03.07.09 · modern-geometry / gauge-theory

Moduli space of ASD connections $M_{k} (S^{4})$

shipped3 tiersLean: none

Anchor (Master): Atiyah-Hitchin-Singer 1978 *Proc. R. Soc. A* 362, 425-461; Donaldson-Kronheimer §4-7; Freed-Uhlenbeck §3-8; Donaldson 1983 *J. Diff. Geom.* 18, 279-315; Uhlenbeck 1982a/b *Comm. Math. Phys.* 83, 84

Intuition [Beginner]

A single instanton on the four-sphere is a single localised lump of curvature that wraps the gauge group around space exactly once. The previous unit gave the explicit one-instanton: pick a centre point and a positive scale, and you get a gauge field. Different choices of centre and scale give different gauge fields. So the one-instanton is not just one field; it is a whole five-parameter family.

The moduli space is the geometric shape of that family. It is a space whose points are gauge fields, where two fields count as the same point if they differ by a gauge change. For charge one on the four-sphere, the family is parametrised by a four-dimensional centre and a one-dimensional scale, giving a five-dimensional shape. For higher charge, the family has more parameters.

The big surprise is that the dimension follows a clean formula: for charge $k$ , the moduli space has real dimension $8 k - 3$ . That formula was first written down by Atiyah, Hitchin, and Singer in 1978.

Visual [Beginner]

Picture a single instanton as a bump in four-dimensional space. The position of the bump is four numbers; the width of the bump is one number. That is five numbers total. Each instanton in the charge-one family is one point in a five-dimensional shape.

When the scale shrinks to zero, the bump spikes into a point. When the scale grows huge, the bump spreads out and the gauge field looks flat. Both limits push the instanton out of the smooth family. To close up the shape, you have to attach those limit configurations as ideal objects.

Worked example [Beginner]

The charge-one moduli space on the four-sphere is the simplest example. Each point is described by two pieces of data: a centre point in four-dimensional space (four numbers, written $x_{0}$ ) and a positive scale (one number, written $λ$ ). The total parameter count is five.

Step 1. Start with the BPST instanton centred at the origin with scale one. The bump has its peak at zero and falls off as the fourth power of the distance.

Step 2. Translate the centre to a new point $x_{0} = (1, 2, 0, 0)$ . The bump moves; the gauge field changes; the resulting connection is a different point in the moduli space.

Step 3. Now change only the scale, to $λ = 3$ . The bump becomes three times wider; its peak height shrinks. The action stays at the same value $8 π^{2}$ by topology.

Step 4. The five-dimensional shape parametrising all such configurations is the upper half-space: four real coordinates for the centre and one positive coordinate for the scale. This is the smooth moduli space at charge one.

What this tells us: the moduli space is not a finite list of solutions but a continuous shape with a definite dimension. The dimension equals five by the formula $8 k - 3$ at $k = 1$ . The shape has a boundary at zero scale where the bump becomes a delta-like spike — those limit configurations are exactly what Uhlenbeck's compactification adds back as ideal points.

Check your understanding [Beginner]

Formal definition [Intermediate+]

Let $P \to S^{4}$ be a principal $SU (2)$ -bundle with second Chern number $c_{2} (P) = k \geq 1$ . Such bundles are classified up to isomorphism by $π_{3} (SU (2)) = Z$ via the clutching construction over the equatorial $S^{3}$ 03.08.04, so for each $k$ there is a unique isomorphism class. Fix an $Ad$ -invariant inner product on $su (2)$ as in 03.07.07, and fix the round metric on $S^{4}$ .

Let $A_{P}$ denote the affine space of smooth connections on $P$ (modelled on $Ω^{1} (S^{4}; ad P)$ 03.05.07). The gauge group is $$ \mathcal{G}_P = \Gamma(\operatorname{Aut}(P)) = {g: P \to P \mid g \text{ is fibre-preserving and $G$ -equivariant}}, $$ isomorphic to $C^{\infty} (S^{4}; SU (2)^{Ad})$ via sections of the associated adjoint bundle. The group $G_{P}$ acts on $A_{P}$ by pullback, $$ g \cdot A = g^{-1} A g + g^{-1} dg, \qquad F_{g \cdot A} = g^{-1} F_A g. $$ The action is smooth in the appropriate Banach-space completion (typically Sobolev $L_{l}^{2}$ for $l \geq 3$ to ensure $C^{1}$ regularity of $A$ ).

Definition (moduli space of ASD connections). The moduli space of anti-self-dual connections on $P$ is the quotient $$ \mathcal{M}_k(S^4) = {A \in \mathcal{A}_P : F_A^+ = 0} / \mathcal{G}P, $$ where $F_A^+ = \pi+(F_A) $i s t h ese l f - d u a l p a r t o f t h ec u r v a t u r e w i t h r es p ec tt o t h er o u n d - m e t r i cH o d g es t a r, an d t h e q u o t i e n t i s b y t h e g a ug e a c t i o n . T w o A S D co nn ec t i o n s a r e i d e n t i f i e d i f an d o n l y i f t h ey d i f f er b y an e l e m e n t o f$ \mathcal{G}_P$ 03.05.09.

The moduli space splits as $M_{k} = M_{k}^{*} ⊔ M_{k}^{red}$ , where $M_{k}^{*}$ is the locus of irreducible connections (those whose stabiliser in $G_{P}$ is the centre $Z (SU (2)) = {\pm I}$ ) and $M_{k}^{red}$ is the locus of reducible connections (whose stabiliser is strictly larger). On $S^{4}$ all ASD connections with $k \geq 1$ are irreducible.

Definition (framed moduli space). Fix a base point $x_{\infty} \in S^{4}$ and a frame in the fibre $P_{x_{\infty}}$ . The framed moduli space $M_{k} (S^{4})$ is the quotient by the subgroup $G_{P}^{0} \subset G_{P}$ of gauge transformations equal to the identity at $x_{\infty}$ . The forgetful map $M_{k} \to M_{k}$ is a principal $SU (2)$ -bundle on the irreducible part, increasing the dimension by $dim SU (2) = 3$ .

Sign convention. We use the Donaldson-Kronheimer convention: $*^{2} = + 1$ on two-forms in dimension four, $F_{A}^{+} = 0$ means anti-self-dual, and the orientation of $S^{4}$ is chosen so that $k = c_{2} (P) \geq 0$ for ASD connections. Reversing the orientation swaps self-dual and anti-self-dual and negates $k$ ; the resulting moduli space $M_{- k} (\overset{ˉ}{S}^{4})$ is canonically identified with $M_{k} (S^{4})$ .

Counterexamples to common slips

The condition $F_{A}^{+} = 0$ defines anti-self-dual connections, not self-dual ones. The reverse condition $F_{A}^{-} = 0$ defines self-dual connections; on a fixed orientation, the moduli spaces are different (one is empty unless $k \leq 0$ ).
The moduli space $M_{k} (S^{4})$ is not the space of solutions of the ASD equation — it is the quotient of that space by the gauge group. A gauge family of solutions is one point in $M_{k}$ , not a line.
The dimension formula $8 k - 3$ counts real dimensions, not complex. The framed moduli space has dimension $8 k$ ; the difference is the three-dimensional gauge stabiliser at the framing point.
On a general four-manifold $M$ , the dimension is $8 k - 3 (1 + b^{+} (M))$ . The $S^{4}$ case is special only because $b^{+} (S^{4}) = 0$ ; the formula reduces to $8 k - 3$ . Reading "the dimension is $8 k - 3$ " out of context as the universal answer is a frequent slip.

Key theorem with proof [Intermediate+]

Theorem (Atiyah-Hitchin-Singer dimension formula; 1978). Let $P \to M$ be a principal $SU (2)$ -bundle with second Chern number $k$ over a compact oriented Riemannian four-manifold $M$ , and let $A$ be an irreducible anti-self-dual connection on $P$ . Then the moduli space $M_{k} (M)$ is smooth of real dimension $$ \dim \mathcal{M}_k(M) = 8k - 3(1 + b^+(M)) $$ in a neighbourhood of $[A]$ , provided the obstruction space vanishes. For $M = S^{4}$ , $b^{+} (S^{4}) = 0$ and so $dim M_{k} (S^{4}) = 8 k - 3$ ^{[Atiyah-Hitchin-Singer 1978]}.

Proof. Linearise the ASD equation $F_{A}^{+} = 0$ at the connection $A$ . A small variation $a \in Ω^{1} (M; ad P)$ produces $F_{A + a} = F_{A} + d_{A} a + a \land a$ , whose self-dual part is $$ F^+{A+a} = F^+A + d_A^+ a + (a \wedge a)^+, $$ where $d_A^+ = \pi+ \circ d_A $an d$ \pi+ \colon \Omega^2(M; \operatorname{ad} P) \to \Omega^2_+(M; \operatorname{ad} P) $i s t h eor t h o g o na l p r o j ec t i o n . A t an A S D co nn ec t i o n,$ F^+_A = 0$, and the linearised ASD equation reads $$ d_A^+ a = 0. \tag{ $†$ } $$ The infinitesimal gauge action on connections is $a \mapsto a + d_{A} ξ$ for $ξ \in Ω^{0} (M; ad P)$ . Modding out by the gauge action means restricting to a slice transverse to the orbit, conventionally the Coulomb-gauge slice $ker d_{A}^{*} \subset Ω^{1} (M; ad P)$ .

Putting these together yields the deformation complex of $A$ : $$ \mathcal{D}A: \quad 0 \to \Omega^0(M; \operatorname{ad} P) \xrightarrow{d_A} \Omega^1(M; \operatorname{ad} P) \xrightarrow{d_A^+} \Omega^2+(M; \operatorname{ad} P) \to 0. \tag{ $† †$ } $$ This complex is elliptic: its assembled symbol is $$ \sigma(d_A + d_A^) = \sigma(d + d^) \otimes \operatorname{id}{\operatorname{ad} P}, $$ which is the elliptic symbol of the operator $d + d^{*}$ acting on $\Omega^0 \oplus \Omega^2+ $f r o m$ \Omega^1 $. T h eco h o m o l o g y o f$ \mathcal{D}A$ has three pieces: $$ H^0 = \ker(d_A), \qquad H^1 = \frac{\ker(d_A^+)}{\operatorname{im}(d_A)}, \qquad H^2 = \frac{\Omega^2+}{\operatorname{im}(d_A^+)} = \operatorname{coker}(d_A^+). $$ The interpretations are: $H^{0}$ is the Lie algebra of the stabiliser of $A$ in $G_{P}$ (zero for irreducible $A$ ); $H^{1}$ is the formal tangent space $T_{[A]} M_{k}$ ; $H^{2}$ is the obstruction space (vanishing $H^{2}$ means the moduli space is smooth at $[A]$ ).

By the Atiyah-Singer index theorem 03.09.10 applied to the elliptic complex $D_{A}$ , the alternating sum of cohomology dimensions equals a topological index: $$ \dim H^0 - \dim H^1 + \dim H^2 = -\operatorname{ind}(\mathcal{D}_A). $$ A standard Chern-character computation (Atiyah-Hitchin-Singer §6) gives $$ \operatorname{ind}(\mathcal{D}_A) = 8k - 3(1 + b^+(M)). $$ The arithmetic: the symbol class of $d_{A} + d_{A}^{*}$ in K-theory is the symbol of $d + d^{*}$ tensored with the adjoint bundle of $P$ , and the index formula evaluates to $- rk (ad P) \cdot χ (M) - 3 b^{+} (M) + 8 p_{1} (ad P) /4 = 8 k - 3 (1 + b^{+} (M))$ after applying the signature theorem and the Chern-Weil identity $p_{1} (ad P) = 4 c_{2} (P)$ for $SU (2)$ bundles (the adjoint bundle has rank three, so the rank contribution is $3 χ (M) = 3 (b^{0} - b^{1} + b^{+} + b^{-} - b^{3} + b^{4})$ ; on a simply-connected four-manifold with $b^{1} = 0$ this reduces to $3 (2 + b^{+} + b^{-})$ , and combining with the signature $τ (M) = b^{+} - b^{-}$ recovers $- 3 (1 + b^{+})$ after cancellation).

For irreducible $A$ (so $H^{0} = 0$ ) and vanishing obstruction $H^{2} = 0$ , we obtain $$ \dim \mathcal{M}_k(M) = \dim H^1 = -\operatorname{ind}(\mathcal{D}_A) = 8k - 3(1 + b^+(M)). $$ Setting $M = S^{4}$ , $b^{+} (S^{4}) = 0$ yields $dim M_{k} (S^{4}) = 8 k - 3$ . The vanishing of $H^{2}$ on $S^{4}$ for the round metric is a consequence of the Weitzenböck identity applied to $d_{A}^{+}$ on the positively-curved sphere (Atiyah-Hitchin-Singer 1978 §7). $□$

Bridge. The dimension formula builds toward 03.07.07 (BPST instanton), where the Synthesis paragraph already announced $dim M_{1} (S^{4}) = 5$ as the five-parameter family of centre and scale; this unit upgrades that announcement to a proof and identifies the general $k$ formula. The foundational reason a dimension count is available is exactly the elliptic theory of the deformation complex (§ $† †$ ): ellipticity gives finite-dimensional cohomology, and the Atiyah-Singer index theorem 03.09.10 computes the alternating sum from topology alone — the central insight is that the moduli-space dimension is a Fredholm index, not a geometric count. This pattern appears again in 03.09.10 (Atiyah-Singer), where the index of the chiral Dirac operator on a four-manifold likewise equals an integer-valued characteristic number; the bridge is the recognition that ASD moduli and chiral spinor indices are governed by the same K-theoretic machinery. Putting these together, the bridge is the identification of the moduli-space dimension with $- ind (d_{A} + d_{A}^{*})$ , where the right-hand side is computed by Chern-Weil and the left-hand side has geometric content. The pattern recurs in Donaldson polynomial invariants 03.07.10 (the planned successor unit) where the moduli space's intersection theory produces diffeomorphism invariants — generalises Bogomolny saturation in 03.07.07 from "minimisers exist" to "minimisers form a smooth manifold whose topology is computable from $M$ alone". This is exactly the bridge from gauge theory to four-manifold topology.

Exercises [Intermediate+]

Exercise 3 (medium, symbolic).

Show that the deformation complex $D_{A}$ of an ASD connection on a compact four-manifold is elliptic.

Hint

Compute the symbol sequence and show it is exact off the zero section of $T^{*} M$ .

Answer

The principal symbols of $d_{A}$ and $d_{A}^{+}$ are independent of the connection (they only depend on the leading-order derivatives), so $σ (d_{A}) (ξ) = ξ \land \cdot$ and $σ (d_{A}^{+}) (ξ) = π_{+} (ξ \land \cdot)$ for $ξ \in T^{*} M$ . The symbol sequence $$ 0 \to \operatorname{ad} P_x \xrightarrow{\xi \wedge} T^x M \otimes \operatorname{ad} P_x \xrightarrow{\pi+ \circ \xi \wedge} \Lambda^2_+ T^x M \otimes \operatorname{ad} P_x \to 0 $$ is exact for every $ξ \neq = 0$ . For exactness at the middle: $\ker(\pi+ \xi \wedge) = {a : \pi_+(\xi \wedge a) = 0} $, an d a c a l c u l a t i o nin t h es t an d a r d ba s i s$ {e_1, \ldots, e_4} $w i t h$ \xi = e_1 $s h o w s t hi s k er n e l i ss p ann e d b y$ \xi \wedge \xi = 0 $pl u se l e m e n t s p r o p or t i o na l t o$ \xi $. T h e ima g eo f$ \xi \wedge $f r o m$ \operatorname{ad} P $in t o$ T^*M \otimes \operatorname{ad} P $a t$ \xi = e_1 $i s t h e l in e$ e_1 \otimes \operatorname{ad} P $, w hi c hma t c h es t h e k er n e l o f$ \pi_+ \circ \xi \wedge $in t h es am e d i r ec t i o n . E x a c t n ess a tt h er i g h t i s d im e n s i o n - co u n t in g : d o maini s$ 4 \cdot 3 $, k er n e l i s$ 1 \cdot 3 $, ima g e i s$ 3 \cdot 3 = 9 $, co d o maini s$ 3 \cdot 3 = 9 $. T h ese q u e n ce i se x a c t a t e v er y$ \xi \neq 0 $, h e n ce$ \mathcal{D}_A$ is elliptic.

Exercise 4 (medium, symbolic).

Identify the geometric meaning of the three cohomology groups $H^{0}, H^{1}, H^{2}$ of the deformation complex.

Hint

$H^{0}$ measures gauge stabilisers; $H^{1}$ measures moduli tangents; $H^{2}$ measures obstructions.

Answer

$H^{0} = ker (d_{A}) \subset Ω^{0} (ad P)$ consists of covariantly constant sections of $ad P$ ; these exponentiate to gauge transformations fixing $A$ , so $H^{0}$ is the Lie algebra of the stabiliser of $A$ in $G_{P}$ . For irreducible connections on a compact connected four-manifold, $H^{0} = 0$ when the bundle is non-flat (any covariantly constant section reduces the structure group).

$H^{1} = ker (d_{A}^{+}) / im (d_{A})$ is the Coulomb-slice tangent space at $[A]$ modulo infinitesimal gauge equivalence; it equals the formal tangent space $T_{[A]} M_{k}$ .

$H^{2} = coker (d_{A}^{+})$ is the obstruction space. The Kuranishi map sends a tangent vector $a \in H^{1}$ to the obstruction $(a \land a)^{+} mod im (d_{A}^{+}) \in H^{2}$ ; the zero set of this map (locally near $A$ ) is the actual moduli space. Vanishing $H^{2}$ implies smoothness at $[A]$ of dimension $dim H^{1}$ .

Exercise 5 (medium, symbolic).

Show that the parameter count $(x_{0}, λ) \in R^{4} \times R_{> 0}$ for the BPST family gives a manifold diffeomorphic to the open upper half-space $R^{4} \times R_{> 0}$ , and explain why this is the full $M_{1} (S^{4})$ .

Hint

Match the parameter count to the dimension formula $8 k - 3$ and use uniqueness of $SU (2)$ -bundles on $S^{4}$ in each topological class.

Answer

The BPST family $$ A^{(x_0, \lambda)}\mu(x) = \frac{(x - x_0)^\nu}{|x - x_0|^2 + \lambda^2} \bar\eta^a{\mu\nu} T_a $$ is parametrised by $(x_{0}, λ) \in R^{4} \times R_{> 0}$ , a five-dimensional open manifold. By the dimension formula $dim M_{1} (S^{4}) = 8 \cdot 1 - 3 = 5$ , the BPST family fills the entire moduli space dimension-wise. By the AHDM construction 03.07.07, every ASD $SU (2)$ -connection on $S^{4}$ with $c_{2} = 1$ is gauge-equivalent to a BPST connection for some $(x_{0}, λ)$ . Two BPST connections with different parameters are not gauge-equivalent (the centre is determined by where the curvature density peaks; the scale by the width of the bump). So $(x_{0}, λ) \mapsto [A^{(x_{0}, λ)}]$ is a smooth bijection from $R^{4} \times R_{> 0}$ onto $M_{1} (S^{4})$ , and standard elliptic regularity makes it a diffeomorphism. Hence $M_{1} (S^{4}) ≅ R^{4} \times R_{> 0}$ , the open upper half-space.

Exercise 6 (hard, short-answer).

Describe what happens to a sequence of charge-one BPST connections $A^{(0, λ_{n})}$ with $λ_{n} \to 0$ . In what sense does the sequence converge, and what is the limit in the Uhlenbeck compactification?

Hint

The pointwise curvature density concentrates; the total energy is preserved; the bundle "bubbles off".

Answer

The pointwise squared curvature of $A^{(0, λ_{n})}$ is $96 λ_{n}^{4} / (∣ x ∣^{2} + λ_{n}^{2})^{4}$ . As $λ_{n} \to 0$ , this density vanishes uniformly on compact subsets of $R^{4} ∖ {0}$ but its integral over $R^{4}$ remains constant at $16 π^{2}$ . So the curvature measure $∣ F_{A_{n}} ∣^{2} d vol$ converges in the distributional sense to $16 π^{2} δ_{0}$ , a delta-measure at the origin. The limiting connection on $S^{4} ∖ {0}$ is gauge-equivalent to the flat connection on the product $SU (2)$ -bundle.

In the Uhlenbeck compactification $\overline{M}_{k}$ , this limit is recorded as an ideal instanton: a pair $(A_{\infty}, x_{0})$ where $A_{\infty} \in M_{0}$ is a flat connection and $x_{0} \in S^{4}$ is the bubbling point. The total energy is conserved: $8 π^{2} \cdot 1 = 8 π^{2} \cdot 0 + 8 π^{2} \cdot 1$ , with the latter term carried by the delta at $x_{0}$ . The ideal-instanton stratum has dimension $dim M_{0} + dim S^{4} = 0 + 4 = 4$ , one less than the bulk dimension five, as expected for a codimension-one boundary stratum. Rubric: full credit for identifying the distributional limit, the product-bundle gauge equivalence away from $x_{0}$ , and the Uhlenbeck record $(A_{\infty}, x_{0}) \in M_{0} \times S^{4}$ .

Exercise 7 (hard, proof).

Prove that the moduli space $M_{k} (S^{4})$ is orientable.

Hint

Use the determinant-line-bundle construction: orientability of the index bundle of a family of Fredholm operators.

Answer

The tangent bundle of $M_{k}^{*}$ at a point $[A]$ is the index space $H^{1} (D_{A})$ , the first cohomology of the deformation complex. As $A$ varies over $M_{k}^{*}$ , these cohomology spaces fit together into a smooth real vector bundle on $M_{k}^{*}$ — the index bundle of the family ${d_{A} + d_{A}^{*} : A \in M_{k}^{*}}$ of Fredholm operators.

The orientability of $M_{k}^{*}$ is equivalent to the orientability of this index bundle, which in turn is governed by the parity of the first Stiefel-Whitney class. Donaldson's argument (Donaldson 1990 §5, Donaldson-Kronheimer §5.4) shows that on a simply-connected compact oriented four-manifold $M$ , the moduli space $M_{k} (M)$ is orientable, with a canonical orientation determined by a choice of homology orientation on $M$ . The proof reduces to showing that the determinant line bundle of the family of Dirac-like operators $d_{A} + d_{A}^{*}$ is canonically trivialised by the integer-cohomology structure of $M$ . The simply-connected case applies to $M = S^{4}$ .

For irreducible connections on $S^{4}$ , the index bundle has its determinant line equal to the product real line bundle, so $M_{k}^{*} (S^{4})$ is orientable. Rubric: full credit for invoking the index-bundle construction and the Donaldson determinant-line-bundle theorem.

Exercise 8 (hard, short-answer).

Explain why $b^{+} (S^{4}) = 0$ implies the vanishing of the obstruction $H^{2}$ for the round-metric ASD connection on $S^{4}$ , and how the picture changes on a four-manifold with $b^{+} > 0$ .

Hint

Use the Weitzenböck formula for $d_{A}^{+}$ on a positively-curved manifold and compare to the harmonic-form picture.

Answer

The obstruction space is $H^{2} = coker (d_{A}^{+})$ , isomorphic (by elliptic regularity and Hodge theory) to $ker ((d_{A}^{+})^{*}) \subset Ω_{+}^{2} (M; ad P)$ . On a compact four-manifold $M$ with the round metric on $S^{4}$ (or any positively-curved Einstein metric), the Weitzenböck formula gives $$ (d_A^+)^* d_A^+ = \tfrac{1}{2}\nabla_A^* \nabla_A + \mathcal{R}+ + [F_A^+, \cdot], $$ where $\mathcal{R}+ $i s a c u r v a t u r eo p er a t or o n$ \Omega^2_+ $p r o p or t i o na l t o t h esc a l a r c u r v a t u r e, an d t h e l a s t co mm u t a t or t er m v ani s h es w h e n$ A $i s A S D . F or$ \mathrm{SU}(2) $o n$ S^4 $w i t h t h er o u n d m e t r i c,$ \mathcal{R}+ > 0 $p o in tw i se . T h e W e i t z e nb \overset{o}{¨} c k f or m u l a f or ces an y e l e m e n t$ \omega \in \ker((d_A^+)^) $t os a t i s f y$ 0 = \int \langle \omega, \nabla_A^\nabla_A \omega + 2 \mathcal{R}+ \omega \rangle = |\nabla_A \omega|^2 + 2 \int \langle \mathcal{R}_+ \omega, \omega \rangle $, b o t h t er m s n o n - n e g a t i v e, so$ \omega = 0 $. H e n ce$ H^2 = 0 $o n$ S^4$ at every irreducible ASD connection.

On a four-manifold $M$ with $b^{+} (M) > 0$ , harmonic self-dual two-forms exist (they generate $H_{+}^{2} (M; R)$ of dimension $b^{+}$ ), and these obstruct deformations: the obstruction $H^{2}$ contains the constant self-dual harmonic forms tensored with the Lie algebra. The Atiyah-Hitchin-Singer formula $dim M_{k} (M) = 8 k - 3 (1 + b^{+} (M))$ encodes this: each extra unit of $b^{+}$ subtracts three (the dimension of $ad P_{x}$ ) from the dimension. Generic metric perturbations on $M$ resolve this obstruction; this is Donaldson's transversality argument (Donaldson 1990 §4). Rubric: full credit for the Weitzenböck argument on $S^{4}$ plus the harmonic-form interpretation of the $b^{+}$ correction.

Lean formalization [Intermediate+]

lean_status: none. Mathlib has no infrastructure for principal $G$ -bundles with Sobolev-class connections, gauge groups acting on connection spaces, or the elliptic deformation complex of an ASD connection. The closest existing pieces are in Mathlib.Geometry.Manifold.* (smooth manifolds, tangent bundles) and Mathlib.Analysis.Fourier.* (some Sobolev-space results), neither of which support the gauge-theoretic constructions needed here. The required formalisation reads schematically:

import Mathlib.Geometry.Manifold.SmoothManifoldWithCorners

-- Placeholder. Mathlib lacks principal bundles, gauge groups,
-- elliptic deformation complexes, and Atiyah-Singer index theory.
axiom PrincipalBundle (G M : Type) : Type
axiom Connection {G M : Type} (P : PrincipalBundle G M) : Type
axiom GaugeGroup {G M : Type} (P : PrincipalBundle G M) : Type
axiom ASDConnection {G M : Type} (P : PrincipalBundle G M) (A : Connection P) : Prop

-- The moduli space, as a quotient of ASD connections by gauge.
axiom ModuliSpace {G M : Type} (P : PrincipalBundle G M) : Type

-- The Atiyah-Hitchin-Singer dimension formula, on S^4.
axiom ahs_dimension_s4 (k : Nat) (P : PrincipalBundle (SpecialUnitary 2) (Sphere 4))
    (hk : c2 P = k) : dim (ModuliSpace P) = 8 * k - 3

The substantive gap encompasses the entire deformation-complex framework: the elliptic complex $D_{A}$ , its cohomology, the Kuranishi map governing the local model of $M_{k}$ , the Atiyah-Singer computation of $ind (D_{A})$ in K-theory, and the Uhlenbeck compactification. Each of these is a separate formalisation project.

Advanced results [Master]

Theorem (Atiyah-Hitchin-Singer 1978). Let $P \to M$ be a principal $SU (2)$ -bundle with $c_{2} (P) = k$ on a compact oriented Riemannian four-manifold $M$ . The space $\mathcal{M}_k^(M) $o f i r r e d u c ib l e an t i - se l f - d u a l co nn ec t i o n s m o d u l o g a ug e i s a s m oo t h r e a l mani f o l d (inan e i g hb o u r h oo d o f e a c h$ [A] $w h er e t h eo b s t r u c t i o n s p a ce$ H^2(\mathcal{D}_A)$ vanishes) of dimension* $$ \dim \mathcal{M}_k(M) = 8k - 3(1 + b^+(M)). $$ For $M = S^{4}$ with the round metric, the obstruction vanishes at every irreducible ASD connection and $M_{k} (S^{4})$ is a $(8 k - 3)$ -dimensional smooth manifold ^{[Atiyah-Hitchin-Singer 1978]}.

The case $k = 1$ on $S^{4}$ yields the open five-dimensional half-space $R^{4} \times R_{> 0}$ parametrised by centre and scale, the BPST family. For $k = 2$ on $S^{4}$ , $dim M_{2} = 13$ ; the family is the ADHM moduli of pairs of $SU (2)$ -instantons, encompassing all relative positions, relative gauge rotations, and individual scales.

Theorem (Uhlenbeck compactness; 1982). Let ${A_{n}}$ be a sequence of ASD connections on $P \to M$ with $c_{2} (P) = k$ , all of action bounded above by $C < \infty$ . After passing to a subsequence and applying gauge transformations, ${A_{n}}$ converges weakly in $L_{1}^{2}$ on $M ∖ {x_{1}, \dots, x_{l}}$ to an ASD connection $A_{\infty}$ on a possibly different principal bundle $P^{'} \to M$ with $c_{2} (P^{'}) = k^{'} \leq k$ , and the curvature densities $∣ F_{A_{n}} ∣^{2} d vol$ converge as measures to $∣ F_{A_{\infty}} ∣^{2} d vol + \sum_{i} 8 π^{2} m_{i} δ_{x_{i}}$ with $m_{i} \geq 1$ integers and $k = k^{'} + \sum m_{i}$ ^{[Uhlenbeck 1982a]}^{[Uhlenbeck 1982b]}.

This is the bubbling phenomenon. A sequence of charge- $k$ instantons can lose charge to one or more points $x_{i}$ , where small concentrated instantons of charge $m_{i}$ "bubble off". The total topological charge is conserved across the degeneration.

Theorem (Uhlenbeck removable singularities; 1982). Let $A$ be an ASD connection on the product $SU (2)$ -bundle over $R^{4} ∖ {0}$ with finite Yang-Mills action. There exists a smooth gauge transformation $g$ on $R^{4} ∖ {0}$ and a smooth ASD connection $\tilde{A}$ on a possibly nontrivial principal $SU (2)$ -bundle $\tilde{P} \to R^{4}$ such that $g \cdot A = \tilde{A} ∣_{R^{4} ∖ {0}}$ ^{[Uhlenbeck 1982a]}.

Removable singularities mean: a finite-action ASD connection on a punctured neighbourhood extends, after gauge transformation, across the puncture. Combined with Yang-Mills's conformal invariance in dimension four, this packages all finite-action ASD connections on $R^{4}$ as ASD connections on $S^{4} = R^{4} \cup {\infty}$ .

Definition (Uhlenbeck compactification). The Uhlenbeck compactification of $M_{k} (M)$ is the topological space $$ \overline{\mathcal{M}}k(M) = \bigsqcup{l \geq 0} \mathcal{M}_{k - l}(M) \times \operatorname{Sym}^l(M) ,/, \sim, $$ where $Sym^{l} (M)$ is the $l$ -fold symmetric product of $M$ and the equivalence $\sim$ patches together the strata via the bubbling correspondence. The top stratum is $M_{k} (M)$ (no bubbling); the codimension- $4 l$ strata $M_{k - l} (M) \times Sym^{l} (M)$ record degenerations where $l$ units of charge have bubbled off at $l$ points.

Theorem (Donaldson 1983 — polynomial invariants). For a compact oriented simply-connected smooth four-manifold $M$ with $b^{+} (M) \geq 1$ odd, the moduli space $M_{k} (M)$ admits an orientation and (for generic metric) is a smooth manifold of dimension $d_{k} = 8 k - 3 (1 + b^{+} (M))$ . The Donaldson polynomial $q_{M, k} : H_{2} (M; Z)^{\otimes d_{k} /2} \to Z$ is defined by intersection theory on $\overline{M}_{k} (M)$ paired with the $μ$ -map classes from the universal bundle. The polynomial $q_{M, k}$ is a smooth-structure invariant of $M$ — invariant under diffeomorphisms but in general not under homeomorphisms of $M$ ^{[Donaldson 1983]}.

Donaldson 1983 used $q_{M, k}$ to exhibit pairs of compact simply-connected smooth four-manifolds with the same intersection form (hence homeomorphic by Freedman's theorem) but distinct $q_{M, k}$ , hence not diffeomorphic. The smooth Poincaré conjecture in dimension four remains open as of this writing; Donaldson invariants and their successors (Seiberg-Witten) supply the strongest known smooth-structure invariants.

Theorem (ADHM parametrisation of $M_{k} (S^{4})$ ; Atiyah-Hitchin-Drinfeld-Manin 1978). For $G = SU (2)$ and $k \geq 1$ , the framed moduli space $M_{k} (S^{4})$ is diffeomorphic to the algebraic variety $$ {(B_1, B_2, I, J) : [B_1, B_1^] + [B_2, B_2^] + II^* - J^*J = 0, \ [B_1, B_2] + IJ = 0} ,/, \mathrm{U}(k), $$ where $B_{1}, B_{2} \in Mat_{k \times k} (C)$ , $I \in Mat_{k \times 2} (C)$ , $J \in Mat_{2 \times k} (C)$ , and $U (k)$ acts by simultaneous conjugation. This variety has real dimension $4 k^{2} + 4 k + 4 k - k^{2} - 3 k^{2} = 8 k$ , matching $dim M_{k} (S^{4})$ ^{[Atiyah-Hitchin-Drinfeld-Manin 1978]}.

The ADHM construction reduces the analytic problem of classifying ASD connections on $S^{4}$ to a finite-dimensional linear-algebra problem on quadruples of matrices modulo a unitary action. Each ADHM datum produces an ASD connection via a Penrose-twistor recipe (Donaldson-Kronheimer §6); the construction is fully constructive.

Synthesis. The moduli space $M_{k} (S^{4})$ is the central object of four-dimensional gauge theory: a finite-dimensional smooth manifold of dimension $8 k - 3$ whose points are gauge-equivalence classes of absolute minimisers of the Yang-Mills action 03.07.05 in topological class $k$ . The central insight of Atiyah-Hitchin-Singer 1978 is that this dimension is computable from topology alone — the moduli-space dimension equals the negative of the Atiyah-Singer index 03.09.10 of an elliptic complex assembled from the ASD-equation linearisation and the Coulomb gauge condition, identifying the geometric count with a Fredholm-analytic index. The foundational reason a smooth-manifold structure exists at irreducible ASD connections is the vanishing of the second cohomology $H^{2} (D_{A})$ , which on $S^{4}$ with the round metric follows from the Weitzenböck-Bochner positivity argument.

Putting these together, the bridge from analysis to topology is the Uhlenbeck compactification: $M_{k} (S^{4})$ is open and non-compact (bubbling allows curvature to concentrate to delta-spikes and lower-charge configurations to absorb the remaining topology), but its closure $\overline{M}_{k} (S^{4})$ is compact, with boundary strata $M_{k - l} (S^{4}) \times Sym^{l} (S^{4})$ that combinatorially record charge loss. This pattern appears again in 03.07.07 (BPST instanton), where the explicit one-instanton family $R^{4} \times R_{> 0}$ at $λ \to 0$ exits the open moduli space through a delta-singularity at the centre — and this is exactly the codimension-one stratum $M_{0} (S^{4}) \times S^{4} ≅ S^{4}$ that closes the moduli space.

The same pattern recurs in Donaldson's 1983 application to four-manifold topology: the intersection theory on $\overline{M}_{k} (M)$ paired with $μ$ -classes generalises the BPST-and-bubble picture from $S^{4}$ to an arbitrary compact simply-connected four-manifold $M$ , producing polynomial invariants $q_{M, k}$ that distinguish smooth structures inside fixed homeomorphism classes. The Bogomolny bound supplies the energy floor; the ASD equation cuts out the minimisers; the moduli space inherits dimension from Atiyah-Singer; the Uhlenbeck compactification glues in ideal instantons; and the Donaldson polynomial extracts the smooth-structure invariant. The bridge is the recognition that all four of these mechanisms are the same four-dimensional phenomenon $Λ^{2} = Λ^{+} \oplus Λ^{-}$ , viewed at four different scales: pointwise (Bogomolny), local (ASD), global (moduli), and topological (Donaldson). This is exactly the bridge from gauge theory to differential topology, and it identifies the moduli space with the Donaldson invariant.

Full proof set [Master]

Proposition (deformation complex is elliptic). The complex $D_{A} : 0 \to Ω^{0} (M; ad P) d_{A} Ω^{1} (M; ad P) d_{A}^{+} Ω_{+}^{2} (M; ad P) \to 0$ is elliptic at every point of $M$ .

Proof. Ellipticity of a complex of differential operators means the symbol sequence is exact off the zero section of $T^{*} M$ . The principal symbols of $d_{A}$ and $d_{A}^{+}$ depend only on the leading derivatives and are independent of $A$ : $σ (d_{A}) (ξ) = ξ \land$ on $ad P$ -valued forms, and $σ (d_{A}^{+}) (ξ) = π_{+} \circ (ξ \land)$ . Fix $ξ \in T_{x}^{*} M$ , $ξ \neq = 0$ . The symbol sequence at $ξ$ is $$ 0 \to \operatorname{ad} P_x \xrightarrow{\xi \wedge} T^x M \otimes \operatorname{ad} P_x \xrightarrow{\pi+ \circ (\xi \wedge)} \Lambda^2_+ T^_x M \otimes \operatorname{ad} P_x \to 0. $$ Exactness at $ad P_{x}$ : if $ξ \land a = 0$ for $a \in ad P_{x}$ , write $a = a^{j} \otimes T_{j}$ ; then $ξ \land a = (ξ \otimes a^{j}) \otimes T_{j} = 0$ forces $a^{j} = 0$ for all $j$ , so $a = 0$ . Injectivity holds.

Exactness at $T_{x}^{*} M \otimes ad P_{x}$ : take an orthonormal coframe ${ξ, e_{2}, e_{3}, e_{4}}$ at $x$ . A general element is $a = a^{1} ξ + \sum_{i = 2}^{4} a^{i} e_{i}$ with $a^{i} \in ad P_{x}$ . Then $ξ \land a = \sum_{i = 2}^{4} a^{i} ξ \land e_{i}$ . The two-form $ξ \land e_{i}$ has self-dual part $\frac{1}{2} (ξ \land e_{i} + * (ξ \land e_{i}))$ , and the three two-forms $ξ \land e_{2}, ξ \land e_{3}, ξ \land e_{4}$ have linearly independent self-dual projections (their self-dual parts form a basis of $Λ_{+}^{2}$ , modulo dualisation). So $π_{+} (ξ \land a) = 0$ forces $a^{i} = 0$ for $i = 2, 3, 4$ , leaving $a = a^{1} ξ = ξ \land a^{1}$ in the image of the previous map.

Exactness at $Λ_{+}^{2} \otimes ad P_{x}$ : the map $π_{+} \circ (ξ \land) : T_{x}^{*} M \otimes ad P_{x} \to Λ_{+}^{2} \otimes ad P_{x}$ has image ${π_{+} (ξ \land a) : a \in T_{x}^{*} M} \otimes ad P_{x}$ . The map $a \mapsto π_{+} (ξ \land a)$ from $T_{x}^{*} M$ to $Λ_{+}^{2}$ has rank $3$ (the three independent self-dual projections of $ξ \land e_{i}$ for $i = 2, 3, 4$ ) and $Λ_{+}^{2}$ has dimension $3$ , so the map is surjective. Surjectivity holds.

Hence the symbol sequence is exact for every $ξ \neq = 0$ , and $D_{A}$ is elliptic. $□$

Proposition (index of $D_{A}$ on $S^{4}$ ). For a principal $SU (2)$ -bundle $P \to S^{4}$ with $c_{2} (P) = k$ and any ASD connection $A$ on $P$ , $$ \operatorname{ind}(\mathcal{D}_A) = 3 - 8k. $$

Proof. The Euler characteristic of $D_{A}$ is $$ \operatorname{ind}(\mathcal{D}_A) = \dim H^0 - \dim H^1 + \dim H^2. $$ By the Atiyah-Singer index theorem 03.09.10 applied to $D_{A}$ , the index equals a topological characteristic-number integral computed in K-theory from the symbol class. The Atiyah-Hitchin-Singer computation (1978 §6) packages the answer as $$ \operatorname{ind}(\mathcal{D}_A) = -2 p_1(\operatorname{ad} P)[M] + \tfrac{1}{2}(\chi(M) + \tau(M)) \cdot \operatorname{rk}(\operatorname{ad} P) $$ for a general compact oriented Riemannian four-manifold $M$ , where $χ$ is the Euler characteristic, $τ$ is the signature, $p_{1}$ is the first Pontryagin class, and $rk (ad P) = 3$ for $SU (2)$ . Using the identity $p_{1} (ad P) = 4 c_{2} (P)$ for $SU (2)$ bundles (a standard Chern-Weil identity 03.06.04) and the values $χ (S^{4}) = 2$ , $τ (S^{4}) = 0$ , $c_{2} (P) [S^{4}] = k$ : $$ \operatorname{ind}(\mathcal{D}_A) = -2 \cdot 4k + \tfrac{1}{2}(2 + 0) \cdot 3 = -8k + 3. $$ For general $M$ the formula reads $ind (D_{A}) = - 8 k + \frac{3}{2} (χ (M) + τ (M))$ , with $χ + τ = 2 (1 - b^{1} (M) + b^{+} (M))$ , giving the AHS dimension formula $- ind = 8 k - 3 (1 - b^{1} (M) + b^{+} (M))$ . On a simply-connected manifold $b^{1} = 0$ , recovering $8 k - 3 (1 + b^{+})$ . $□$

Proposition ( $M_{1} (S^{4}) ≅ R^{4} \times R_{> 0}$ ). The moduli space of irreducible ASD $SU (2)$ -connections on $S^{4}$ with $c_{2} = 1$ is diffeomorphic to the open upper half-space $R^{4} \times R_{> 0}$ , with coordinates $(x_{0}, λ)$ recording the centre and scale of the BPST family.

Proof. Construct the map $Φ : R^{4} \times R_{> 0} \to M_{1} (S^{4})$ by $Φ (x_{0}, λ) = [A^{(x_{0}, λ)}]$ , where $A_{μ}^{(x_{0}, λ)} (x) = \frac{( x - x _{0} ) ^{ν}}{∣ x - x _{0} ∣ ^{2} + λ ^{2}} \overset{η}{ˉ}_{μν}^{a} T_{a}$ is the BPST connection (see Proposition "BPST anti-self-duality" in 03.07.07).

Smoothness of $Φ$ . The BPST formula is jointly smooth in $(x, x_{0}, λ)$ for $λ > 0$ , so $A^{(x_{0}, λ)}$ depends smoothly on $(x_{0}, λ)$ as a connection on $S^{4}$ (after stereographic compactification and gauge-extension across $\infty$ ). Smoothness of the projection to the gauge quotient follows from the local-slice theorem at irreducible connections.

Injectivity of $Φ$ . Suppose $A^{(x_{0}, λ)}$ and $A^{(x_{0}^{'}, λ^{'})}$ are gauge-equivalent. The pointwise curvature density of $A^{(x_{0}, λ)}$ is $$ |F_{A^{(x_0, \lambda)}}|^2(x) = \frac{96 \lambda^4}{(|x - x_0|^2 + \lambda^2)^4}, $$ a gauge-invariant scalar function on $S^{4}$ . This function attains its maximum at $x = x_{0}$ and the maximum value is $96/ λ^{4}$ , determining both the centre and the scale uniquely. Hence $(x_{0}, λ) = (x_{0}^{'}, λ^{'})$ and $Φ$ is injective.

Surjectivity of $Φ$ . By the AHDM construction 03.07.07, every ASD $SU (2)$ -connection on $S^{4}$ with $c_{2} = 1$ is gauge-equivalent to a BPST connection for some $(x_{0}, λ)$ . (The ADHM data at $k = 1$ reduces to $B_{1}, B_{2} \in C$ , $I \in C^{1 \times 2}$ , $J \in C^{2 \times 1}$ with $∣ I ∣^{2} - ∣ J ∣^{2} = 0$ and $I J = 0$ , modulo $U (1)$ ; solving these constraints recovers $R^{4} \times R_{> 0}$ .)

Diffeomorphism. Both source and target are smooth manifolds of the same dimension (five, by AHS). A smooth bijection between manifolds of equal dimension is a diffeomorphism iff its differential is everywhere injective. By the index formula and irreducibility of every BPST connection, $H^{0} (D_{A}) = H^{2} (D_{A}) = 0$ at every $(x_{0}, λ)$ , so the differential $D Φ$ identifies tangent spaces $T_{(x_{0}, λ)} (R^{4} \times R_{> 0}) = R^{5}$ and $H^{1} (D_{A^{(x_{0}, λ)}}) = T_{[A]} M_{1} (S^{4}) = R^{5}$ . Injectivity of $D Φ$ at every point follows from the explicit zero-mode analysis (the BPST family carries five linearly independent zero modes: four translation modes $\partial A / \partial x_{0}^{i}$ and one dilation mode $\partial A / \partial λ$ ). Hence $Φ$ is a diffeomorphism. $□$

Proposition ( $M_{k} (S^{4})$ is non-compact and the closure boundary is bubbling). For every $k \geq 1$ , the moduli space $M_{k} (S^{4})$ is non-compact. Its Uhlenbeck compactification $\overline{M}_{k} (S^{4})$ has top stratum $M_{k} (S^{4})$ and boundary strata $M_{k - l} (S^{4}) \times Sym^{l} (S^{4})$ for $1 \leq l \leq k$ , with $\operatorname{Sym}^0(S^4) = {} $an d$ \mathcal{M}_0(S^4) = {} $g i v in g t h e d ee p es t s t r a t u m$ \operatorname{Sym}^k(S^4)$ for the totally-bubbled configurations.

Proof. Non-compactness at $k = 1$ is exhibited by the sequence $A^{(0, 1/ n)}$ as $n \to \infty$ : the curvature density $96 (1/ n)^{4} / (∣ x ∣^{2} + 1/ n^{2})^{4}$ converges pointwise to zero off the origin and to infinity at the origin, with bounded total integral $16 π^{2}$ . The sequence has no convergent subsequence in $M_{1} (S^{4})$ because every limit candidate would have to be a smooth connection with charge one, but the limit measure $16 π^{2} δ_{0}$ is supported at a single point and corresponds to no smooth ASD connection.

For general $k$ , the analogous sequences arise by shrinking the scales of one or more component instantons within a multi-instanton configuration. Uhlenbeck's compactness theorem identifies precisely the limit configurations: a Cauchy-like criterion in the weak $L_{1}^{2}$ topology, modulo gauge and modulo a finite set of bubbling points. The limit datum is an ASD connection $A_{\infty}$ on a bundle of lower charge $k^{'} = k - \sum m_{i}$ together with a finite set of points $(x_{1}, m_{1}), \dots, (x_{l}, m_{l})$ with $m_{i} \geq 1$ and $\sum m_{i} = k - k^{'}$ . Recording this datum gives the Uhlenbeck compactification stratification $$ \overline{\mathcal{M}}k(S^4) = \bigsqcup{k' + |\mathbf{m}| = k} \mathcal{M}_{k'}(S^4) \times \operatorname{Sym}^{|\mathbf{m}|}(S^4), $$ where $Sym^{∣ m ∣} (S^{4})$ records the multi-set of bubbling points (with multiplicity if multiple units of charge bubble at the same point).

On $S^{4}$ , $M_{0} (S^{4})$ is a single point (the flat connection on the product bundle, unique up to gauge), so the deepest stratum at $k = l$ is just $Sym^{k} (S^{4})$ , an unordered $k$ -tuple of bubbling points. The full compactified moduli is the open manifold $M_{k} (S^{4})$ with these closed bubbling strata glued in along their natural collar neighbourhoods. $□$

Proposition (Donaldson polynomial well-defined). On a compact oriented simply-connected four-manifold $M$ with $b^{+} (M) \geq 1$ odd and $c_{2} (P) = k$ chosen so that $d_{k} = 8 k - 3 (1 + b^{+} (M))$ is non-negative and even, the Donaldson polynomial $$ q_{M, k} \in \mathrm{Sym}^{d_k/2} H_2(M; \mathbb{Z})^* $$ is well-defined, independent of generic metric, and an invariant of the underlying smooth structure on $M$ .

Proof sketch. The polynomial $q_{M, k}$ is constructed as a pairing of intersection-theory classes in $\overline{M}_{k} (M)$ via the $μ$ -map $μ : H_{2} (M) \to H^{2} (M_{k}^{*})$ , which is the slant product against the second Chern class of the universal bundle $E \to M_{k}^{*} \times M$ . Pairing $d_{k} /2$ copies of $μ (Σ_{i})$ over the fundamental class of $\overline{M}_{k} (M)$ gives the integer $$ q_{M,k}(\Sigma_1 \cdots \Sigma_{d_k/2}) = \int_{[\overline{\mathcal{M}}k(M)]} \mu(\Sigma_1) \cup \cdots \cup \mu(\Sigma{d_k/2}). $$ Well-definedness requires: (a) orientability of $M_{k} (M)$ , established by the index-bundle determinant-line argument; (b) the Uhlenbeck compactification has a fundamental class in homology of the correct degree, established by analytical-stratification arguments in Donaldson-Kronheimer §7; (c) the integral is independent of generic metric, established by a cobordism argument: two generic metrics are connected by a path through the metric space, and the parametric moduli space gives a cobordism between the two evaluation results.

The smooth-structure invariance follows because the construction depends only on the choice of orientation of $M$ and a homology orientation; both are determined by the smooth structure on $M$ . Donaldson 1983 used this construction to distinguish smooth structures: for a connected sum $M_{1} # M_{2}$ of two simply-connected smooth four-manifolds with positive intersection forms, the Donaldson polynomial vanishes by a stretching-the-neck argument, while for many manifolds with the same intersection form the polynomial is nonzero — giving examples of homeomorphic but non-diffeomorphic four-manifolds. $□$

Connections [Master]

BPST instanton and the Bogomolny bound 03.07.07. The BPST one-instanton is the explicit five-parameter family $R^{4} \times R_{> 0}$ realising $M_{1} (S^{4})$ . The Bogomolny bound exhibits ASD connections as absolute minimisers of the Yang-Mills action in each topological class; the present unit assembles the gauge-equivalence classes of those minimisers into a finite-dimensional smooth manifold at each charge $k$ . The Synthesis paragraph of the prereq unit already announced $dim M_{k} = 8 k - 3$ ; this unit promotes that announcement to a proof via the deformation complex and the Atiyah-Singer index theorem.
Yang-Mills action 03.07.05. The moduli space $M_{k}$ is by construction the quotient by gauge of the absolute minimiser set of the Yang-Mills action functional restricted to topological class $c_{2} = k$ . The action takes the constant value $8 π^{2} k$ on $M_{k} (S^{4})$ by the Bogomolny saturation, so the moduli space is a level set of the Yang-Mills function modulo gauge equivalence.
Atiyah-Singer index theorem 03.09.10. The dimension formula $dim M_{k} (M) = 8 k - 3 (1 + b^{+} (M))$ is the negative Atiyah-Singer index of the deformation complex $D_{A}$ applied to an ASD connection $A$ . The index theorem computes this as a Chern-character / signature integral on $M$ ; specialising to $SU (2)$ on $S^{4}$ yields the clean answer $8 k - 3$ . The same index machinery computes the dimension of moduli spaces of all elliptic-equation solutions in differential geometry, of which ASD moduli is a paradigmatic case.
Principal-bundle connection 03.05.07. The objects parameterised in $M_{k}$ are connections on a principal $SU (2)$ -bundle modulo the bundle's gauge group. The smooth structure on the connection space and the action of the gauge group rest on the foundational picture established in the prereq unit; the moduli space is the local-slice quotient.
Curvature of a connection 03.05.09. The defining equation $F_{A}^{+} = 0$ of the moduli space is a first-order PDE on the curvature. The Hodge-star eigenspace decomposition $Ω^{2} = Ω_{+}^{2} \oplus Ω_{-}^{2}$ on a four-manifold makes this equation algebraically meaningful; the prereq unit on curvature supplies the linear-algebra framework on which the present unit builds.
Pontryagin and Chern classes 03.06.04. The topological label $k = c_{2} (P) \in Z$ on $M_{k} (S^{4})$ is the second Chern class of the principal $SU (2)$ -bundle. The integrality of $k$ is the foundational reason the moduli space comes in a discrete family indexed by $Z_{\geq 0}$ , and the dimension formula $8 k - 3$ is a linear function of this characteristic number — the index theorem's expression of dimension as a Chern-character integral.
Chern-Weil homomorphism 03.06.06. The de Rham representative $\frac{1}{8 π ^{2}} tr (F \land F)$ of $c_{2} (P)$ is what makes the topological charge in $M_{k}$ computable from any single connection in the family. Chern-Weil is invoked both in the AHS index computation (via $p_{1} (ad P) = 4 c_{2} (P)$ ) and in the topological-charge labelling of the moduli space.
Classifying space $B G$ 03.08.04. Principal $SU (2)$ -bundles over $S^{4}$ are classified by $π_{3} (SU (2)) = Z$ via the clutching construction, so $M_{k} (S^{4})$ for distinct $k$ live on isomorphism-distinct bundles. The moduli space $M_{k}$ is a connected component (when non-empty) of the full space $A / G$ of gauge-equivalence classes of all connections; the components are indexed by the homotopy class in $π_{3} (SU (2))$ .
Penrose twistor space and the Ward correspondence 03.07.11. The moduli space $M_{k} (S^{4})$ has a parallel twistor-side incarnation: under the Ward correspondence, ASD bundles on $S^{4}$ correspond to holomorphic rank-two bundles on $CP^{3}$ plain on twistor lines with $c_{2} = k$ , and the moduli of such holomorphic bundles is a complex-algebraic variety whose complex-analytic dimension agrees with the gauge-theoretic real dimension $8 k - 3$ after accounting for the framing fixed at infinity. The dimension agreement is not a coincidence: the Atiyah-Singer index calculation of the elliptic deformation complex on the gauge side translates through the Penrose fibration into a Riemann-Roch computation on $CP^{3}$ for the moduli of holomorphic bundles, and both compute the same integer. The Uhlenbeck-bubbling stratification on the gauge side corresponds on the twistor side to the jumping-locus stratification, where the restriction $E ∣_{τ^{- 1} (x)}$ ceases to be the plain bundle on $CP^{1}$ and acquires a jumping Grothendieck splitting $⨁_{i} O (a_{i})$ with some $a_{i} \neq = 0$ . Connection type: parallel computation — the moduli-space dimension is the analytic index on the gauge side and the Riemann-Roch dimension on the twistor side, with the Penrose fibration as the bridge.

Historical & philosophical context [Master]

The dimension formula $dim M_{k} (M) = 8 k - 3 (1 + b^{+} (M))$ was established by Michael Atiyah, Nigel Hitchin, and Isadore Singer in their 1978 Proceedings of the Royal Society paper "Self-duality in four-dimensional Riemannian geometry" ^{[Atiyah-Hitchin-Singer 1978]}, a thirty-six page article that constructed the deformation complex $D_{A}$ of an ASD connection, computed its index via Atiyah-Singer, and identified the cohomology groups $H^{0}, H^{1}, H^{2}$ with stabiliser, tangent space, and obstruction respectively. The same year, Atiyah, Hitchin, Drinfeld, and Manin published their four-page note "Construction of instantons" in Physics Letters A ^{[Atiyah-Hitchin-Drinfeld-Manin 1978]}, reducing the classification of $SU (2)$ -instantons on $S^{4}$ to a finite-dimensional matrix problem (the ADHM construction). Together these two papers gave both the dimension count and an explicit parametrisation of $M_{k} (S^{4})$ for every $k$ .

The analytic foundations of the moduli theory were supplied by Karen Uhlenbeck in two 1982 papers in Communications in Mathematical Physics: "Removable singularities in Yang-Mills fields" ^{[Uhlenbeck 1982a]} and "Connections with $L^{p}$ bounds on curvature" ^{[Uhlenbeck 1982b]}. The first establishes that a finite-action ASD connection on a punctured neighbourhood extends smoothly across the puncture after gauge transformation; the second establishes the compactness theorem giving the bubbling description of the closure of $M_{k}$ . Without Uhlenbeck's theorems, the moduli space would be analytically intractable.

Simon Donaldson's 1983 Journal of Differential Geometry paper "An application of gauge theory to four-dimensional topology" ^{[Donaldson 1983]} used the moduli space $M_{k} (M)$ on a compact oriented simply-connected smooth four-manifold $M$ to construct invariants — the Donaldson polynomial $q_{M, k}$ — that distinguish smooth structures within a fixed homeomorphism class. Combined with Freedman's topological classification of simply-connected four-manifolds (also 1982), Donaldson's polynomials produced the first examples of compact simply-connected four-manifolds with the same intersection form (hence homeomorphic) but distinct smooth structures. Donaldson received the 1986 Fields Medal in part for this work.

The mathematical reach of the moduli-space machinery extended further in subsequent decades: Donaldson-Kronheimer's 1990 monograph The Geometry of Four-Manifolds ^{[Donaldson-Kronheimer 1990]} gave the canonical book-length treatment; Freed-Uhlenbeck's Instantons and Four-Manifolds ^{[Freed-Uhlenbeck 1991]} supplied a parallel pedagogical account; and the introduction of Seiberg-Witten invariants in 1994 simplified many of the analytic difficulties while preserving the topological content. The ASD moduli space remains the prototypical example of a finite-dimensional geometric quotient arising from an infinite-dimensional gauge-theory problem, and its dimension formula remains the cleanest application of the Atiyah-Singer index theorem to non-linear differential geometry.

Bibliography [Master]

@article{AtiyahHitchinSinger1978,
  author    = {Atiyah, M. F. and Hitchin, N. J. and Singer, I. M.},
  title     = {Self-duality in four-dimensional {R}iemannian geometry},
  journal   = {Proceedings of the Royal Society of London. Series A, Mathematical and Physical Sciences},
  volume    = {362},
  number    = {1711},
  pages     = {425--461},
  year      = {1978}
}

@article{AHDM1978,
  author    = {Atiyah, M. F. and Hitchin, N. J. and Drinfeld, V. G. and Manin, Yu. I.},
  title     = {Construction of instantons},
  journal   = {Physics Letters A},
  volume    = {65},
  pages     = {185--187},
  year      = {1978}
}

@article{Uhlenbeck1982a,
  author    = {Uhlenbeck, Karen K.},
  title     = {Removable singularities in {Y}ang-{M}ills fields},
  journal   = {Communications in Mathematical Physics},
  volume    = {83},
  pages     = {11--29},
  year      = {1982}
}

@article{Uhlenbeck1982b,
  author    = {Uhlenbeck, Karen K.},
  title     = {Connections with {$L^p$} bounds on curvature},
  journal   = {Communications in Mathematical Physics},
  volume    = {83},
  pages     = {31--42},
  year      = {1982}
}

@article{Donaldson1983,
  author    = {Donaldson, S. K.},
  title     = {An application of gauge theory to four-dimensional topology},
  journal   = {Journal of Differential Geometry},
  volume    = {18},
  number    = {2},
  pages     = {279--315},
  year      = {1983}
}

@book{DonaldsonKronheimer1990,
  author    = {Donaldson, S. K. and Kronheimer, P. B.},
  title     = {The Geometry of Four-Manifolds},
  publisher = {Oxford University Press},
  year      = {1990}
}

@book{FreedUhlenbeck1991,
  author    = {Freed, Daniel S. and Uhlenbeck, Karen K.},
  title     = {Instantons and Four-Manifolds},
  edition   = {2nd},
  series    = {MSRI Publications},
  volume    = {1},
  publisher = {Springer-Verlag},
  year      = {1991}
}

@book{AtiyahPisa1979,
  author    = {Atiyah, Michael F.},
  title     = {Geometry of {Y}ang-{M}ills Fields},
  series    = {Lezioni Fermiane},
  publisher = {Accademia Nazionale dei Lincei / Scuola Normale Superiore, Pisa},
  year      = {1979}
}

@article{BelavinPolyakovSchwartzTyupkin1975,
  author    = {Belavin, A. A. and Polyakov, A. M. and Schwartz, A. S. and Tyupkin, Yu. S.},
  title     = {Pseudoparticle solutions of the {Y}ang-{M}ills equations},
  journal   = {Physics Letters B},
  volume    = {59},
  pages     = {85--87},
  year      = {1975}
}

@article{Freedman1982,
  author    = {Freedman, Michael Hartley},
  title     = {The topology of four-dimensional manifolds},
  journal   = {Journal of Differential Geometry},
  volume    = {17},
  number    = {3},
  pages     = {357--453},
  year      = {1982}
}

P1 unit #2 from the Atiyah Yang-Mills audit (Cycle 6). Direct sequel to 03.07.07 (BPST instanton); assembles the BPST five-parameter family into the general moduli space $M_{k} (S^{4})$ of charge- $k$ instantons and proves the AHS dimension formula $dim M_{k} (S^{4}) = 8 k - 3$ .

Prerequisites

03.07.07
03.07.05
03.05.07
03.05.09
03.06.04
03.06.06
03.09.10

Tier anchors

beginner: Atiyah *Geometry of Yang-Mills Fields* Ch. 2-3 (level pacing); 3Blue1Brown / Strogatz visual family-of-solutions account
intermediate: Atiyah-Hitchin-Singer 1978 *Proc. R. Soc. A* 362; Donaldson-Kronheimer §4; Freed-Uhlenbeck §3-4
master: Atiyah-Hitchin-Singer 1978 *Proc. R. Soc. A* 362, 425-461; Donaldson-Kronheimer §4-7; Freed-Uhlenbeck §3-8; Donaldson 1983 *J. Diff. Geom.* 18, 279-315; Uhlenbeck 1982a/b *Comm. Math. Phys.* 83, 84

References

tong
md/pages/gaugetheory.md · Gauge Theory lectures, §2.4 moduli of instantons
TODO_REF
Atiyah-Hitchin-Singer — Self-duality in four-dimensional Riemannian geometry · *Proc. R. Soc. London A* 362 (1978), 425-461, originator paper for the deformation complex and dimension formula
TODO_REF
Donaldson — An application of gauge theory to four-dimensional topology · *Journal of Differential Geometry* 18 (1983), 279-315, polynomial invariants and exotic smooth structures
TODO_REF
Uhlenbeck — Removable singularities in Yang-Mills fields · *Communications in Mathematical Physics* 83 (1982), 11-29
TODO_REF
Uhlenbeck — Connections with $L^p$ bounds on curvature · *Communications in Mathematical Physics* 83 (1982), 31-42; the compactness theorem underlying ideal-instanton compactification
TODO_REF
Atiyah — Geometry of Yang-Mills Fields · Scuola Normale Superiore (Pisa), 1979, Ch. 2-3 moduli space description
TODO_REF
Donaldson-Kronheimer — The Geometry of Four-Manifolds · Oxford 1990, §4 the moduli space, §6 ADHM, §7 Uhlenbeck compactification
TODO_REF
Freed-Uhlenbeck — Instantons and Four-Manifolds · MSRI Pub. 1, Springer 1991, §3-§8 analytic structure of the moduli space
TODO_REF
Atiyah-Hitchin-Drinfeld-Manin — Construction of instantons · *Physics Letters A* 65 (1978), 185-187, ADHM parametrisation of $\mathcal{M}_k(S^4)$

Reviewer

TBD

Estimated time

beginner: 18m
intermediate: 45m
master: 90m