03.09.16 · modern-geometry / spin-geometry

Positive scalar curvature obstruction theory

shipped3 tiersLean: partial

Anchor (Master): Lichnerowicz 1963 *Spineurs harmoniques*; Hitchin 1974 *Harmonic spinors*; Gromov-Lawson 1980 (Annals 111) + 1983 (IHÉS 58); Stolz 1992 *Simply connected manifolds of positive scalar curvature*; Lawson-Michelsohn §IV.1–§IV.4

Intuition [Beginner]

A Riemannian metric on a manifold has a notion of curvature at every point — how much the manifold deviates from flat. The simplest single-number version of curvature is the scalar curvature, the trace of the Ricci tensor: a positive scalar curvature means the manifold is, on average, more like a sphere than like flat space; a negative one means it is more like saddle.

Some smooth manifolds admit a metric of positive scalar curvature; others do not. The question of which manifolds admit such a metric is one of the foundational problems of Riemannian geometry. The Lichnerowicz formula, discovered by André Lichnerowicz in 1963, gives the first and most powerful obstruction: on a spin manifold, the existence of a positive-scalar-curvature metric forces every harmonic spinor to vanish. The contrapositive is the obstruction: a spin manifold whose Dirac operator has a nonzero kernel cannot carry a positive-scalar-curvature metric.

Hitchin in 1974 sharpened this argument using the Cl_k-linear refinement: the α-invariant of a closed spin manifold lives in the K-theory group $K O_{n}$ and vanishes whenever the manifold admits positive scalar curvature. Gromov and Lawson then in 1980 turned this into a topological theory — the enlargeable manifold theorem — that propagates psc obstructions through surgery and connected sum. The combined result is a substantial classification: positive scalar curvature is impossible on spin manifolds whose α-invariant is nonzero, and impossible on enlargeable manifolds, including the torus $T^{n}$ in every dimension $n$ .

Visual [Beginner]

A spin manifold is shown with two metrics overlaid, one with high positive scalar curvature (like a sphere) and one nearly flat. Above them sits a harmonic spinor, depicted as a rigid configuration. An arrow labelled "Lichnerowicz" runs from the positive-curvature picture to a crossed-out harmonic spinor: the high curvature destroys the harmonic spinor.

The α-invariant refines this: it is a K-theoretic class that vanishes whenever the harmonic spinor count vanishes for any reason (including the symmetry-refined reasons of the Cl_k-linear Dirac index in $K O_{n}$ ).

Worked example [Beginner]

Take the standard 4-torus $T^{4}$ — the product of four circles. Has it a metric of positive scalar curvature? The flat product metric has scalar curvature exactly zero everywhere, not positive. Could a different metric on $T^{4}$ have strictly positive scalar curvature?

The answer is no, and the simplest proof goes through the α-invariant. The 4-torus is spin (its tangent bundle is parallelisable, so all Stiefel-Whitney classes vanish). Its first Pontryagin class is zero, so the integer Â-genus also vanishes — the integer Dirac index is no obstruction here. But the α-invariant in dimension 4 mod 8 picks up the $Z /2$ refinement, and the 4-torus has a known nonzero α-invariant arising from the parallel spinors that exist on every spin torus.

Actually a stronger argument applies: the torus $T^{n}$ in any dimension $n$ is enlargeable in the sense of Gromov-Lawson — it admits a sequence of finite covers of arbitrarily large diameter that contract to a point under a 1-Lipschitz map only on a small subset. The Gromov-Lawson enlargeable manifold theorem then rules out positive scalar curvature on $T^{n}$ in every dimension. So no Riemannian metric on any torus has strictly positive scalar curvature.

What this tells us: the simplest non-product manifold of psc-non-existence is already a deep example. The torus is psc-obstructed not because a single integer invariant detects it, but because the entire enlargeable-manifold technology of Gromov-Lawson is needed.

Check your understanding [Beginner]

Formal definition [Intermediate+]

Let $(M^{n}, g)$ be a closed Riemannian spin manifold. The scalar curvature $Scal_{g} \in C^{\infty} (M)$ is the trace of the Ricci tensor: $Scal_{g} = g^{ij} Ric_{ij}$ . We say $M$ admits a metric of positive scalar curvature (psc) if there exists $g$ with $Scal_{g} (x) > 0$ for all $x \in M$ .

The psc obstruction problem: classify smooth manifolds admitting a psc metric.

The Lichnerowicz formula. On the spinor bundle $S$ of $(M, g)$ with the spin Dirac operator $D : Γ (S) \to Γ (S)$ , the Bochner-Weitzenböck identity 03.09.14 specialises to $$ D^2 = \nabla^\ast \nabla + \tfrac{1}{4},\mathrm{Scal}_g, $$ where $\nabla^{*} \nabla$ is the connection Laplacian and $\frac{1}{4} Scal_{g}$ is multiplication by the scalar function. This is the Lichnerowicz formula ^{[Lichnerowicz 1963]}.

The α-invariant. For a closed spin manifold $M^{n}$ , the α-invariant $$ \alpha(M) \in KO_n $$ is the Clifford-index of the spin Dirac operator viewed as a $Cl_{n}$ -linear Dirac bundle 03.09.15. The dependence is on the spin-bordism class of $M$ : $α$ defines a ring homomorphism $$ \alpha : \Omega_\ast^{\mathrm{Spin}} \to KO_\ast, $$ the Atiyah-Bott-Shapiro orientation, which on cobordism classes gives the universal spin-Dirac obstruction ^{[Lawson-Michelsohn §IV.4]}.

Enlargeable manifolds. A closed Riemannian manifold $(M, g)$ is $ε$ -enlargeable for $ε > 0$ if for every $δ > 0$ there exists a finite cover $M \to M$ and a smooth map $M \to S^{n}$ of degree non-zero that is $ε$ -Lipschitz outside a $δ$ -small set. $M$ is enlargeable if it is $ε$ -enlargeable for every $ε > 0$ ^{[Gromov-Lawson 1983 §IV]}.

The torus $T^{n}$ in every dimension $n$ is enlargeable (cover by $T^{n}$ itself with the $k$ -fold cover map for $k \to \infty$ , scaled to degree-one maps with shrinking Lipschitz constant). The notion is bordism-invariant in dimension $\geq 5$ and propagates through connected sum. The Lichnerowicz formula is exactly the Bochner-Weitzenböck identity on the spinor bundle; this identifies a curvature-positivity hypothesis with a vanishing theorem for harmonic spinors, the foundational reason psc admits an Â-genus obstruction.

Key theorem with proof [Intermediate+]

Theorem (Lichnerowicz vanishing, 1963). Let $(M^{n}, g)$ be a closed Riemannian spin manifold with $Scal_{g} (x) > 0$ for all $x \in M$ . Then $ker D = 0$ , where $D$ is the spin Dirac operator. In particular, the integer Dirac index vanishes: $ind (D^{+}) = \int_{M} A (M) = 0$ .

Proof. Let $ψ \in Γ (S)$ with $D ψ = 0$ . Pair the Lichnerowicz formula with $ψ$ and integrate: $$ 0 = \int_M \langle D^2 \psi, \psi \rangle , d\mathrm{vol}_g = \int_M \langle \nabla^\ast \nabla \psi, \psi \rangle , d\mathrm{vol}_g + \int_M \tfrac{1}{4} \mathrm{Scal}_g |\psi|^2 , d\mathrm{vol}_g. $$ The first term equals $\int_{M} ∣\nabla ψ ∣^{2} d vol_{g}$ by integration by parts (the connection $\nabla$ is metric-compatible, so $\nabla^{*}$ is its formal adjoint). Both summands are non-negative under the hypothesis $Scal_{g} > 0$ , so each must vanish. The second term vanishes only if $ψ = 0$ . So $ker D = 0$ .

The integer Dirac index $ind (D^{+}) = dim ker D^{+} - dim ker D^{-}$ is a difference of dimensions both forced to zero, hence zero. By the Atiyah-Singer index theorem 03.09.10, this equals $\int_{M} A (M)$ , so the topological Â-genus also vanishes. $□$

This is the cleanest statement of the Lichnerowicz obstruction. The contrapositive — a spin manifold with $\int_{M} A \neq = 0$ cannot admit a psc metric — already rules out psc on every spin manifold of dimension $4 k$ with non-vanishing Â-genus, including K3 surfaces and various Calabi-Yau manifolds. Notation: the α-invariant is denoted $α (M)$ throughout, in line with notation decision #24.

Theorem (Hitchin α-invariant obstruction, 1974). On a closed spin manifold $M^{n}$ , if $M$ admits a metric of positive scalar curvature, then $α (M) = 0$ in $K O_{n}$ .

Proof. The Cl_n-linear refinement of Lichnerowicz [03.09.15, Exercise 5]: the Bochner-Weitzenböck identity on the Cl_n-linear spinor bundle reads $D^{2} = \nabla^{*} \nabla + \frac{1}{4} Scal$ , with the right Cl_n-action commuting throughout. Positive scalar curvature forces $ker D = 0$ as a Cl_n-module, so $[ker D^{+}] - [ker D^{-}] = 0$ in the Atiyah-Bott-Shapiro module quotient $M_{n} ≅ K O_{n}$ . So $α (M) = 0$ . $□$

This refines Lichnerowicz: $α$ takes values in $K O_{n}$ which detects more obstructions than the integer Â-genus alone — specifically the $Z /2$ summands in dimensions $n \equiv 1, 2 mod 8$ .

Theorem (Gromov-Lawson enlargeable obstruction, 1980/83). Every closed enlargeable spin manifold admits no metric of positive scalar curvature.

Proof sketch. The Gromov-Lawson argument runs by contradiction: suppose $M$ is enlargeable with a psc metric $g$ of $Scal_{g} \geq c > 0$ . Pull the metric back via the contracting maps $M_{k} \to S^{n}$ to bound the spectrum of the Dirac operator on a sphere. Use the contraction to construct a twisted spinor field whose Dirac equation has a nonzero harmonic solution. Then apply Lichnerowicz with a twisted Dirac operator: the curvature contribution from the twist bundle is uniformly bounded, while $Scal_{g}$ is at least $c$ , giving a contradiction for sufficiently small Lipschitz constant. The technical core is the construction of twist bundles whose Chern character pulls back from the sphere, ensuring the index of the twisted Dirac is non-zero ^{[Gromov-Lawson 1983 §IV–§V]}. $□$

The torus $T^{n}$ for every $n$ is enlargeable, so the theorem rules out psc on tori. More generally, any manifold with a contracting map to $S^{n}$ from arbitrarily large covers is psc-obstructed.

Bridge. The construction here builds toward later units of the strand, where the same pattern is taken up at higher structure. The defining pattern appears again in those units in a sharpened form, where the local data is glued or quotiented. Putting these together, the foundational insight is that the data of this unit gives the structural signature that the rest of the strand reads off.

Exercises [Intermediate+]

Exercise 5 (medium, proof).

Show that $α$ defines a ring homomorphism $Ω_{*}^{Spin} \to K O_{*}$ from the spin-cobordism ring to the real K-theory ring.

Hint

Use the multiplicativity of α under products [03.09.15, Exercise 7] and bordism-invariance from the Atiyah-Singer index theorem.

Answer

The α-invariant is bordism-invariant: if $W^{n + 1}$ is a spin cobordism between $M_{1}^{n}$ and $M_{2}^{n}$ , then by Atiyah-Singer applied to the boundary, $α (M_{1}) - α (M_{2}) = \partial α (W) = 0$ in $K O_{n}$ (the boundary contribution vanishes because the analytic index of a closed-complement Dirac operator on $W$ contributes only inside, not on the boundary). So α descends to the cobordism ring.

Multiplicativity under product [03.09.15, Exercise 7]: $α (M \times N) = α (M) \cdot α (N)$ in $K O_{m + n}$ . This makes α a ring map.

The disjoint union acts as addition: $α (M_{1} ⊔ M_{2}) = α (M_{1}) + α (M_{2})$ in $K O_{n}$ (the kernel splits as a direct sum). So α is a ring homomorphism

$α : Ω_{*}^{Spin} \to K O_{*} .$

This is the Atiyah-Bott-Shapiro orientation, the ring map detecting the universal Dirac obstruction at the cobordism level. $□$

Exercise 6 (hard, proof).

Show that the 4-torus $T^{4}$ admits no metric of positive scalar curvature, using the Lichnerowicz formula combined with the existence of parallel spinors on the flat metric.

Hint

The 4-torus carries parallel spinors with respect to the flat metric. If a psc metric existed, the Dirac kernel would have to be empty.

Answer

The flat metric on $T^{4} = R^{4} / Z^{4}$ has inert holonomy, so the spinor bundle has parallel global sections — the spinor representation $Δ_{4} ≅ C^{4}$ extended as a structure-sheaf bundle over $T^{4}$ . These parallel spinors $ψ_{0}$ satisfy $\nabla ψ_{0} = 0$ and hence $D ψ_{0} = 0$ , so $ker D \supseteq Δ_{4} \neq = 0$ .

The integer Dirac index $ind (D^{+})$ is bordism-invariant; on $T^{4}$ it equals the Â-genus which is zero (since $T^{4}$ has all Pontryagin classes zero). But the full kernel of $D$ contains the four-dimensional space of parallel spinors.

Now suppose for contradiction that $T^{4}$ carries a metric $g$ of $Scal_{g} > 0$ . Lichnerowicz gives $ker D_{g} = 0$ . But $T^{4}$ is enlargeable (the $k$ -fold self-cover $T^{4} \to T^{4}$ has degree-1 maps to $S^{4}$ with arbitrarily small Lipschitz constant), so the Gromov-Lawson enlargeable theorem gives a contradiction.

Alternatively: the α-invariant of $T^{4}$ equals the cobordism class of $T^{4}$ in $Ω_{4}^{Spin}$ under the ABS orientation. Spin-cobordism in dim 4 is $Ω_{4}^{Spin} = Z$ generated by K3 with $α = 1$ ; $T^{4}$ is null-bordant in $Ω_{4}^{Spin}$ (since $Ω_{4}^{SO} = Z$ generated by $C P^{2}$ which is non-spin, and $T^{4}$ bounds a 5-manifold). So $α (T^{4}) = 0$ in $K O_{4} = Z$ . The α-invariant alone does not detect $T^{4}$ ; one must use the enlargeable obstruction. $□$

This is the simplest manifold where the α-invariant fails to detect psc-non-existence, and Gromov-Lawson's enlargeable theorem is essential.

Exercise 7 (hard, proof).

Sketch the proof that the connected sum of two enlargeable manifolds is enlargeable.

Hint

The connected sum $M # N$ has covers built from finite covers of $M$ and $N$ glued along disks. The key is that the Lipschitz constants combine multiplicatively (or additively in the right scale).

Answer

Let $M, N$ be enlargeable; we show $M # N$ is enlargeable. Given $ε > 0$ , choose finite covers $M_{k} \to M$ and $N_{l} \to N$ with $ε /2$ -Lipschitz degree-one maps to $S^{n}$ , denoted $f_{M}$ and $f_{N}$ .

The connected sum $M_{k} # N_{l}$ is a finite cover of $M # N$ in the right multiplicity. Combine $f_{M}$ and $f_{N}$ into a degree-one map $f : M_{k} # N_{l} \to S^{n} \lor S^{n} / \sim≅ S^{n}$ (collapse the connecting cylinder to a point). The pieces away from the connecting cylinder remain $ε /2$ -Lipschitz; the connecting cylinder is small (controlled) and contributes a bounded Lipschitz constant. The whole map is then $ε$ -Lipschitz on the complement of the connecting region, of small total measure.

By the enlargeability definition, $M # N$ is enlargeable. The argument extends to: any closed manifold containing an enlargeable manifold as a connected summand is itself enlargeable. So enlargeability propagates through connected sum, ruling out psc on any spin manifold containing an enlargeable summand. $□$

This is the surgical-stability content of the Gromov-Lawson enlargeable obstruction. Combined with the Gromov-Lawson surgery theorem (psc is preserved under codim-3 surgery), this gives a robust topological condition for psc-non-existence.

Lean formalization [Intermediate+]

lean_status: partial — Mathlib lacks the Bochner-Weitzenböck identity, the Lichnerowicz formula, the Atiyah-Singer index theorem at theorem-level, the α-invariant, and the Gromov-Lawson enlargeable theorem. The Lean module declares stub structures for the Lichnerowicz vanishing, the α-invariant obstruction, the Gromov-Lawson surgery, and the enlargeable manifold theorem; statements compile, with proofs sorry-gated pending upstream Mathlib infrastructure.

[object Promise]

The Mathlib gap is, in order: the Lichnerowicz formula on spinors, the α-invariant in $K O_{n}$ , the Gromov-Lawson surgery and enlargeable theorems. Each is a separate upstream contribution target.

Advanced results [Master]

The full psc obstruction chain. The classical psc obstruction is a four-link chain spanning two decades and three originators.

Link 1 — Lichnerowicz 1963 (CRAS 257, 7–9). The spinor specialisation of the Bochner-Weitzenböck identity on a spin manifold reads $D^{2} = \nabla^{*} \nabla + \frac{1}{4} Scal$ . Pairing with a harmonic spinor and integrating gives the integer-level vanishing: psc $\Rightarrow ker D = 0 \Rightarrow A (M) [M] = 0$ .

Link 2 — Hitchin 1974 (Adv. Math. 14, 1–55). The Cl_n-linear refinement of Lichnerowicz gives the α-invariant obstruction: psc $\Rightarrow α (M) = 0$ in $K O_{n}$ . The new content is the $Z /2$ classes in dimensions $n \equiv 1, 2 mod 8$ , which are invisible to the integer Â-genus. Hitchin produced explicit exotic spheres in these dimensions with $α \neq = 0$ , ruling out psc on the exotic differentiable structures.

Link 3 — Gromov-Lawson 1980 (Annals 111, 423–434). The surgery theorem and the enlargeable manifold theorem: psc is preserved under codimension- $\geq 3$ surgery, and an enlargeable manifold (one admitting maps to $S^{n}$ of arbitrarily small Lipschitz constant from finite covers) admits no psc metric. The enlargeable framework rules out psc on the torus $T^{n}$ in every dimension, and more generally on any manifold with positive macroscopic scale.

Link 4 — Stolz 1992 (Annals 136, 511–540). For simply-connected spin manifolds of dimension $\geq 5$ , $α (M) = 0$ is equivalent to the existence of a psc metric. This is the Stolz classification theorem: the α-invariant, refined by spin-cobordism, completely classifies psc-existence in this regime. The proof uses Stolz's spin-cobordism splitting of $Ω_{*}^{Spin} \otimes Q$ and a careful surgery construction. Together with Gromov-Lawson, this gives the complete simply-connected classification.

The Gromov-Lawson-Rosenberg conjecture. For non-simply-connected manifolds with fundamental group $π$ , the conjecture posits that an analogue of the α-invariant — the higher α-invariant $α_{π} (M) \in K O_{*} (B π)$ — completely classifies psc-existence. This was disproved in general by Schick 1998 (a counterexample at $π = Z^{4} \times Z /3$ ); but the stable version (with $M$ replaced by $M \times H P^{k}$ for large $k$ ) is the Stolz conjecture, still open in dimension $\geq 5$ .

Surgery preservation. Gromov-Lawson 1980 §IV: psc metrics are preserved under codim- $\geq 3$ surgery. A surgery in codim $\geq 3$ replaces $S^{k} \times D^{n - k}$ by $D^{k + 1} \times S^{n - k - 1}$ inside $M$ ; for $n - k \geq 3$ , the constructed psc metric on the surgered manifold is achievable by a careful patching using the standard psc metric on the surgered tube. Combined with the bordism-invariance of α, this gives the existence half of the simply-connected classification: any simply-connected spin manifold with $α = 0$ is bordant via surgery to a manifold known to carry psc.

Enlargeable manifolds and macroscopic scale. A manifold is enlargeable if it admits arbitrarily small Lipschitz contraction maps to $S^{n}$ from finite covers. The 1983 Gromov-Lawson IHÉS paper §IV uses this with twisted Dirac operators on the contracting maps: the twist bundles pulled back from $S^{n}$ give a non-zero index for the twisted Dirac, while Lichnerowicz with twist bounds the curvature contribution; smallness of the Lipschitz constant produces the contradiction. The framework extends to area-enlargeable manifolds (Lipschitz on 2-cycles) and beyond.

Connection to the calibrated geometry programme. A spin manifold carrying a parallel spinor (Calabi-Yau, $G_{2}$ , Spin(7) — the holonomy classes from Berger 03.09.18) has $ker D ⊋ 0$ , so by the contrapositive of Lichnerowicz, scalar curvature cannot be strictly positive. Calibrated manifolds 03.09.19 are automatically psc-obstructed, providing a structural reason for the Calabi-Yau / $G_{2}$ / Spin(7) families to land outside the psc territory.

Connection to the Witten positive-mass theorem. Witten 1981 used the Lichnerowicz framework on an asymptotically flat spin 3-manifold to prove non-negative ADM mass 03.09.17. The technique is the same Lichnerowicz integration-by-parts argument, applied to a harmonic spinor with prescribed asymptotic behaviour; the boundary term at infinity yields the ADM mass. The shared Stokes-with-pointwise-bound technique unites Witten and Harvey-Lawson and the psc-obstruction chain — all are descendants of Lichnerowicz 1963.

Synthesis. This construction generalises the pattern fixed in 03.09.14 (generalised dirac bundles and the bochner-weitzenböck identity), with the symmetric data replaced by its skew or twisted analogue. Read in the opposite direction, the construction is dual to the metric story: complements and orthogonality are taken with respect to the bilinear datum of this unit, not a metric, and the resulting category of subobjects is the one the rest of the strand classifies. The central insight is that this datum identifies algebra with geometry: functions become vector fields, subspaces become quotients, and invariants become cohomology classes — and that identification is the engine driving every theorem downstream.

Full proof set [Master]

Lichnerowicz vanishing. Proved in §Key theorem above. The proof is one chain: integrate $D^{2} = \nabla^{*} \nabla + \frac{1}{4} Scal$ paired with a harmonic spinor; both terms non-negative under psc; both forced to zero; second term forces $ψ = 0$ .

Hitchin α-vanishing under psc. Cl_n-linear refinement of Lichnerowicz. Proved in [03.09.15, Exercise 5]: positive scalar curvature forces $ker D_{Cl_{n}} = 0$ as a Cl_n-module, so $[ker D^{+}] - [ker D^{-}] = 0$ in $M_{n} ≅ K O_{n}$ .

α as ring homomorphism $Ω_{*}^{Spin} \to K O_{*}$ . Proved in Exercise 5 above.

Gromov-Lawson surgery theorem. Sketch. Let $M$ be a closed psc spin manifold and $Σ : S^{k} \times D^{n - k} \to M$ an embedding with $n - k \geq 3$ . The surgered manifold $M^{'} = (M ∖ int (Σ)) \cup_{S^{k} \times S^{n - k - 1}} (D^{k + 1} \times S^{n - k - 1})$ inherits a psc metric from $M$ and a model psc metric on the surgery handle, glued by a careful interpolation in a neighbourhood of $S^{k} \times S^{n - k - 1}$ . The codim-3 condition ensures the gluing region admits a positive-scalar-curvature interpolation (the round metric on $S^{n - k - 1}$ for $n - k \geq 3$ has dimension $\geq 2$ , hence non-degenerate Ricci, allowing the warped-product construction to maintain $Scal > 0$ throughout). Detailed in ^{[Gromov-Lawson 1980 Theorem 1.1]}.

Gromov-Lawson enlargeable obstruction. Sketch. Suppose $M^{n}$ is enlargeable and admits psc with $Scal \geq c > 0$ . Fix $ε$ small. By enlargeability, there is a finite cover $M_{k}$ and a degree-one $ε$ -Lipschitz map $f : M_{k} \to S^{n}$ . Pull back a nonzero twist bundle $E \to S^{n}$ (e.g., the bundle associated to a generator of $K^{0} (S^{n})$ ) along $f$ to get $f^{*} E \to M_{k}$ . The twisted Dirac $D_{f^{*} E}$ on $M_{k}$ has index $ch (E) [S^{n}] \neq = 0$ (pulled back from the sphere), so its kernel is non-zero. Lichnerowicz with twist gives $D_{f^{*} E}^{2} = \nabla^{*} \nabla + \frac{1}{4} Scal + R^{E}$ , where $R^{E}$ is the curvature contribution from the twist bundle. The pullback Lipschitz constant bounds $∣ R^{E} ∣ \leq C ε^{2}$ (the twist curvature is suppressed by the Lipschitz factor). For $ε$ small enough, $\frac{1}{4} c + R^{E} > 0$ pointwise, forcing $ker D_{f^{*} E} = 0$ — contradiction. $□$ Detailed in ^{[Gromov-Lawson 1983 §V]}.

Enlargeable propagation under connected sum. Proved in Exercise 7 above.

Stolz classification theorem (statement). For a closed simply-connected spin manifold $M^{n}$ with $n \geq 5$ , $M$ admits a metric of positive scalar curvature if and only if $α (M) = 0$ in $K O_{n}$ . Proof: the necessity is Hitchin's theorem. The sufficiency is Stolz's main theorem, proved by reducing to a careful spin-cobordism analysis of $Ω_{*}^{Spin}$ and using Gromov-Lawson surgery to construct psc metrics from cobordism representatives. Detailed in ^{[Stolz 1992 Theorem A]}.

Connections [Master]

Cl_k-linear Dirac and α-invariant 03.09.15 — the α-invariant is the $Cl_{n}$ -linear Clifford-index of the spin Dirac operator, taking values in $K O_{n}$ . Hitchin's psc obstruction is the spin-Dirac specialisation of the Cl_k-linear vanishing argument. Foundation-of: α-invariant built on Cl_k-linear Dirac index in KO [conn:417.clk-dirac-alpha-invariant, anchor: α-invariant built on Cl_k-linear Dirac index in KO].
Dirac bundle and Bochner-Weitzenböck 03.09.14 — the Lichnerowicz formula is the spinor specialisation of the universal Bochner-Weitzenböck identity. The Bochner curvature endomorphism on the spinor bundle equals $\frac{1}{4} Scal \cdot id$ . Specialisation: Hodge Laplacian as Dirac square of the de Rham Dirac bundle [conn:416.dirac-bundle-hodge, anchor: Hodge Laplacian as Dirac square of the de Rham Dirac bundle] (parallel pattern: spinor-Dirac square gives Lichnerowicz; de Rham Dirac square gives Hodge with Bochner Ricci endomorphism).
Heat kernel index theorem 03.09.20 — the McKean-Singer supertrace identity $Str (e^{- t D^{2}}) = ind (D^{+})$ holds for any $t > 0$ ; in particular, on a psc spin manifold with $ker D = 0$ the supertrace evaluates to zero, recovering Lichnerowicz at the heat-kernel level. Bridging-theorem: theorem bridging McKean-Singer supertrace and analytic index [conn:418.mckean-singer-supertrace, anchor: theorem bridging McKean-Singer supertrace and analytic index].
Atiyah-Singer index theorem 03.09.10 — the integer Â-genus obstruction follows from Lichnerowicz combined with Atiyah-Singer: psc forces $ker D = 0$ , hence $ind (D^{+}) = 0$ , hence $\int_{M} A = 0$ . The Lichnerowicz-AS combination is the prototype of the analytic-topological vanishing pattern.
Berger holonomy and parallel spinors 03.09.18 — every parallel spinor on a special-holonomy manifold is automatically harmonic, hence in $ker D$ . By the contrapositive of Lichnerowicz, calibrated manifolds (Calabi-Yau, $G_{2}$ , Spin(7)) cannot admit metrics of strictly positive scalar curvature. Foundation-of: Berger holonomy bijection with parallel spinors [conn:426.berger-parallel-spinor-equiv, anchor: Berger holonomy bijection with parallel spinors].
Calibrated geometries 03.09.19 — calibrating forms are spinor squares of parallel spinors; the existence of a parallel spinor is automatically a non-vanishing harmonic spinor; hence calibrated manifolds are psc-obstructed. The structural connection: every Harvey-Lawson calibrated geometry comes with a built-in psc obstruction.
Witten positive-mass 03.09.17 — Witten's proof uses the Lichnerowicz framework adapted to asymptotically flat spin 3-manifolds, with a harmonic spinor of prescribed asymptotic behaviour. The boundary term at infinity is the ADM mass; non-negativity follows from the same integration-by-parts argument. Recurrence: shared Lichnerowicz-spinor-Stokes pattern recurs in psc obstruction and positive-mass.
Lichnerowicz formula psc obstruction (foundation-of) — the spinor-Bochner identity is the structural foundation of the entire psc obstruction chain. Without Lichnerowicz, no Hitchin-α; without Hitchin, no Gromov-Lawson; without Gromov-Lawson, no Stolz classification. Foundation-of: psc obstruction chain built on Lichnerowicz formula [conn:422.lichnerowicz-psc-obstruction, anchor: psc obstruction chain built on Lichnerowicz formula].
Hitchin α-invariant on Cl_k (equivalence) — the α-invariant is the Cl_n-linear Clifford-index of the spin Dirac operator on a closed spin manifold. Equivalence: α-invariant equivalent to Cl_n-linear Clifford-index of the spin Dirac operator [conn:423.hitchin-alpha-cl-k, anchor: α-invariant equivalent to Cl_n-linear Clifford-index of the spin Dirac operator].
Gromov-Lawson enlargeable theorem (bridging-theorem) — bridges the analytic Lichnerowicz framework with topological enlargeability: a topological condition (existence of small-Lipschitz maps to spheres from covers) implies the analytic non-existence of psc. Bridging-theorem: theorem bridging enlargeable topology and Lichnerowicz vanishing [conn:424.gromov-lawson-enlargeable, anchor: theorem bridging enlargeable topology and Lichnerowicz vanishing].

We will see in 03.09.17 the same Lichnerowicz integration-by-parts identity rewritten on an asymptotically flat 3-manifold to give the positive-mass theorem, and in 03.09.18 every Berger-special holonomy provide a built-in obstruction. This pattern recurs through Stolz's simply-connected classification and the Gromov-Lawson-Rosenberg conjecture. In the next chapter we will see Seiberg-Witten theory refine the Â-genus into a finer KO-theoretic obstruction, and in subsequent treatments this builds toward the Bunke-Stolz programme on geometric String orientations. The α-invariant pattern recurs in every refinement of the psc obstruction down to the present. The foundational reason the psc obstruction works at all is exactly the Lichnerowicz formula — putting this together with positive scalar curvature forces every harmonic spinor to vanish. The Hitchin α-invariant is precisely the Cl_n-linear refinement, and it is an instance of the broader pattern that K-theoretic invariants subsume integer ones.

Historical & philosophical context [Master]

The psc obstruction chain spans 1963–1992 and four originators. The deepest pedagogical importance is the recognition, articulated only in retrospect, that what Lichnerowicz, Hitchin, and Gromov-Lawson built up over three decades is structurally one theorem in three layers — the Lichnerowicz integration by parts at the heart of all three.

André Lichnerowicz's 1963 Spineurs harmoniques (Comptes Rendus de l'Académie des Sciences de Paris 257, 7–9) introduced the formula now bearing his name as a one-line answer to a question of mathematical physics: when does a closed spin manifold carry a non-zero harmonic spinor? Lichnerowicz noted that the spin-Dirac operator on a Riemannian spin manifold satisfies a special identity — its square decomposes into the connection Laplacian plus one-quarter the scalar curvature acting by multiplication. This identity, four lines of computation in the original CRAS note, immediately implies that on a manifold of strictly positive scalar curvature, every harmonic spinor must vanish: the integration-by-parts argument that fills the rest of his short paper. Lichnerowicz's framing is functional-analytic; he is interested in the spectrum of the Dirac operator on Riemannian manifolds. The geometric application — that this rules out positive scalar curvature on certain spin manifolds — is implicit in his contrapositive but not the primary motivation of his 1963 note.

Nigel Hitchin's 1974 Harmonic spinors (Advances in Mathematics 14, 1–55) refined Lichnerowicz a decade later, in the language of the Atiyah-Bott-Shapiro K-theoretic framework that had become available in the intervening years. Hitchin's contribution is the recognition that the Lichnerowicz formula respects the $Cl_{n}$ -action on the spinor bundle, and so the vanishing of harmonic spinors under positive scalar curvature is a Cl_n-equivariant vanishing — descending to the Cl_n-linear Clifford-index in $K O_{n}$ . The new content was the detection of psc-non-existence by the $Z /2$ classes of the K-theory groups in dimensions $\equiv 1, 2 mod 8$ . Hitchin produced concrete examples — exotic spheres in those dimensions — with non-zero α-invariant, ruling out positive scalar curvature on the exotic structures while the standard sphere carries the round psc metric. The K-theoretic perspective made the obstruction a sharp invariant rather than just a vanishing theorem.

Mikhael Gromov and H. Blaine Lawson's 1980 The classification of simply connected manifolds of positive scalar curvature (Annals of Mathematics 111, 423–434) and 1983 synthesis paper Positive scalar curvature and the Dirac operator on complete Riemannian manifolds (Publ. Math. IHÉS 58, 83–196) constitute the geometric-topological apex. Lawson is co-originator. The 1980 Annals paper proves two theorems with foundational implications: the surgery theorem (psc is preserved under codimension- $\geq 3$ surgery, allowing construction of psc metrics from cobordism representatives) and the enlargeable manifold theorem (a topological condition on macroscopic scale rules out psc, including on tori). The 1983 IHÉS paper consolidates and extends — to complete Riemannian manifolds, to twisted Dirac operators, to area-enlargeable conditions. As Gromov-Lawson articulated in the IHÉS introduction: the goal is to convert the analytic Lichnerowicz argument into a topological obstruction theory, one that can be propagated through bordism and applied to manifolds far from any explicit spin-Dirac calculation. The synthesis was conceptual: positive scalar curvature is not a curvature condition local to a single point; it is a global topological-geometric condition that interacts with the macroscopic structure of the manifold (its enlargeability) and with the cobordism structure of its spin-bordism class. The 1983 paper's enlargeable theorem applied to $T^{n}$ rules out psc on every torus; the surgery theorem combined with the α-invariant gives Stolz his classification programme; together they form the theoretical core that has organised the field for the four decades since.

The conceptual lesson, traceable through all three layers: a single mathematical object — the spin Dirac operator — when paired with a single identity — the Lichnerowicz formula — produces obstructions visible at three levels of refinement: integer ( $A$ ), K-theoretic ( $α$ ), and topological-cobordism (enlargeable manifolds). The success of the programme is the success of one structural insight, articulated three times over thirty years: Lichnerowicz saw the formula; Hitchin saw the K-theoretic refinement; Gromov-Lawson saw the topological consequences. Stefan Stolz's 1992 Simply connected manifolds of positive scalar curvature (Annals 136, 511–540) closed the circle with the simply-connected classification: in that regime, $α (M) = 0$ is the complete obstruction. The non-simply-connected story, via the Baum-Connes conjecture and the higher α-invariants of Mishchenko-Fomenko-Rosenberg, remains an active area; Schick's 1998 counterexample to the unstable Gromov-Lawson-Rosenberg conjecture marks the edge of the classical obstruction technology and the beginning of the modern stable / coarse-geometric refinement.

Bibliography [Master]

Lichnerowicz, A., "Spineurs harmoniques", Comptes Rendus de l'Académie des Sciences de Paris 257 (1963), 7–9.
Hitchin, N., "Harmonic spinors", Advances in Mathematics 14 (1974), 1–55.
Gromov, M. & Lawson, H. B., "The classification of simply connected manifolds of positive scalar curvature", Annals of Mathematics 111 (1980), 423–434. (Lawson is co-originator.)
Gromov, M. & Lawson, H. B., "Positive scalar curvature and the Dirac operator on complete Riemannian manifolds", Publications Mathématiques de l'IHÉS 58 (1983), 83–196. (Lawson is co-originator; the synthesis paper.)
Lawson, H. B. & Michelsohn, M.-L., Spin Geometry, Princeton University Press, 1989. §IV.1–§IV.4.
Stolz, S., "Simply connected manifolds of positive scalar curvature", Annals of Mathematics 136 (1992), 511–540.
Schick, T., "A counterexample to the (unstable) Gromov-Lawson-Rosenberg conjecture", Topology 37 (1998), 1165–1168.
Rosenberg, J., "C*-algebras, positive scalar curvature, and the Novikov conjecture", Publ. Math. IHÉS 58 (1983), 197–212.
Atiyah, M. F. & Singer, I. M., "The index of elliptic operators IV", Annals of Mathematics 93 (1971), 119–138.
Bär, C., Lecture notes on Dirac operators, 2011 (available online; comprehensive modern textbook treatment of the obstruction chain).

Prerequisites

03.09.14
03.09.15
03.09.05
03.09.04

Tier anchors

beginner: Strogatz / 3Blue1Brown analogous level for the curvature-as-an-obstruction story
intermediate: Lawson-Michelsohn §IV.1–§IV.4, with Gromov-Lawson 1980 (Annals 111) as parallel reading
master: Lichnerowicz 1963 *Spineurs harmoniques*; Hitchin 1974 *Harmonic spinors*; Gromov-Lawson 1980 (Annals 111) + 1983 (IHÉS 58); Stolz 1992 *Simply connected manifolds of positive scalar curvature*; Lawson-Michelsohn §IV.1–§IV.4

References

TODO_REF
Lichnerowicz — Spineurs harmoniques · Comptes Rendus Acad. Sci. Paris 257 (1963), 7–9 — the original formula
TODO_REF
Hitchin — Harmonic spinors · Advances in Mathematics 14 (1974), 1–55 — the α-invariant and exotic-sphere obstructions
TODO_REF
Gromov-Lawson — The classification of simply connected manifolds of positive scalar curvature · Annals of Mathematics 111 (1980), 423–434 — the surgery theorem; the enlargeable manifold theorem; the simply-connected classification programme
TODO_REF
Gromov-Lawson — Positive scalar curvature and the Dirac operator on complete Riemannian manifolds · Publ. Math. IHÉS 58 (1983), 83–196 — the synthesis paper, complete-manifold extensions, enlargeable manifolds
TODO_REF
Lawson-Michelsohn — Spin Geometry · §IV.1–§IV.4 (Lichnerowicz, vanishing, Hitchin, surgery, enlargeable)
TODO_REF
Stolz — Simply connected manifolds of positive scalar curvature · Annals of Mathematics 136 (1992), 511–540 — refinement and classification beyond α
TODO_REF
Schick — A counterexample to the (unstable) Gromov-Lawson-Rosenberg conjecture · Topology 37 (1998), 1165–1168 — boundary of the α-invariant programme

Lean module

Codex.SpinGeometry.PSCObstruction

Mathlib gap

Mathlib has the basic Riemannian-geometry library and a partial Clifford-
algebra layer, but does not yet have the Bochner-Weitzenböck identity
(statable only on a Dirac bundle, absent), the Lichnerowicz formula
(specialisation to spinors, absent), the Atiyah-Singer index theorem
(statement-level only), the α-invariant of Hitchin (requires KO-theory
and the Cl_k-linear refinement), or the Gromov-Lawson surgery and
enlargeable manifold theorems (require both psc-stability under
surgery and a metric-perturbation argument absent from current Mathlib).
The Lean stub records placeholder structures for the Lichnerowicz
vanishing theorem, the α-invariant obstruction, the Gromov-Lawson
surgery theorem, and the enlargeable manifold obstruction; the upstream
contribution roadmap targets `RiemannianManifold.scalarCurvature`,
`LichnerowiczFormula`, `AlphaInvariant`, `EnlargeableManifold`.

Reviewer

TBD

Estimated time

beginner: 25m
intermediate: 65m
master: 110m