9 · yang-mills theory, instantons, and twistors

gauge fields as connections on principal bundles

Let $G$ be a compact Lie group with Lie algebra $\mathfrak{g}$. A classical gauge field is a connection $A$ on a principal $G$-bundle $P\to M$. In a local trivialization the connection is represented by a $\mathfrak{g}$-valued one-form, and its curvature $$ F_A=dA+A\wedge A $$ is a $\mathfrak{g}$-valued two-form, the nonabelian field strength. Under a gauge transformation $g:M\to G$ one has $$ A\mapsto g^{-1}Ag+g^{-1}dg,\qquad F_A\mapsto g^{-1}F_Ag. $$ The Bianchi identity $D_A F_A=0$ is the nonabelian analog of the homogeneous Maxwell equation and is an automatic consequence of the structure equation.

definition (principal connection and curvature). a connection one-form $\omega\in\Omega^1(P,\mathfrak{g})$ is equivariant and reproduces fundamental vector fields (chapter 2). a local gauge potential is $A=s^*\omega$ for a local section $s$. the curvature is $\Omega=d\omega+\frac12[\omega,\omega]$, written locally $F_A=dA+A\wedge A$.

theorem (gauge transformation law). if $s'=s\cdot g$ for a smooth map $g:U\to G$, then $$ A'=g^{-1}Ag+g^{-1}dg,\qquad F_{A'}=g^{-1}F_A g. $$

proof. equivariance of $\omega$ under the right $G$-action yields $(s\cdot g)^*\omega=\operatorname{Ad}_{g^{-1}}(s^*\omega)+g^{-1}dg$ on matrix groups. exterior differentiation and the Maurer-Cartan structure equation convert this into the adjoint action on $F_A$. $\square$

theorem (bianchi identity). for any connection, $$ D_A F_A:=dF_A+[A,F_A]=0. $$

proof. differentiate $F=dA+A\wedge A$ to obtain $dF=dA\wedge A-A\wedge dA$ (graded signs). substitute $dA=F-A\wedge A$ and use the Jacobi identity for the Lie bracket; all cubic terms cancel, leaving $dF+[A,F]=0$. $\square$

remark (abelian reduction). if $G=\mathrm{U}(1)$ then $A\wedge A=0$, so $F=dA$ and $D_A F=dF=0$, recovering Maxwell's homogeneous equation as pure geometry (chapter 4).

the yang-mills lagrangian and instantons

The Yang-Mills action on a Riemannian four-manifold is the $L^2$-norm of the curvature: $$ S_{\mathrm{YM}}[A]=\frac12\int_M\|F_A\|^2\,d\mathrm{vol}. $$ By the Chern-Weil theorem the second Chern number (instanton number) $$ k=-\frac{1}{8\pi^2}\int_M\operatorname{Tr}(F_A\wedge F_A)\in\mathbb{Z} $$ is a topological invariant (sign fixed so that ASD connections have $k\ge 0$ or $k\le 0$ consistently with the conventions of chapter 5). The pointwise identity relating $\|F\|^2$ to the Chern-Weil density implies the topological lower bound $$ \int_M\|F_A\|^2\,d\mathrm{vol}\ge 8\pi^2|k|, $$ with equality if and only if $F_A=\pm{*F_A}$. Solutions of the anti-self-dual (ASD) equation $F_A=-{*F_A}$ (or the self-dual equation) are called instantons. They are absolute minima of the Yang-Mills energy within each topological sector and represent finite-action tunneling configurations between distinct vacua of the classical theory.

definition (yang-mills action and equation). $$ S_{\mathrm{YM}}[A] =\frac12\int_M\|F_A\|^2\,d\mathrm{vol} =-\frac12\int_M\operatorname{Tr}(F_A\wedge{*F_A}). $$ critical points satisfy $D_A^* F_A=0$, the Yang-Mills equation, together with Bianchi $D_A F_A=0$.

definition (instanton number). for a principal $G$-bundle on a closed oriented four-manifold with $G\subset\mathrm{U}(N)$ (or $\mathrm{SU}(N)$), $$ k =-\frac{1}{8\pi^2}\int_M\operatorname{Tr}(F_A\wedge F_A) $$ equals the second Chern number $c_2$ of the associated complex vector bundle (sign convention fixed once and for all so that $k\in\mathbb{Z}$ matches chapter 5).

theorem (self-duality implies yang-mills). if $\dim M=4$ and $*F_A=\pm F_A$, then $D_A^* F_A=0$ automatically.

proof. on an oriented Riemannian four-manifold, $*:\Lambda^2\to\Lambda^2$ satisfies $*^2=\mathrm{id}$ on two-forms. for $\mathfrak{g}$-valued two-forms one has $D_A^*\alpha=-*D_A{*\alpha}$ up to the standard overall sign fixed by Hodge dual conventions. if $*F=\pm F$ and $D_A F=0$ (Bianchi), then $D_A^* F=\mp *D_A F=0$. $\square$

theorem (topological energy bound). for every connection $A$ on a closed oriented Riemannian four-manifold, $$ \int_M\|F_A\|^2\,d\mathrm{vol} \ge 8\pi^2|k|, $$ with equality if and only if $F_A=*F_A$ or $F_A=-*F_A$ according to the sign of $k$.

proof. decompose $F=F^++F^-$ into self-dual and anti-self-dual parts, the $\pm 1$ eigenspaces of $*$ on $\Lambda^2$. then $$ \|F\|^2=\|F^+\|^2+\|F^-\|^2, $$ while $$ \operatorname{Tr}(F\wedge F) =\bigl(\|F^+\|^2-\|F^-\|^2\bigr)\,d\mathrm{vol} $$ up to the identification of the pointwise inner product with the exterior product against the volume form. hence $$ 0\le\|F\mp *F\|^2 =2\|F\|^2\mp 2\operatorname{Tr}(F\wedge F)/\mathrm{vol}, $$ and integrating with the Chern-Weil formula for $k$ produces the bound, with equality precisely when $F^\mp=0$. $\square$

remark (vacua and tunneling). flat connections on $S^3$ (or $\mathbb{R}^3$ with decay) are classified by $\pi_3(G)$; an instanton interpolates between such vacua as a finite-action path in temporal gauge, the classical precursor of theta vacua and tunneling amplitudes.

penrose twistor space and ward's theorem

On the conformal compactification $S^4$ of Euclidean four-space the ASD equation becomes conformally invariant. Penrose's twistor correspondence identifies the space of oriented lines (complex structures on tangent spaces) in $S^4$ with the complex projective space $\mathbb{CP}^3$. Under this identification an ASD connection on a bundle $E\to S^4$ corresponds to a holomorphic vector bundle $\mathcal{E}\to\mathbb{CP}^3$ that is holomorphically trivial on every real projective line (the twistor lines corresponding to points of $S^4$). This is Ward's theorem: the nonlinear PDE of anti-self-duality is transformed into the linear problem of classifying holomorphic bundles with prescribed trivializations on a family of rational curves.

definition (twistor space of $S^4$). the twistor space of Euclidean $S^4$ is $\mathbb{CP}^3$, fibring over $S^4$ with fiber $\mathbb{CP}^1$ parametrizing orthogonal complex structures $J$ with $J^2=-\mathrm{id}$ on each $T_x S^4$ compatible with the conformal metric. real structures on $\mathbb{CP}^3$ encode Euclidean reality.

definition (twistor line). each $x\in S^4$ determines a projective line $L_x\simeq\mathbb{CP}^1\subset\mathbb{CP}^3$ consisting of all complex structures at $x$. these are the real twistor lines.

theorem (ward correspondence). there is a natural bijection between: (i) gauge-equivalence classes of ASD $\mathrm{GL}(N,\mathbb{C})$-connections on open sets of $S^4$ (or $\mathbb{R}^4$); and (ii) holomorphic rank-$N$ vector bundles on the corresponding open sets of twistor space that are holomorphically trivial on every twistor line $L_x$ in that set, up to holomorphic isomorphism. real structures compatible with $\mathrm{SU}(N)$ or $\mathrm{U}(N)$ select the corresponding real gauge groups.

proof. given an ASD connection $\nabla$ on $E\to U\subset S^4$, form the pullback bundle $\pi^*E$ on the twistor space $Z_U=\pi^{-1}(U)\subset\mathbb{CP}^3$. at each twistor point $z\in L_x$ the almost-complex structure $J_z$ determines a splitting $T_\mathbb{C}^*U=\Lambda^{1,0}\oplus\Lambda^{0,1}$. anti-self-duality of $F$ is equivalent to $F^{0,2}=0$ for every such complex structure (the self-dual two-forms are of type $(1,1)$ for every compatible $J$). thus the $(0,1)$ part of $\nabla$ defines a holomorphic structure $\bar\partial_{\mathcal{E}}$ on $\pi^*E$. restriction to $L_x$ is the restriction of a connection on a point (no directions tangent to $U$), so $\mathcal{E}|_{L_x}$ is holomorphically trivial.

conversely, given $\mathcal{E}\to Z_U$ holomorphic and trivial on each $L_x$, the fiber $E_x:=H^0(L_x,\mathcal{E}|_{L_x})$ defines a smooth bundle $E\to U$. differentiating the trivializations along directions in $U$ produces a connection whose $(0,2)$ curvature vanishes for every compatible complex structure, hence $F$ is anti-self-dual. functoriality under isomorphism and gauge account for equivalence. $\square$

remark (conformal invariance). $*:\Lambda^2\to\Lambda^2$ is conformally invariant in four dimensions, so the ASD equation and Ward's correspondence depend only on the conformal class of the metric.

the horrocks construction and linear fields

Explicit holomorphic bundles on $\mathbb{CP}^3$ may be constructed by the monad method of Horrocks: one exhibits a complex of vector bundles $$ 0\to\mathcal{O}(-1)^{\oplus k} \to\mathcal{O}^{\oplus(2k+N)} \to\mathcal{O}(1)^{\oplus k} \to 0 $$ whose middle cohomology sheaf is the desired bundle of rank $N$ and second Chern class $k$. Linear field equations in an instanton background (the massless Dirac equation and the self-dual Maxwell equation) are solved by evaluating sheaf cohomology groups of appropriate twists of $\mathcal{E}$ over twistor space. Barth's theorems guarantee that every bundle arising from an $\mathrm{SU}(N)$ instanton of charge $k$ arises in this way (in the appropriate stable range / cohomological vanishing regime).

definition (monad / horrocks display). a monad on $\mathbb{CP}^3$ is a complex $$ 0\to A\xrightarrow{\alpha} B\xrightarrow{\beta} C\to 0 $$ of holomorphic vector bundles with $\alpha$ injective, $\beta$ surjective, and $\beta\circ\alpha=0$, such that the cohomology $E=\ker\beta/\operatorname{im}\alpha$ is a holomorphic vector bundle. the Horrocks form above takes $A=\mathcal{O}(-1)^{\oplus k}$, $C=\mathcal{O}(1)^{\oplus k}$, and $B$ trivial of rank $2k+N$.

theorem (cher classes from monads). for a Horrocks monad of the displayed type, the middle cohomology bundle $\mathcal{E}$ has $\operatorname{rank}\mathcal{E}=N$ and $c_2(\mathcal{E})=k$ (with $c_1=0$ for the $\mathrm{SU}(N)$ case after the usual determinant reduction).

proof. in $K$-theory of $\mathbb{CP}^3$, $[\mathcal{E}]=[B]-[A]-[C]$. taking Chern characters, $$ \operatorname{ch}(\mathcal{E}) =(2k+N)-\,k\,\operatorname{ch}(\mathcal{O}(-1)) -k\,\operatorname{ch}(\mathcal{O}(1)). $$ with $\operatorname{ch}(\mathcal{O}(m))=e^{mh}$ and $h=c_1(\mathcal{O}(1))$, the expansion yields rank $N$, vanishing first Chern class after matching, and second Chern character component producing $c_2=k$. $\square$

theorem (linear fields via cohomology). massless solutions of spin $s$ in an ASD background $A$ on $\mathbb{R}^4$ or $S^4$ are in natural bijection with cohomology groups $H^1(\mathbb{CP}^3,\mathcal{E}(m))$ (or duals $H^2$) for twists $m$ determined by $s$ and the Penrose-Ward transform of the field equation.

proof (idea). the Penrose transform identifies sheaf cohomology on twistor space with kernels of massless field operators on spacetime. Ward's holomorphic bundle $\mathcal{E}$ encodes the ASD background; twisting by $\mathcal{O}(m)$ adjusts spin weight and conformal weight. a representative cocycle, restricted to twistor lines and differentiated, produces a spacetime spinor field annihilated by the twistor-transformed Dirac or Maxwell operator in the background $A$. $\square$

the adhm construction

The Atiyah-Drinfeld-Hitchin-Manin construction reduces the entire problem to finite-dimensional linear algebra. One starts with complex vector spaces $V$ and $W$ of dimensions $k$ and $N$ respectively, together with linear maps $$ B_1,B_2\in\operatorname{End}(V),\qquad I\in\operatorname{Hom}(W,V),\qquad J\in\operatorname{Hom}(V,W). $$ These data are required to satisfy the real and complex moment-map (or ADHM) equations $$ \begin{aligned} \mu_r&=[B_1,B_1^\dagger]+[B_2,B_2^\dagger]+II^\dagger-J^\dagger J=0,\\ \mu_c&=[B_1,B_2]+IJ=0. \end{aligned} $$ From any solution one builds a Dirac-type operator $$ D_x^\dagger =\begin{pmatrix} B_1-x_1 & B_2-x_2 & I \\ -(B_2-x_2)^\dagger & (B_1-x_1)^\dagger & J^\dagger \end{pmatrix} $$ parametrized by $x\in\mathbb{H}\simeq\mathbb{R}^4$. The kernel of $D_x^\dagger$ is an $N$-dimensional subspace of $V\oplus V\oplus W$; orthogonal projection onto this kernel yields a connection $A$ on the trivial bundle of rank $N$ whose curvature is anti-self-dual and whose instanton number is $k$. Every instanton arises uniquely in this manner (up to the natural action of $U(k)$).

definition (adhm data). an ADHM datum of charge $k$ and rank $N$ is a quadruple $(B_1,B_2,I,J)$ as above satisfying $\mu_r=\mu_c=0$ and a nondegeneracy condition ensuring that $D_x^\dagger$ has full rank $2k$ for every $x\in\mathbb{R}^4$ (so $\dim\ker D_x^\dagger=N$).

definition (adhm connection). let $U(x):\mathbb{C}^N\xrightarrow{\sim}\ker D_x^\dagger \subset\mathbb{C}^{2k+N}$ be a smooth unitary frame of the kernel. the ADHM connection is the compression of the trivial connection: $$ A=U^\dagger dU, $$ a $\mathfrak{u}(N)$-valued one-form on $\mathbb{R}^4$.

theorem (adhm produces instantons). for nondegenerate ADHM data of charge $k$, the connection $A=U^\dagger dU$ is an ASD connection of instanton number $k$ on the trivial bundle of rank $N$ over $\mathbb{R}^4$ (completing smoothly over $S^4$ after compactification for $\mathrm{SU}(N)$ data). gauge-equivalent connections arise from the residual $U(k)$ action $$ (B_i,I,J)\mapsto (gB_ig^{-1},gI,Jg^{-1}),\qquad g\in U(k). $$

proof. write the trivial connection on the ambient bundle $\mathbb{C}^{2k+N}$ and project orthogonally onto $\ker D_x^\dagger$. the curvature of a projected connection is $F=P\,dP\wedge dP\,P$ (Grassmannian formula). the ADHM moment-map equations are precisely the condition that the $(0,2)$ part of this curvature vanishes for every complex structure in the hyperkahler family on $\mathbb{R}^4$ (equivalently, that $F$ is anti-self-dual). the topological charge equals $\dim V=k$ by the asymptotic behavior of $U(x)$ as $|x|\to\infty$, matching Chern-Weil. the $U(k)$ action conjugates $D_x^\dagger$ without changing $\ker D_x^\dagger$ up to unitary rotation of the ambient space that fixes the projection $P$, hence yields the same $A$ up to gauge. $\square$

theorem (completeness of adhm). every ASD $\mathrm{SU}(N)$ connection of charge $k$ on $S^4$ (or finite-action ASD connection on $\mathbb{R}^4$ with suitable framing at infinity) arises from ADHM data unique up to $U(k)$.

proof (outline). by Ward, the connection determines a holomorphic bundle $\mathcal{E}$ on $\mathbb{CP}^3$ trivial on lines. Barth and others compute $H^1(\mathcal{E}(-1))=\mathbb{C}^k$ and vanishing of $H^0(\mathcal{E}(-1))$, producing a monad presentation of type $\mathcal{O}(-1)^k\to\mathcal{O}^{2k+N}\to\mathcal{O}(1)^k$. choosing bases and imposing the real structure from Euclidean conjugation yields matrices $(B_1,B_2,I,J)$ satisfying the ADHM equations. reconstruction of $A$ from the monad recovers the original connection. $\square$

The moduli space of irreducible $\mathrm{SU}(2)$ instantons of charge $k$ is a smooth manifold of dimension $8k-3$. This dimension coincides exactly with the prediction of the Atiyah-Singer index theorem applied to the deformation complex of the ASD equation, confirming the global consistency of the analytic, topological, and algebraic descriptions.

theorem (dimension of instanton moduli). the moduli space of irreducible framed $\mathrm{SU}(2)$ instantons of charge $k\ge 1$ on $S^4$ is a smooth manifold of dimension $8k-3$ (or $8k$ before quotienting by the residual $\mathrm{SU}(2)$ gauge action at infinity, depending on framing conventions).

proof. the deformation complex of the ASD equation is $$ 0\to\Omega^0(\operatorname{ad} P) \xrightarrow{D_A} \Omega^1(\operatorname{ad} P) \xrightarrow{D_A^+} \Omega^{2,+}(\operatorname{ad} P) \to 0, $$ where $D_A^+$ projects $D_A$ onto self-dual two-forms. ellipticity at irreducible connections implies that the tangent space to moduli is $H^1$ of this complex, and $\mathrm{index}=\dim H^1-\dim H^0-\dim H^2$. vanishing theorems ($H^0=0$ for irreducibility, $H^2=0$ by ASD Weitzenbock) reduce this to the analytical index. Atiyah-Singer yields $\mathrm{index}=8k$ for $\mathrm{SU}(2)$ charge $k$ (equivalently $4\cdot 2k$ from the real representation dimension count). quotienting by free $\mathrm{SU}(2)$ residual gauge (dimension $3$) for unframed moduli gives $8k-3$. ADHM parameters are $4k^2+4kN$ real entries for general $N$, with $3k^2$ real ADHM equations ($k^2$ for $\mu_c$ complex and $k^2$ for $\mu_r$) and $k^2$ gauge parameters in $U(k)$; for $N=2$ the count collapses to $8k$ framed degrees of freedom, matching. $\square$

Yang-Mills theory, viewed through the lens of twistor geometry and the ADHM construction, therefore realizes the full program of the treatise: local differential geometry of connections, global topological invariants supplied by characteristic classes and $K$-theory, spinorial and index-theoretic control of deformations, and an explicit algebraic solution of the resulting nonlinear field equations. The instanton moduli spaces that emerge are themselves rich geometric objects, linking classical field theory to the deepest structures of modern geometry and topology.

remark (closure). the program that began with fields as sections of bundles on a manifold ends with gauge fields as connections, topological charges as Chern numbers, fermionic and deformation indices as topology, and nonlinear ASD dynamics solved by matrices. what remains is the rich geometry of the moduli spaces themselves.

exercises

exercise 1 (self-dual split). for a real two-form $F$ on an oriented Euclidean four-manifold, prove $\|F\|^2=\|F^+\|^2+\|F^-\|^2$ and $F\wedge F=\bigl(\|F^+\|^2-\|F^-\|^2\bigr)\,d\mathrm{vol}$. deduce the energy bound of the text.

exercise 2 (bpst charge). the basic BPST instanton on $\mathbb{R}^4$ has $A_i=\frac{\sigma_i\bar x-x\sigma_i}{|x|^2+\rho^2}$ (up to identification of $\mathfrak{su}(2)$ with imaginary quaternions). verify that $k=1$ by evaluating $\int\operatorname{Tr}(F\wedge F)$ at infinity, or accept the asymptotic pure-gauge winding and conclude $k=1$ from $\pi_3(\mathrm{SU}(2))$.

exercise 3 (adhm dim count for $\mathrm{SU}(2)$). count real parameters in $(B_1,B_2,I,J)$ for $N=2$, subtract the real and complex ADHM equations and the $U(k)$ action, and recover the framed dimension $8k$.

exercise 4 (asd implies yang-mills). if $F_A=\pm *F_A$ on an oriented Riemannian four-manifold, deduce $d_A{}^*F_A=0$ from the Bianchi identity $d_AF_A=0$. conclude that anti-self-dual connections solve the Yang-Mills equations.