Deep Dive: Binomial Expansion

Provenance (2026-09-29). Read in full for this pass: research/veridian/WMA12-verified-facts.md (§2's verbatim transcription of spec item 4.5; §3's formula-booklet section, including the booklet-page rendered-as-image check of the binomial-series anomaly; §4's mark-scheme conventions; §6.5 in full; §7's "topics with enough verified material" entry for this topic; and the binomial-expansion.ts mark-scheme-bullet-coverage audit entry dated 2026-09-14, plus its bx-term-not-raised-as-a-whole citation-accuracy follow-up); the complete lesson file content/veridian/curriculum/wma12/binomial-expansion.ts (all 843 lines, read to confirm exactly what it already teaches — its two derivations of ⁿCᵣ, its method-comparison, worked chain, chain-drill, 8-mark marked-solution and six trap-taxonomy entries — so nothing below restates them); content/veridian/technique-hub/wma12/codex-formula-and-method-bank.md (§5, its binomial-expansion formula entry); content/veridian/technique-hub/wma12/forge-exam-technique-guide.md (grepped for "binomial," then read in context: the June 2025 real topic-order note, Part 3's "binomial/log/series questions with a choice of formula" trigger, and Part 4's binomial trap-taxonomy locator line); content/veridian/technique-hub/wma12/signal-frequency-and-boundaries.md (the binomial-expansion frequency row and the June 2025 per-question mark table); and content/veridian/question-bank/wma12.ts (grepped for "binomial" — zero matches, confirming no worked past-paper item exists there for this topic, so nothing to avoid duplicating on that front). Not affiliated with or endorsed by Pearson Edexcel. Every quotation below is short (under ~15 words) and attributed to a named document; no full question, mark-scheme extract, or examiner-report passage is reproduced at length.

#CODEX Deep-Dive — Binomial Expansion of (a + bx)ⁿ for Positive Integer n (WMA12, spec 4.5)

#What this document is, and isn't

The lesson file already teaches this topic in real depth: two independent derivations of ⁿCᵣ (the combinatorial headcount argument and Pascal's addition-rule argument), a method-comparison between the general-term formula and Pascal's triangle, a fully worked chain and chain-drill on a negative-b expansion, an 8-mark marked-solution pairing a "first three terms" opener with an unknown-parameter part and a sign-based rejection, six MCQs, and six trap-taxonomy entries — four of them tied to a real, named examiner-report or mark-scheme finding. codex-formula-and-method-bank.md already holds the compact formula statement and the booklet's two-formula anomaly (§5). forge-exam-technique-guide.md already covers the exam-craft angle (the formula-choice trigger, the trap-taxonomy locator). signal-frequency-and-boundaries.md already states the real, floor-level frequency count.

None of those four documents does what this one does: prove, rather than merely assert, why the two booklet formulas must agree for a positive integer n (the lesson's own PQ2 explanation states the fact — "the general series happens to terminate at exactly the same (n+1) terms" — without deriving it); name and work through the one genuinely distinct sub-case this spec item's real exam history contains that the lesson does not cover — recovering two unknown parameters (n and k) simultaneously, rather than one; and close with a single consolidated ledger of every real exam angle this specific topic has actually been tested from, in one place.

The honest evidentiary picture, stated up front. Spec 4.5 is one flat, undotted line — the thinnest kind of spec item this course's lesson structure recognises — and signal-frequency-and-boundaries.md records it as confirmed present in 2 series / 3 papers of the 20 series checked for a question paper and mark scheme (Jan 2024; June 2025 main paper; June 2025's 01R regional variant), carrying 3–7 marks where found. That is a real floor, not a rarity signal — spec 4.5 is compulsory every series — but it means this document's synthesis ledger has exactly three real anchors to draw on, not dozens, and it does not stretch that into more than it is.

---

#Part 1 — The complete theory model: one counting idea, two equivalent routes to it

$(a+bx)^n$ for positive integer $n$ expands to exactly $n+1$ terms, $\sum_{r=0}^{n}\binom{n}{r}a^{n-r}(bx)^r$, where $\binom{n}{r}=\dfrac{n!}{r!(n-r)!}$ counts the number of ways to choose which $r$ of the $n$ bracket-factors contribute $bx$ rather than $a$. The lesson proves this two independent ways — directly, as a headcount with overcounted orderings divided out, and via Pascal's triangle's addition rule, $\binom{n}{r}=\binom{n-1}{r-1}+\binom{n-1}{r}$ — and this document does not re-derive either; see binomial-expansion.ts's two mechanism blocks.

A fresh check that the two routes are the same headcount, with numbers the lesson never uses. Coefficient of $x^2$ in $(4+3x)^5$: general term gives $\binom{5}{2}\times4^3\times3^2 = 10\times64\times9 =5760$. Independently, a full expansion of $(4+3x)^5$ (multiplied out term by term, not via the formula) gives ascending coefficients $1024,\,3840,\,5760,\,4320,\,1620,\,243$ — the third entry, $5760$, matches exactly. Both routes count the identical set of bracket-selections; they cannot disagree, and this is why.

---

#Part 2 — Sub-case: proving the "harder" booklet formula still gives the identical answer

The P2 formula booklet prints two binomial formulas on the same page (WMA12-verified-facts.md §3, independently re-confirmed by codex-formula-and-method-bank.md §5): the combinatorial $(a+b)^n=a^n+\binom{n}{1}a^{n-1}b+\ldots+b^n$ for positive integer $n$, and directly beneath it the general series $(1+x)^n=1+nx+\dfrac{n(n-1)}{2!}x^2+\ldots$, valid for real $n$ and $|x|<1$ — P4 content, not required at P2. The lesson correctly warns students away from the second formula on tractability grounds, and states in passing that for a positive integer $n$ it "happens to terminate at exactly the same $(n+1)$ terms" as the first — but never proves that claim. It is worth proving, because it is the mathematical fact underneath the entire formula-choice trap.

The general series' $r$-th coefficient is the descending product $\dfrac{n(n-1)(n-2)\cdots(n-r+1)}{r!}$ — $r$ factors on top, $r!$ on the bottom. For $r\le n$, that descending product is exactly $\dfrac{n!}{(n-r)!}$, so the coefficient is $\dfrac{n!}{(n-r)!\,r!}=\binom{n}{r}$ — identical to the combinatorial form, term by term, not merely "also correct." The series terminates, rather than running forever, for exactly one reason: once $r=n+1$, the descending product $n(n-1)\cdots(n-n)$ contains the factor $n-n=0$ — one of its $r$ factors is zero — so that coefficient, and every one after it, is forced to $0$. Checked directly: for $n=4$, the $r=5$ term's numerator is $4\times3\times2\times1\times0=0$. This is the real reason the "more difficult" formula is not actually wrong on P2 — it is the identical polynomial, just reached by a route that keeps producing terms (all of them correctly zero past $r=n$) instead of stopping cleanly at $r=n$ the way the combinatorial form does by construction.

Why the choice still matters in practice, despite being mathematically equivalent: the general series form requires computing $n(n-1)$, $n(n-1)(n-2)$, and so on as separate running products, which is strictly more arithmetic than reading $\binom{n}{r}$ off a single factorial ratio or a Pascal's row — and once a question adds a second unknown, that extra arithmetic becomes the difference between a tractable and an intractable equation. This is exactly what the real evidence in Part 5 below confirms.

---

#Part 3 — Sub-case: the sign of $(bx)^r$, briefly, with fresh numbers

Already fully derived and drilled in the lesson (sign-lost-on-odd-powers-of-a-negative-b, worked-chain on $(3-2x)^7$, chain-drill on $(1-4x)^6$, MCQ on $(5-2x)^6$) — not re-derived here. Stated in one line for completeness of the theory model: $(bx)^r=b^rx^r$ carries the sign of $b^r$, negative exactly when $r$ is odd and $b<0$.

A fresh check, with numbers the lesson does not use: coefficient of $x^5$ in $(2-3x)^8$. General term: $\binom{8}{5}\times2^{3}\times(-3)^5=56\times8\times(-243)=-108\,864$; $r=5$ is odd, so the sign is negative, consistent with the rule. A full expansion of $(2-3x)^8$ confirms the same $x^5$-coefficient, $-108\,864$, independently. The real trap this connects to, confirmed on the lesson's own anchor question and not merely derived: examiners recorded candidates "using invisible brackets that were not recovered in further work" (Jan 2024, Q2) — the identical failure to raise the entire bracketed term, numeral and sign included, that produces a wrong magnitude on a positive-b question and a wrong sign on a negative-b one.

---

#Part 4 — Sub-case: one unknown parameter from a given coefficient (pointer only)

The lesson's own 8-mark marked-solution already covers this in full — $(3+kx)^6$, one unknown $k$, found from a given $x^2$-coefficient, with a sign-based rejection of one root in part (c) — and is not reproduced here. What that marked-solution does not cover, and what genuinely distinct exam evidence exists for beyond it, is Part 5 below.

---

#Part 5 — Sub-case: two unknowns recovered simultaneously — the real, harder variant

The real evidence. WMA12-verified-facts.md §6.5 documents a regional variant paper (June 2025, "01R," Q7) built around $(1+kx)^n$ with both $n$ and $k$ unknown — the examiner report states directly that "there were two unknowns, n and k" and that candidates who reached for the general series form "made little progress," while those using the combinatorial form fared better. This is a real, distinct sub-case: the lesson's own marked-solution only ever has one unknown (with $n$ given), and no other WMA12 document in this technique hub works through the two-unknown mechanism explicitly.

The mechanism. For $(1+kx)^n$: coefficient of $x^1$ is $\binom{n}{1}k=nk$; coefficient of $x^2$ is $\binom{n}{2}k^2=\dfrac{n(n-1)}{2}k^2$. Two given coefficients produce two equations in the two unknowns $n$ and $k$ — solvable in principle by straightforward substitution (isolate $k$ from the first equation, substitute into the second), which is the method a real mark scheme would credit via the standard M1 (form both equations from the general term)/dM1 (solve simultaneously) structure this course's own marking conventions use elsewhere (WMA12-verified-facts.md §4).

A genuinely faster route, reasoned directly from the two coefficient-expressions above (not sourced to any examiner report — flagged here on that honest basis, exactly as this document's own discipline requires): dividing the square of the $x$-coefficient by the $x^2$-coefficient eliminates $k$ outright, since $\dfrac{(nk)^2}{\frac{n(n-1)}{2}k^2}=\dfrac{n^2}{\frac{n(n-1)}{2}}=\dfrac{2n}{n-1}$ — a function of $n$ alone. This turns a genuinely two-unknown problem into a one-unknown equation before any value of $k$ is ever found, without touching the harder general-series arithmetic the real examiner report shows costing candidates progress.

Original micro-example (VERIDIAN-original — not a reproduction of the real Q7, whose exact wording this document does not have access to; only the examiner report's paraphrase of its structure): in the expansion of $(1+kx)^n$, the coefficient of $x$ is $18$ and the coefficient of $x^2$ is $135$; find $n$ and $k$. Ratio: $\dfrac{18^2}{135}=\dfrac{324}{135}=\dfrac{12}{5}$. Set $\dfrac{2n}{n-1}=\dfrac{12}{5}$: $10n=12n-12$, so $n=6$. Then from $nk=18$: $k=3$. Check against the second equation directly: $\binom{6}{2}\times3^2=15\times9=135$ — matches. (Verified independently by computer algebra for this document: $n=6,k=3$ is the unique consistent solution.)

---

#Part 6 — Sub-case: isolating the requested value once several are on the page

Not new mathematics — a mark-scheme discipline specific to this topic, confirmed on the lesson's own anchor question and already taught in full there (required-term-not-isolated-from-a-list). Stated here only because a synthesis of "every real exam angle" is incomplete without it: the real mark scheme states "the term or coefficient must be identified if they have more than one term" (Jan 2024, Q2), and the examiner report adds that the final mark "was commonly not awarded to candidates who failed to isolate the required term from a list." This is the same paper-wide "quote it, don't bury it" discipline forge-exam-technique-guide.md's self-mark checklist states generally — confirmed here as a real, named cost on this exact topic, not a generic exam-skills platitude.

---

#Every real exam angle this topic has actually been tested from — consolidated ledger

Read every row as "confirmed seen in the series named," never as a probability. No likelihood percentage and no grade boundary specific to this topic is stated anywhere in this document, per this course's standing discipline; signal-frequency-and-boundaries.md carries the paper-wide grade boundaries and is the place to look for those.

Cross-cutting observations, each traceable to a specific row above, none invented:

• This topic's real difficulty gradient is confirmed, not assumed: the identical formula-choice decision is "generally done very well" with one unknown (row 2) and demonstrably harder with two (row 3) — real evidence that the general-series form's extra arithmetic, harmless when everything else is known, becomes the actual obstacle once it has to be solved alongside a second unknown.
• The one anchor with a detailed error breakdown (row 1) shows the dominant real failure mode is mechanical indexing, not conceptual: matching $r$ to the wrong power, or reading the wrong row of Pascal's triangle, confirmed ahead of any misunderstanding of what the theorem itself states.
• Two of this topic's six lesson-level trap-taxonomy entries lack any real per-question anchor (pascal-triangle-row-miscounted and the sign-alternation trap, both reasoned directly from the mechanism rather than sourced to a named finding, per the lesson file's own header note) — the other four are each tied to a real, quoted mark-scheme or examiner-report line, which is still an unusually well-evidenced ratio for a topic this narrow.
• No grade-boundary or likelihood claim is made anywhere in this document. The topic's own confirmed floor is "2 series / 3 papers of the 20 checked" (signal-frequency-and-boundaries.md) — stated as a floor on what has been individually cited, never converted into a percentage or a prediction.