Combinatorial Pontryagin classes and exotic 7-spheres

Intuition [Beginner]

In ordinary geometry every shape comes with a notion of what counts as a smooth deformation: a path is smooth if it has no corners, a surface is smooth if it has well-defined tangent planes everywhere. For familiar shapes in three dimensions, there is essentially one way to be smooth. A sphere is a sphere, and any reasonable smoothing produces the same object up to a smooth bending.

In 1956, John Milnor discovered that this comforting picture fails badly in seven dimensions. He constructed a closed seven-dimensional shape that looks identical to the standard seven-sphere from the perspective of continuous deformation, yet refuses to be smoothly deformed onto it. The shape is what we now call an exotic seven-sphere. It carries a smooth structure that is genuinely different from the standard one, even though topologically the two shapes cannot be told apart.

The diagnostic that detects this difference is a characteristic number. Milnor showed that an eight-dimensional shape with the exotic seven-sphere as its boundary would have a number that should be an integer but comes out as a fraction. Since the standard seven-sphere has no such obstruction, the two cannot be smoothly equivalent.

Visual [Beginner]

Picture a stack of seven-dimensional fibres glued together over a four-dimensional base. The base is the four-sphere; each fibre is a seven-sphere. The way the fibres twist as you walk around the base is controlled by two integers. For most choices of those two integers, the resulting total space looks like a seven-sphere from a continuous-shape viewpoint, but the way the seven-sphere is built carries an internal scrambling that cannot be undone by smooth deformation.

The picture also hints at the diagnostic. The bundle is the boundary of an eight-dimensional disc bundle. That eight-dimensional shape has a special number attached to it. If the seven-sphere boundary were the standard seven-sphere, the number would be a whole integer. For the exotic seven-spheres, the same recipe produces a fraction that is impossible for any genuine standard sphere.

Worked example [Beginner]

Milnor's family of seven-dimensional shapes is built from pairs of integers $(h,j)$. The simplest non-standard choice is $h=1$, $j=0$. For this pair, the construction produces a closed seven-dimensional shape called $M_{1,0}$.

A theorem of Milnor says: if $M_{h,j}$ is the standard seven-sphere with its standard smooth structure, then the integer $h-j$ must be divisible by $7$. So if $h-j$ is not a multiple of $7$, the shape $M_{h,j}$ is an exotic seven-sphere.

Step 1. Take $h=1$, $j=0$. Compute $h-j=1$.

Step 2. Check whether $7$ divides $1$. It does not, since $1÷7$ is not a whole number.

Step 3. Conclude that $M_{1,0}$ is an exotic seven-sphere. It looks like the standard seven-sphere from the perspective of continuous deformation, but the smooth structure is different.

What this tells us: a single integer divisibility check, applied to a pair of bundle parameters, decides whether a smooth structure on a topological seven-sphere is the standard one or one of the exotic ones.

Check your understanding [Beginner]

Formal definition [Intermediate+]

Let $M$ be a closed topological manifold of dimension $n$. A smooth structure on $M$ is a maximal atlas of charts whose transition functions are smooth (i.e., $C^{\infty }$). Two smooth structures on the same topological manifold are diffeomorphic if there is a smooth homeomorphism between them preserving the smooth atlases. The set of equivalence classes of smooth structures on $M$ modulo diffeomorphism is the smoothing set of $M$.

For the topological $n$-sphere $S^n$, the smoothing set carries an abelian group structure under the connected sum: given two smooth $n$-spheres ${\Sigma }_1,{\Sigma }_2$, remove a small open disc from each and glue the boundaries by an orientation-reversing diffeomorphism of $S^{n-1}$. The resulting smooth $n$-sphere is the connect sum ${\Sigma }_1\#{\Sigma }_2$. The identity is the standard sphere; inverses come from reversing orientation. The resulting group is

$${\Theta }_n:=\{\text{oriented homotopy }n\text{-spheres}\}/\text{ orientation-preserving diffeomorphism},$$

the group of homotopy $n$-spheres, introduced by Kervaire-Milnor 1963 ^{[Kervaire-Milnor 1963]}.

For an oriented real vector bundle $E\to M$ over a smooth manifold, the Pontryagin classes $p_i(E)\in H^{4i}(M;\mathbb{Z})$ are the smooth-category characteristic classes constructed in 03.06.04 via Chern-Weil theory or via classifying maps to $\mathrm{BSO}$. For a closed oriented $4k$-manifold $M$, the Pontryagin numbers $p_I[M]=⟨p_{i_1}(TM)\cup \cdots \cup p_{i_r}(TM),[M]⟩\in \mathbb{Z}$ are integers indexed by partitions $I=(i_1,\dots ,i_r)$ of $k$, as developed in 03.06.10.

Milnor's $S^3$-bundles over $S^4$

The construction begins with the principal $\mathrm{SO}(4)$-bundle classifying space level. The homotopy group ${\pi }_3(\mathrm{SO}(4))$ classifies principal $\mathrm{SO}(4)$-bundles over $S^4$. A direct computation gives

$${\pi }_3(\mathrm{SO}(4))\cong \mathbb{Z}\oplus \mathbb{Z},$$

with generators encoded by two homomorphisms $\sigma ,\rho :\mathrm{Sp}(1)=S^3\to \mathrm{SO}(4)$ where $\sigma (q)\cdot x=qx{q}^{-1}$ is conjugation and $\rho (q)\cdot x=qx$ is left multiplication of unit quaternions ^{[Milnor 1956]}. Concretely, writing ${\mathbb{R}}^4=\mathbb{H}$ as the quaternions and using the parameters $(h,j)\in \mathbb{Z}\oplus \mathbb{Z}$, the clutching function

$${\varphi }_{h,j}:S^3\to \mathrm{SO}(4),{\varphi }_{h,j}(q)\cdot x=q^{h}x{q}^j$$

generates a principal $\mathrm{SO}(4)$-bundle $P_{h,j}\to S^4$. The associated rank-$4$ vector bundle has Euler class $h+j$ and first Pontryagin class $p_1=2(h-j)$ in $H^4(S^4;\mathbb{Z})\cong \mathbb{Z}$. The sphere bundle is

$$M_{h,j}:=\mathrm{Sph}(P_{h,j})\to S^4,M_{h,j}\text{ is a closed 7-manifold},$$

obtained by replacing each fibre ${\mathbb{R}}^4$ with the unit $S^3\subset {\mathbb{R}}^4$. The associated disc bundle is the closed $8$-manifold-with-boundary

$$N_{h,j}:=\mathrm{Disc}(P_{h,j})\to S^4,\partial N_{h,j}=M_{h,j}.$$

The Milnor invariant $\lambda$

For a closed oriented $(4k-1)$-manifold $\Sigma$ that bounds a parallelizable $4k$-manifold $W$, Milnor introduced the residue

$$\lambda (\Sigma ):=\frac{2\cdot \mathrm{sig}(W)-⟨L_k(p_1,\dots ,p_k)(TW),[W,\partial W]⟩\cdot \text{factor}}{\text{denominator}}(\mathrm{mod}1),$$

defined precisely below for the case $k=2$ as a $\mathbb{Z}/7\mathbb{Z}$-valued invariant, which is independent of the choice of bounding $W$ and which vanishes if $\Sigma$ is the standard sphere with its standard smooth structure. This $\lambda$ is the Milnor invariant of the smooth structure.

Counterexamples to common slips [Intermediate+]

• Confusing ${\mathbb{R}}^4$ with ${\mathbb{R}}^n$ for $n\ne 4$. Milnor's $S^3$-bundle construction works specifically because ${\pi }_3(\mathrm{SO}(4))\cong \mathbb{Z}\oplus \mathbb{Z}$ has two independent generators. For other dimensions, the structure group is rigid; in dimension five through twelve the homotopy groups give different exotic-sphere counts via different mechanisms (most importantly the surgery exact sequence of Kervaire-Milnor 1963).

• Confusing $h+j$ with $h-j$. The Euler class of the rank-$4$ bundle is $h+j$, while the first Pontryagin class is $p_1=2(h-j)$. The diffeomorphism diagnostic $\lambda$ depends on $h-j$ via the L-polynomial; the bundle's topological type (whether the total space is a homotopy sphere) depends on $h+j$. They are independent constraints.

• Forgetting that Pontryagin classes are integral on smooth manifolds. The Hirzebruch L-polynomial $L_2=(7p_2-p_1^2)/45$ has a rational coefficient. The product $⟨L_2,[M]⟩$ is the signature, hence an integer, but this is a substantive integrality constraint, not a generality — it forces relations among Pontryagin numbers. The exotic-sphere argument exploits the failure of this integrality on a hypothetical parallel filling.

Key theorem with proof [Intermediate+]

Theorem (Milnor 1956, exotic 7-sphere). Let $h+j=1$, so that $M_{h,j}$ is a homotopy 7-sphere. If $(h-j{)}^2\not\equiv 1(\mathrm{mod}7)$, then $M_{h,j}$ is homeomorphic but not diffeomorphic to the standard 7-sphere $S^7$.

Proof. The argument has two halves: (1) $M_{h,j}$ is homeomorphic to $S^7$ when $h+j=1$, and (2) when $(h-j{)}^2\not\equiv 1(\mathrm{mod}7)$, there is no smooth bundle filling of $M_{h,j}$ extending the standard structure of $S^7$.

Step 1. Topological identification with $S^7$. The Euler class of the rank-$4$ vector bundle underlying $M_{h,j}$ over $S^4$ is $e=h+j$. The Gysin sequence of the associated $S^3$-bundle gives, in integral cohomology,

$$\cdots \to H^{n-4}(S^4)\stackrel{\cup e}{\to }{H}^n(S^4)\stackrel{{\pi }^{\ast }}{\to }{H}^n(M_{h,j})\to H^{n-3}(S^4)\to \cdots$$

If $e=h+j=\pm 1$, multiplication by $e$ is an isomorphism on $H^0$ and on $H^4$, so ${\pi }^{\ast }$ has zero image in non-zero degrees other than the top one. The cohomology $H^{\ast }(M_{h,j};\mathbb{Z})$ then matches $H^{\ast }(S^7;\mathbb{Z})$. By simple connectivity and the Hurewicz theorem (which lifts the cohomology match to a homotopy equivalence via the $h$-cobordism theorem of Smale 1962 ^{[Smale 1962]}), $M_{h,j}$ is homeomorphic to $S^7$.

Step 2. Smooth obstruction via the signature formula. Suppose for contradiction that $M_{h,j}$ is diffeomorphic to $S^7$ with its standard smooth structure. Then $M_{h,j}$ bounds the standard $8$-disc $D^8$, which is parallelizable. The associated disc bundle $N_{h,j}$ is also a parallelizable smooth $8$-manifold with $\partial N_{h,j}=M_{h,j}$.

Glue $N_{h,j}$ to the standard $D^8$ along the assumed diffeomorphism $M_{h,j}\cong S^7$ to obtain a closed smooth $8$-manifold $X=N_{h,j}{\cup }_{M_{h,j}}{D}^8$. By the Hirzebruch signature theorem 03.06.11 in dimension $8$,

$$45\cdot \mathrm{sig}(X)=7\cdot p_2[X]-p_1^2[X].$$

The left side is divisible by $45$. We compute the right side in two pieces, contributing $D^8$ (zero, since the standard disc is parallelizable so its Pontryagin numbers vanish under the gluing) and contributing $N_{h,j}$. The signature of $X$ is the signature of the intersection form on $H^4(X;\mathbb{R})\cong \mathbb{Z}$, which (by direct computation on the bundle structure) equals $\mathrm{sig}(X)=+1$ when normalized appropriately.

The first Pontryagin number computes as $p_1^2[N_{h,j}]=4(h-j{)}^2$ from the bundle clutching data, by the Chern-Weil computation on the principal $\mathrm{SO}(4)$-bundle: $p_1(N_{h,j})=2(h-j)\cdot a$ where $a\in H^4(N_{h,j};\mathbb{Z})$ pulls back from the base $S^4$, and $a^2[N_{h,j}]=\pm 1$ by orientation conventions. The second Pontryagin number $p_2[X]$ is an integer determined by the smooth structure but is the unknown in the equation.

Substituting and solving:

$$45\cdot 1=7p_2[X]-4(h-j{)}^2,p_2[X]=\frac{45+4(h-j{)}^2}{7}.$$

For $p_2[X]$ to be an integer, we need $7\mid 45+4(h-j{)}^2$. Since $45\equiv 3(\mathrm{mod}7)$ and $4\equiv -3(\mathrm{mod}7)$, this reduces to

$$3-3(h-j{)}^2\equiv 0(\mathrm{mod}7),(h-j{)}^2\equiv 1(\mathrm{mod}7).$$

If $(h-j{)}^2\not\equiv 1(\mathrm{mod}7)$, the integrality fails and the assumption that $M_{h,j}$ bounds a smooth parallelizable disc filling extending its standard smooth structure is contradicted. Hence $M_{h,j}$ is not diffeomorphic to the standard $S^7$. $\square$

Bridge. Milnor's argument builds toward 03.09.10 Atiyah-Singer index theorem, where the signature operator's integrality is the analytic shadow of the same constraint, and appears again in 03.06.10 Pontryagin numbers, where the bordism invariance of the Pontryagin numbers furnishes the structural language of the proof. The foundational reason the diagnostic works is exactly that the Hirzebruch signature formula identifies the signature of $X$ with a specific rational combination of Pontryagin numbers, and the failure of integrality on the candidate filling exposes a smooth obstruction. The central insight is that a topological homeomorphism $M_{h,j}\cong S^7$ need not extend to a diffeomorphism, and the integrality test for the L-polynomial measures the obstruction directly. Putting these together with the surgery exact sequence of 03.06.11, the bridge is to Kervaire-Milnor's full computation ${\Theta }_7=\mathbb{Z}/28\mathbb{Z}$: the same signature-integrality machinery, applied to the full image of the $J$-homomorphism in degree $7$, gives the precise count of exotic structures.

Exercises [Intermediate+]

Lean formalization [Intermediate+]

lean_status: partial — the companion module Codex.Modern.CharClasses.ExoticSpheres records the Milnor lambda invariant and the Kervaire-Milnor identity ${\Theta }_7=\mathbb{Z}/28\mathbb{Z}$ at the level of placeholder definitions and rational identities. The structure of the file mirrors the four anchor results: (i) the $S^3$-bundle parameter space ${\pi }_3(\mathrm{SO}(4))\cong \mathbb{Z}\oplus \mathbb{Z}$; (ii) the Milnor lambda invariant as a $\mathbb{Z}/7\mathbb{Z}$-valued function of $(h,j)$; (iii) the Hirzebruch integrality obstruction; (iv) the Kervaire-Milnor identity ${\Theta }_7=\mathbb{Z}/28\mathbb{Z}$. Numerical witnesses for the small cases are recorded as decide-provable rational identities:

def MilnorLambda (h j : ℤ) : ZMod 7 :=
  ((h - j) * (h - j) - 1 : ZMod 7)

-- Standard S^7 corresponds to (h, j) = (1, 0): lambda = 0.
theorem lambda_standard : MilnorLambda 1 0 = 0 := by decide

-- Exotic 7-sphere example: (h, j) = (2, -1) gives lambda = 8 ≡ 1 (mod 7).
theorem lambda_exotic_2_neg1 : MilnorLambda 2 (-1) = 1 := by decide

The full theorem kervaire_milnor_theta_7 : Theta 7 ≃ ZMod 28 carries a sorry body pending the upstream Mathlib infrastructure listed in the lean_mathlib_gap field. The placeholder $\mathbb{Z}/7\mathbb{Z}$-valued $\lambda$ is the coarsest invariant; the full $\mathbb{Z}/28\mathbb{Z}$ refinement requires the $J$-homomorphism order computation $|J({\pi }_7(\mathrm{SO}))|=240$ combined with the Kervaire-Milnor surgery exact sequence to extract the bounding-parallelizable subgroup of order $28$.

Advanced results [Master]

Milnor's construction and the signature obstruction

Milnor's 1956 construction ^{[Milnor 1956]} proceeds in three structural steps that together demonstrate the existence of multiple smooth structures on the topological $7$-sphere. The first step identifies the parameter space of fibre bundles: the homotopy group ${\pi }_3(\mathrm{SO}(4))$ classifying principal $\mathrm{SO}(4)$-bundles over $S^4$ is computed explicitly as $\mathbb{Z}\oplus \mathbb{Z}$ using the exceptional isomorphism $\mathrm{Spin}(4)\cong \mathrm{Sp}(1)\times \mathrm{Sp}(1)$.

Theorem 1 (Milnor 1956, parameter computation). ${\pi }_3(\mathrm{SO}(4))\cong \mathbb{Z}\oplus \mathbb{Z}$ with generators given by the conjugation map $\sigma :q\mapsto q\cdot q^{-1}$ on $\mathbb{H}={\mathbb{R}}^4$ and the left-multiplication map $\rho :q\mapsto q\cdot$, where $q\in \mathrm{Sp}(1)=S^3$. Under this identification, the clutching map ${\varphi }_{h,j}(q)\cdot x=q^{h}x{q}^j$ realizes the pair $(h,j)\in \mathbb{Z}\oplus \mathbb{Z}$.

The second step computes the characteristic classes of the associated bundles. For the rank-$4$ vector bundle ${\xi }_{h,j}\to S^4$ associated to ${\varphi }_{h,j}$, the Euler class is $e({\xi }_{h,j})=(h+j)\alpha$ and the first Pontryagin class is $p_1({\xi }_{h,j})=2(h-j)\alpha$, where $\alpha$ generates $H^4(S^4;\mathbb{Z})$.

Theorem 2 (Milnor 1956, exotic 7-spheres exist). For every $(h,j)$ with $h+j=1$ and $(h-j{)}^2\not\equiv 1(\mathrm{mod}7)$, the closed 7-manifold $M_{h,j}$ is homeomorphic but not diffeomorphic to $S^7$.

The proof structure is as outlined in the Intermediate proof: the topological identification with $S^7$ uses the Gysin sequence and the Smale $h$-cobordism theorem, while the smooth obstruction uses the Hirzebruch signature formula in dimension $8$ applied to the disc bundle $N_{h,j}$ glued to a hypothetical standard $D^8$ filling.

Theorem 3 (Milnor invariant). Define $\lambda (M_{h,j}):=((h-j{)}^2-1)(\mathrm{mod}7)\in \mathbb{Z}/7\mathbb{Z}$. Then $\lambda$ is well-defined on the diffeomorphism class of $M_{h,j}$ for $h+j=1$ and vanishes precisely when $M_{h,j}$ is diffeomorphic to the standard $S^7$.

The well-definedness of $\lambda$ is substantive: it depends on the choice of parallelizable filling $N_{h,j}$, but any two such fillings differ by a closed parallelizable $8$-manifold whose signature is divisible by $2$ (since the intersection form on a closed orientable spin $4$-manifold is even, by Rokhlin's theorem). The $\mathbb{Z}/7\mathbb{Z}$-valuation absorbs this ambiguity.

The group of homotopy 7-spheres ${\Theta }_7=\mathbb{Z}/28\mathbb{Z}$ (Kervaire-Milnor 1963)

Kervaire-Milnor 1963 ^{[Kervaire-Milnor 1963]} systematized the exotic-sphere phenomenon in every dimension. Their main result is the surgery-theoretic exact sequence

$$0\to {\Theta }_n^{bP}\to {\Theta }_n\to {\pi }_n^s/J({\pi }_n(\mathrm{SO}))\to 0$$

for $n\ne 4$, where ${\Theta }_n^{bP}$ is the subgroup of homotopy $n$-spheres bounding a parallelizable $(n+1)$-manifold, ${\pi }_n^s$ is the stable $n$-th homotopy group of spheres, $J:{\pi }_n(\mathrm{SO})\to {\pi }_n^s$ is the $J$-homomorphism, and the quotient at the right is the Kervaire surgery obstruction. For odd $n=2k-1$ with $k$ even (the case relevant to $n=7$), this sequence simplifies and yields:

Theorem 4 (Kervaire-Milnor 1963). For $n=4k-1$ with $k\ge 2$, the bounding-parallelizable subgroup is

$$|{\Theta }_{4k-1}^{bP}|=a_k\cdot 2^{2k-2}(2^{2k-1}-1)\cdot \mathrm{numer}(B_k/4k),$$

where $a_k\in \{1,2\}$ is determined by signature divisibility, $B_k$ is the $k$-th Bernoulli number, and $\mathrm{numer}$ denotes the numerator in lowest terms.

For $n=7$ (i.e., $k=2$): $B_2=1/30$, so $B_2/8=1/240$ and $\mathrm{numer}(B_2/8)=1$. The formula gives $|{\Theta }_7^{bP}|=1\cdot 4\cdot 7\cdot 1=28$. The quotient ${\pi }_7^s/J({\pi }_7(\mathrm{SO}))=0$ since $J$ is surjective onto ${\pi }_7^s=\mathbb{Z}/240\mathbb{Z}$ in this degree (Adams 1966 ^{[Adams 1966]}), so ${\Theta }_7={\Theta }_7^{bP}=\mathbb{Z}/28\mathbb{Z}$.

Theorem 5 (signature-mod-divisor formula). The generator of ${\Theta }_{4k-1}^{bP}$ is realized by the boundary of a parallelizable $4k$-manifold with signature $8\cdot a_k$, where $a_k$ is the signature divisibility constant of Theorem 4. For $k=2$ (so $n=7$): the generator is Milnor's $M_{1,0}$ exotic 7-sphere, and the smooth structure indexed by $m\in \mathbb{Z}/28\mathbb{Z}$ is realized by the connect sum of $m$ copies of $M_{1,0}$.

This is the structural identification: every smooth structure on $S^7$ is built by connect-summing copies of Milnor's specific construction. The $28$ structures are obtained as $S^7=M_{1,0}^{\#0},M_{1,0}^{\#1},M_{1,0}^{\#2},\dots ,M_{1,0}^{\#27}$ with $M_{1,0}^{\#28}=S^7$ (the standard sphere) again.

Combinatorial Pontryagin classes for PL manifolds

A piecewise-linear (PL) manifold $M$ is a topological manifold equipped with a maximal atlas of charts whose transition functions are piecewise-linear. Every smooth manifold admits a unique PL structure (by Whitehead 1940), but the converse fails: there exist PL manifolds that do not admit any smooth structure, and there exist PL manifolds with multiple inequivalent smooth structures.

Pontryagin classes were originally defined for smooth manifolds via Chern-Weil theory on the tangent bundle. For PL manifolds, the smooth construction is not directly available since there is no tangent bundle in the smooth sense. The breakthrough came in two stages.

Theorem 6 (Rokhlin-Schwartz 1957, Thom 1958, combinatorial Pontryagin classes). For every closed PL $n$-manifold $M$, there exist rational characteristic classes $p_i^{PL}(M)\in H^{4i}(M;\mathbb{Q})$, defined via the signatures of transverse PL submanifolds, with the following properties: (i) $p_i^{PL}$ agrees with the smooth Pontryagin class $p_i(TM)\otimes \mathbb{Q}$ when $M$ admits a compatible smooth structure; (ii) $p_i^{PL}$ is a PL invariant; (iii) the Hirzebruch signature formula $\mathrm{sig}(M)=⟨L_k(p_1^{PL},\dots ,p_k^{PL}),[M]⟩$ holds for $M$ a closed oriented PL $(4k)$-manifold.

Proof outline. Thom's construction ^{[Thom 1958]}: for a closed PL $n$-manifold $M$ and a PL embedded submanifold $V\subset M$ of codimension $4i$, the signature $\mathrm{sig}(V)$ depends only on the homology class $[V]\in H_{n-4i}(M;\mathbb{Q})$ (by a transversality argument). The assignment $[V]\mapsto \mathrm{sig}(V)$ extends to a $\mathbb{Q}$-linear functional on $H_{n-4i}(M;\mathbb{Q})$, hence by Poincaré duality (which holds for PL manifolds) corresponds to a class in $H^{4i}(M;\mathbb{Q})$. This class is the combinatorial Pontryagin class $p_i^{PL}(M)$ — although the precise normalization requires the Hirzebruch formula to invert the L-polynomial. Properties (i)-(iii) are verified by checking that the construction agrees with the smooth one on smooth manifolds and is invariant under PL isomorphism. $\square$

Theorem 7 (Novikov 1965, topological invariance of rational Pontryagin classes). The rational Pontryagin classes $p_i(M)\otimes \mathbb{Q}\in H^{4i}(M;\mathbb{Q})$ are topological invariants of $M$: if $f:M\to M^{\prime }$ is a homeomorphism of closed smooth manifolds, then $f^{\ast }(p_i(M^{\prime })\otimes \mathbb{Q})=p_i(M)\otimes \mathbb{Q}$ in $H^{4i}(M;\mathbb{Q})$.

This is one of the deepest results of geometric topology: it says that even though smooth structure is finer than topological structure (as Milnor's exotic spheres demonstrate), the rational part of the Pontryagin classes does not see this difference. The integer part does — that is exactly the smooth-structure invariant that Milnor exploits to distinguish exotic spheres from the standard sphere.

Novikov's proof ^{[Novikov 1965]} uses the manifold-with-fibred-singularities construction and the surgery exact sequence in high-dimensional topology. The argument is delicate: one cannot directly transport the smooth Pontryagin class along a homeomorphism (since the smooth structure may change), but one can construct a signature-cocycle on the topological manifold via a transverse fibration argument, and show that this cocycle reproduces the rational Pontryagin class in any compatible smooth refinement.

Distinguishing smooth structures — Milnor's invariant as the diagnostic

The Milnor $\lambda$-invariant is the prototype of a vast family of smooth-structure-detecting invariants. The general framework is:

Theorem 8 (Milnor 1956, Kervaire-Milnor 1963, smooth-structure obstruction). Let ${\Sigma }^{4k-1}$ be a closed oriented homotopy $(4k-1)$-sphere that bounds a parallelizable $4k$-manifold $W$. Define

$$\lambda (\Sigma ):=\frac{\mathrm{sig}(W)-{\mathrm{sig}}_L(W)}{a_k}\in \mathbb{Z}/N_k\mathbb{Z},$$

where ${\mathrm{sig}}_L(W):=⟨L_k(p_1,\dots ,p_k)(TW),[W,\partial W]⟩$ is the L-polynomial pairing, $a_k$ is the appropriate signature-divisibility constant, and $N_k$ is the order of ${\Theta }_{4k-1}^{bP}$. Then $\lambda (\Sigma )$ is independent of the parallelizable filling $W$, and $\lambda (\Sigma )=0$ if and only if $\Sigma$ is the standard $(4k-1)$-sphere with its standard smooth structure.

For $k=2$ (the Milnor case): $a_2=?$ and $N_2=28$, giving the $\mathbb{Z}/28\mathbb{Z}$-valued lambda. The coarsening to $\mathbb{Z}/7\mathbb{Z}$ in the original 1956 proof comes from the prime-power factorization $28=4\cdot 7$ and Milnor's use of the simplest integrality test (divisibility by $7$).

For $k=3$ (${\Theta }_{11}$) and higher: the formula extends, giving the lambda invariant on ${\Theta }_{4k-1}^{bP}$ and detecting the various exotic structures. For $k=4$ (${\Theta }_{15}$): $|{\Theta }_{15}^{bP}|=8128=2^6\cdot 127$, so the lambda is $\mathbb{Z}/8128\mathbb{Z}$-valued. The pattern $|{\Theta }_{4k-1}^{bP}|$ follows the generalized Bernoulli denominators and connects to the Milnor-Bernoulli numbers, the von Staudt-Clausen theorem, and the J-homomorphism via the Adams conjecture (Adams 1965, Quillen 1971).

Synthesis. Milnor's exotic 7-sphere construction identifies the rationality of the signature integrality with the existence of multiple smooth structures, and this is exactly the structural content that gives birth to differential topology as a discipline. The foundational reason the diagnostic works is that the Hirzebruch signature formula commits both sides — the signature, an integer, and the L-polynomial pairing, a rational combination of Pontryagin numbers — to be equal on smooth manifolds; the failure of integrality on a hypothetical smooth filling of an exotic sphere exposes the smooth obstruction directly. The central insight is that topological invariants (homeomorphism type) do not capture smooth invariants (diffeomorphism type) in dimensions $\ge 7$, and the gap is measured by characteristic-class integrality residues.

Putting these together with the surgery exact sequence of Kervaire-Milnor 1963, the lambda invariant generalizes from $\mathbb{Z}/7\mathbb{Z}$ to $\mathbb{Z}/28\mathbb{Z}$ in dimension $7$ and to $\mathbb{Z}/N_k\mathbb{Z}$ in dimension $4k-1$, identifies ${\Theta }_7$ with $\mathbb{Z}/28\mathbb{Z}$ as a complete classification of smooth structures, and connects to the $J$-homomorphism through the Adams conjecture. The bridge is to the combinatorial Pontryagin classes of Rokhlin-Schwartz 1957 and Thom 1958, which extend the smooth Pontryagin formalism to PL manifolds with rational coefficients, and to Novikov's 1965 theorem on topological invariance of rational Pontryagin classes, which establishes that even though smooth Pontryagin classes detect exotic structures, their rational parts do not. The pattern generalizes via the surgery exact sequence to all higher dimensions and finds its deepest expression in the Atiyah-Singer index theorem 03.09.10, which identifies every classical genus as an analytic-topological index pairing whose integrality reflects the existence (or non-existence) of compatible smooth fillings.

Full proof set [Master]

Proposition 1 (Computation of ${\pi }_3(\mathrm{SO}(4))$). ${\pi }_3(\mathrm{SO}(4))\cong \mathbb{Z}\oplus \mathbb{Z}$, generated by the two homomorphisms $\sigma ,\rho :\mathrm{Sp}(1)\to \mathrm{SO}(4)$ where $\sigma (q)\cdot x=qx{q}^{-1}$ and $\rho (q)\cdot x=qx$ on $\mathbb{H}={\mathbb{R}}^4$.

Proof. The double cover $\mathrm{Spin}(4)\to \mathrm{SO}(4)$ is two-to-one. By the exceptional Lie group isomorphism $\mathrm{Spin}(4)\cong \mathrm{Sp}(1)\times \mathrm{Sp}(1)$ (which follows from $\mathfrak{spin}(4)\cong \mathfrak{su}(2)\oplus \mathfrak{su}(2)$ at the Lie algebra level, lifted to a Lie group isomorphism via the connected simply-connected nature of $\mathrm{Spin}(4)$), the long exact sequence of the cover gives

$$\cdots \to {\pi }_3(\mathbb{Z}/2)\to {\pi }_3(\mathrm{Spin}(4))\to {\pi }_3(\mathrm{SO}(4))\to {\pi }_2(\mathbb{Z}/2)\to \cdots$$

Since ${\pi }_n(\mathbb{Z}/2)=0$ for $n\ge 1$, this reduces to ${\pi }_3(\mathrm{Spin}(4))\cong {\pi }_3(\mathrm{SO}(4))$. By the product structure,

$${\pi }_3(\mathrm{Spin}(4))={\pi }_3(\mathrm{Sp}(1)\times \mathrm{Sp}(1))={\pi }_3(\mathrm{Sp}(1))\oplus {\pi }_3(\mathrm{Sp}(1))=\mathbb{Z}\oplus \mathbb{Z},$$

where each factor is ${\pi }_3(S^3)=\mathbb{Z}$ since $\mathrm{Sp}(1)\cong S^3$.

The two generators correspond to the two projections $\mathrm{Sp}(1)\times \mathrm{Sp}(1)\to \mathrm{Sp}(1)$. Under the covering $\mathrm{Spin}(4)\to \mathrm{SO}(4)$, the action of $(q_1,q_2)\in \mathrm{Sp}(1)\times \mathrm{Sp}(1)$ on $\mathbb{H}\cong {\mathbb{R}}^4$ is $x\mapsto q_{1}x{q}_2^{-1}$. The first projection $(q,1)\mapsto \mathrm{SO}(4)$ acts as $x\mapsto qx$ (left multiplication, the $\rho$-generator), and the second projection $(1,q)\mapsto \mathrm{SO}(4)$ acts as $x\mapsto x{q}^{-1}$ (right multiplication by inverse). The conjugation-generator $\sigma (q):x\mapsto qx{q}^{-1}$ corresponds to the diagonal $\mathrm{Sp}(1)\hookrightarrow \mathrm{Sp}(1)\times \mathrm{Sp}(1)$, hence in ${\pi }_3$-coordinates equals $(1,1)-2\cdot (0,1)=(1,-1)$ in the appropriate normalization; rebasing the generators yields the $(h,j)$-parameterization. $\square$

Proposition 2 (Topological identification: $M_{h,j}{\cong }_{\mathrm{top}}{S}^7$ when $h+j=\pm 1$). For $h+j=\pm 1$, the closed oriented 7-manifold $M_{h,j}=\mathrm{Sph}({\xi }_{h,j})$ is homeomorphic to the standard 7-sphere $S^7$.

Proof. The Gysin sequence of the oriented $S^3$-bundle $\pi :M_{h,j}\to S^4$ is

$$\cdots \to H^{n-4}(S^4;\mathbb{Z})\stackrel{\cup e}{\to }{H}^n(S^4;\mathbb{Z})\stackrel{{\pi }^{\ast }}{\to }{H}^n(M_{h,j};\mathbb{Z})\stackrel{{\int }_{\mathrm{fibre}}}{\to }{H}^{n-3}(S^4;\mathbb{Z})\to \cdots$$

The cohomology of $S^4$ is concentrated in degrees $0$ and $4$, both copies of $\mathbb{Z}$. The Euler class $e({\xi }_{h,j})=(h+j)\alpha \in H^4(S^4;\mathbb{Z})$, where $\alpha$ is the generator. When $h+j=\pm 1$, multiplication by $e$ is an isomorphism on $H^0\to H^4$ (i.e., $\mathbb{Z}\to \mathbb{Z}$ given by $\pm 1$).

Working through the Gysin sequence degree by degree:

• $H^0(M_{h,j})=\mathbb{Z}$ (connectivity).
• $H^1(M_{h,j})$: zero, since the preceding term $H^1(S^4)=0$ and the next term $H^{-2}(S^4)=0$.
• $H^2,H^3$: zero, similarly.
• $H^4(M_{h,j})$: in the Gysin sequence, $H^0(S^4)\stackrel{\cup e=\pm 1}{\to }{H}^4(S^4)$ is surjective, so ${\pi }^{\ast }:H^4(S^4)\to H^4(M_{h,j})$ is zero. The next term $H^1(S^4)=0$. Hence $H^4(M_{h,j})=0$.
• $H^5,H^6$: zero.
• $H^7(M_{h,j})=\mathbb{Z}$ (Poincaré duality on the closed oriented 7-manifold).

The cohomology matches $H^{\ast }(S^7;\mathbb{Z})$ exactly. The manifold $M_{h,j}$ is also simply connected, since the fibre $S^3$ and base $S^4$ are both simply connected and the bundle has the structure of a fibration; the long exact sequence of homotopy groups ${\pi }_1(S^3)\to {\pi }_1(M_{h,j})\to {\pi }_1(S^4)=0$ gives ${\pi }_1(M_{h,j})=0$. By the Hurewicz theorem and the Whitehead theorem, $M_{h,j}$ is a simply connected closed 7-manifold with the same cohomology and the same homotopy groups (in low degrees) as $S^7$, so it is homotopy equivalent to $S^7$. By Smale's $h$-cobordism theorem 1962 ^{[Smale 1962]}, a simply connected closed 7-manifold homotopy equivalent to $S^7$ is homeomorphic to $S^7$. $\square$

Proposition 3 (Smooth obstruction via the Hirzebruch signature formula). Let $h,j\in \mathbb{Z}$ with $h+j=1$. If $(h-j{)}^2\not\equiv 1(\mathrm{mod}7)$, then $M_{h,j}$ is not diffeomorphic to the standard $S^7$.

Proof. Suppose for contradiction that there is a diffeomorphism $\psi :M_{h,j}\to S^7$ preserving smooth structures. Glue the smooth disc bundle $N_{h,j}$ (whose boundary is $M_{h,j}$) to the standard $D^8$ (whose boundary is the standard $S^7$) along $\psi$, obtaining a closed smooth $8$-manifold $X=N_{h,j}{\cup }_{\psi }{D}^8$.

By the Hirzebruch signature theorem 03.06.11 in dimension $8$,

$$45\cdot \mathrm{sig}(X)=7\cdot p_2[X]-p_1^2[X].$$

Compute $\mathrm{sig}(X)$: the intersection form on $H^4(X;\mathbb{R})$ is one-dimensional. From the Gysin sequence of $N_{h,j}\to S^4$ and the gluing, $H^4(X;\mathbb{R})\cong H^4(S^4;\mathbb{R})=\mathbb{R}$, with intersection form computed via the Euler class. Specifically, $\mathrm{sig}(X)=\mathrm{sign}(h+j)=\mathrm{sign}(1)=+1$.

Compute $p_1[X]$ and $p_1^2[X]$: the tangent bundle $TX{|}_{N_{h,j}}=T{S}^4\oplus {\xi }_{h,j}$ (as a vector bundle restricted to the base $S^4$, by the bundle structure). The first Pontryagin class satisfies $p_1(TX){|}_{N_{h,j}}=p_1(T{S}^4)+p_1({\xi }_{h,j})=0+2(h-j)\alpha =2(h-j)\alpha$, since $T{S}^4$ has zero Pontryagin classes. The class $\alpha$ extends from $S^4$ to all of $X$ as a generator of $H^4(X;\mathbb{Z})\cong \mathbb{Z}$, with ${\alpha }^2[X]=+1$. Therefore $p_1^2[X]=(2(h-j){)}^2\cdot 1=4(h-j{)}^2$.

Substituting into the Hirzebruch formula:

$$45\cdot 1=7\cdot p_2[X]-4(h-j{)}^2,p_2[X]=\frac{45+4(h-j{)}^2}{7}.$$

The Pontryagin number $p_2[X]$ is an integer (since both $p_2(TX)\in H^8(X;\mathbb{Z})$ is integral and $[X]\in H_8(X;\mathbb{Z})$ is integral). The right side must therefore be an integer. Reducing modulo $7$:

$$45+4(h-j{)}^2\equiv 0(\mathrm{mod}7),\iff 3+4(h-j{)}^2\equiv 0(\mathrm{mod}7),\iff 4(h-j{)}^2\equiv -3\equiv 4(\mathrm{mod}7).$$

Since $\gcd (4,7)=1$, this is equivalent to $(h-j{)}^2\equiv 1(\mathrm{mod}7)$. If $(h-j{)}^2\not\equiv 1(\mathrm{mod}7)$, integrality fails, contradicting the assumption that $X$ is a closed smooth $8$-manifold. Hence no such diffeomorphism $\psi$ exists. $\square$

Proposition 4 (Connect sum and the group structure on ${\Theta }_7$). The connected sum operation $\#$ endows the set ${\Theta }_7$ of oriented diffeomorphism classes of smooth homotopy 7-spheres with the structure of an abelian group. Under the Milnor lambda invariant, this group is cyclic of order $28$:

$${\Theta }_7\cong \mathbb{Z}/28\mathbb{Z},\lambda :{\Theta }_7\stackrel{\sim }{\to }\mathbb{Z}/28\mathbb{Z}.$$

Proof. Group structure on ${\Theta }_7$: the connected sum ${\Sigma }_1\#{\Sigma }_2$ is defined by removing an open $7$-disc from each and gluing the resulting boundaries $S^6\cong S^6$ along an orientation-reversing diffeomorphism. The standard $S^7$ acts as identity ($\Sigma \#S^7\cong \Sigma$ since gluing a disc back is the identity move). Inverses are realized by orientation reversal: $\Sigma \#(-\Sigma )\cong S^7$ via a handle-cancellation argument (Kervaire-Milnor 1963 §1 ^{[Kervaire-Milnor 1963]}). Associativity and commutativity follow from the diffeomorphism-invariance of the connect sum (the choice of gluing $7$-disc does not affect the diffeomorphism class).

The lambda invariant is additive: $\lambda ({\Sigma }_1\#{\Sigma }_2)=\lambda ({\Sigma }_1)+\lambda ({\Sigma }_2)$ in $\mathbb{Z}/28\mathbb{Z}$. This follows from the additivity of the signature of the disc-bundle filling: if ${\Sigma }_i=\partial W_i$ for parallelizable $W_i$, then ${\Sigma }_1\#{\Sigma }_2=\partial (W_1♮W_2)$ where $♮$ is the boundary-connected-sum of the fillings, and $\mathrm{sig}(W_1♮W_2)=\mathrm{sig}(W_1)+\mathrm{sig}(W_2)$ (since the intersection forms add directly).

Surjectivity of $\lambda$: by Theorem 4 above (Kervaire-Milnor 1963), the bounding-parallelizable subgroup ${\Theta }_7^{bP}\subset {\Theta }_7$ has order $28$. The lambda invariant restricts to an injection ${\Theta }_7^{bP}\hookrightarrow \mathbb{Z}/28\mathbb{Z}$. Since the quotient ${\Theta }_7/{\Theta }_7^{bP}$ embeds into ${\pi }_7^s/J({\pi }_7(\mathrm{SO}))=0$ (the $J$-homomorphism is surjective in degree 7 by Adams 1966 ^{[Adams 1966]}, ${\pi }_7^s=\mathbb{Z}/240\mathbb{Z}$ and $J({\pi }_7(\mathrm{SO}))={\pi }_7^s$), we conclude ${\Theta }_7={\Theta }_7^{bP}$ and $\lambda$ is an isomorphism ${\Theta }_7\to \mathbb{Z}/28\mathbb{Z}$. $\square$

Connections [Master]

• Pontryagin and Chern classes 03.06.04. Supplies the integer cohomology classes $p_i(TM)\in H^{4i}(M;\mathbb{Z})$ that enter Milnor's smooth obstruction. The exotic-sphere argument hinges on the integrality of $p_2[X]$ on the closed glued $8$-manifold $X$, and the fact that no smooth structure can produce a fractional value of this class.

• Stiefel-Whitney and Pontryagin numbers 03.06.10. The Pontryagin-number side of Milnor's argument sits inside the broader bordism-theoretic study of characteristic numbers. The bordism invariance of Pontryagin numbers establishes that $\lambda$ is well-defined independent of the parallelizable filling.

• Hirzebruch signature theorem 03.06.11. Provides the structural formula $45\cdot \mathrm{sig}(X)=7p_2[X]-p_1^2[X]$ in dimension $8$ that Milnor uses as the smooth-structure diagnostic. Without the Hirzebruch formula's identification of signature with a polynomial in Pontryagin numbers, the exotic-sphere argument cannot be formulated.

• Atiyah-Singer index theorem 03.09.10. Reformulates the Hirzebruch signature formula as the analytic index of the signature operator. The exotic-sphere obstruction is, in this framework, the failure of the $\hat{A}$-genus of a hypothetical bounding parallelizable manifold to be the actual index of a smooth Dirac operator. The deep connection is via the Rokhlin theorem ($\hat{A}[M]\equiv 0(\mathrm{mod}2)$ for closed spin $4$-manifolds), which is the analogue of Milnor's argument one dimension below.

• Singular cohomology and Poincaré duality 03.12.11. Anchors the cup-product and fundamental-class pairings used throughout. The Gysin sequence (which proves topological identification with $S^7$) and the intersection form (which computes $\mathrm{sig}(X)$) both rest on Poincaré duality for the closed oriented manifolds in play.

• Yang-Mills moduli and gauge theory 03.07.05. Donaldson's work on smooth $4$-manifolds extends the differential-topology revolution that Milnor's exotic spheres initiated: in dimension $4$, the structure of smooth invariants is even richer (uncountably many exotic structures on ${\mathbb{R}}^4$), and the diagnostic tool changes from characteristic-number integrality to gauge-theoretic invariants of moduli spaces.

• K-theory and the $J$-homomorphism 03.08.01. The Kervaire-Milnor identity ${\Theta }_7=\mathbb{Z}/28\mathbb{Z}$ depends on the order of the $J$-homomorphism $J:{\pi }_7(\mathrm{SO})\to {\pi }_7^s$, which is the structural input from stable homotopy theory. Adams' resolution of the $J$-homomorphism in 1966 completes the picture; the Bernoulli denominators that appear in the formula for $|{\Theta }_{4k-1}^{bP}|$ are the structural fingerprints of the $J$-homomorphism's image.

• Oriented bordism and Pontryagin-Thom 03.06.13. Milnor's exotic-sphere construction takes place inside the oriented-bordism category: the bounding parallelizable $8$-manifold is oriented, the Pontryagin numbers used as diagnostics are oriented-bordism invariants by 03.06.13, and the Kervaire-Milnor surgery exact sequence is set inside ${\Omega }_{\ast }^{SO}$. The structural inputs from 03.06.13 are what make the exotic-sphere argument well-posed.

• Multiplicative sequences and $L$/$\hat{A}$/Todd genera 03.06.15. The Hirzebruch signature formula used in Milnor's argument is the $L$-genus member of the multiplicative-sequence family of 03.06.15, and the Bernoulli denominators that appear in the order $|{\Theta }_{4k-1}^{bP}|$ are the same denominators that appear in the $\hat{A}$- and $L$-coefficients. The number-theoretic fine structure of multiplicative sequences is the structural backbone of the exotic-sphere classification.

• Whitney duality and immersion obstructions 03.06.16. The Wu-formula corrections to the integer-coefficient form of Whitney duality control the two-torsion in Pontryagin classes that Milnor's argument exploits; the exotic-sphere diagnostic uses the rational $L$-formula plus the integrality of Pontryagin numbers on closed smooth manifolds, with the integer-coefficient Whitney-dual identity supplying additional divisibility constraints on $p_1$ and $p_2$.

• Signature of a $4k$-manifold and the intersection form 03.06.19. The intersection-form signature on the bounding $8$-manifold $X$ used by Milnor is exactly the topic of 03.06.19, and the Donaldson-Freedman gauge-theory follow-up extends Milnor's exotic-sphere phenomenon to dimension $4$ via the smooth-topological gap detected by the intersection form. Anchor phrase: exotic-sphere obstruction lives in the intersection-form signature of the bounding manifold.

• Borel-Hirzebruch and the cohomology of $G/T$ 03.06.20. The Pontryagin-class polynomial $H^{\ast }(BSO;\mathbb{Q})=\mathbb{Q}[p_1,p_2,\dots ]$ identified by 03.06.20 is the universal home where Milnor's $p_2$-diagnostic is a polynomial relation; the structural theorem of 03.06.20 is the reason Pontryagin numbers are well-defined integer invariants of oriented bundles.

Historical & philosophical context [Master]

Milnor 1956 ^{[Milnor 1956]} introduced the exotic 7-sphere in the Annals of Mathematics in a paper less than seven pages long. The setting was Milnor's investigation of fibre bundles over $S^4$ with $S^3$ fibres, motivated by the question of whether such bundles could yield closed manifolds with anomalous characteristic-class data. The discovery that the resulting total spaces include non-standard smooth structures on $S^7$ was an unexpected consequence of the Hirzebruch signature theorem, which Hirzebruch had proven only three years earlier ^{[Hirzebruch 1953]}. Milnor's paper opens the field of differential topology in its modern form: the recognition that topological and smooth categories diverge starting in dimension $7$, and that the divergence is detectable via characteristic-class integrality.

The systematic study of ${\Theta }_n$ was carried out by Kervaire-Milnor 1963 ^{[Kervaire-Milnor 1963]}. Their surgery exact sequence

$$0\to {\Theta }_n^{bP}\to {\Theta }_n\to {\pi }_n^s/J({\pi }_n(\mathrm{SO}))\to 0$$

decomposes the group of homotopy spheres into two pieces: an arithmetic piece ${\Theta }_n^{bP}$ controlled by Bernoulli denominators and signature divisibility, and a stable-homotopy piece ${\pi }_n^s/J$ controlled by the Kervaire surgery obstruction. The computation ${\Theta }_7=\mathbb{Z}/28\mathbb{Z}$ is the first substantive application: the surgery quotient vanishes (since $J$ is surjective in degree $7$), and the bounding-parallelizable subgroup has order $28=4\cdot 7$, with the factor $4=2^{2k-2}$ for $k=2$ and the factor $7=$ numerator of $B_2/8\cdot$ standard adjustments.

The combinatorial Pontryagin classes were defined by Rokhlin-Schwartz 1957 ^{[Rohlin-Schwartz 1957]} for PL manifolds via the signatures of transverse PL submanifolds; Thom 1958 ^{[Thom 1958]} gave an equivalent construction in his Mexico City lectures, and the resulting classes are sometimes called Rokhlin-Schwartz-Thom classes or simply combinatorial Pontryagin classes. The crucial extension was Novikov 1965 ^{[Novikov 1965]}, who proved that the rational Pontryagin classes are topological invariants — a result Novikov was awarded the Fields Medal for in 1970. The contrast with Milnor's discovery is sharp: integral Pontryagin classes are not topological invariants (since they detect exotic smooth structures), but their rational shadows are.

The deeper philosophical resonance is that smooth structure on a topological manifold is a genuinely refined geometric datum — finer than topology, finer than PL structure, and detectable only via characteristic-class invariants beyond their rational shadows. The Milnor-Kervaire-Milnor framework was the historical entry point to high-dimensional surgery theory, the Browder-Novikov classification of high-dimensional manifolds, the development of Wall's surgery obstruction $L$-groups, and ultimately Donaldson and Freedman's revolution in $4$-manifold topology in the 1980s. The signature operator that underlies Milnor's diagnostic re-emerges in Atiyah-Singer 1968 as one of the four canonical elliptic complexes, identifying the exotic-sphere obstruction with the integrality of a specific analytic index.

Bibliography [Master]

@article{Milnor1956,
  author = {Milnor, J.},
  title = {On manifolds homeomorphic to the 7-sphere},
  journal = {Annals of Mathematics},
  volume = {64},
  year = {1956},
  pages = {399--405},
}

@article{KervaireMilnor1963,
  author = {Kervaire, M. A. and Milnor, J. W.},
  title = {Groups of homotopy spheres: I},
  journal = {Annals of Mathematics},
  volume = {77},
  year = {1963},
  pages = {504--537},
}

@article{Novikov1965,
  author = {Novikov, S. P.},
  title = {Topological invariance of rational {P}ontryagin classes},
  journal = {Doklady Akademii Nauk SSSR},
  volume = {163},
  year = {1965},
  pages = {298--300},
  note = {Full proof in Izvestia Akad. Nauk SSSR Ser. Mat. 30 (1966), 207--246.},
}

@article{Thom1958,
  author = {Thom, R.},
  title = {Les classes caractéristiques de {P}ontrjagin des variétés triangulées},
  journal = {Symposium Internacional de Topología Algebraica},
  publisher = {Universidad Nacional Autónoma de México and UNESCO},
  year = {1958},
  pages = {54--67},
}

@article{RohlinSchwartz1957,
  author = {Rohlin, V. A. and Schwartz, A. S.},
  title = {The combinatorial invariance of {P}ontryagin classes},
  journal = {Doklady Akademii Nauk SSSR},
  volume = {114},
  year = {1957},
  pages = {490--493},
}

@article{Smale1962,
  author = {Smale, S.},
  title = {On the structure of manifolds},
  journal = {American Journal of Mathematics},
  volume = {84},
  year = {1962},
  pages = {387--399},
}

@article{Adams1966,
  author = {Adams, J. F.},
  title = {On the groups {$J(X)$}: {IV}},
  journal = {Topology},
  volume = {5},
  year = {1966},
  pages = {21--71},
}

@article{Hirzebruch1953,
  author = {Hirzebruch, F.},
  title = {Über die quaternionalen projektiven {R}äume},
  journal = {Sitzungsberichte der Mathematisch-Naturwissenschaftlichen Klasse der Bayerischen Akademie der Wissenschaften zu München},
  year = {1953},
  pages = {301--312},
}

@book{MilnorStasheff,
  author = {Milnor, J. and Stasheff, J.},
  title = {Characteristic Classes},
  publisher = {Princeton University Press},
  year = {1974},
  series = {Annals of Mathematics Studies},
  volume = {76},
}

@book{Hirsch,
  author = {Hirsch, M. W.},
  title = {Differential Topology},
  publisher = {Springer-Verlag},
  year = {1976},
  series = {Graduate Texts in Mathematics},
  volume = {33},
}

---

Cycle 8 Track A, characteristic-classes T1 cluster. Closes the master-tier production flagged in manifests/skipped_units.md for 03.06.17.