Orientations on instanton trajectory moduli

Intuition Beginner

Floer differentials count instanton trajectories. But they do not merely count how many trajectories exist. Each isolated trajectory gets a sign, plus or minus one, and the signed sum is the coefficient in the differential.

Orientations are the rule that assigns those signs consistently. A moduli space is like a curve or collection of points with a direction attached. When a one-dimensional family has boundary points, the signs of its boundary points add to zero. This is the geometric reason the Floer differential squares to zero.

Without coherent orientations, the theory could still count modulo two. Integer-valued instanton Floer homology needs the sign information.

Visual Beginner

An oriented interval has one positive end and one negative end. Floer theory uses the same sign principle for compactified trajectory spaces.

Worked example Beginner

Suppose there are three isolated trajectories from $\alpha$ to $\beta$. If all were counted without signs, the count would be three. With orientations, the signs might be $+1$, $+1$, and $-1$.

The signed count is then $$ (+1)+(+1)+(-1)=1. $$ This signed count is the coefficient of $\beta$ in the differential of $\alpha$.

What this tells us: signs can create cancellation. That cancellation is essential when boundary points of a one-dimensional moduli space come in oppositely oriented pairs.

Check your understanding Beginner

Formal definition Intermediate+

Let $\mathcal{M}(\alpha ,\beta )$ be a regular moduli space of ASD trajectories from a flat connection $\alpha$ to a flat connection $\beta$ on a closed oriented three-manifold $Y$. At a trajectory $A$, the gauge-fixed linearized ASD operator is a Fredholm operator $$ D_A: \mathcal X_A\longrightarrow \mathcal Y_A $$ between suitable weighted Sobolev spaces. Its determinant line is $$ \det(D_A)=\Lambda^{\max}\ker D_A\otimes\left(\Lambda^{\max}\operatorname{coker}D_A\right)^*. $$ If $D_A$ is surjective, then $\det (D_A)={\Lambda }^{\max }{T}_A\mathcal{M}(\alpha ,\beta )$, so orienting the determinant line orients the moduli space at $A$.

A coherent orientation system is a choice of orientation for all determinant lines of trajectory operators such that the gluing isomorphism $$ \det(D_{A_1})\otimes\det(D_{A_2})\otimes\mathbb R_{\mathrm{glue}} \cong \det(D_{A_1# A_2}) $$ preserves the prescribed orientations. The one-dimensional factor ${\mathbb{R}}_{\mathrm{glue}}$ is the neck-length direction from 03.07.21.

For an isolated regular trajectory, the determinant line is a one-dimensional real line after quotienting by translation and gauge. Its orientation determines a sign $$ \epsilon(A)\in{+1,-1}. $$ The signed count is $$ #\mathcal M^0(\alpha,\beta)=\sum_{[A]\in\mathcal M^0(\alpha,\beta)}\epsilon(A). $$

Counterexamples to common slips

• Orienting the base three-manifold is not enough. One must orient determinant lines of Fredholm operators.
• A choice for one moduli space is not enough. The choices must be coherent under gluing.
• Over $\mathbb{Z}/2$, signs disappear. Over $\mathbb{Z}$, the determinant-line orientation data are part of the definition.

Key theorem with proof Intermediate+

Theorem (coherent orientations give signed boundary cancellation). Suppose the regular instanton trajectory moduli spaces are equipped with coherent orientations. Then every compact oriented one-dimensional moduli space has boundary signed count zero, and its boundary signs agree with the product signs of the two broken trajectories appearing in the Floer composition.

Proof. A compact oriented one-dimensional smooth manifold is a finite union of circles and closed intervals. Circles have no boundary. Each closed interval has two boundary points, one positive and one negative according to the outward-normal-first convention. Therefore the total signed boundary count is zero.

For instanton moduli spaces, Uhlenbeck compactness 03.07.20 identifies boundary points of a one-dimensional compactified moduli space with broken pairs of trajectories $$ \alpha\to\gamma\to\beta. $$ The gluing theorem 03.07.21 identifies a neighborhood of such a boundary point with a half-open gluing parameter times the two lower-dimensional moduli spaces. Coherence of determinant-line orientations says the boundary orientation induced by the glued moduli space matches the product of the signs assigned to the two component trajectories, with the standard boundary sign.

Thus the algebraic sum over all broken boundary points is exactly the signed sum of products of trajectory counts. Since the total signed boundary count of the compact oriented one-manifold is zero, these products cancel in pairs globally. $\square$

Bridge. Orientation theory builds toward 03.07.23 because the Floer differential over $\mathbb{Z}$ uses signed counts of zero-dimensional trajectory moduli, and it appears again in gluing because determinant-line orientations must be compatible with boundary collars. The foundational reason is that signs turn geometric boundary cancellation into algebraic cancellation; this is exactly what identifies the boundary of a compact one-manifold with the coefficient of ${\partial }^2$. Putting these together, coherent orientations generalise the sign rule from finite-dimensional Morse theory, and the bridge is dual to the gluing collar in 03.07.21.

Exercises Intermediate+

Advanced results Master

The determinant line of a Fredholm operator is functorial under exact triples. Gluing produces an approximate exact sequence relating the kernels and cokernels of the two component operators, the gluing parameter, and the glued operator. Passing to determinant lines yields the gluing isomorphism used in coherent orientation theory.

For instanton Floer theory over an integer homology sphere, the relevant homology orientation is canonical because $H^1(Y;\mathbb{R})=0$ and the remaining orientation data have a standard choice. For more general three-manifolds or for four-dimensional cobordism maps, one must choose a homology orientation, usually an orientation of a real vector space built from $H^1$ and self-dual harmonic two-forms. Donaldson's orientation theorem identifies these finite-dimensional choices with orientations of the gauge-theoretic determinant lines.

Translation symmetry adds a small bookkeeping point. A parametrized trajectory has a one-dimensional direction generated by translating the cylinder coordinate. Passing to the unparametrized moduli space removes this direction. The sign convention must specify whether the translation direction is placed before or after the unparametrized orientation; this choice fixes the boundary sign in the gluing formula.

The determinant-line viewpoint also explains why mod-two instanton Floer homology is easier. If coefficients are reduced modulo two, every sign is identified with $1$, so the determinant-line orientation choices disappear. Integer coefficients retain more information and require the coherent orientation system.

Synthesis. The foundational reason orientations enter is that the Floer differential is a signed count of points in zero-dimensional moduli spaces. This is exactly what determinant lines provide for regular Fredholm zero sets. The central insight identifies gluing compatibility of determinant lines with boundary-orientation compatibility, while mod-two counting is dual to forgetting that sign data. Putting these together, coherent orientations supply the bridge from analytic moduli spaces to chain complexes over $\mathbb{Z}$.

Full proof set Master

Proposition 1 (determinant line of a surjective Fredholm operator). If $D:X\to Y$ is Fredholm and surjective, then an orientation of $\det (D)$ is the same as an orientation of $\ker D$.

Proof. By definition, $$ \det(D)=\Lambda^{\max}\ker D\otimes(\Lambda^{\max}\operatorname{coker}D)^*. $$ If $D$ is surjective, then $\mathrm{coker}D=0$. The top exterior power of the zero vector space is canonically $\mathbb{R}$ with its standard orientation. Therefore $\det (D)$ canonically identifies with ${\Lambda }^{\max }\ker D$, and orienting one is the same as orienting the other. $\square$

Proposition 2 (oriented boundary count of a compact one-manifold). The signed count of the boundary of a compact oriented one-dimensional manifold is zero.

Proof. Every compact one-dimensional manifold is a finite disjoint union of circles and closed intervals. Circles have no boundary. A closed interval has two endpoints; with the boundary orientation convention, one endpoint is positive and the other is negative. Hence each interval contributes $+1-1=0$, and the whole disjoint union has total boundary count zero. $\square$

Proposition 3 (product signs in a broken boundary). Under a coherent orientation system, the sign of a broken boundary point $\alpha \to \gamma \to \beta$ equals the boundary sign times the product of the signs of the two component trajectories.

Proof. The gluing theorem gives a collar near the broken point with coordinates consisting of the two component trajectories and a gluing parameter. The determinant-line gluing isomorphism identifies $$ \det(D_{A_1})\otimes\det(D_{A_2})\otimes\mathbb R_{\mathrm{glue}} $$ with the determinant line of the glued operator. Coherence means this isomorphism preserves the chosen orientations. The boundary orientation is obtained by placing the outward normal to the gluing parameter first. Removing that normal leaves exactly the product orientation on the two component determinant lines, up to the standard boundary sign. Thus the broken boundary sign is the boundary sign times the product of the two trajectory signs. $\square$

Connections Master

• Gluing theorem for instanton trajectories 03.07.21. Coherent orientations are defined through the determinant-line isomorphism induced by gluing, and gluing supplies the collar where boundary signs are read.

• Uhlenbeck compactness 03.07.20. Compactness identifies the broken trajectories that form boundaries; orientations assign the signs needed for their algebraic cancellation.

• Spectral flow and Floer grading 03.07.19. The same Fredholm operators whose determinant lines are oriented have indices determined by spectral flow, so grading and signs are two faces of one linearized operator.

• Instanton Floer homology 03.07.23. The integer-valued Floer differential uses signed counts of zero-dimensional trajectory moduli, and ${\partial }^2=0$ depends on coherent boundary signs.

Historical & philosophical context Master

Donaldson's 1987 orientation theorem established the determinant-line orientation framework for Yang-Mills moduli spaces and linked analytic orientations to finite-dimensional homology-orientation data ^{[Donaldson 1987]}. This made integer-valued Donaldson invariants and signed instanton counts possible.

Floer's original instanton invariant used signed trajectory counts, while Donaldson's later tract and Furuta's appendix made the orientation choices and gluing compatibility part of the standard construction ^{[Donaldson 2002]}. The same determinant-line formalism remains the orientation language in Seiberg-Witten and symplectic Floer theories.

Bibliography Master

@article{Donaldson1987Orientations,
  author = {Donaldson, Simon K.},
  title = {The orientation of Yang-Mills moduli spaces and 4-manifold topology},
  journal = {Journal of Geometry and Physics},
  volume = {4},
  pages = {369--428},
  year = {1987}
}

@book{Donaldson2002FloerOrientations,
  author = {Donaldson, Simon K.},
  title = {Floer Homology Groups in Yang-Mills Theory},
  series = {Cambridge Tracts in Mathematics},
  volume = {147},
  publisher = {Cambridge University Press},
  year = {2002}
}

@book{DonaldsonKronheimer1990Orientations,
  author = {Donaldson, Simon K. and Kronheimer, Peter B.},
  title = {The Geometry of Four-Manifolds},
  publisher = {Oxford University Press},
  year = {1990}
}

@article{Floer1988Orientations,
  author = {Floer, Andreas},
  title = {An instanton-invariant for 3-manifolds},
  journal = {Communications in Mathematical Physics},
  volume = {118},
  pages = {215--240},
  year = {1988}
}