Deep Dive: Hidden Higher Degree Equations From Differentiation

#CODEX Deep-Dive: Hidden Higher-Degree Equations From Differentiation (WMA11, spec 4.1–4.3)

Provenance (2026-09-29). Built from a direct read of research/veridian/WMA11-verified-facts.md §4.4 ("Differentiation") and its own "Lesson-audit log" entry for hidden-higher-degree-equations-from-differentiation.ts (dated 2026-09-14); the full text of the existing lesson content/veridian/curriculum/wma11/hidden-higher-degree-equations-from-differentiation.ts (806 lines); the existing signal-topic-brief-hidden-higher-degree-equations-from-differentiation.md in this same folder; codex-formula-and-method-bank.md §2.4.3 (checked for overlap before writing a line of this document); and content/veridian/specCoverage.ts for both the WMA11 §4 entries (items 1–3) and the WMA12 §2 item-1 entry (Factor/Remainder Theorem), used below to establish a curriculum-boundary claim. content/veridian/question-bank/wma11.ts was checked directly and contains no item tagged to this topic. No new primary-source PDF was opened for this pass — every exam-derived fact below is inherited from the facts bank's own two verified anchors (Oct 2021 Q2, Jan 2025 Q5(c)) and its one verified ruled-out case (Jan 2019 Q6); nothing here claims a new primary-source extraction. Not affiliated with or endorsed by Pearson Edexcel.

#0. What this document adds, and what it deliberately does not repeat

The existing lesson already carries a full teach sequence, a worked-chain, a chain-drill, an 11-mark marked solution, four MCQs and a five-item trap taxonomy, all built on real WMA11 findings. The SIGNAL brief already states the frequency honestly ("seen in 2 of 5 series checked"), the two anchors' mark-tariff shapes, and the confirmed trap set with sourcing. Neither document derives the underlying reason the pattern exists in the first place, or where its own boundaries are. This document is additive in four specific ways:

1. A derived necessary-and-sufficient condition on the original function f(x) for "differentiate, set equal to a value" to produce this exact disguise — the lesson and formula bank both assert the test ("no x³ or x¹ term") but neither derives why differentiation of a WMA11-scope function lands there.
2. A curriculum-boundary argument for why the disguise is always a quadratic, never a cubic, grounded in where the Factor Theorem actually sits in this course's own spec data (WMA12, not WMA11) — not previously stated anywhere in this topic's existing material.
3. A structural account of why Jan 2019 Q6 fails to produce the disguise — the facts bank and lesson both record that it was ruled out, but neither explains, in general terms, what property of that question's f(x) rules it out, or gives the test a name that would let a student recognise the same non-example in a different disguise.
4. A small number of fresh, hand-checked micro-examples, using numbers not already used anywhere in the lesson, illustrating each branch of the condition in (1) — not full past-paper-length questions.

Everything below is either a direct algebraic derivation (independently checkable) or a claim traceable to the sources named in the provenance line above. Where a claim is my own reasoning rather than a sourced exam finding, it is labelled as such, matching this lesson's own convention of separating reasoning from citation.

#1. The mechanism, restated precisely

Spec 4.2 covers differentiating $x^n$ for rational $n$; spec 4.1 explicitly excludes the chain rule at this stage. So every term this topic can ever produce is a single power of $x$, differentiated on its own: $\frac{d}{dx}(ax^n) = anx^{n-1}$. Given $f(x) = \sum_i a_i x^{n_i}$ (a finite sum of such terms, possibly plus a constant), the question sets $f'(x) = k$ for some target $k$, which rearranges to

$$f'(x) - k = 0, \quad \text{i.e.} \quad \sum_i a_i n_i \, x^{n_i - 1} - k = 0.$$

If any $n_i - 1$ is negative, this is not yet a polynomial equation. Multiplying through by $x^m$ for the smallest even $m$ that clears every negative exponent turns it into one — and it is this multiplication, not the differentiation itself, that can jump the visible degree of the equation up sharply, which is the entire reason a "differentiation" question can end up needing a "solve the quartic" skill with no warning in its own name.

#2. The parity theorem: exactly when the result hides a quadratic in x²

Claim. Suppose every power of $x$ appearing in $f(x)$ is an odd integer (positive or negative), with the possible addition of a constant term. Then every term of $f'(x)$ has an even integer power of $x$, and multiplying $f'(x) - k = 0$ through by any even power of $x$ preserves that — so the resulting equation, however high its visible degree, is expressible as a polynomial in $u = x^2$.

Derivation. For a term $a x^n$ with $n$ an odd integer, $\frac{d}{dx}(ax^n) = an\,x^{n-1}$, and $n-1$ is even whenever $n$ is odd — this holds for positive odd $n$ (5→4, 3→2, 1→0) exactly as it does for negative odd $n$ ($-1 \to -2$, $-3 \to -4$). A constant term differentiates to $0$ and contributes nothing. So $f'(x)$ is a sum of terms with even (possibly negative) exponents. The target $-k$ sits at exponent $0$, itself even. Multiplying the whole equation by $x^m$ for even $m$ shifts every exponent by $m$ — even plus even is even — so parity survives the fraction-clearing step. A polynomial in which every surviving exponent of $x$ is even is, by definition, a polynomial in $u = x^2$ (since $x^{2j} = u^j$ for every integer $j$, positive or negative once cleared).

Checked against both real anchors. Oct 2021 Q2's $f(x) = 3x^5 + 4x^3 - x + 5$ has powers $5, 3, 1$ — all odd — plus a constant; $f'(x) = 15x^4 + 12x^2 - 1$ has powers $4, 2, 0$, all even, exactly as the theorem predicts, and no fraction-clearing was even needed. Jan 2025 Q5's $f(x) = 4x^3 + 2x^{-1} + 9$ has powers $3, -1$ — both odd — plus a constant; $f'(x) = 12x^2 - 2x^{-2}$ has powers $2, -2$, both even; clearing by $x^2$ (even) preserves that, landing on $12x^4 - kx^2 - 2 = 0$. Both of this topic's only two confirmed real anchors satisfy the odd-powers-plus-constant condition exactly, and both land on an even-powers-only derivative for exactly the reason the derivation above gives.

A refinement worth naming: how many terms survive. The theorem above says when an equation is expressible as a polynomial in $u$; it does not by itself guarantee a genuine three-term quadratic. If $f'(x)$ contributes only one nonzero even-power term (say a single $x^4$ term), then $f'(x) = k$ is a direct "$x^4 = \text{constant}$" statement — solvable by taking a fourth root, with no quadratic step and nothing to disguise. The disguise only bites when $f'(x)$ contributes two distinct nonzero even-power terms (as both real anchors do — $x^4 \,\&\, x^2$, or $x^2 \,\&\, x^{-2}$): combined with the constant $-k$, that is exactly three genuinely different powers of $u$ ($u^2, u^1, u^0$), which is what a three-term quadratic in $u$ actually requires. This term-count refinement is my own reasoning, built on the theorem above; it is not itself stated in the facts bank, but it is what makes the difference between a real anchor and the non-example in §4 precise rather than just observed.

#3. Why the disguise is always a quadratic, and never a cubic, at WMA11

Both confirmed real anchors hide a quadratic in $u$. Nothing in the parity derivation above actually caps the degree at a quadratic — a function built from powers $x^7, x^5, x^3, x$ plus a constant would differentiate to a genuine cubic in $u = x^2$ (a degree-6 polynomial in $x$), by the identical mechanism. No such question exists in the two verified anchors, and the reason is not arithmetic difficulty — it is a hard curriculum boundary, checkable directly in this project's own spec data. specCoverage.ts places "Simple algebraic division; use of the Factor Theorem and the Remainder Theorem" at WMA12 §2, item 1 — the Pure 2 paper — not anywhere in the WMA11 (Pure 1) table of contents that governs this topic's own spec refs (4.1–4.3). A general three-term quadratic is solvable by factorising, the quadratic formula, or completing the square — all P1 tools — but a general cubic in $u$ has no such guaranteed closed method without the Factor Theorem, which this paper's students have not yet been taught. A WMA11 "hidden higher-degree equation" question can only ever be designed to bottom out at something P1's own toolkit can finish, which means the quartic-hiding-a-quadratic shape is not a coincidence shared by the two real anchors — it is the only shape this topic's spec position permits. This is my own structural reasoning from the two documents' spec placements, not a fact either verified-facts document states outright, and it predicts (without any real anchor yet confirming it) that a hidden-cubic-in-x² variant is not a legitimate WMA11 question shape, however plausible it looks algebraically.

#4. The one confirmed non-example, generalised: why fractional powers don't hide anything

The facts bank's Lesson-audit log records that an earlier pass had listed Jan 2019 Q6 as a third anchor for this topic, and that a direct re-fetch of the real question overturned that: $f(x) = 2x^{5/2} - 40x + 8$ ($x>0$), where $f'(x) = 5x^{3/2} - 40$ and $f''(x) = \tfrac{15}{2}x^{1/2}$, both solved directly by raising to a reciprocal power (e.g. $x^{3/2}=8 \Rightarrow x = 8^{2/3} = 4$) — never producing a polynomial in only even powers of $x$. The examiner report's own diagnosis: candidates struggled with "the inverse process of indices," not with recognising a disguised quadratic.

Section 2's refinement explains exactly why, in general terms the facts bank does not itself state. $f'(x) = 5x^{3/2}-40$ contributes exactly one nonzero power-of-$x$ term ($x^{3/2}$) once the constant target is moved across — there is no second, independent power of $x$ for it to pair with under a substitution, so $f'(x)=0$ is a single-term index equation by construction, not a disguised quadratic that happens to look like one. The fractional exponent is almost incidental to why this fails to disguise anything; a function like $f(x)=6x^4-7$ (a single even-power term, no negative or fractional index anywhere) would fail for exactly the same one-term reason. The real, generalisable test this non-example teaches is: count the distinct nonzero powers surviving in $f'(x)$. One → a direct index equation (a different WMA11 skill entirely, spec 4.2's "raise both sides to the reciprocal power," taught elsewhere in this course). Two, both even → this topic's disguised quadratic in $x^2$.

#5. Three short, fresh, hand-checked micro-examples

None of the numbers below appear anywhere in the existing lesson.

5.1 Route A — pure odd-power polynomial, no fraction-clearing. $f(x) = x^5 - 2x^3 + 4x$. Find all $x$ for which $f'(x) = 3$. $f'(x) = 5x^4 - 6x^2 + 4$. Set equal to 3: $5x^4 - 6x^2 + 1 = 0$. As a quadratic in $u=x^2$: $5u^2-6u+1=0$, factorising as $(5u-1)(u-1)=0$ [check: $5u^2-5u-u+1=5u^2-6u+1$ ✓], giving $u=\tfrac15$ or $u=1$ — both positive, so both survive. $x = \pm\tfrac{1}{\sqrt5}$ or $x=\pm1$. Check at $x=1$: $5-6+4=3$ ✓. Check at $x=1/\sqrt5$: $5(\tfrac15)^2-6(\tfrac15)+4 = 0.2-1.2+4=3$ ✓.

5.2 Route B — a negative-power term, fraction-clearing required. $f(x) = x^3 - \dfrac{3}{x}$ ($x \neq 0$). Find all $x$ for which $f'(x) = 10$. Rewrite $-3x^{-1}$ first: $f'(x) = 3x^2 + 3x^{-2}$. Set equal to 10: $3x^2+\tfrac{3}{x^2}=10$. Multiply through by $x^2$: $3x^4+3=10x^2$, i.e. $3x^4-10x^2+3=0$. As a quadratic in $u$: $3u^2-10u+3=0$, factorising as $(3u-1)(u-3)=0$ [check: $3u^2-9u-u+3=3u^2-10u+3$ ✓], giving $u=\tfrac13$ or $u=3$ — both positive. $x=\pm\tfrac{1}{\sqrt3}$ or $x=\pm\sqrt3$. Check at $x^2=3$: $3(3)+\tfrac33=9+1=10$ ✓. Check at $x^2=\tfrac13$: $3(\tfrac13)+\tfrac{3}{1/3}=1+9=10$ ✓.

5.3 The non-example — a single surviving power, nothing to disguise. $f(x) = 6x^4 - 7$. Find $x$ for which $f'(x)=48$. $f'(x)=24x^3$. Set equal to 48: $24x^3=48 \Rightarrow x^3=2 \Rightarrow x=\sqrt[3]2$ — one term, one power, one real solution, solved directly. There is no second power of $x$ present for $u=x^2$ to pair against, so no quadratic step exists to hide — exactly the structural reason Jan 2019 Q6 (§4) resolves the same way, without needing the fractional index to be the explanation.

#6. The ± step, made precise: why a positive u always gives a pair, and why a domain statement overrides that

The map $x \mapsto x^2$ is two-to-one on $\mathbb{R}\setminus\{0\}$: every positive $u$ has exactly two real square roots, $+\sqrt u$ and $-\sqrt u$, and nothing about solving $u=x^2$ for $x$ can recover which sign the original $x$ had, because the squaring step that got to $u$ in the first place destroyed that information. This is why both real anchors' mark schemes require "both values" once $u$ is found: it is not a scheme convention, it is the actual multiplicity of a non-injective function's inverse. The only way to legitimately discard one branch is if the original question restricts the domain (typically to $x>0$) — a restriction that makes $x\mapsto x^2$ injective on the smaller domain, so the inverse becomes single-valued there. Jan 2025 Q5 states $x>0$ in its opening line, which is exactly why the same examiner report penalises candidates who kept both signs anyway ("despite 'x > 0' in the wording of the question," per that report). Oct 2021 Q2 states no such restriction, so both signs are the only correct answer set. The rule that follows from this is stronger than "check the question": it is that the default is always both signs, and a domain restriction is what removes marks for using it, never what adds marks for guessing to include it — there is no symmetric error of wrongly restricting to one sign when no restriction was stated, because $\mathbb{R}$ has no such restriction to be missed.

#7. The companion trap, and the one thread connecting it to the equation-solving trap

The lesson's own second half documents a distinct but related trap: reaching $f'(a)=0$ at a genuine stationary point and then failing to write the resulting horizontal tangent, $y=f(a)$, at all — confirmed directly, "thrown by the zero gradient" (Oct 2021 Q6(c) examiner report). This is not the same skill as §6's forgetting-to-square-root trap — one is an equation-solving completeness error, the other a differentiation-application completeness error — but both documented traps in this one topic share an identical shape, worth naming as a single idea rather than two unrelated ones: in both cases, the student reaches a correct intermediate value that looks too small or too trivial to be a real answer ($x^2=\tfrac14$ looking like it can't be "the" answer; a gradient of exactly $0$ looking like a sign the method broke) and stops or second-guesses rather than finishing the one remaining, genuinely easy step. Neither the facts bank nor the lesson states this connection explicitly; it is offered here as a reading of the two already-cited findings side by side, not as an independent sourced claim.

#8. Exam-angle synthesis — everything real this topic has actually been tested from

Drawing together the SIGNAL brief's own honest count and the facts bank's citations, without adding anything beyond them:

• Frequency: seen in 2 of 5 fully-mined WMA11 series (Oct 2021, Jan 2025), out of five series read question-by-question in the pass this course's facts bank is built on (Jan 2019, Oct 2020, Oct 2021, Jan 2023, Jan 2025). A third series once suspected (Jan 2019 Q6) was directly checked and ruled out (§4).
• Two confirmed route-ins: a pure odd-degree polynomial needing no fraction-clearing (Oct 2021), and a negative-power term needing the fraction cleared by multiplying through by $x^2$ (Jan 2025) — §2's theorem explains why both routes land on the identical even-powers-only shape despite looking different on the page.
• Two confirmed mark-tariff shapes: a short, self-contained ~5-mark question that is the whole question (Oct 2021 Q2), and a single dependent method mark embedded inside a longer tangent-line-and-curve question, where the disguised quadratic is only the first of four marks (Jan 2025 Q5(c)).
• Confirmed dependency depth: Oct 2021 Q2's real mark scheme runs M1 A1 (differentiate) → dM1 (form the 3TQ, dependent on the first M) → ddM1 (solve it, dependent on both prior method marks) → A1 (both x-values) — a doubly-dependent structure, not a flat sequence of independent marks.
• Domain handling is inconsistent between the two real anchors, and that inconsistency is the actual trap (§6): Oct 2021 has no restriction and needs both signs; Jan 2025 states $x>0$ and penalises keeping both.
• A rigour trap distinct from the mathematics: the real Oct 2021 mark scheme refuses a factorisation that does not visibly follow from the undivided equation — "failed to show that they had divided through by 3," per that report — a presentation requirement, not an arithmetic one.
• A substituted-variable-confusion trap, smaller-scale and distinct from forgetting to convert back: a real minority "chose to replace x² by x" outright, per the same Oct 2021 report, rather than treating $u$ as a genuinely new variable.
• The companion calculus-end trap (§7): a correctly-reached zero gradient treated as a dead end rather than the valid, simpler answer it is.

#9. What remains genuinely open

This topic's real evidence base is thin by this course's own standard — two confirmed anchors, not three, after the Jan 2019 correction — and two further WMA11 series (Oct 2022, Jan 2024) are confirmed genuine in the facts bank but have never been mined question-by-question, so it is honestly unknown whether either contains a third real instance of this pattern. Nothing above should be read as predicting a fixed frequency going forward; "2 of 5 series checked" describes what has been verified, not a rate. The hidden-cubic-in-x² boundary argument in §3 is my own structural reasoning from this project's spec data, not a finding independently confirmed against a real mark scheme that tests it — because, per the argument itself, no such WMA11 question should exist to test it against.