanding by many." The fix is structural, not a matter of care: always write out both cases before solving either.

critical-values-correct-but-wrong-region-selected

Confirmed by the mark scheme's own design, not just examiner commentary: Pearson attaches a dedicated, separately-earned mark to the single step of choosing which region satisfies a modulus inequality — distinct from, and dependent on, the earlier mark(s) that found the critical values in the first place. Verbatim from a plain linear modulus-inequality question of exactly the shape this lesson teaches: "dM1: Selects outside region for their critical values... It is dependent upon having attempted to solve one correct equation" (Jan 2022, Q7(b)). On a harder WMA13 question that combines the modulus with a second technique — so that TWO separate method marks, not one, precede this same step — the dependency escalates and the mark scheme uses a doubly-dependent ddM1 instead, verified verbatim elsewhere in this paper's mark schemes: "ddM1: Chooses the outside region for their values... It is dependent on both previous method marks" (Oct 2023, Q9(c)). Either way, the region-selection step is never folded into the mark that found the values — it is always its own separate checkpoint, exactly because getting the two boundary values right and then picking the wrong side of them is common enough to need one.

combined-transformation-only-partially-applied

Confirmed on a real combined-transformation coordinate question: common errors included giving "an error of (−8, −3) or (−8, −6)" and, separately, that "a common error was to subtract 1 from the x coordinate to give (−5, −9)" (Jan 2024, Q1). Both are the same underlying failure: applying only one of the two combined moves correctly, or applying the right move to the wrong coordinate. The check from the mechanism above is the direct fix — track one point through both moves independently and confirm each one only touched the coordinate it was supposed to.

calculator-used-with-no-algebraic-method-shown

WMA13 carries an explicit no-calculator-methods rubric on specific questions, verified verbatim from the January 2023 question paper: "In this question you must show all stages of your working. Solutions relying entirely on calculator technology are not acceptable." The facts bank confirms this same requirement recurred specifically on the January 2024 modulus simultaneous-equation question (Q8) — the exact topic this lesson covers. Treat the rubric as a cue that the algebra itself, not just the final numbers, is what earns the marks on that question.

The card

sec θ = 1/cos θ; cosec θ = 1/sin θ; cot θ = cos θ/sin θ = 1/tan θ. Not the other pairing.
A coefficient in front stays outside the reciprocal: 3 cosec θ = 3/sin θ, never 1/(3 sin θ) — a real, mark-scheme-penalised WMA13 slip.
sec²θ ≡ 1 + tan²θ; cosec²θ ≡ 1 + cot²θ — both from dividing cos²θ+sin²θ≡1 by cos²θ or sin²θ. Memorise; not in the booklet.
sec θ, cosec θ: |value| ≥ 1 always; undefined where cos θ, sin θ = 0. cot θ: any real value; undefined where sin θ = 0.
arcsin: domain [−1,1], range −90°–90°. arccos: domain [−1,1], range 0°–180°. arctan: domain ℝ, range −90°<y<90°, open.
Inverse exists only where the original is one-one — that's why each range above is a restriction, not the full curve.

Why it works — Where sec²θ = 1 + tan²θ actually comes from

Start from the one identity that is genuinely foundational — $\cos^2\theta + \sin^2\theta \equiv 1$ — and divide every term by $\cos^2\theta$ (valid wherever $\cos\theta \neq 0$, which is exactly where $\sec\theta$ is defined in the first place, so nothing is lost). $\frac{\cos^2\theta}{\cos^2\theta} + \frac{\sin^2\theta}{\cos^2\theta} \equiv \frac{1}{\cos^2\theta}$. The first term is $1$. The second term is $\left(\frac{\sin\theta}{\cos\theta}\right)^2 = \tan^2\theta$. The right-hand side is $\left(\frac{1}{\cos\theta}\right)^2 = \sec^2\theta$. So the division produces, without any further assumption, $1 + \tan^2\theta \equiv \sec^2\theta$ — the identity was already sitting inside $\cos^2\theta + \sin^2\theta \equiv 1$, waiting to be uncovered by dividing through by the right thing. Divide the SAME starting identity by $\sin^2\theta$ instead, and the companion identity falls out the same way: $\frac{\cos^2\theta}{\sin^2\theta} + \frac{\sin^2\theta}{\sin^2\theta} \equiv \frac{1}{\sin^2\theta}$, giving $\cot^2\theta + 1 \equiv \text{cosec}^2\theta$. Both identities are the same starting fact, viewed through two different divisions — which is also why they are not independent things to memorise separately so much as one thing to know how to re-derive twice. This matters for exam technique specifically: general marking guidance for this paper records that where a method uses a learned formula or identity, examiners advise quoting it BEFORE substituting values, because the method mark can otherwise only be inferred from correct working and is then vulnerable to any slip inside that working. Writing "$\sec^2\theta \equiv 1 + \tan^2\theta$" as its own line, before using it, is the safer habit — and it is genuinely the same advice this course's own quadratic-functions lesson gives for quoting the quadratic formula before substituting into it.

Traps — 7

cosec-written-as-reciprocal-of-cosine

Confirmed directly on a real WMA13 identity question: candidates "struggled writing cosec θ = 1/cos θ" (Jan 2022, Q2) — reaching for cosine because of the shared first syllable, when the correct pairing is cosec θ = 1/sin θ. The same report records the flip side too: "even some of the weakest candidates were able to earn a mark for stating the identity cosec θ = 1/sin θ" (Jan 2022, Q2) — this is a genuinely accessible mark, and losing it to a mispairing rather than a harder step is the most avoidable loss in this whole topic. Check any pairing by confirming it multiplies to 1: cosec θ · sin θ = 1, not cosec θ · cos θ.

coefficient-folded-into-the-reciprocal

A second, distinct error on the exact same real WMA13 question named above, re-verified directly against the primary mark scheme (not just the examiner-report prose): the same series' report records candidates "sometimes" writing "3 cosecθ = 1/ 3sinθ" (Jan 2022, Q2) — folding the coefficient into the denominator alongside sin θ, rather than leaving it multiplying the finished reciprocal. The real mark scheme is explicit that this scores nothing for the identity mark: "Note that 3cosecθ = 1/(3sinθ) is B0 unless there is an aside that does state cosecθ = 1/sinθ." The fix is mechanical, not a new idea: reciprocate the trig function alone first — cosec θ = 1/sin θ — then multiply by whatever coefficient sits in front, e.g. 3 cosec θ = 3 × (1/sin θ) = 3/sin θ. The coefficient never moves inside the denominator next to sin θ.

quadratic-in-tan-or-cot-missing-a-second-solution

Documented on a different WMA13 trig equation, not one reducing via sec²θ or cosec²θ specifically, but the exact same shape of failure: "it was disappointing that many candidates failed to identify the second solution here to gain the full marks" (Jun 2022, Q7). Every equation in this lesson that reduces to a quadratic in tan θ or cot θ produces TWO values of the trig ratio, and each of those typically produces TWO angles in a full 0°–360° range — up to four solutions from one equation. Stopping after the first branch, or after one solution within a branch, is this trap in miniature.

solved-ratio-mistaken-for-the-angle-itself

Confirmed on a WMA13 question that solved for sin x rather than x directly: "a few interpreted the solution to their equation as being the value for x rather than for sin x, thereby losing both marks" (Jan 2024, Q6(c)). The identical failure is available here with tan θ or cot θ: solving cot²θ − cotθ − 6 = 0 gives values of cot θ (3 and −2), not values of θ. The step from "cot θ = 3" to "θ = 18.4° or 198.4°" is not optional bookkeeping — it is a required, markable step that a genuine number of candidates skip on the analogous sin-based question.

identity-not-quoted-before-use

General marking guidance for this paper states the advice explicitly: "where a method involves using a formula that has been learnt, the advice given in recent examiners' reports is that the formula should be quoted first... where the formula is not quoted, the method mark can be gained by implication from correct working with values but may be lost if there is any mistake in the working" (Jan 2023 general marking guidance, cross-checked against Oct 2023 and Jun 2022). Applied here: writing "sec²θ ≡ 1 + tan²θ" as its own line before substituting is what makes the method mark secure even if the next line contains a slip — skipping straight to the substituted equation makes the method mark depend on everything downstream staying correct.

restricted-domain-of-the-inverse-left-unstated

Confirmed as the single most consistently dropped mark on real inverse-function questions on this paper: "despite being a standard question, the majority of candidates still failed to state the domain for their inverse function and did not achieve the B mark" (Jun 2022, Q2), and independently, "a majority of candidates were not aware that the domain was required, and therefore by far the most common score seen was 2/3" (Jan 2022, Q6). Those quotes are about inverse functions generally (spec 1.2), not about arcsin/arccos/arctan specifically — but the failure is identical in shape: a question asking for the domain OR range of arcsin, arccos or arctan is asking for exactly the kind of stated restriction these reports record candidates omitting, even when the rest of the working is correct.

missing-intermediate-line-in-a-prove-that

A general pattern confirmed across a WMA13 series' full set of reports: "it was noticeable in this series that many candidates omitted important lines when proceeding to the given solution resulting in the loss of some vital marks" (Jan 2022, general summary). Applied to this topic specifically: a "show that sec²θ ≡ 1 + tan²θ" question is asking for the division-by-cos²θ derivation shown above written out, not just asserted — jumping from cos²θ + sin²θ ≡ 1 straight to the answer, with the dividing-through step invisible, is exactly the omission these reports repeatedly penalise.

The card

sin(A±B) ≡ sinAcosB ± cosAsinB; cos(A±B) ≡ cosAcosB ∓ sinAsinB; tan(A±B) ≡ (tanA±tanB)/(1∓tanAtanB) — all on the formula sheet.
Set B=A in each: sin2A ≡ 2sinAcosA; cos2A ≡ cos²A−sin²A ≡ 2cos²A−1 ≡ 1−2sin²A; tan2A ≡ 2tanA/(1−tan²A) — none on the sheet.
Pick the cos2A form matching what's already there: sin² present → 1−2sin²A; cos² present → 2cos²A−1; factoring a difference of squares → cos²A−sin²A.
Solving sin2θ=k or cos2θ=k directly: widen the interval to match 2θ before solving, then halve every answer and re-check against the ORIGINAL interval.
Proving an identity: show every line, including the double-angle substitution itself — a matching final line is not evidence it was actually reached.

Why it works — Deriving sin2A and cos2A — the same substitution, twice

Take the compound angle formula for sine, $\sin(A+B) \equiv \sin A\cos B + \cos A\sin B$, and substitute $B=A$ — not a new formula, the exact same one with both letters told to be the same angle: $\sin(A+A) \equiv \sin A\cos A + \cos A\sin A = 2\sin A\cos A$. So $\sin 2A \equiv 2\sin A\cos A$ — the double angle formula for sine is nothing but the sum formula with $B$ set equal to $A$, and every line of the derivation is already available from a formula printed on the sheet. Do the identical substitution to the compound-angle formula for cosine, $\cos(A+B) \equiv \cos A\cos B - \sin A\sin B$: $\cos(A+A) \equiv \cos A\cos A - \sin A\sin A = \cos^2 A - \sin^2 A$, so $\cos 2A \equiv \cos^2 A - \sin^2 A$ is the first, most direct double-angle form for cosine — and unlike sine, cosine doesn't stop there, because $\cos^2 A - \sin^2 A$ can be rewritten twice more using the one identity relating the two squares, $\cos^2 A + \sin^2 A \equiv 1$ (spec 2.2, memorised). Eliminate $\sin^2 A$ by substituting $\sin^2 A \equiv 1-\cos^2 A$: $\cos 2A \equiv \cos^2 A - (1-\cos^2 A) = 2\cos^2 A - 1$. Eliminate $\cos^2 A$ instead, substituting $\cos^2 A \equiv 1-\sin^2 A$: $\cos 2A \equiv (1-\sin^2 A) - \sin^2 A = 1-2\sin^2 A$. All three — $\cos^2 A - \sin^2 A$, $2\cos^2 A - 1$, $1-2\sin^2 A$ — are the SAME quantity, not three formulae to memorise separately: one formula wearing three algebraically equivalent outfits, and which one is useful depends entirely on what the rest of the question already has sitting in it (the next block gives the actual rule for choosing). None of $\sin 2A$, any form of $\cos 2A$, or $\tan 2A$ is printed in the formula booklet — every one of them has to be rebuildable from the compound-angle formula that IS printed, on demand, exactly as just shown.

Traps — 5

wrong-double-angle-cosine-form

Confirmed verbatim on a real WMA13 exam question of exactly this lesson's worked-chain shape — "Solve, for 0° ⩽ x < 360°, the equation 2cos2x = 7cosx" (Oct 2020, Q1, 5 marks; NOT a "prove"/"show that" question — an equation to solve by the same substitute-then-3TQ route as this lesson's own worked chain). The examiner report records it as "generally well done with about two-thirds of the candidates scoring full marks" — so this is a minority-but-real error, not the modal one — "where errors occurred, these were mainly writing cos 2x as cos²x − 1 or 1 − cos²x, or incorrectly stating 4 cos²x − 2 = 2 cos x as the first step, implying an incorrect identity due to a lack of brackets": half-remembering one of the three real forms (cos²x−sin²x, 2cos²x−1, 1−2sin²x) and dropping the factor of 2, or — the second, distinct error — writing 2×2cos²x−1 without the bracket around (2cos²x−1) it needs, which the real mark scheme flags by name: "2 × 2cos²x − 1 = 7cosx is M0 unless the correct identity has been previously stated or recovery occurs." Both errors cost the same mark for the same underlying reason: the coefficient of 2 outside the bracket has to survive the substitution, and it is exactly that 2 — as a factor, or as a bracket protecting it — that goes missing.

incomplete-working-on-a-prove-that

The single most repeated failure mode on this topic across multiple series, and not a maths error at all: "It was noticeable in this series that many candidates omitted important lines when proceeding to the given solution resulting in the loss of some vital marks" (Jan 2022, general summary). The identical discipline is documented independently outside this topic too — a Jan 2022 report on a differentiation "show that" question notes "many good candidates lost marks here for merely writing down the given answer from a correct dx/dy without any intermediate lines," cited here only as evidence the same general marking principle (every line on the way to a GIVEN answer is part of what is marked, not optional scaffolding) recurs across different question types, not as a trig-specific incident in its own right. On a "prove" or "show that" instruction, the double-angle substitution line itself is one of the lines being marked.

dividing-away-a-common-trig-factor

A genuinely common algebra trap once a double-angle form has been substituted into an equation like sin2θ = sinθ: 2sinθcosθ = sinθ looks, on sight, like something to simplify by dividing both sides by sinθ, leaving 2cosθ = 1. That step is illegal exactly when sinθ could be zero, and it silently discards every solution where sinθ = 0 actually satisfies the ORIGINAL equation. Rearrange to zero and factorise instead — sinθ(2cosθ − 1) = 0 — so both branches, sinθ = 0 and cosθ = ½, survive. The rule isn't specific to trig: never divide an equation by an expression that could itself be zero, only ever factor it out.

unhelpful-cos2a-form-chosen

All three forms of cos2A are equally TRUE, so picking the 'wrong' one never produces a wrong answer — it produces an equation that still has two different trig functions mixed together, when a better choice would have collapsed it to one. Substituting cos²A − sin²A into an equation that is otherwise entirely in sinA doesn't simplify anything; it introduces a fresh cos²A term that then has to be eliminated with sin²A + cos²A ≡ 1 anyway — the same identity a better initial choice would have used once, not twice. Before substituting, check what function the rest of the equation is already written in, and match the cos2A form to it.

half-angle-substitution-misread

"Half the angle" and "half the value" are not the same operation, and the notation makes them easy to blur: sin(θ/2) means the sine of half the angle, an entirely different number from (sinθ)/2, half the sine's value. The half-angle identities in this lesson — sin²(θ/2) ≡ (1−cosθ)/2 and its cosine equivalent — only come out right if the /2 is read as dividing the ANGLE going into the substitution A=θ/2, not as an operation performed on sinθ or cosθ afterwards. Write the substitution out explicitly ('let A = θ/2') before touching the double-angle formula, rather than trying to halve an angle and a trig function in the same mental step.

The card

a cosθ + b sinθ ≡ R cos(θ∓α): expand, compare coefficients, then R = √(a²+b²), tanα = b/a (or a/b for a sin-led form).
Signs pick the form: a,b both + → cos(θ−α); a+ b− → cos(θ+α). Check Rcosα and Rsinα both come out positive.
Max value R at θ=α (cos-led) or where the argument is 90° (sin-led); min value −R half a turn later.
cos(θ−α)=k, −R<k<R, always has TWO solutions in a 360° sweep — find both, then convert back to θ.
Give exactly what's asked: if the question wants the exact value of R, a decimal loses that mark even if correct; if it wants α in radians, degrees loses that mark too, same angle or not.
Not on the formula sheet: build it fresh from cos(A∓B), which is.

Why it works — Deriving R and α — matching coefficients, not memorising a formula

Take the target directly: write $a\cos\theta + b\sin\theta \equiv R\cos(\theta - \alpha)$ for some $R > 0$ and $\alpha$ still to be found, and expand the right-hand side using the compound angle formula recapped above: $R\cos(\theta - \alpha) = R\cos\theta\cos\alpha + R\sin\theta\sin\alpha = (R\cos\alpha)\cos\theta + (R\sin\alpha)\sin\theta$. This has to equal $a\cos\theta + b\sin\theta$ for every value of $\theta$, which is only possible if the two expressions have the same coefficient of $\cos\theta$ and the same coefficient of $\sin\theta$ separately — an identity, not an equation to solve for one value of $\theta$. That gives two simultaneous equations in $R$ and $\alpha$: $R\cos\alpha = a$ and $R\sin\alpha = b$. Square both and add: $R^2\cos^2\alpha + R^2\sin^2\alpha = a^2 + b^2$, and since $\cos^2\alpha + \sin^2\alpha \equiv 1$ (memorised, not on the sheet), the left side collapses to $R^2$ — so $R = \sqrt{a^2 + b^2}$, taking the positive root because $R > 0$ was fixed as a condition from the start. Divide instead of adding: $\frac{R\sin\alpha}{R\cos\alpha} = \frac{b}{a}$, so $\tan\alpha = \frac{b}{a}$, giving $\alpha = \arctan\left(\frac{b}{a}\right)$ once a quadrant is chosen. Nothing here is a new formula to learn — $R\cos\alpha = a$ and $R\sin\alpha = b$ are exactly the Pythagoras-and-tangent pair used to convert between a point $(a, b)$ and its distance-and-angle description, applied to the two coefficients rather than to a point on a graph.

Traps — 7

missing-second-solution-in-range

Confirmed verbatim on a real "solve in a given range" trig question: "It was disappointing that many candidates failed to identify the second solution here to gain the full marks" (Jun 2022, Q7). Cosine (and sine) are two-to-one over a full 360° turn for any target value strictly between −1 and 1, so an equation of the form cos(θ−α) = k always has a second solution in a 360° interval — finding the first one algebraically correctly is not the same as finding the answer.

degrees-radians-mismatch

A documented pattern across five separate series (Oct 2020, Jan 2022, Jun 2022, Jun 2023, Jan 2024): candidates lose the final accuracy mark on a 'solve in a given range' trig equation by giving the answer in degrees when the question specified radians, or vice versa. The interval itself is the tell — 0° ≤ θ < 360° wants degrees, 0 ≤ θ < 2π wants radians — and it is worth checking that match before writing a single answer down, not after.

exact-R-decimal-rounded-or-alpha-given-in-wrong-angle-unit

This trap sits earlier than the one above — on the R and α accuracy marks THEMSELVES, while building the harmonic form, not on the final solved θ. A real WMA13 harmonic-form question can specify both a required FORM for R (exact, not decimal) and a required UNIT for α (radians, not degrees), and a numerically correct value in the wrong form or unit earns nothing on that mark. Confirmed verbatim on a real question testing exactly this spec point, Oct 2020 Q7(a): "Express cos x + 4 sin x in the form R cos(x − α) where R > 0 and 0 < α < π/2. Give the exact value of R and give the value of α, in radians, to 3 decimal places." The real mark scheme is explicit on both counts. For R = √17: "Condone R = ±√17 (Do not allow decimals for this mark...)" — a correct decimal such as 4.123 scores zero on that mark, because the question asked for the EXACT value, and √17 is not the same answer as its rounded decimal for marking purposes. For α = awrt 1.326: "Note that the degree equivalent α = awrt 75.96° is A0" — the identical angle, correctly computed, in the wrong unit, still earns nothing. Read the question’s own instructions on form and unit before writing R and α down, not after computing them the way that feels automatic.

extra-out-of-range-solutions-kept

The same five-series pattern's third face: including a solution generated by the ±360n general form that actually falls outside the stated interval once it is written out. Generating θ = α ± cos⁻¹(k) + 360n is the right method; the last step — checking every value produced against the stated bounds before it goes in the answer — is not optional bookkeeping, it is the mark.

stops-after-the-shifted-angle-never-recovers-theta

The specific error this lesson's marked solution builds around — solving for (θ−α) or (θ+α) and reporting that as if it answered the question — is one instance of a broader, independently documented failure: stopping at an intermediate variable instead of converting back to the one actually asked about. A different WMA13 equation type shows the identical pattern one layer earlier, at the point of substitution rather than shifting: "A few interpreted the solution to their equation as being the value for x rather than for sin x, thereby losing both marks" (Jan 2024, Q6(c)). Different equation, same discipline missing: know which variable you have just solved for, and know which one the question asked for.

minimising-the-original-expression-confused-with-minimising-the-wave

When a harmonic-form term sits inside a denominator (or is being subtracted from something), minimising the WHOLE expression can require MAXIMISING the wave, not minimising it — the two optimisations point in opposite directions. Confirmed directly on a real paper, where a fraction of the shape (constant)/(constant + wave) had to be minimised: "this did cause some confusion for many in that they did not realise that for the fraction to be a minimum the denominator had to be a maximum" (Oct 2020, Q7). Before optimising anything built out of R cos(θ−α), ask explicitly which direction — max or min of the WAVE — actually produces the extreme of the thing the question asked about.

incomplete-working-on-a-prove-that-identity

Spec 2.3's own guidance names an example identity students should be able to prove — "cos x cos 2x + sin x sin 2x ≡ cos x" — and identity proofs of this kind are marked on the presence of every intermediate line, not just a correct method plus a correct final line. Confirmed, a general summary from a real series: "many candidates omitted important lines when proceeding to the given solution resulting in the loss of some vital marks" (Jan 2022). On a 'show that' or 'prove' instruction, the compound-angle expansion line itself is not optional working to skip past — it is one of the lines being marked.

The card

f'(x)/f(x) → ln|f(x)|+c. f'(x)[f(x)]ⁿ → [f(x)]ⁿ⁺¹/(n+1)+c. Both are the chain rule, reversed.
Check: does the integrand contain something's derivative sitting beside a function of that same something?
Coefficient mismatch: pull out or introduce a constant first, so the numerator becomes exactly f'(x).
sin²A, cos²A, tan²A: rewrite first — cos2A ≡ 1−2sin²A ≡ 2cos²A−1 ≡ cos²A−sin²A, or sec²A ≡ 1+tan²A. Never integrate a squared trig function directly.
Keep the modulus unless f(x) provably never changes sign. Always add +c.

Why it works — Where $\int \frac{f'(x)}{f(x)}\,dx = \ln|f(x)| + c$ actually comes from

Start from the claimed answer and differentiate it, because that is the fastest way to see why the result is true rather than just that it is. Take $y = \ln|f(x)|$ and apply the chain rule: the outer function is $\ln|u|$, the inner function is $u = f(x)$. $\frac{d}{du}[\ln|u|] = \frac{1}{u}$ is itself one of the standard results memorised from spec 5.1 (it is on the 'must memorise' list, not the sheet), so $\frac{dy}{dx} = \frac{1}{f(x)} \cdot f'(x) = \frac{f'(x)}{f(x)}$. That is the entire theorem: differentiating $\ln|f(x)|$, using nothing beyond the chain rule already known before this lesson, produces exactly $\frac{f'(x)}{f(x)}$ — so integrating $\frac{f'(x)}{f(x)}$ has to return $\ln|f(x)| + c$, because integration is defined as the operation that undoes differentiation, not as a separate technique that happens to agree with it. The modulus is not decoration: $\ln|u|$ is defined and differentiates to $\frac{1}{u}$ for every $u \neq 0$, negative values included, while $\ln(u)$ alone is undefined the moment $u$ turns negative. The modulus is what keeps the result true on the whole domain where $f(x)$ can be negative — not a convention layered on top of the theorem, but part of what makes the theorem correct in the first place.

Traps — 5

modulus-dropped-in-log-integration

Confirmed directly on a real recognition-integration question involving negative values: "many did not have the modulus and left the values as 5ln(-2) and 5ln(-4) not realising these are undefined" (Oct 2020, Q9(c)); independently confirmed in Jun 2023, Q9(c). The mechanism this lesson's own derivation makes explicit: $\ln|f(x)|$ is defined and differentiates correctly for every $f(x) \neq 0$, including negative values, while $\ln(f(x))$ alone breaks the moment $f(x)$ turns negative. Whether the modulus is essential or just safe depends entirely on whether $f(x)$ can actually be negative on the domain in question — check that before deciding it can be dropped, never assume.

constant-of-integration-omitted

Confirmed across at least two series on this exact topic: "A fair number of candidates could not be awarded the final mark, despite having obtained the correct algebraic expression, as they had forgotten to include the constant of integration" (Oct 2021, Q5(ii)); independently confirmed, "a substantial number of candidates neglected to include the constant of integration and were penalised" (Jan 2022, Q3(i)). It costs one mark on an otherwise perfect answer, on an indefinite integral specifically — a definite integral (like part (d) of the marked solution above) has no constant to lose, which is worth noticing precisely so the two cases are not confused.

double-angle-form-confused

A real double-angle "show that" question records candidates who "commonly wrote cos 2x as cos²x − 1 or 1 − cos²x" (Oct 2020, Q1) — misapplying or confusing the three equivalent double-angle-for-cosine forms with each other, or with the plain Pythagorean identity. This lesson's identity-rewrite technique for $\sin^2 x$, $\cos^2 x$ and $\tan^2 x$ depends entirely on quoting the RIGHT rearranged form with the RIGHT coefficient — get the identity wrong (as in the marked solution's own common wrong path above) and the integration method that follows can be flawless and still reach the wrong answer.

intermediate-lines-omitted-on-hence-or-show-that

Confirmed as a general pattern across multiple series, not tied to one question: "many candidates omitted important lines when proceeding to the given solution resulting in the loss of some vital marks" (Jan 2022, general summary); the same report separately records "many good candidates lost marks here for merely writing down the given answer from a correct dx/dy without any intermediate lines." A "hence, show that" part with a printed target answer — exactly the shape of part (d) in the marked solution above — is where this costs the most: every line from the antiderivative to the printed target has to actually appear.

booklet-result-not-recognised

Confirmed on a real recognition-type integral: "some students did not realise the result could be found from their Formula Booklet, and instead used a substitution" (Jun 2023, Q9(c), on ∫cosec x dx). Substitution is a fully valid, mark-scheme-credited alternative route — the method-comparison block above shows exactly that — but re-deriving from scratch under time pressure, when a result is already printed and the question does not demand a derivation, spends time and adds a line where a sign or arithmetic slip can happen that a direct quote could not have introduced.

The card

eˣ: domain ℝ, range >0, asymptote y=0 as x→−∞. ln x: domain >0, range ℝ, asymptote x=0. ln(eˣ)=x, e^(ln x)=x.
Solve e^(ax+b)=p: take ln both sides. Solve ln(ax+b)=q: apply e^(...) both sides.
y=axⁿ → log y = n log x + log a: plot log y vs log x, gradient=n, intercept=log a.
y=kbˣ → log y = x log b + log k: plot log y vs x, gradient=log b, intercept=log k.
Un-log every intercept (10^intercept) before it's a genuine constant. Gradient never needs un-logging.

Why it works — Why ln x exists at all — built on the one-one/domain-range mechanism from the functions lesson

The functions lesson established a chain of reasoning that applies to $e^x$ exactly as it applies to any other function: an inverse exists as a function if and only if the original is one-one, because a many-one function would need its inverse to hand back two different outputs for one input, which is precisely what disqualifies something from being a function at all. $e^x$ is strictly increasing on all of $\mathbb{R}$ — established two blocks above — so no two different $x$-values ever give the same $e^x$: it is one-one on its whole domain, no restriction needed (unlike the quadratics in the earlier WMA11 lesson, which needed a domain restriction before an inverse could exist at all). So $e^x$ has an inverse function, defined on all of $\mathbb{R}$, and it is given its own name rather than left as "$e^x$, walked backwards": $\ln x$. The functions lesson's domain/range-swap fact applies immediately: the domain of $e^x$ is $\mathbb{R}$ and its range is $\{y : y > 0\}$, so the domain of $\ln x$ is $\{x : x > 0\}$ and its range is all of $\mathbb{R}$ — exactly what the first prequestion question above asked for, before this mechanism had been written out. The defining cancellation follows the same pattern the spec's own guidance states for inverses in general, applied here: $\ln(e^x) = x$ for every real $x$, and $e^{\ln x} = x$ for every $x > 0$. This is not a separate rule to memorise — it is what "inverse function" means, applied to this particular pair. And it is the whole method for solving the two equation types spec 3.2 names. To solve $e^{ax+b} = p$: apply $\ln$ to both sides, and the left side collapses by the cancellation above to exactly $ax+b$, leaving $ax+b = \ln p$ to rearrange for $x$ — an ordinary linear equation, once the exponential has been undone. To solve $\ln(ax+b) = q$: apply $e^{(\ldots)}$ to both sides instead, and the left side collapses to $ax+b$, leaving $ax+b = e^q$. Concretely: $e^{5-2x} = 7 \Rightarrow 5-2x = \ln 7 \Rightarrow x = \dfrac{5-\ln 7}{2} \approx 1.53$ (3 s.f.); and $\ln(4-x) = 2 \Rightarrow 4-x = e^2 \Rightarrow x = 4-e^2 \approx -3.39$ (3 s.f.). Both moves are the same one move, in opposite directions — apply whichever function undoes the one already sitting in the equation.

Traps — 6

modelling-answer-missing-units-or-context-word

The single most consistently reported error anywhere in the WMA13 exponentials/logs archive, confirmed near-identically across at least four separate series: "the units (tonnes) were often omitted" (Oct 2021, Q3(b)); "only about half of them remembered that... they needed to also state the units" (Jan 2022, Q4(c)); "many lost the accuracy mark by either omitting the units '£' or not referring to a 'loss'" (Oct 2022, Q5(a)); "many did not gain the mark as they did not include the units (m2)" (Jan 2025, Q2(a)). A correct number with no unit or no context word attached is treated as an incomplete answer, not a minor presentation slip — write the unit or the contextual word every time a modelling question's answer is a quantity, not just an $x$.

calculator-equation-solver-used-with-no-algebra-shown

WMA13 papers carry an explicit, verbatim, paper-wide rubric on specific parts — "In this question you must show all stages of your working. Solutions relying entirely on calculator technology are not acceptable" (verified verbatim from the question paper itself) — and equations of the form $e^{ax+b}=p$ or $\ln(ax+b)=q$ are exactly where it lands. An examiner report confirms the consequence directly: "candidates should make sure that they do not resort to using a general equation solver... as this may not necessarily score full marks... it may be that full marks will not be awarded in the future, despite a correct answer" (Jun 2022, §5.3). Show the $\ln$ or $e^{(\ldots)}$ line explicitly, every time, on a question carrying this instruction.

log-base-confused-on-a-log-log-graph

Verified real context, Jun 2023 Q2: a log-log graph set explicitly in base $6$ — the examiner report records "being confused by log base 6, with some attempts to use log base 10 or 'e' and 'ln'" as the main documented error. The mechanism block above proves the gradient of a log-log graph is the same whatever base is chosen (true because the exponent $n$ in $y=ax^n$ doesn't depend on which base the logs are taken in — this is NOT true for a log-linear graph's gradient, $\log_c b$ for $y=kb^x$, which does change with the base $c$) — but the INTERCEPT is not, because it equals $\log_c(\text{constant})$ for whichever base $c$ the graph was actually plotted in. Un-logging an intercept with the wrong base — reading a base-6 intercept as if it were base 10 — produces a wrong constant even when every other step was correct.

exact-value-required-but-answer-rounded-early

Verified verbatim, from the general marking guidance itself, applying across every WMA13 topic but landing especially hard here because solving $e^{ax+b}=p$ so often produces an answer that is only exact as a logarithm: "Examiners' reports have emphasised that where, for example, an exact answer is asked for... marks will normally be lost if the candidate resorts to using rounded decimals". Where a question asks for an exact value of $k$ or $x$, stop at the $\ln(\ldots)$ or $\dfrac{\ln(\ldots)}{\ldots}$ form — a rounded decimal answers a different question than the one asked, even if a later part of the same question does then want 3 significant figures.

domain-restriction-from-a-non-positive-argument-not-explained

Confirmed real context: a temperature-decay model (Jan 2024, Q5) carries a hard lower-bound domain restriction that has to be explained using log-of-a-non-positive-number reasoning — the facts bank describes this context without quoting the examiner report verbatim on this specific point, so it is presented here as a described real question type, not a direct quote. The underlying mechanism is exact, though: $\ln(ax+b)$ is only defined where $ax+b>0$, so a model built around a $\ln(\ldots)$ term is only valid on the part of its domain where that argument stays positive — and a question that asks WHY a model breaks down, or for how long it remains valid, is asking for exactly this restriction to be identified and explained, not for more algebra on the equation itself.

log-notation-dropped-or-misplaced-when-reading-a-log-log-graph

Verified verbatim, Jun 2023 Q2(a)(i) examiner report — a distinct notation failure from the base-confusion trap above, on the very same real question: writing down the linear equation from a given log-log graph "proved to be straight forward and was generally done correctly for one mark. If this was not achieved, for example, a few students simply wrote $y = 4 - 2x$" — dropping every $\log_6$ entirely and treating the two labelled axes as if they were plain, unlogged $x$ and $y$. A second, separate slip the same report names: "the argument was often written as a superscript within the log" — writing something that reads as $\log 6^x$ rather than $\log_6 x$, putting the base in the wrong position. Neither is a calculation error — the axis labels on the graph already state the base and which variable is logged; the failure is not reading them onto the equation correctly, not anything that happens afterward.

The card

eˣ: domain ℝ, range >0, asymptote y=0 as x→−∞. ln x: domain >0, range ℝ, asymptote x=0. ln(eˣ)=x, e^(ln x)=x.
Solve e^(ax+b)=p: take ln both sides. Solve ln(ax+b)=q: apply e^(...) both sides.
y=axⁿ → log y = n log x + log a: plot log y vs log x, gradient=n, intercept=log a.
y=kbˣ → log y = x log b + log k: plot log y vs x, gradient=log b, intercept=log k.
Un-log every intercept (10^intercept) before it's a genuine constant. Gradient never needs un-logging.

Why it works — Why ln x exists at all — built on the one-one/domain-range mechanism from the functions lesson

The functions lesson established a chain of reasoning that applies to $e^x$ exactly as it applies to any other function: an inverse exists as a function if and only if the original is one-one, because a many-one function would need its inverse to hand back two different outputs for one input, which is precisely what disqualifies something from being a function at all. $e^x$ is strictly increasing on all of $\mathbb{R}$ — established two blocks above — so no two different $x$-values ever give the same $e^x$: it is one-one on its whole domain, no restriction needed (unlike the quadratics in the earlier WMA11 lesson, which needed a domain restriction before an inverse could exist at all). So $e^x$ has an inverse function, defined on all of $\mathbb{R}$, and it is given its own name rather than left as "$e^x$, walked backwards": $\ln x$. The functions lesson's domain/range-swap fact applies immediately: the domain of $e^x$ is $\mathbb{R}$ and its range is $\{y : y > 0\}$, so the domain of $\ln x$ is $\{x : x > 0\}$ and its range is all of $\mathbb{R}$ — exactly what the first prequestion question above asked for, before this mechanism had been written out. The defining cancellation follows the same pattern the spec's own guidance states for inverses in general, applied here: $\ln(e^x) = x$ for every real $x$, and $e^{\ln x} = x$ for every $x > 0$. This is not a separate rule to memorise — it is what "inverse function" means, applied to this particular pair. And it is the whole method for solving the two equation types spec 3.2 names. To solve $e^{ax+b} = p$: apply $\ln$ to both sides, and the left side collapses by the cancellation above to exactly $ax+b$, leaving $ax+b = \ln p$ to rearrange for $x$ — an ordinary linear equation, once the exponential has been undone. To solve $\ln(ax+b) = q$: apply $e^{(\ldots)}$ to both sides instead, and the left side collapses to $ax+b$, leaving $ax+b = e^q$. Concretely: $e^{5-2x} = 7 \Rightarrow 5-2x = \ln 7 \Rightarrow x = \dfrac{5-\ln 7}{2} \approx 1.53$ (3 s.f.); and $\ln(4-x) = 2 \Rightarrow 4-x = e^2 \Rightarrow x = 4-e^2 \approx -3.39$ (3 s.f.). Both moves are the same one move, in opposite directions — apply whichever function undoes the one already sitting in the equation.

Traps — 6

modelling-answer-missing-units-or-context-word

The single most consistently reported error anywhere in the WMA13 exponentials/logs archive, confirmed near-identically across at least four separate series: "the units (tonnes) were often omitted" (Oct 2021, Q3(b)); "only about half of them remembered that... they needed to also state the units" (Jan 2022, Q4(c)); "many lost the accuracy mark by either omitting the units '£' or not referring to a 'loss'" (Oct 2022, Q5(a)); "many did not gain the mark as they did not include the units (m2)" (Jan 2025, Q2(a)). A correct number with no unit or no context word attached is treated as an incomplete answer, not a minor presentation slip — write the unit or the contextual word every time a modelling question's answer is a quantity, not just an $x$.

calculator-equation-solver-used-with-no-algebra-shown

WMA13 papers carry an explicit, verbatim, paper-wide rubric on specific parts — "In this question you must show all stages of your working. Solutions relying entirely on calculator technology are not acceptable" (verified verbatim from the question paper itself) — and equations of the form $e^{ax+b}=p$ or $\ln(ax+b)=q$ are exactly where it lands. An examiner report confirms the consequence directly: "candidates should make sure that they do not resort to using a general equation solver... as this may not necessarily score full marks... it may be that full marks will not be awarded in the future, despite a correct answer" (Jun 2022, §5.3). Show the $\ln$ or $e^{(\ldots)}$ line explicitly, every time, on a question carrying this instruction.

log-base-confused-on-a-log-log-graph

Verified real context, Jun 2023 Q2: a log-log graph set explicitly in base $6$ — the examiner report records "being confused by log base 6, with some attempts to use log base 10 or 'e' and 'ln'" as the main documented error. The mechanism block above proves the gradient of a log-log graph is the same whatever base is chosen (true because the exponent $n$ in $y=ax^n$ doesn't depend on which base the logs are taken in — this is NOT true for a log-linear graph's gradient, $\log_c b$ for $y=kb^x$, which does change with the base $c$) — but the INTERCEPT is not, because it equals $\log_c(\text{constant})$ for whichever base $c$ the graph was actually plotted in. Un-logging an intercept with the wrong base — reading a base-6 intercept as if it were base 10 — produces a wrong constant even when every other step was correct.

exact-value-required-but-answer-rounded-early

Verified verbatim, from the general marking guidance itself, applying across every WMA13 topic but landing especially hard here because solving $e^{ax+b}=p$ so often produces an answer that is only exact as a logarithm: "Examiners' reports have emphasised that where, for example, an exact answer is asked for... marks will normally be lost if the candidate resorts to using rounded decimals". Where a question asks for an exact value of $k$ or $x$, stop at the $\ln(\ldots)$ or $\dfrac{\ln(\ldots)}{\ldots}$ form — a rounded decimal answers a different question than the one asked, even if a later part of the same question does then want 3 significant figures.

domain-restriction-from-a-non-positive-argument-not-explained

Confirmed real context: a temperature-decay model (Jan 2024, Q5) carries a hard lower-bound domain restriction that has to be explained using log-of-a-non-positive-number reasoning — the facts bank describes this context without quoting the examiner report verbatim on this specific point, so it is presented here as a described real question type, not a direct quote. The underlying mechanism is exact, though: $\ln(ax+b)$ is only defined where $ax+b>0$, so a model built around a $\ln(\ldots)$ term is only valid on the part of its domain where that argument stays positive — and a question that asks WHY a model breaks down, or for how long it remains valid, is asking for exactly this restriction to be identified and explained, not for more algebra on the equation itself.

log-notation-dropped-or-misplaced-when-reading-a-log-log-graph

Verified verbatim, Jun 2023 Q2(a)(i) examiner report — a distinct notation failure from the base-confusion trap above, on the very same real question: writing down the linear equation from a given log-log graph "proved to be straight forward and was generally done correctly for one mark. If this was not achieved, for example, a few students simply wrote $y = 4 - 2x$" — dropping every $\log_6$ entirely and treating the two labelled axes as if they were plain, unlogged $x$ and $y$. A second, separate slip the same report names: "the argument was often written as a superscript within the log" — writing something that reads as $\log 6^x$ rather than $\log_6 x$, putting the base in the wrong position. Neither is a calculation error — the axis labels on the graph already state the base and which variable is logged; the failure is not reading them onto the equation correctly, not anything that happens afterward.

The card

"Initial" = value at t=0, for any base: N₀aᵗ gives N₀ directly, since a⁰=1.
d/dx(aˣ) = aˣ ln a. Via e^(x ln a) + chain rule, or direct — same answer either way.
Rate of change = differentiate, THEN substitute. Never substitute alone; never a finite difference. Both score zero.
State the unit and the direction (increasing/decreasing) with every rate — a bare number is an incomplete answer.
A model breaks down where its equation needs an exponential term to be zero or negative — that's never a slip, it's the model's own predicted range running out.
A negative exponent (a⁻ᵗ) needs the chain rule too: d/dt[a⁻ᵗ] = -a⁻ᵗln a. Losing that minus sign is a real, repeated way to lose the method mark itself, not just an accuracy mark.

Why it works — Deriving d/dx(aˣ) = aˣ ln a — using both prerequisite lessons for real

Two blocks in the two prerequisite lessons hand this derivation everything it needs, used for real rather than restated. First, from the exponentials/logs lesson's own log-law recap: the power law, $\ln(x^n)=n\ln x$, applied to $y=a^x$ by taking $\ln$ of both sides — exactly the second prequestion above — gives $\ln y = \ln(a^x) = x\ln a$. Second, from that same lesson's mechanism block deriving $\ln x$ as the inverse of $e^x$: $e^{\ln y}=y$ for any $y>0$, so raising both sides of $\ln y = x\ln a$ back to a power of $e$ gives $y=e^{x\ln a}$ — a genuine rewriting of $a^x$ entirely in terms of $e$, valid because $\ln a$ is simply a fixed real number once $a$ is fixed ($a>0$, so $\ln a$ exists, whatever its own sign). Now the differentiation-rules lesson's own memorised result applies directly, with no new rule needed: $e^{kx}$ differentiates to $ke^{kx}$ for any constant $k$, and $x\ln a$ is exactly the linear form $kx$ this needs, with $k=\ln a$. So $\frac{d}{dx}\left[e^{x\ln a}\right] = \ln a \cdot e^{x\ln a}$. The last step translates back into the original variable: since $e^{x\ln a}=a^x$ — the identity that started this derivation — $\frac{d}{dx}[a^x] = a^x\ln a$, exactly the result spec 4.4 names, reached without inventing a single new rule. For a model with a fixed coefficient in front, $N=N_0a^t$, the ordinary constant-multiple rule from WMA11 carries straight through: $\frac{dN}{dt} = N_0a^t\ln a$ — the coefficient $N_0$ is simply carried along unchanged, exactly as it would be differentiating $N_0e^{kt}$.