03.12.45 · modern-geometry / homotopy

Arithmetic square and integral fracture theorems

shipped3 tiersLean: partial

Anchor (Master): Sullivan 1970 *Geometric Topology: Localization, Periodicity, and Galois Symmetry* (MIT notes; revised K-Monographs in Mathematics 8, 2005) — originator; Bousfield-Kan 1972 LNM 304 §VI (modern reframing as a homotopy-limit construction); Dwyer-Greenlees-Iyengar 2006 *Duality in algebra and topology* (Adv. Math. 200) §4 (modern arithmetic-fracture treatment); May-Ponto 2012 *More Concise Algebraic Topology* Ch. 10-13 (textbook synthesis); Hess 2007 *Rational homotopy theory: a brief introduction* arXiv:math/0604626; Neisendorfer 2010 *Algebraic Methods in Unstable Homotopy Theory* (Cambridge) Ch. 12 (nilpotent and finite-type assumptions)

Intuition [Beginner]

Imagine you want to understand an integer like $42$ . One angle is to look at it through rational glasses: as a fraction it's just $42/1$ , a single point on the rational number line. Another angle is to look at it prime-by-prime: $42 = 2 \cdot 3 \cdot 7$ , so it has one factor of $2$ , one factor of $3$ , one factor of $7$ , and no others. The full integer $42$ is reconstructed from its rational image and its behaviour at each prime, glued by the consistency condition that the two views agree when you forget enough information.

The arithmetic square is the same idea for spaces. Instead of an integer, the object is a topological space $X$ . The "rational glasses" replace $X$ with a rational version $X_{Q}$ that records the rational homotopy and forgets all the torsion. The "prime-by-prime" view replaces $X$ with a family of $p$ -local completions $X_{p}$ , one for each prime $p$ , each recording the $p$ -primary information and forgetting everything coprime to $p$ . The square then asks: is $X$ recovered from $X_{Q}$ , the family of $X_{p}$ , and the way they overlap when both are rationalised again?

The remarkable answer is yes, when $X$ is sufficiently nice. The space is reconstructed as a pullback diagram from its rational and $p$ -local pieces, the same way an integer is reconstructed from its rational image and its $p$ -adic completions. The result lets algebraic topologists work one prime at a time and then assemble.

Visual [Beginner]

A square diagram with the space $X$ at the top-left, the product of all $p$ -completions (one factor for each prime $p$ ) at the top-right, the rationalisation of $X$ at the bottom-left, and the rationalisation of the product-of-completions at the bottom-right. The horizontal arrows go from the prime-by-prime side to its rationalisation; the vertical arrows go from the integral side to the rational side. The whole picture commutes, and the fracture theorem says $X$ is the homotopy pullback of the other three corners.

The picture captures the heart of the fracture pattern: four objects, four arrows, a commuting square, and a comparison map from $X$ to the pullback that the theorem promises to be a weak equivalence.

Worked example [Beginner]

Take the $2$ -sphere $X = S^{2}$ and walk through the four corners of its arithmetic square.

Step 1. The top-left corner is $X = S^{2}$ itself, a familiar sphere with homotopy groups $π_{2} (S^{2}) = Z$ , $π_{3} (S^{2}) = Z$ (generated by the Hopf map), $π_{4} (S^{2}) = Z /2$ , $π_{5} (S^{2}) = Z /2$ , and so on with a complicated mixture of $Z$ summands and finite torsion groups.

Step 2. The bottom-left corner is the rationalisation $S_{Q}^{2}$ . By Serre's rational-homotopy computation, $S_{Q}^{2}$ has two non-zero rational homotopy groups: $π_{2}$ tensored with the rationals gives $Q$ , and $π_{3}$ tensored with the rationals gives another copy of $Q$ (the rationalised Hopf map). All higher rational homotopy groups vanish. The torsion has been crushed; the rational structure is much simpler.

Step 3. The top-right corner is the product (over primes) of the $p$ -completions $S_{p}^{2}$ . Each $p$ -completion $S_{p}^{2}$ has the same rational homotopy groups as $S^{2}$ but tensored with the $p$ -adic integers, and records the $p$ -primary part of every torsion group. The product over primes records all of this in a single bookkeeping device.

Step 4. The bottom-right corner is the rationalisation of the same product-over-primes object. Because rationalisation kills $p$ -primary torsion, this corner looks like a rational space — specifically, an enlarged version of $S_{Q}^{2}$ that records the "rational glue" between the prime-by-prime views.

What this tells us: the fracture theorem promises that the original $S^{2}$ is recovered by taking a homotopy pullback of the bottom-left, top-right, and bottom-right corners — the sphere is reassembled from its rational image plus its family of $p$ -completions plus a gluing constraint. This is the integral version of "recover an integer from its rational image and its $p$ -adic completions".

Check your understanding [Beginner]

Exercise (easy, multiple choice).

In the arithmetic square for a space $X$ , which corner records the rational homotopy of $X$ ?

A. The top-left corner $X$ . B. The top-right corner: the product over primes of the $p$ -completions $X_{p}$ . C. The bottom-left corner $X_{Q}$ . D. The bottom-right corner: the rationalisation of the product-over-primes of $p$ -completions.

Hint

The rationalisation of a space is denoted with the $Q$ subscript. The corner asked about is the one with just one such subscript on $X$ alone.

Answer

C. The bottom-left corner $X_{Q}$ . This is the rationalisation of $X$ — the space obtained by tensoring all homotopy groups of $X$ with the rationals (under the standing nilpotence assumption that this operation is well-defined and corresponds to a Bousfield localisation). The top-left corner is the integral $X$ . The top-right corner is the product of $p$ -completions, recording prime-by-prime information. The bottom-right corner is the rationalisation of the product-of-completions, which serves as the gluing data.

Exercise (easy, true-false).

True or false: the arithmetic square is a commutative square in the homotopy category.

Hint

A commutative square has four corners and four arrows; the two paths from the top-left corner to the bottom-right corner must give the same arrow after rationalisation and $p$ -completion are both applied.

Answer

True. The arithmetic square commutes in the homotopy category: the two paths from $X$ to the bottom-right corner — first rationalise $X$ and then go to the rationalised product, or first go to the product of completions and then rationalise — agree because rationalisation and $p$ -completion commute with each other up to natural equivalence. The fracture theorem then asks the stronger question: is $X$ the homotopy pullback of the other three corners? That is the substantive content of the theorem.

Exercise (easy, numeric).

For how many primes $p$ does the $2$ -sphere $S^{2}$ have a non-vanishing $p$ -completion $S_{p}^{2}$ ? State the answer as a single number, or as "infinity" if it is infinite.

Hint

The $p$ -completion of a connected space with nontrivial homotopy groups records the $p$ -primary part of each group. The homotopy groups of $S^{2}$ contain torsion at every prime starting from $p = 2$ .

Answer

Infinity. The $p$ -completion $S_{p}^{2}$ is non-vanishing for every prime $p$ , because the homotopy groups of $S^{2}$ contain $p$ -primary torsion for every prime $p$ (a deep theorem; the first instance is the Hopf map giving $Z \subset π_{3} (S^{2})$ , which contributes at every prime once enough torsion has accumulated in higher homotopy). The product over primes therefore runs over the infinite set of all primes; this is one of the reasons the lim^1 hypothesis enters the fracture theorem.

Formal definition [Intermediate+]

Let $H = Ho (sSet_{*})$ denote the pointed homotopy category (equivalently, the homotopy category of pointed simply connected — or, more generally, nilpotent — CW complexes). Let $E$ denote a generalised homology theory, presented by a spectrum.

Definition (Bousfield localisation at $E$ ). A morphism $f : X \to Y$ in $H$ is an $E$ -equivalence if $E_{*} (f) : E_{*} (X) \to E_{*} (Y)$ is an isomorphism of graded abelian groups. An object $Z \in H$ is $E$ -local if for every $E$ -equivalence $f : X \to Y$ the induced map $f^{*} : Hom_{H} (Y, Z) \to Hom_{H} (X, Z)$ is a bijection. The $E$ -localisation of $X$ is an object $L_{E} X$ together with a morphism $η_{X} : X \to L_{E} X$ such that $L_{E} X$ is $E$ -local and $η_{X}$ is an $E$ -equivalence. Bousfield 1975 proved that $L_{E} X$ exists and is functorial in $X$ , with $L_{E} : H \to H$ an idempotent endofunctor whose image is the full subcategory of $E$ -local objects.

Definition (rationalisation $L_{Q}$ and $p$ -completion $L_{p}$ ). The rationalisation $L_{Q} := L_{H Q}$ is the Bousfield localisation at the rational Eilenberg-MacLane spectrum $H Q$ . Equivalently, by Serre's class theory, a map $f : X \to Y$ between simply connected spaces is an $H Q$ -equivalence if and only if it induces a rational-homotopy isomorphism $π_{n} (X) \otimes Q \to π_{n} (Y) \otimes Q$ for every $n$ . The $p$ -completion $L_{p} := L_{H F_{p}}$ is the Bousfield localisation at the mod- $p$ Eilenberg-MacLane spectrum $H F_{p}$ ; for simply connected $X$ of finite type, $L_{p} X$ has homotopy groups $π_{n} (X) \otimes_{Z} Z_{p}$ (the $p$ -adic integers, equivalently the Ext-completion of $π_{n} (X)$ at $p$ ).

Definition (nilpotent space, finite type). A connected space $X$ is nilpotent if its fundamental group $π_{1} (X)$ is a nilpotent group and acts nilpotently on each higher homotopy group $π_{n} (X)$ for $n \geq 2$ . The space has finite type if every homotopy group $π_{n} (X)$ is finitely generated over $Z$ (or, equivalently for nilpotent spaces, every singular cohomology group $H^{n} (X; Z)$ is finitely generated). Simply connected spaces are automatically nilpotent; the nilpotent class is the natural setting for the arithmetic square because Bousfield-Kan completion is well-behaved on it.

Definition (Sullivan arithmetic square). The Sullivan arithmetic square at $X \in H$ is the commutative square $$ \begin{array}{ccc} X & \xrightarrow{\eta_{\mathrm{comp}}} & \prod_p L_p X \ \eta_\mathbb{Q} \downarrow & & \downarrow L_\mathbb{Q}(\eta_{\mathrm{comp}}) \ L_\mathbb{Q} X & \xrightarrow{L_\mathbb{Q}(\eta_{\mathrm{comp}})*} & L\mathbb{Q}!\left(\prod_p L_p X\right) \end{array} $$ where $η_{Q} : X \to L_{Q} X$ is rationalisation, $η_{comp} : X \to \prod_{p} L_{p} X$ is the product of $p$ -completion maps, and the bottom and right arrows are obtained by applying $L_{Q}$ . The square commutes by the universal property of localisation.

Definition (homotopy pullback and comparison map). The homotopy pullback $P (X) := L_{Q} X \times_{L_{Q} (\prod_{p} L_{p} X)}^{h} \prod_{p} L_{p} X$ is the standard homotopy fibre product of the three lower-and-right corners. The arithmetic square induces a canonical comparison map $α_{X} : X \to P (X)$ in $H$ . The integral fracture theorem asks: when is $α_{X}$ a weak equivalence?

Counterexamples to common slips

The square need not be a strict pullback in $\mathbf{sSet}_ $. * T h e in t e g r a l f r a c t u r e t h eor e mi s a s t a t e m e n t in$ \mathcal{H} $, n o t a tt h e m o d e l - c a t e g or y l e v e l . T h ecor r es p o n d in g s t r i c tp u l l ba c k in$ \mathbf{sSet}_*$ is generally wrong; one must take the homotopy pullback, equivalently a fibrant replacement of the cospan before pulling back.
Non-nilpotent fundamental groups break fracture. If $π_{1} (X)$ is not nilpotent (for example, a non-abelian free group), the action on higher homotopy groups can mix prime information across primes in a way that $\prod_{p} L_{p} X$ does not detect, and $α_{X}$ fails to be a weak equivalence.
Finite-type matters for the $lim^{1}$ term. Without finite-generation of the homotopy groups, the Milnor short exact sequence relating $π_{n} (P (X))$ to the inverse-limit and $lim^{1}$ data over the prime tower can produce a non-vanishing $lim^{1}$ contribution, breaking the comparison.
Rationalisation is not always tensoring with $Q$ . For non-simply-connected $X$ , the rational homotopy theory needs Sullivan's PL-de Rham apparatus or Quillen's dg-Lie-algebra model. The naive "tensor with $Q$ " of $π_{*} (X)$ is correct only under nilpotence.

Key theorem with proof [Intermediate+]

Theorem (Sullivan 1970; Bousfield-Kan 1972, Integral Fracture Theorem). Let $X$ be a connected nilpotent space of finite type. Then the comparison map $α_{X} : X \to P (X) = L_{Q} X \times_{L_{Q} (\prod_{p} L_{p} X)}^{h} \prod_{p} L_{p} X$ is a weak equivalence in $H$ .

Proof. We prove the theorem by showing that $α_{X}$ induces an isomorphism on every homotopy group $π_{n}$ for $n \geq 1$ . By the Whitehead theorem (Hatcher §4.5; recovered here as a consequence of the Quillen-Serre model structure of 03.12.31), inducing isomorphisms on all $π_{n}$ between connected spaces of the homotopy types under consideration is enough to conclude that $α_{X}$ is a weak equivalence.

Step 1. The Milnor short exact sequence for the homotopy pullback. For any homotopy pullback diagram $$ \mathrm{P} \to B \to D \leftarrow C $$ the long exact sequence of the homotopy fibre identifies $π_{n} (P)$ as the pullback (in the category of groups, abelian for $n \geq 2$ ) of $π_{n} (B)$ and $π_{n} (C)$ over $π_{n} (D)$ , modulo a connecting term measuring failure of the fibre sequence to split. Applied to the arithmetic square, this gives a short exact sequence $$ 0 \to \lim^1!p (\pi{n+1}(L_p X)) \to \pi_n(\mathrm{P}(X)) \to \pi_n(L_\mathbb{Q} X) \times_{\pi_n(L_\mathbb{Q}(\prod_p L_p X))} \pi_n(\prod_p L_p X) \to 0 $$ for each $n \geq 1$ , where the $lim^{1}$ term is the derived inverse limit over the family of primes.

Step 2. The $lim^{1}$ term vanishes under finite-type. For $X$ of finite type, each $π_{n + 1} (X)$ is a finitely generated abelian group, hence a direct sum of finite cyclic groups and copies of $Z$ . The $p$ -completion $π_{n + 1} (L_{p} X) = π_{n + 1} (X) \otimes Z_{p}$ is a finitely generated $Z_{p}$ -module, hence compact in the $p$ -adic topology. The inverse system ${π_{n + 1} (L_{p} X)}_{p}$ over the family of primes (with the product structure) has $lim^{1} = 0$ by the Mittag-Leffler criterion: the maps in the inverse system are surjections at each step (in fact, projections in the product structure), and surjective inverse systems have vanishing $lim^{1}$ (the standard inverse-limit lemma; Bousfield-Kan 1972 §IX.3).

Step 3. The remaining pullback computes $π_{n} (X)$ . The pullback $π_{n} (L_{Q} X) \times_{π_{n} (L_{Q} (\prod_{p} L_{p} X))} π_{n} (\prod_{p} L_{p} X)$ is identified by the local-global principle for finitely generated abelian groups. Concretely, $π_{n} (L_{Q} X) = π_{n} (X) \otimes Q$ , $π_{n} (\prod_{p} L_{p} X) = \prod_{p} (π_{n} (X) \otimes Z_{p}) = π_{n} (X) \otimes \hat{Z}$ where $\hat{Z} = \prod_{p} Z_{p}$ is the profinite completion of $Z$ , and $π_{n} (L_{Q} (\prod_{p} L_{p} X)) = π_{n} (X) \otimes \hat{Z} \otimes Q = π_{n} (X) \otimes A_{f}$ where $A_{f} = \hat{Z} \otimes Q$ is the ring of finite adeles.

The pullback is then the abelian group fitting into the short exact sequence $$ 0 \to \pi_n(X) \to \pi_n(X) \otimes \mathbb{Q} \oplus \pi_n(X) \otimes \hat{\mathbb{Z}} \to \pi_n(X) \otimes \mathbb{A}_f \to 0, $$ which is the local-global short exact sequence for finitely generated abelian groups (a consequence of the structure theorem and the standard adelic short exact sequence $0 \to Z \to Q \oplus \hat{Z} \to A_{f} \to 0$ of number-theoretic provenance). The pullback corner $π_{n} (P (X))$ therefore equals $π_{n} (X)$ .

Step 4. Conclusion. Steps 1-3 combine to give $π_{n} (P (X)) ≅ π_{n} (X)$ for every $n \geq 1$ . The comparison map $α_{X}$ induces this isomorphism (it is the natural map from $X$ to the pullback, which agrees with the structure map of the local-global short exact sequence under the identifications above). By the Whitehead theorem, $α_{X}$ is a weak equivalence in $H$ . $□$

Bridge. This theorem builds toward the chromatic-fracture machinery of stable homotopy theory and appears again in 03.12.31 (model categories) where the Bousfield localisations $L_{Q}$ and $L_{p}$ are the canonical examples of Bousfield-localised model structures on $sSet_{*}$ . The foundational reason fracture works is the local-global short exact sequence for finitely generated abelian groups, and this is exactly the homotopical lift of the algebraic statement that an integer is determined by its rational image and its profinite completion. The central insight is that the Milnor $lim^{1}$ term vanishes under finite-type, so the homotopy pullback reduces to an ordinary pullback at the level of homotopy groups, and the integral assembly proceeds prime-by-prime. Putting these together, the arithmetic square identifies the homotopy category of nilpotent finite-type spaces with the homotopy category of arithmetic squares: every such $X$ is recoverable from its rational and $p$ -local pieces, and every consistent system of pieces glues to a space. The bridge is between integral and rational/local computations — every classical integral homotopy calculation can be reorganised as a rational computation plus a family of $p$ -local computations, and the pattern generalises to chromatic homotopy theory where the prime-by-prime assembly becomes height-by-height assembly via Morava $K$ -theory localisations. This pattern recurs throughout modern stable homotopy: the central thesis of chromatic homotopy is that the stable homotopy category is built from its chromatic strata by exactly this fracture pattern.

Exercises [Intermediate+]

Exercise 1 (easy, multiple choice).

The Bousfield localisation $L_{E}$ at a homology theory $E$ inverts which class of morphisms?

A. The morphisms that are isomorphisms on every homotopy group. B. The morphisms $f$ such that $E_{*} (f)$ is an isomorphism of graded abelian groups. C. The morphisms that induce surjections on $E$ -cohomology. D. The morphisms that are weak equivalences in $Top$ .

Hint

The defining property of Bousfield localisation at $E$ is that it inverts the maps detected by $E$ on homology.

Answer

B. The morphisms $f$ such that $E_{*} (f)$ is an isomorphism. This is the class of $E$ -equivalences. Option A defines weak homotopy equivalence (the un-localised case). Option C is one-sided: surjection on cohomology is not enough; one needs isomorphism on homology in every degree. Option D conflates Bousfield localisation with the underlying model category.

Exercise 3 (medium, symbolic).

Compute the homotopy groups of the rationalisation $S_{Q}^{3}$ of the $3$ -sphere. State which $π_{n} (S_{Q}^{3})$ are nonzero, and identify them with explicit rational vector spaces.

Hint

Serre's rational-homotopy theorem: for odd $n$ , the rationalisation $S_{Q}^{n}$ has one nontrivial rational homotopy group (in degree $n$ ). For even $n$ , there are two.

Answer

By Serre's rational-homotopy computation, $S_{Q}^{3}$ is an odd-sphere rationalisation, so it has a single non-zero rational homotopy group: $π_{3} (S_{Q}^{3}) = Q$ , and $π_{n} (S_{Q}^{3}) = 0$ for $n \neq = 3$ .

Equivalently, $S_{Q}^{3}$ is the rational Eilenberg-MacLane space $K (Q, 3)$ . The fact that odd-dimensional spheres rationalise to $K (Q, n)$ is a consequence of the Hopf-Whitney theorem applied to the rational sphere: every odd-dimensional sphere is rationally a $K (Q, n)$ .

For contrast, $S_{Q}^{2}$ has $π_{2} (S_{Q}^{2}) = Q$ and $π_{3} (S_{Q}^{2}) = Q$ (from the rationalised Hopf fibration $S_{Q}^{3} \to S_{Q}^{2} \to S_{Q}^{2}$ identifying $S_{Q}^{2}$ as a rational fibration), with all higher rational homotopy groups vanishing. Even spheres have two non-zero rational homotopy groups, in degrees $n$ and $2 n - 1$ . Rubric: full credit for the odd-sphere computation; partial credit for stating only $π_{3} = Q$ without noting the vanishing of other degrees.

Exercise 4 (medium, short-answer).

State the local-global short exact sequence for finitely generated abelian groups and explain how it underwrites Step 3 of the proof of the integral fracture theorem.

Hint

The sequence relates $A$ , $A \otimes Q$ , $A \otimes \hat{Z}$ , and $A \otimes A_{f}$ for a finitely generated abelian group $A$ . It is a homotopical lift of the number-theoretic short exact sequence $0 \to Z \to Q \oplus \hat{Z} \to A_{f} \to 0$ .

Answer

The local-global short exact sequence for a finitely generated abelian group $A$ reads $$ 0 \to A \to A \otimes \mathbb{Q} \oplus A \otimes \hat{\mathbb{Z}} \to A \otimes \mathbb{A}_f \to 0, $$ where $\hat{Z} = \prod_{p} Z_{p}$ is the profinite completion and $A_{f} = \hat{Z} \otimes Q$ is the ring of finite adeles. The first map sends $a \mapsto (a, a)$ (the two natural localisations) and the second sends $(q, z) \mapsto q - z$ (the difference of their images in the adelic completion).

The sequence is exact because it is the tensor product of $A$ with the integer short exact sequence $0 \to Z \to Q \oplus \hat{Z} \to A_{f} \to 0$ (the latter is the standard number-theoretic exact sequence; over $Z$ the modules involved are flat in the relevant senses, so tensoring with a finitely generated $A$ preserves exactness).

In Step 3 of the fracture-theorem proof, this short exact sequence appears at the level of the $n$ -th homotopy group $A = π_{n} (X)$ . The pullback corner $π_{n} (L_{Q} X) \times_{π_{n} (L_{Q} (\prod_{p} L_{p} X))} π_{n} (\prod_{p} L_{p} X)$ is exactly the kernel of the difference map $A \otimes Q \oplus A \otimes \hat{Z} \to A \otimes A_{f}$ , which by exactness equals $A = π_{n} (X)$ . Rubric: full credit for the sequence statement, the integer short exact sequence underneath, and the identification of the pullback corner with the kernel.

Exercise 5 (medium, short-answer).

Show that the $lim^{1}$ term in Step 2 of the fracture-theorem proof vanishes for the inverse system of $p$ -completions ${π_{n + 1} (L_{p} X)}_{p}$ when $X$ is of finite type.

Hint

For a finitely generated abelian group $A$ , the inverse system ${A \otimes Z_{p}}_{p}$ over primes (with the product structure) has surjective transition maps. Use the Mittag-Leffler criterion.

Answer

The Mittag-Leffler criterion states: for an inverse system ${G_{p}}_{p}$ of abelian groups with transition maps $G_{p} \to G_{q}$ , the derived inverse limit $lim^{1} G_{p}$ vanishes if for every $p$ there exists $q \geq p$ such that the image of $G_{r} \to G_{p}$ stabilises for all $r \geq q$ . A sufficient condition is that all transition maps are surjective (the image is already the full $G_{p}$ at the first step).

For the $p$ -completion family of a finite-type nilpotent space $X$ : the indexing here is over the set of primes with the discrete order, and the relevant inverse system is the partial products ${\prod_{p \leq N} π_{n + 1} (L_{p} X)}_{N}$ where the transition map from $N + 1$ to $N$ projects away the last factor. These projection maps are surjective by construction (projection onto a direct factor of a product), so the Mittag-Leffler criterion applies and $lim^{1} = 0$ .

Alternatively, finiteness of generation of $π_{n + 1} (X)$ gives that each $π_{n + 1} (L_{p} X) = π_{n + 1} (X) \otimes Z_{p}$ is a finitely generated $Z_{p}$ -module, hence compact in the $p$ -adic topology. The inverse-system structure makes this a Mittag-Leffler tower of compact abelian groups, for which $lim^{1}$ always vanishes (a stronger statement than the surjective-transition-map case). Rubric: full credit for either the surjective-projection argument or the compactness argument; both are standard.

Exercise 6 (medium, symbolic).

Verify the integral fracture theorem on a concrete homotopy group computation: compute $π_{3} (S^{2})$ by reassembling it from $π_{3} (S_{Q}^{2})$ and ${π_{3} (S_{p}^{2})}_{p}$ .

Hint

$π_{3} (S^{2}) = Z$ generated by the Hopf map. Compute its rationalisation and its $p$ -completions, then reassemble via the local-global short exact sequence.

Answer

The Hopf map $η : S^{3} \to S^{2}$ generates $π_{3} (S^{2}) = Z$ (Hopf 1931; also reproved by Pontryagin via framed cobordism). The rationalisation gives $π_{3} (S_{Q}^{2}) = Z \otimes Q = Q$ . The $p$ -completion gives $π_{3} (S_{p}^{2}) = Z \otimes Z_{p} = Z_{p}$ for every prime $p$ . The product is $\prod_{p} Z_{p} = \hat{Z}$ , the profinite completion of $Z$ . The rationalised product is $\hat{Z} \otimes Q = A_{f}$ , the ring of finite adeles.

The integral fracture theorem says $π_{3} (S^{2})$ is the pullback $$ \mathbb{Q} \times_{\mathbb{A}_f} \hat{\mathbb{Z}} = {(q, z) \in \mathbb{Q} \times \hat{\mathbb{Z}} : q = z \text{ in } \mathbb{A}_f}. $$ This pullback is exactly $Z$ : a pair $(q, z)$ has $q$ rational and $z$ profinite-integral, and equality in $A_{f}$ forces $q \in Z \subset Q$ (the integers are the elements of $Q$ that lie in $\hat{Z}$ inside $A_{f}$ , by the standard adelic characterisation of $Z$ as $Q \cap \hat{Z}$ ).

Hence $π_{3} (S^{2}) = Z$ as expected, reconstructed correctly from its rationalisation $Q$ , its profinite completion $\hat{Z}$ , and the rational-adelic gluing constraint. Rubric: full credit for the four-step calculation (rationalisation, $p$ -completions, product, pullback) culminating in the adelic identification of $Z$ .

Exercise 7 (hard, short-answer).

State the integral fracture theorem for nilpotent spaces with finite-type relaxed. Construct an explicit counterexample showing that the comparison map can fail to be a weak equivalence when $X$ is nilpotent but not of finite type.

Hint

Take an infinite wedge or product of spheres with no upper bound on the dimensions, or a space whose homotopy groups are infinite-dimensional rational vector spaces, and check whether the Milnor $lim^{1}$ vanishes.

Answer

Counterexample. Take $X = ⋁_{i \in N} S^{2}$ , an infinite wedge of $2$ -spheres. This space is simply connected, hence nilpotent. However, $π_{2} (X) = ⨁_{i \in N} Z$ , the free abelian group on countably many generators — not finitely generated, so $X$ is not of finite type.

For each prime $p$ , $π_{2} (L_{p} X) = ⨁_{N} Z_{p}$ (the $p$ -completion of the direct sum is the direct sum of $p$ -completions when the index set is countable, but the inverse limit over primes does not commute with the infinite direct sum). Concretely, $π_{2} (\prod_{p} L_{p} X) = \prod_{p} ⨁_{N} Z_{p}$ , while $\prod_{p} L_{p} (⨁_{N} S^{2}) \to \prod_{p} ⨁_{N} L_{p} S^{2}$ involves a non-vanishing $lim^{1}$ contribution from the failure of $⨁$ to commute with $\prod$ .

The $lim^{1}$ obstruction term gives $π_{2} (P (X))$ a non-zero piece not accounted for by $π_{2} (X) = ⨁_{N} Z$ . The comparison map $α_{X}$ therefore fails to be a weak equivalence on $π_{2}$ alone.

Statement of the relaxed theorem. For a connected nilpotent space $X$ (without finite-type), the comparison map $α_{X} : X \to P (X)$ fits into a Milnor short exact sequence on each homotopy group: $$ 0 \to \lim^1!p \pi{n+1}(L_p X) \to \pi_n(\mathrm{P}(X)) \to \pi_n(L_\mathbb{Q} X) \times_{\pi_n(L_\mathbb{Q}(\prod_p L_p X))} \pi_n(\prod_p L_p X) \to 0. $$ The comparison map $α_{X}$ is a weak equivalence if and only if $lim^{1}_{p} π_{n + 1} (L_{p} X) = 0$ for every $n$ . Finite-type is one sufficient condition (the one used in the main proof); other sufficient conditions include nilpotent-by-Mittag-Leffler towers. Rubric: full credit for the wedge counterexample plus the conditional statement of the relaxed theorem.

Exercise 8 (hard, short-answer).

Sketch how the integral fracture theorem extends to the stable homotopy category via the chromatic-fracture square: state the height- $n$ fracture relating $L_{K (n)}$ -localisation, $L_{E (n - 1)}$ -localisation, and the chromatic-pullback diagram.

Hint

Replace the family of primes by the chromatic height filtration, the rationalisation by $L_{E (n - 1)}$ , and the product of $p$ -completions by $L_{K (n)}$ .

Answer

The chromatic-fracture square (Hovey-Strickland 1999, building on Devinatz-Hopkins-Smith 1988 nilpotence and periodicity work) is the stable-homotopy analogue of the arithmetic square. Fix a prime $p$ and a chromatic height $n \geq 1$ . There is a commutative square in the $p$ -local stable homotopy category $Sp_{(p)}$ : $$ \begin{array}{ccc} L_{E(n)} X & \to & L_{K(n)} X \ \downarrow & & \downarrow \ L_{E(n-1)} X & \to & L_{E(n-1)} L_{K(n)} X \end{array} $$ where $E (n)$ is the Johnson-Wilson spectrum at height $n$ , $K (n)$ is Morava $K$ -theory at height $n$ , and the localisations $L_{E (n)}$ , $L_{K (n)}$ , $L_{E (n - 1)}$ are the corresponding Bousfield localisations of spectra. The chromatic-fracture theorem states that this square is a homotopy pullback: $L_{E (n)} X$ is recovered from $L_{K (n)} X$ , $L_{E (n - 1)} X$ , and the rational-gluing corner $L_{E (n - 1)} L_{K (n)} X$ .

Iterating downward from a high height, the chromatic-fracture tower assembles the $E (\infty)$ -local (= integral $p$ -local) sphere spectrum from its monochromatic strata $L_{K (n)} S^{0}$ . The pattern is the exact stable analogue of the arithmetic square's assembly of $X$ from its rational and $p$ -local pieces, with "prime" replaced by "chromatic height" and " $p$ -completion" replaced by "Morava $K$ -theory localisation".

The vanishing $lim^{1}$ hypothesis in the integral fracture becomes a chromatic-convergence hypothesis (Hopkins-Ravenel 1992 Annals of Math. 137; chromatic convergence at finite type implies the chromatic-fracture tower converges). Rubric: full credit for the square diagram, the prime-to-height analogy, and the convergence remark.

Lean formalization [Intermediate+]

The companion Lean module Codex.Modern.Homotopy.ArithmeticSquare at lean/Codex/Modern/Homotopy/ArithmeticSquare.lean states the central theorem and the sphere corollary; the proofs are sorry pending the upstream Mathlib infrastructure listed in the frontmatter lean_mathlib_gap. The schematic shape of the formalisation is:

import Mathlib.CategoryTheory.Category.Basic
import Mathlib.CategoryTheory.Functor.Category

namespace Codex.Modern.Homotopy.ArithmeticSquare

open CategoryTheory

/-- The ambient homotopy category. Placeholder for Ho(sSet_*). -/
structure HoSpaces where
  carrier : Type
  cat : Category carrier

/-- A homology theory used as the localising spectrum. -/
structure HomologyTheory where
  name : String

def HQ : HomologyTheory := { name := "HQ" }
def HFp (p : ℕ) (_hp : p.Prime) : HomologyTheory := { name := "HF_" ++ toString p }

/-- Bousfield localisation of Ho at E. Placeholder. -/
structure BousfieldLocalization (Ho : HoSpaces) (_E : HomologyTheory) where
  obj : Ho.carrier → Ho.carrier
  eta : ∀ X : Ho.carrier, Unit
  universal : Unit

/-- The Sullivan arithmetic square. -/
structure ArithmeticSquare (Ho : HoSpaces) (X : Ho.carrier) where
  productCompletion : Ho.carrier
  rationalCorner : Ho.carrier
  gluingCorner : Ho.carrier
  commutes : Unit

/-- Integral Fracture Theorem (Sullivan 1970; Bousfield-Kan 1972).
    Under nilpotence + finite-type + lim^1 vanishing, the comparison
    map X → homotopy-pullback is a weak equivalence. -/
theorem fracture_theorem
    (Ho : HoSpaces) (X : Ho.carrier)
    (_nilp : IsNilpotent Ho X)
    (_ft : HasFiniteType Ho X)
    (_lim1 : Lim1Vanishes Ho X)
    (sq : ArithmeticSquare Ho X) :
    arithmeticComparison Ho X sq = () := by
  sorry

end Codex.Modern.Homotopy.ArithmeticSquare

The four-stage proof of the integral fracture theorem (Milnor short exact sequence, $lim^{1}$ vanishing, local-global short exact sequence, Whitehead conclusion) sits behind four sorry calls in the current state. Each sorry resolves once the corresponding Mathlib infrastructure ships: the Lim1Vanishes predicate and the Mittag-Leffler criterion (small piece, formalisable from Mathlib.CategoryTheory.Limits.Shapes); the BousfieldLocalization API (medium piece, requiring the localisation construction at a spectrum); the homotopy-pullback API in the model-categorical setting (large piece, interlocking with the open ModelCategory formalisation target); and the Whitehead-theorem identification of weak equivalences in $H$ (medium piece, achievable from the Quillen-Serre model structure once it is verified). The arithmetic-square API is therefore a downstream consumer of several Mathlib formalisation targets that are independently desirable.

Advanced results [Master]

Bousfield-Kan p-completion and rationalisation

Theorem (Bousfield 1975, Localisation of spaces with respect to homology). For any connected spectrum $E$ and any pointed simplicial set $X$ , the Bousfield localisation $L_{E} X$ exists and is functorial in $X$ . The functor $L_E : \mathcal{H}_ \to \mathcal{H}_ $i s i d e m p o t e n t, an d i t s ima g e i s t h e f u l l s u b c a t e g or y o f$ E$-local objects.

The construction proceeds in two steps. First, one defines the class of $E$ -local objects as those $Z$ for which every $E$ -equivalence $f : X \to Y$ induces a bijection $f^{*} : [Y, Z] \to [X, Z]$ on homotopy classes of maps. Second, one constructs $L_{E} X$ as the small-object-argument fibrant replacement of $X$ in a model structure on $sSet_{*}$ whose weak equivalences are the $E$ -equivalences and whose fibrant objects are the $E$ -local spaces. This is the $E$ -local model structure, and the construction is a Quillen-Bousfield localisation of the standard model structure on $sSet_{*}$ in the sense of 03.12.31.

Theorem (Bousfield-Kan 1972, $p$ -completion and rationalisation as cosimplicial limits). For $X$ a nilpotent space of finite type, the $p$ -completion $L_{H F_{p}} X$ admits a cosimplicial-resolution description as the homotopy limit $$ L_p X = \operatorname{holim}_{\Delta} \big[ X \xrightarrow{} \mathbb{F}_p X \xrightarrow{} \mathbb{F}_p^2 X \xrightarrow{} \cdots \big] $$ *where $F_{p} X := map (X, K (F_{p}, *))$ is the $F_{p}$ -cochain functor.*

The construction is Bousfield-Kan §III; it produces a tower of partial completions $X \to τ_{\leq n} L_{p} X$ converging to $L_{p} X$ at finite type, and is the source of the Bousfield-Kan spectral sequence relating $E_{*} (X)$ to $π_{*} (L_{E} X)$ (Bousfield-Kan §IX) — the spectral sequence appears as a sibling unit in this chapter.

The rationalisation $L_{Q} X$ has a parallel cosimplicial description, and for simply connected $X$ admits the Sullivan PL-de Rham model: $L_{Q} X$ is recoverable from the commutative differential graded algebra $A_{P L} (X)$ of polynomial de Rham forms on $X$ , via the Sullivan minimal model construction of 03.12.06. The Quillen 1969 Annals of Math. 90 alternative construction uses differential graded Lie algebras instead of cdga's; the two are Quillen-equivalent.

The Sullivan arithmetic-square pullback diagram

Theorem (Sullivan 1970; modern reformulation Bousfield-Kan 1972 §VI; Dwyer-Greenlees-Iyengar 2006 §4). Let $X$ be a connected nilpotent space of finite type. The Sullivan arithmetic square $$ \begin{array}{ccc} X & \xrightarrow{\eta_{\mathrm{comp}}} & \prod_p L_p X \ \eta_\mathbb{Q} \downarrow & & \downarrow \ L_\mathbb{Q} X & \xrightarrow{} & L_\mathbb{Q}!\left(\prod_p L_p X\right) \end{array} $$ is a homotopy pullback square in the pointed homotopy category $\mathcal{H}_$.*

This is the central content of the chapter. The square is the homotopical incarnation of the local-global principle in number theory: an integer is recovered from its rational image and its profinite completion, modulo the adelic gluing constraint; an integral nilpotent finite-type space is recovered from its rationalisation and its product of $p$ -completions, modulo the rationalisation of the product. The proof was given as the Key theorem above.

Theorem (Hilton-Mislin-Roitberg 1975). The localisation framework extends to the un-pointed nilpotent category: if $X$ is a nilpotent space, then the arithmetic square statement is functorial in $X$ , and the functor $X \mapsto P (X)$ on the nilpotent-space category is naturally equivalent to the identity functor in the homotopy category.

This is the categorical assertion underlying the fracture theorem: the comparison map $α : id \to P$ is a natural weak equivalence on the full subcategory of nilpotent finite-type spaces. The result is the foundational reason that the rational and $p$ -local fragments of a space carry the full information.

Convergence and the $lim^{1}$ -vanishing hypothesis

Theorem (Bousfield-Kan 1972 §IX, $lim^{1}$ vanishing under finite-type). For $X$ a connected nilpotent space of finite type, the derived inverse limit $$ \lim^1!_p \pi_n(L_p X) = 0 \quad \text{for every } n \geq 1. $$

The argument was given in Step 2 of the Key theorem proof: at finite type, each $π_{n} (L_{p} X) = π_{n} (X) \otimes Z_{p}$ is a finitely generated $Z_{p}$ -module, and the inverse system over primes has surjective transition maps in the product structure, so the Mittag-Leffler criterion gives $lim^{1} = 0$ .

Theorem (May-Ponto 2012 §11; counterexample at non-finite-type). There exist connected nilpotent spaces $X$ of non-finite-type for which $lim^{1}_{p} π_{n} (L_{p} X) \neq = 0$ and the comparison map $α_{X}$ fails to be a weak equivalence.

The counterexample given in Exercise 7 — an infinite wedge of $S^{2}$ 's — exhibits this. The lesson is that finite-type is the essential structural hypothesis for the integral fracture theorem; relaxing it requires a finer convergence analysis (or replacing the inverse limit by an explicit Milnor short exact sequence with the $lim^{1}$ term tracked).

Chromatic-homotopy application — height-by-height assembly

Theorem (Hopkins-Ravenel 1992 Annals of Math. 137, Chromatic Convergence). Let $X$ be a finite $p$ -local spectrum. Then the chromatic tower $$ \cdots \to L_{E(n)} X \to L_{E(n-1)} X \to \cdots \to L_{E(0)} X = X_\mathbb{Q} $$ has homotopy inverse limit canonically equivalent to $X$ itself: $$ X \simeq \operatorname{holim}n L{E(n)} X. $$

This is the spectrum-level integral-assembly theorem analogous to the Sullivan arithmetic square. The chromatic-fracture squares at each height (Exercise 8) provide the inductive step: $L_{E (n)} X$ is recovered as a homotopy pullback of $L_{E (n - 1)} X$ and $L_{K (n)} X$ over $L_{E (n - 1)} L_{K (n)} X$ , with the analogy to the arithmetic-square corners being prime → chromatic height, rationalisation → $L_{E (n - 1)}$ , $p$ -completion → $L_{K (n)}$ , gluing corner → $L_{E (n - 1)} L_{K (n)}$ .

Theorem (Devinatz-Hopkins-Smith 1988, Nilpotence Theorem). A self-map $f : Σ^{d} X \to X$ of a finite $p$ -local spectrum is nilpotent in the stable homotopy category if and only if it is nilpotent under every Morava $K (n)$ -homology functor.

This is the foundational theorem underwriting the chromatic-fracture framework: it says the Morava $K$ -theories detect nilpotence in the stable homotopy category, so the height-by-height fracture decomposition has no "extra" information beyond what the $K (n)$ 's see. Combined with chromatic convergence, this gives that the chromatic-fracture squares fully reconstruct integral stable homotopy.

Theorem (Hovey-Strickland 1999, Morava- $K$ -theory localisation as a Bousfield localisation). The Morava $K (n)$ -theory localisation $L_{K (n)}$ is a Bousfield localisation in the sense of 03.12.31, and the monochromatic categories $L_{K (n)} Sp_{(p)}$ are the height- $n$ strata of the stable homotopy category. The chromatic-fracture square at height $n$ is the homotopy-pullback diagram $$ \begin{array}{ccc} L_{E(n)} X & \to & L_{K(n)} X \ \downarrow & & \downarrow \ L_{E(n-1)} X & \to & L_{E(n-1)} L_{K(n)} X \end{array} $$

The proof uses the relation $E (n)_{*} = E (n - 1)_{*} \times K (n)_{*}$ at the level of homology of the standard generators, combined with the chromatic-tower convergence to set up the homotopy pullback. The square is the height- $n$ assembly: it identifies $L_{E (n)}$ as the homotopy pullback of one step up the chromatic tower and the monochromatic stratum at the current height.

Theorem (Devinatz-Hopkins 1995 Topology 34, Higher Chromatic Galois Group). The action of the Morava stabiliser group $G_{n}$ on the height- $n$ monochromatic category $L_{K (n)} Sp_{(p)}$ identifies $L_{K (n)} S^{0} ≃ E_{n}^{h G_{n}}$ as the homotopy fixed points of the Lubin-Tate spectrum $E_{n}$ under the action of $G_{n}$ .

This is the Galois-theoretic shape of the height- $n$ stratum, the chromatic-homotopy analogue of the Galois-symmetry side of Sullivan's 1970 monograph (whose title explicitly invokes "Localization, Periodicity, and Galois Symmetry"). The arithmetic-square framework's deepest legacy is the realisation that integral stable homotopy is the homotopy-fixed-points of a tower of Galois-theoretic constructions on the Lubin-Tate spectra at each height.

Synthesis. The arithmetic-square framework is the foundational reason that the integral homotopy theory of nilpotent finite-type spaces splits into a rational piece and a family of $p$ -local pieces. The central insight is the local-global short exact sequence for finitely generated abelian groups, which lifts to a homotopy pullback at the space level under the nilpotence + finite-type + $lim^{1}$ -vanishing hypotheses. Putting these together with the Milnor short exact sequence, the comparison map $α_{X}$ becomes a weak equivalence and the integral homotopy theory is recovered from its rational and $p$ -local fragments. This is exactly the structure that identifies the homotopy category of nilpotent finite-type spaces with the homotopy category of arithmetic squares — Sullivan's 1970 monograph identifies an entire category-equivalence on the page where the fracture theorem is proved.

The bridge is between integral and rational/ $p$ -local computations, and it generalises in three directions: to the un-pointed nilpotent category via Hilton-Mislin-Roitberg, to the stable homotopy category via the chromatic-fracture squares of Hovey-Strickland, and to the algebraic-geometry side via Dwyer-Greenlees-Iyengar's reformulation in terms of duality over arbitrary commutative ring spectra. The pattern recurs at every layer of modern stable homotopy: chromatic convergence assembles the integral $p$ -local sphere spectrum from its monochromatic strata $L_{K (n)} S^{0}$ via iterated chromatic-fracture squares, and the assembly is the spectrum-level reincarnation of the arithmetic-square assembly for spaces. The picture is dual to the algebraic-geometry adelic philosophy, where an arithmetic scheme is recovered from its localisations at primes and its rational completion — Sullivan's 1970 invocation of "Galois Symmetry" in the title makes the duality explicit, and the modern chromatic-Galois identification of $L_{K (n)} S^{0}$ with the homotopy fixed points $E_{n}^{h G_{n}}$ closes the loop by exhibiting the height- $n$ stratum as a Galois-theoretic homotopy-fixed-point construction.

Full proof set [Master]

Proposition (Existence of $L_{E}$ as a Quillen-Bousfield localisation). Let $E$ be a connected spectrum. There exists a model structure on $\mathbf{sSet}_ $— t h e$ E $- l oc a l m o d e l s t r u c t u r e — w h ose w e ak e q u i v a l e n ces a r e t h e$ E $- e q u i v a l e n ces an d w h oseco f ib r a t i o n s a r e t h e m o n o m or p hi s m s, an d t h e i d e n t i t y f u n c t or o n$ \mathbf{sSet}_ $i s a l e f tQ u i l l e n f u n c t or f r o m t h es t an d a r d K an - Q u i l l e nm o d e l s t r u c t u r e t o t h e$ E $- l oc a l m o d e l s t r u c t u r e . T h e$ E $- l oc a l m o d e l s t r u c t u r e p r ese n t s t h e B o u s f i e l d l oc a l i s a t i o n$ L_E$ at the homotopy-category level.

Proof. Apply the left Bousfield-localisation construction of Hirschhorn 2003 §3-§4 to the Kan-Quillen model structure on $sSet_{*}$ relative to the class of $E$ -equivalences. The hypothesis required is that the Kan-Quillen model structure is left proper and combinatorial, both of which hold (Hirschhorn §13 verifies left properness for $sSet$ and combinatoriality follows from the standard small-object generation). The construction produces a new model structure whose weak equivalences are the $E$ -equivalences; the cofibrations remain monomorphisms (a Bousfield localisation only enlarges the class of weak equivalences and adjusts fibrations); the fibrant objects are the $E$ -local spaces.

The identity functor is then a left Quillen functor from the original to the localised structure because it preserves cofibrations (the cofibration class is unchanged) and acyclic cofibrations (every Kan-Quillen weak equivalence is automatically an $E$ -equivalence, so acyclic cofibrations of the original structure are acyclic cofibrations of the new). The induced derived left adjoint on homotopy categories is the Bousfield localisation $L_{E}$ , by the universal property of localisation. $□$

Proposition (Rationalisation kills $p$ -primary torsion). Let $X$ be a connected nilpotent space of finite type. The rationalisation $η_{Q} : X \to L_{Q} X$ induces a map on homotopy groups $π_{n} (η_{Q}) : π_{n} (X) \to π_{n} (L_{Q} X)$ whose kernel is the torsion subgroup of $π_{n} (X)$ .

Proof. By Serre's class theory (Serre 1953 Annals of Math. 58), for any class $C$ of abelian groups closed under extensions and quotients, a continuous map $f : X \to Y$ between simply connected spaces is a $C$ -isomorphism on homotopy groups if and only if it is a $C$ -isomorphism on homology groups. Apply with $C = {torsion abelian groups}$ to identify the rationalisation: $η_{Q}$ is the unique map (up to homotopy) inducing an isomorphism on rational homology, and Serre's theorem translates this to: $η_{Q}$ induces $π_{n} (X) \otimes Q \to π_{n} (L_{Q} X)$ which is the identity, with kernel precisely the torsion in $π_{n} (X)$ .

The nilpotent case extends Serre via the nilpotent class theory of Hilton-Mislin-Roitberg 1975: the nilpotent assumption on $π_{1} (X)$ and the action on higher homotopy groups guarantees that the rationalisation is well-defined and behaves as $π_{n} (X) \otimes Q$ . The finite-type assumption guarantees that the tensor product is a meaningful finite-dimensional $Q$ -vector space at each level. $□$

Proposition ( $p$ -completion identifies with $p$ -adic tensor product). Let $X$ be a connected nilpotent space of finite type and let $p$ be a prime. The $p$ -completion map $η_{p} : X \to L_{p} X$ induces on homotopy groups $π_{n} (X) \to π_{n} (L_{p} X) = π_{n} (X) \otimes_{Z} Z_{p}$ , where $Z_{p}$ denotes the $p$ -adic integers.

Proof. The Bousfield-Kan $p$ -completion construction gives $L_{p} X$ as the homotopy limit of the cosimplicial $F_{p}$ -resolution of $X$ (Bousfield-Kan 1972 §III). For nilpotent finite-type $X$ , the cosimplicial resolution converges and the induced map on $π_{n}$ is the $p$ -adic completion of $π_{n} (X)$ : $$ \pi_n(L_p X) = \lim_{k} \pi_n(X) / p^k = \pi_n(X) \otimes_\mathbb{Z} \mathbb{Z}_p. $$ The second equality holds because $π_{n} (X)$ is finitely generated (so its $p$ -adic completion in the $lim$ -sense coincides with the tensor product $\otimes Z_{p}$ by the structure theorem for finitely generated abelian groups). $□$

Proposition (Milnor short exact sequence for the arithmetic-square pullback). For any homotopy-pullback square $P = A \times_{C}^{h} B$ in $\mathbf{sSet}_$, the homotopy groups fit into a Mayer-Vietoris-type exact sequence* $$ \cdots \to \pi_{n+1}(C) \to \pi_n(\mathrm{P}) \to \pi_n(A) \times \pi_n(B) \to \pi_n(C) \to \pi_{n-1}(\mathrm{P}) \to \cdots $$

Proof. The homotopy pullback $P = A \times_{C}^{h} B$ is the fibre of the difference map $A \times B \to C$ , $(a, b) \mapsto f (a) g (b)^{- 1}$ (in the group case; for spaces use the standard fibre-sequence construction). The long exact sequence of homotopy groups for this fibration gives the Mayer-Vietoris sequence claimed.

For the arithmetic-square pullback, applying this with $A = L_{Q} X$ , $B = \prod_{p} L_{p} X$ , $C = L_{Q} (\prod_{p} L_{p} X)$ , the long exact sequence specialises. The pieces of the long exact sequence assemble into the local-global short exact sequence of Step 3 of the Key theorem proof, with the $lim^{1}$ contribution (from the inverse-limit interpretation of $\prod_{p} L_{p} X$ over the family of primes) entering through the $π_{n + 1} (C)$ term. $□$

Proposition (Local-global short exact sequence for $Z$ ). The standard adelic short exact sequence $$ 0 \to \mathbb{Z} \to \mathbb{Q} \oplus \hat{\mathbb{Z}} \to \mathbb{A}_f \to 0 $$ is exact, where $\hat{Z} = \prod_{p} Z_{p}$ is the profinite completion of $Z$ and $A_{f} = \hat{Z} \otimes_{Z} Q$ is the ring of finite adeles.

Proof. Injectivity of $Z \to Q \oplus \hat{Z}$ sending $n \mapsto (n, n)$ is immediate: the first coordinate is injective on $Z$ . Surjectivity onto $A_{f}$ of the difference map $(q, z) \mapsto q - z$ : given $a \in A_{f}$ , the decomposition $A_{f} = Q + \hat{Z}$ in the additive group $A_{f}$ (where the sum is not direct: the intersection $Q \cap \hat{Z} = Z$ inside $A_{f}$ ) gives $a = q + z$ for some $q \in Q$ , $z \in \hat{Z}$ ; then $(q, - z) \in Q \oplus \hat{Z}$ maps to $q + z = a$ .

Exactness in the middle: kernel of $(q, z) \mapsto q - z$ is ${(q, q) : q \in Q \cap \hat{Z} \subset A_{f}}$ . The intersection $Q \cap \hat{Z}$ inside $A_{f}$ is the set of rationals $q$ such that $q \in Z_{p}$ for every prime $p$ — equivalently, the rationals with no negative $p$ -adic valuation at any prime, which by the structure theorem of $Q$ is exactly $Z$ . So the kernel equals the image of $Z$ under the diagonal embedding $n \mapsto (n, n)$ . Exactness is established.

Tensoring with a finitely generated abelian group $A$ preserves the exact sequence (since $\hat{Z}$ is flat over $Z$ on finitely generated modules — flatness in the strict sense fails, but for finitely generated modules the relevant Tor groups vanish — and $Q$ is flat over $Z$ ). $□$

Connections [Master]

Quillen model category 03.12.31. The Bousfield localisations $L_{E}$ at a homology theory $E$ are constructed as left Bousfield localisations of the Kan-Quillen model structure on $sSet_{*}$ , in the sense developed in that unit. The arithmetic square at $X$ is then a homotopy-pullback diagram in the Bousfield-localised homotopy category, and the fracture theorem is the statement that the comparison map $α_{X}$ from $X$ to the homotopy pullback is a weak equivalence — purely a model-categorical statement once the localisations have been set up.
Sullivan minimal models 03.12.06. The rational corner $L_{Q} X$ of the arithmetic square is identified with the Sullivan minimal model of $X$ via the Bousfield-Gugenheim PL-de Rham equivalence. For simply connected $X$ of finite type, the minimal model encodes the rational homotopy type completely, so the bottom-left corner of the arithmetic square is recoverable from algebraic data. The integral fracture theorem then identifies the integral homotopy type as the homotopy pullback of the algebraic rational data and the $p$ -local topological data — a hybrid algebraic-topological reconstruction.
Homotopy colimit (Bousfield-Kan construction) 03.12.37. The Bousfield-Kan completion construction at $p$ is built from a cosimplicial diagram of $F_{p}$ -resolutions, and the $p$ -completion $L_{p} X$ is the homotopy limit of this cosimplicial diagram. The arithmetic square then assembles these holim constructions over the family of primes via the global cosimplicial machinery developed in the sibling unit.
Bousfield-Kan spectral sequence 03.12.38. The Bousfield-Kan spectral sequence relates $E_{*} (X)$ to $π_{*} (L_{E} X)$ as the spectral-sequence-of-a-cosimplicial-resolution applied to the $E$ -completion tower. The arithmetic-square framework provides the global assembly of these spectral sequences across the family of primes: integral homotopy computations can proceed by computing the Bousfield-Kan spectral sequence at $L_{Q}$ and at each $L_{p}$ separately, then assembling via the fracture pullback.
Eilenberg-MacLane spaces 03.12.05. The localising spectra in the arithmetic square — the rational $H Q$ and the mod- $p$ $H F_{p}$ — are the Eilenberg-MacLane spectra. The Bousfield localisations $L_{H Q}$ and $L_{H F_{p}}$ are therefore both presented by Eilenberg-MacLane representability, and the arithmetic square's corners are all built from Eilenberg-MacLane-style targets.
CW complex 03.12.10. The fracture theorem holds for nilpotent finite-type CW complexes; the cofibrant-fibrant replacement underlying the model-categorical statement of the fracture theorem is the CW-replacement framework of the sibling unit. The integral fracture theorem can be re-stated as: every nilpotent finite-type CW complex is recoverable as the homotopy pullback of three CW-complex corners.
Hurewicz theorem 03.12.19. The integral fracture theorem applies most cleanly to spaces whose homotopy groups are accessible via Hurewicz arguments (simply connected, low-dimensional, or under classical hypotheses on the singular cohomology). For such spaces the fracture-theorem proof can be replayed with explicit homotopy-group computations via Hurewicz isomorphisms, providing concrete check-cases for the abstract pullback statement.
Whitehead theorem 03.12.20. The conclusion of the fracture theorem — that $α_{X}$ is a weak equivalence — is verified at the homotopy-group level via Step 4 of the Key theorem proof, which appeals to the Whitehead theorem in the form developed in the sibling unit. The Whitehead theorem is the structural bridge between $π_{*}$ -isomorphism (computable from the local-global short exact sequence) and weak equivalence (the homotopical conclusion).

Historical & philosophical context [Master]

Dennis Sullivan introduced the arithmetic-square framework in his 1970 MIT lecture notes Geometric Topology: Localization, Periodicity, and Galois Symmetry ^{[Sullivan 1970]}. The notes circulated in informal mimeographed form for over three decades, were reprinted with editorial commentary as K-Monographs in Mathematics 8 (Springer 2005), and remain the canonical originator reference. Sullivan's motivation was the unification of two seemingly disparate streams: the rational homotopy theory developed by Quillen in his 1969 Annals of Mathematics paper ^{[Quillen 1969]} using differential graded Lie algebras, and the $p$ -local techniques pioneered by Adams, Hilton, and others in the 1960s for computing stable and unstable homotopy groups. The arithmetic square exhibited these as two sides of a single integral assembly, with the gluing data encoding the rational image of the prime-by-prime tower.

The modern reframing of Sullivan's construction as a homotopy-limit / homotopy-pullback in a Bousfield-localised model category is due to Aldridge Bousfield and Daniel Kan, whose 1972 monograph Homotopy Limits, Completions, and Localizations ^{[Bousfield-Kan 1972]} (Springer Lecture Notes in Mathematics 304) developed the full functorial framework. Bousfield-Kan §VI gives the arithmetic-square statement in the form proved here, with the $lim^{1}$ -vanishing hypothesis explicit and the proof organised around the Milnor short exact sequence. Their construction made the Sullivan framework available to the working homotopy theorist: where Sullivan's original construction was geometric and proof-heavy, Bousfield-Kan turned it into a category-theoretic recipe.

The chromatic-homotopy extension of the arithmetic-square pattern emerged in the 1980s-1990s, beginning with Devinatz-Hopkins-Smith 1988 ^{[Devinatz-Hopkins-Smith 1988]} (the Nilpotence Theorem identifying Morava $K$ -theories as the chromatic detectors), Hopkins-Ravenel 1992 Annals of Mathematics 137 ^{[Hopkins-Ravenel 1992]} (the Chromatic Convergence Theorem assembling finite spectra from their chromatic localisations), and culminating in Hovey-Strickland's 1999 monograph Morava $K$ -theories and Localisation ^{[Hovey-Strickland 1999]} (AMS Memoirs 666), which axiomatised the chromatic-fracture squares as Bousfield-localised pullbacks at each height. The modern arithmetic-fracture framework of Dwyer-Greenlees-Iyengar 2006 Advances in Mathematics 200 ^{[Dwyer-Greenlees-Iyengar 2006]} unifies the space-level and spectrum-level pictures under a single duality-in-algebra-and-topology framework, with the Sullivan arithmetic square and the chromatic-fracture squares as the two foundational examples.

The lineage continues through derived algebraic geometry (Toën-Vezzosi 2008, Lurie 2009-) where the arithmetic-square pattern is the foundational example of derived descent: a derived-algebraic-geometric object is recovered from its rational and $p$ -local pieces by exactly the Sullivan-Bousfield-Kan homotopy-pullback recipe. The unifying philosophical principle — that integral structure is assembled from rational and local-at-each-prime pieces via a homotopy-pullback construction — propagates from number theory (where it is the classical local-global / adelic principle) through algebraic topology (Sullivan-Bousfield-Kan) through derived algebraic geometry (Lurie-Toën-Vezzosi) and is now a standard idiom in modern homotopical mathematics.

Bibliography [Master]

@book{Sullivan1970,
  author    = {Sullivan, Dennis P.},
  title     = {Geometric Topology: Localization, Periodicity, and {G}alois Symmetry},
  series    = {K-Monographs in Mathematics},
  volume    = {8},
  publisher = {Springer},
  year      = {2005},
  note      = {Revised and edited version of the 1970 MIT lecture notes}
}

@book{BousfieldKan1972,
  author    = {Bousfield, Aldridge K. and Kan, Daniel M.},
  title     = {Homotopy Limits, Completions and Localizations},
  series    = {Lecture Notes in Mathematics},
  volume    = {304},
  publisher = {Springer-Verlag},
  year      = {1972}
}

@article{Bousfield1975,
  author    = {Bousfield, Aldridge K.},
  title     = {The localization of spaces with respect to homology},
  journal   = {Topology},
  volume    = {14},
  year      = {1975},
  pages     = {133--150}
}

@article{Quillen1969,
  author    = {Quillen, Daniel G.},
  title     = {Rational homotopy theory},
  journal   = {Annals of Mathematics},
  volume    = {90},
  year      = {1969},
  pages     = {205--295}
}

@article{DwyerGreenleesIyengar2006,
  author    = {Dwyer, William G. and Greenlees, John P. C. and Iyengar, Srikanth B.},
  title     = {Duality in algebra and topology},
  journal   = {Advances in Mathematics},
  volume    = {200},
  year      = {2006},
  pages     = {357--402}
}

@book{MayPonto2012,
  author    = {May, J. Peter and Ponto, Kate},
  title     = {More Concise Algebraic Topology: Localization, Completion, and Model Categories},
  publisher = {University of Chicago Press},
  year      = {2012},
  series    = {Chicago Lectures in Mathematics}
}

@article{DevinatzHopkinsSmith1988,
  author    = {Devinatz, Ethan S. and Hopkins, Michael J. and Smith, Jeffrey H.},
  title     = {Nilpotence and stable homotopy theory {I}},
  journal   = {Annals of Mathematics},
  volume    = {128},
  year      = {1988},
  pages     = {207--241}
}

@article{HopkinsRavenel1992,
  author    = {Hopkins, Michael J. and Ravenel, Douglas C.},
  title     = {Suspension spectra are harmonic},
  journal   = {Bulletin of the Sociedad Matem{\'a}tica Mexicana},
  volume    = {37},
  year      = {1992},
  pages     = {271--279},
  note      = {Companion to the chromatic-convergence theorem of \emph{Annals of Mathematics} 137 (1992)}
}

@book{HoveyStrickland1999,
  author    = {Hovey, Mark and Strickland, Neil P.},
  title     = {Morava {$K$}-theories and Localisation},
  series    = {Memoirs of the American Mathematical Society},
  volume    = {666},
  publisher = {American Mathematical Society},
  year      = {1999}
}

@incollection{HiltonMislinRoitberg1975,
  author    = {Hilton, Peter and Mislin, Guido and Roitberg, Joseph},
  title     = {Localization of Nilpotent Groups and Spaces},
  series    = {North-Holland Mathematics Studies},
  volume    = {15},
  publisher = {North-Holland},
  year      = {1975}
}

@article{Serre1953,
  author    = {Serre, Jean-Pierre},
  title     = {Groupes d'homotopie et classes de groupes ab{\'e}liens},
  journal   = {Annals of Mathematics},
  volume    = {58},
  year      = {1953},
  pages     = {258--294}
}

@article{Hess2007,
  author    = {Hess, Kathryn},
  title     = {Rational homotopy theory: a brief introduction},
  journal   = {arXiv preprint},
  year      = {2006},
  note      = {arXiv:math/0604626}
}

@book{Neisendorfer2010,
  author    = {Neisendorfer, Joseph},
  title     = {Algebraic Methods in Unstable Homotopy Theory},
  series    = {New Mathematical Monographs},
  volume    = {12},
  publisher = {Cambridge University Press},
  year      = {2010}
}

@book{Hatcher2002,
  author    = {Hatcher, Allen},
  title     = {Algebraic Topology},
  publisher = {Cambridge University Press},
  year      = {2002}
}

Prerequisites

03.12.31
03.12.37
03.12.38

Tier anchors

beginner: Hatcher §2.A / §4 informal; May-Ponto *More Concise Algebraic Topology* (Chicago 2012) Ch. 10 introduction; Bousfield-Kan 1972 LNM 304 motivational §I
intermediate: Bousfield-Kan 1972 *Homotopy Limits, Completions and Localizations* LNM 304, §VI; May-Ponto 2012 Ch. 10-12; Hatcher §4.K
master: Sullivan 1970 *Geometric Topology: Localization, Periodicity, and Galois Symmetry* (MIT notes; revised K-Monographs in Mathematics 8, 2005) — originator; Bousfield-Kan 1972 LNM 304 §VI (modern reframing as a homotopy-limit construction); Dwyer-Greenlees-Iyengar 2006 *Duality in algebra and topology* (Adv. Math. 200) §4 (modern arithmetic-fracture treatment); May-Ponto 2012 *More Concise Algebraic Topology* Ch. 10-13 (textbook synthesis); Hess 2007 *Rational homotopy theory: a brief introduction* arXiv:math/0604626; Neisendorfer 2010 *Algebraic Methods in Unstable Homotopy Theory* (Cambridge) Ch. 12 (nilpotent and finite-type assumptions)

References

TODO_REF
Sullivan — Geometric Topology, Localization, Periodicity and Galois Symmetry · MIT notes 1970; revised as K-Monographs in Mathematics 8 (Springer 2005); originator of the arithmetic square; §3-§5 of the 1970 manuscript introduce p-localisation and the integral pullback; the 2005 revision adds an editorial introduction and corrects several typographic errors
TODO_REF
Bousfield-Kan — Homotopy Limits, Completions and Localizations · Springer Lecture Notes in Mathematics 304 (1972); §I-§II Bousfield localisation of a homotopy category at a homology theory, §III p-completion as a Bousfield localisation, §IV-§V completions of simplicial sets, §VI the arithmetic square and fracture theorem with the lim^1 hypothesis explicit
TODO_REF
Dwyer-Greenlees-Iyengar — Duality in algebra and topology · Advances in Mathematics 200 (2006), pp. 357-402; §4 develops the arithmetic-fracture framework over arbitrary commutative ring spectra and identifies the chromatic-tower assembly as the relevant generalisation
TODO_REF
May-Ponto — More Concise Algebraic Topology · University of Chicago Press 2012; Ch. 10 localisations and completions of spaces and spectra, Ch. 11 the arithmetic square in the nilpotent and finite-type cases, Ch. 12 fracture theorems, Ch. 13 chromatic applications
TODO_REF
Hatcher — Algebraic Topology · Cambridge 2002; §4.K p-completion in the simply connected case and the rational homotopy theory chapter §4.J as motivation
TODO_REF
Quillen — Rational homotopy theory · Annals of Mathematics 90 (1969), pp. 205-295; Quillen's original construction of rationalisation via differential graded Lie algebras; foundational for the rational corner of the arithmetic square
TODO_REF
Hess — Rational homotopy theory: a brief introduction · arXiv:math/0604626 (2006); subsequently in the IMA Volumes 152; §3 the rationalisation functor as Bousfield localisation, §4 link to the integral story
TODO_REF
Neisendorfer — Algebraic Methods in Unstable Homotopy Theory · Cambridge 2010; Ch. 12 nilpotent spaces and finite-type assumptions, Ch. 13 the integral fracture in the nilpotent finite-type case, full proofs with lim^1 bookkeeping
TODO_REF
Hovey-Strickland — Morava K-theories and localisation · Memoirs of the AMS 666 (1999); §6-§7 chromatic-fracture squares as the height-by-height fracture analogue, foundational for chromatic homotopy theory
TODO_REF
Adams — Stable Homotopy and Generalised Homology · University of Chicago Press 1974; Part II treats integral assembly from p-local data; the historical motivation for the modern arithmetic square in stable homotopy

Lean module

Codex.Modern.Homotopy.ArithmeticSquare

Mathlib gap

Mathlib has the foundational category theory of limits and colimits
in `Mathlib.CategoryTheory.Limits.*` and the simplicial-set
scaffolding in `Mathlib.AlgebraicTopology.*`, together with the
early-stage model-category infrastructure in
`Mathlib.AlgebraicTopology.ModelCategory.*`, but does not yet ship
the full Bousfield localisation API needed for the arithmetic
square. Specifically missing: (i) a bundled
`BousfieldLocalization` predicate on a model category or
homotopy category, with the localisation functor `L_E` for a
homology theory `E` and the universality witness; (ii) the
p-completion `L_p` and rationalisation `L_Q` endofunctors as
verified instances on the pointed homotopy category; (iii) the
homotopy-pullback construction at the model-categorical level
with the Milnor short exact sequence relating its homotopy groups
to the lim and lim^1 terms; (iv) the nilpotent-space predicate
with its action-on-homotopy-groups encoding; (v) the
finite-type predicate on the homotopy groups; (vi) the formal
statement of Sullivan's fracture theorem as a Quillen equivalence
between the homotopy category of nilpotent finite-type spaces
and the homotopy category of arithmetic squares. The companion
Lean file states the central theorem and its sphere corollary with
`sorry` proofs pending these infrastructure pieces. The aggregated
formalisation target — a complete `ArithmeticSquare` and
`FractureTheorem` API tied to verified `BousfieldLocalization`
and `HomotopyPullback` instances — is a substantial multi-month
Mathlib project; it would interlock with the open formalisation
targets for `ModelCategory`, the small-object argument, and the
Milnor short exact sequence.

Reviewer

TBD

Estimated time

beginner: 22m
intermediate: 50m
master: 100m

Intuition [Beginner]

Visual [Beginner]

Worked example [Beginner]

Check your understanding [Beginner]

Formal definition [Intermediate+]

Counterexamples to common slips

Key theorem with proof [Intermediate+]

Exercises [Intermediate+]

Lean formalization [Intermediate+]

Advanced results [Master]

Bousfield-Kan p-completion and rationalisation

The Sullivan arithmetic-square pullback diagram

Convergence and the lim1-vanishing hypothesis

Chromatic-homotopy application — height-by-height assembly

Full proof set [Master]

Connections [Master]

Historical & philosophical context [Master]

Bibliography [Master]

Convergence and the $lim^{1}$ -vanishing hypothesis