Exposition and Structure (anmath-figures)
When to trigger
- The paper is hard to follow even though the proof is correct
- Notation is inconsistent or introduced after first use
- Theorems and their proofs are interleaved confusingly
- A relation between objects would be far clearer as a commutative diagram
- You are deciding whether the rare case of an actual figure is warranted
In pure mathematics, papers are theorem-and-proof and usually have no experiments and
few or no figures. "Figures" here means exposition and structure: a figure or diagram
is included only when it conveys structure more efficiently than prose.
Exposition principles (Annals readability)
- Statements before proofs. State each definition, lemma, proposition, and theorem in
full before proving it. The reader should always know the target before the argument.
- Notation introduced once, used consistently. Define every symbol at first use; keep
a fixed convention throughout; avoid overloading the same symbol for two things.
- Logical sectioning. Preliminaries → key constructions → main lemmas → proof of the
Main Theorem → consequences. Each section has a clear job and a one-line opener.
- Self-contained where reasonable. Recall the precise external statements you use so a
reader need not reconstruct them from memory.
- Readable by an expert non-specialist. Someone strong in an adjacent area should be
able to follow the architecture; gloss the field-specific shorthand the first time.
Sectioning template
| Section |
Contents |
| Introduction |
Problem, Main Theorem, what is new, method sketch, organization |
| Preliminaries / Notation |
Conventions, recalled definitions, cited external results |
| Constructions / setup |
The objects the proof manipulates |
| Key lemmas |
The intermediate results, stated then proved |
| Proof of Main Theorem |
Assembling the lemmas into the headline result |
| Consequences |
Corollaries and remarks |
| Appendices |
Auxiliary/technical material (see anmath-supplementary) |
Numbering and cross-reference conventions
- Number theorem-like environments in one per-section sequence (Theorem 3.1,
Lemma 3.2, Corollary 3.3) so a referee deep in a long verification can locate any
statement from its number alone.
- Give the headline result a stable early label — Theorem 1.1, or letters
(Theorem A, B, …) in a long paper — and use it everywhere.
- Number displayed equations only when referenced; an unnumbered-but-needed display
forces "the equation above," which breaks silently under revision.
- Open each section with what it proves and what it consumes: "In this section we
prove Proposition 3.1, using Lemma 2.4."
Diagrams and figures (when justified)
- Commutative diagrams: use
tikz-cd (or amscd) when a chain of maps or an exact
sequence is clearer drawn than written. Keep arrows labeled and consistent.
- A genuine figure (a configuration, a region, a graph): include only when it removes
real ambiguity. Use vector output (PDF/EPS), label everything, and reference it in text.
- Tables: occasionally useful for case enumerations or notation summaries; keep clean.
- Most Annals papers have zero figures — do not add one for decoration.
Micro-example: when a diagram earns its place
Prose version: "The map φ factors through the quotient, and the induced map commutes
with the two projections of Section 2."
As a tikz-cd square with labeled arrows, the same content is checkable at a glance.
That is the admission test — the referee verifies commutativity faster from the
picture than from the sentence. A picture that merely restates a relation the prose
already makes precise is decoration.
Checklist
Anti-patterns
- Using a symbol before defining it, or redefining a symbol mid-paper
- Interleaving proof fragments with statements so the target is unclear
- A wall of unsegmented text with no sectioning logic
- Adding a decorative figure that conveys nothing the prose does not
- Field jargon used without a single gloss for the adjacent-area reader
- Sloppy diagram arrows (unlabeled, inconsistent direction) that obscure the maps
Output format
【Section plan】1 Intro · 2 Prelim · 3 ... · n Appendix
【Notation issues fixed】...
【Statements-before-proofs】compliant / fix: ...
【Diagrams】none / commutative diagram in §... via tikz-cd
【Figure justification】none needed / figure in §... because ...
【Next step】anmath-supplementary (appendix triage) or anmath-writing-style
1---2name: anmath-figures3description: Use when organizing the exposition and structure of a pure-mathematics manuscript for Annals of Mathematics — sectioning, notation, statements-before-proofs, commutative diagrams, and readability for an expert non-specialist. Figures are optional and rare; this skill is about exposition first. Does not assess the proof's correctness.4---56# Exposition and Structure (anmath-figures)78## When to trigger910- The paper is hard to follow even though the proof is correct11- Notation is inconsistent or introduced after first use12- Theorems and their proofs are interleaved confusingly13- A relation between objects would be far clearer as a commutative diagram14- You are deciding whether the rare case of an actual figure is warranted1516> In pure mathematics, papers are theorem-and-proof and usually have **no experiments and17> few or no figures**. "Figures" here means *exposition and structure*: a figure or diagram18> is included only when it conveys structure more efficiently than prose.1920## Exposition principles (Annals readability)21221. **Statements before proofs.** State each definition, lemma, proposition, and theorem in23 full before proving it. The reader should always know the target before the argument.242. **Notation introduced once, used consistently.** Define every symbol at first use; keep25 a fixed convention throughout; avoid overloading the same symbol for two things.263. **Logical sectioning.** Preliminaries → key constructions → main lemmas → proof of the27 Main Theorem → consequences. Each section has a clear job and a one-line opener.284. **Self-contained where reasonable.** Recall the precise external statements you use so a29 reader need not reconstruct them from memory.305. **Readable by an expert non-specialist.** Someone strong in an adjacent area should be31 able to follow the architecture; gloss the field-specific shorthand the first time.3233## Sectioning template3435| Section | Contents |36|---------|----------|37| Introduction | Problem, Main Theorem, what is new, method sketch, organization |38| Preliminaries / Notation | Conventions, recalled definitions, cited external results |39| Constructions / setup | The objects the proof manipulates |40| Key lemmas | The intermediate results, stated then proved |41| Proof of Main Theorem | Assembling the lemmas into the headline result |42| Consequences | Corollaries and remarks |43| Appendices | Auxiliary/technical material (see anmath-supplementary) |4445## Numbering and cross-reference conventions4647- Number theorem-like environments in one per-section sequence (Theorem 3.1,48 Lemma 3.2, Corollary 3.3) so a referee deep in a long verification can locate any49 statement from its number alone.50- Give the headline result a stable early label — Theorem 1.1, or letters51 (Theorem A, B, …) in a long paper — and use it everywhere.52- Number displayed equations only when referenced; an unnumbered-but-needed display53 forces "the equation above," which breaks silently under revision.54- Open each section with what it proves and what it consumes: "In this section we55 prove Proposition 3.1, using Lemma 2.4."5657## Diagrams and figures (when justified)5859- **Commutative diagrams**: use `tikz-cd` (or `amscd`) when a chain of maps or an exact60 sequence is clearer drawn than written. Keep arrows labeled and consistent.61- **A genuine figure** (a configuration, a region, a graph): include only when it removes62 real ambiguity. Use vector output (PDF/EPS), label everything, and reference it in text.63- **Tables**: occasionally useful for case enumerations or notation summaries; keep clean.64- Most Annals papers have zero figures — do not add one for decoration.6566## Micro-example: when a diagram earns its place6768Prose version: "The map φ factors through the quotient, and the induced map commutes69with the two projections of Section 2."7071As a `tikz-cd` square with labeled arrows, the same content is checkable at a glance.72That is the admission test — the referee verifies commutativity faster from the73picture than from the sentence. A picture that merely restates a relation the prose74already makes precise is decoration.7576## Checklist7778- [ ] Every definition/lemma/proposition/theorem is stated before it is proved79- [ ] Numbering is one per-section sequence; the headline theorem has a stable label used everywhere80- [ ] Every symbol is defined at first use and used consistently81- [ ] Sections are logically ordered with clear one-line openers82- [ ] External results used are recalled precisely (statement + citation)83- [ ] Commutative diagrams (if any) are typeset with labeled, consistent arrows84- [ ] Any figure is vector, labeled, referenced, and genuinely necessary85- [ ] An expert in an adjacent area could follow the overall architecture8687## Anti-patterns8889- Using a symbol before defining it, or redefining a symbol mid-paper90- Interleaving proof fragments with statements so the target is unclear91- A wall of unsegmented text with no sectioning logic92- Adding a decorative figure that conveys nothing the prose does not93- Field jargon used without a single gloss for the adjacent-area reader94- Sloppy diagram arrows (unlabeled, inconsistent direction) that obscure the maps9596## Output format9798```99【Section plan】1 Intro · 2 Prelim · 3 ... · n Appendix100【Notation issues fixed】...101【Statements-before-proofs】compliant / fix: ...102【Diagrams】none / commutative diagram in §... via tikz-cd103【Figure justification】none needed / figure in §... because ...104【Next step】anmath-supplementary (appendix triage) or anmath-writing-style105```