Topology, Cohomology, and Characteristic Classes

Use this chapter when a QFT question asks what survives an admissible deformation, whether local data glue globally, which periods are quantized, why the cohomology class represented by a curvature expression is independent of the connection, how a stable zero-mode difference becomes topological, or whether a structured background extends across a filling. The chapter is organized by the object being compared and the equivalence relation imposed on it.

Its three main QFT entrances are winding and flux, characteristic classes, and the index bridge. The bordism primer is also directly reachable when bordism language already appears in the problem. No reader needs to complete all six pages before using one of these routes.

The scope is mathematical orientation and reusable calculation. The chapter does not infer a physical sector, anomaly, phase, selection rule, or topological field theory from an invariant alone. Those conclusions require the additional physical assumptions and treatments linked below. Bordism categories, invertible phases, TQFT classifications, and the cobordism hypothesis also remain outside this primer and continue to their exact Symmetry and Gauge Structure or Mathematical QFT treatments.

Enter · Choose a route · Compare invariants · Page guide · Line-bundle thread · Review · Continue

Enter through the invariant you need

This overview has no prerequisite. Start with the object in the calculation, then use the shortest preparation route in the table. The checks are diagnostic rather than sequential: being unready for the index theorem does not block the homotopy, chain, or bordism entrances.

Observable readiness checks and direct repair routes

Check	Ready	Unsure	Repair
Can you name the maps being deformed and what remains fixed during the deformation?	Enter the winding or homotopy route.	Write the source, target, basepoint or boundary data, and one proposed deformation before choosing a page.	Use Homotopy, Degree, Winding, and Covering Spaces to distinguish free, based, boundary-fixed, and gauge-quotiented questions.
Can you compute a boundary, explain why $\partial^2=0$, and name the coefficient group?	Enter the chain, homology, or discrete-topological-data route.	Expand the boundary of one oriented $2$-simplex and evaluate one coboundary on it.	Use Chains, Homology, Cohomology, and Exact Sequences to separate cycles, boundaries, cocycles, and coboundaries.
Can you apply Stokes’ theorem with the orientation and boundary convention visible?	Enter de Rham cohomology, periods, and intersection theory.	Apply Stokes to a disk and its oriented boundary, then distinguish closed from exact.	First repair Differential Forms, Integration, Orientation, and Stokes Theorem.
Can you distinguish a bundle, a connection, its curvature, and the Bianchi identity?	Enter characteristic classes after checking de Rham preparation.	On two overlapping patches, label the transition function, local connection forms, and global curvature.	Use Bundle Connections, Curvature, Gauge Transformations, and Bianchi Identities, then return through de Rham cohomology.
Can you state when an elliptic operator is Fredholm and identify the chiral block whose index is taken?	Enter the analytic-to-topological index bridge.	For one operator, name its domain, target, kernel, cokernel, ellipticity, and chirality convention.	Repair Spectra, Resolvents, Spectral Measures, and Functional Calculus, Characteristic Classes and Chern–Weil Theory, and Spin Structures and Dirac Operators separately.
Can you state which tangential structure, background map, or bundle must extend across a proposed filling?	Enter the bordism primer directly.	Write $\partial W=(-M_0)\sqcup M_1$ and list every structure required to extend over $W$.	Use Bordism and Tangential Structures: a Primer to parse the dimension, manifold category, boundary sign, and carried structure.

Repair only the capability used by the intended route, then return to the same row with a different example.

Choose a route from the QFT question

In this table, requires means that the target page uses an earlier capability in its main argument. Recommended preparation sharpens the comparison but does not block entry. Continue identifies where the mathematical invariant acquires its developed physical meaning.

Goal-to-route choices through topology and cohomology

Reader goal	Route and preparation	Observable result
Classify a circle winding or test whether maps can be deformed into one another	Begin directly: Homotopy, Degree, Winding, and Covering Spaces. There is no required preparation.	Name the admissible homotopy, compute winding or degree under its hypotheses, and say whether the invariant is complete in that example.
Work with cycles, discrete coefficients, torsion, or an exact sequence	Begin directly: Chains, Homology, Cohomology, and Exact Sequences. There is no required preparation.	Specify coefficients, form cycles modulo boundaries or cocycles modulo coboundaries, and interpret a connecting class without replacing it by a real differential form.
Turn a closed current or flux into periods and dual cycles	Requires: differential forms, orientation, integration, and Stokes. Enter de Rham Cohomology, Periods, Duality, and Intersection. Chains and cochains are recommended.	Separate closedness, exactness, real periods, integral periods, duality hypotheses, and intersection signs.
Compute a Chern number or show that a curvature class is connection independent	Requires: bundle connections and de Rham cohomology. Enter Characteristic Classes and Chern–Weil Theory.	State the invariant polynomial, trace representation, coupling and curvature convention, integral lift, and orientation.
Relate a chiral zero-mode difference to topological data	Requires: spectral analysis, spin structures and Dirac operators, and characteristic classes. Enter Fredholm and Dirac Index Theorems and Zero-Mode Counting.	Identify the closed elliptic Fredholm problem, compute the analytic index, and distinguish a net count from the two kernel dimensions.
Decode a bordism expression or test whether background data extend across a filling	Begin directly: Bordism and Tangential Structures: a Primer. Homotopy and characteristic classes are recommended comparisons.	Name the dimension, smooth category, tangential structure, background target, boundary orientation, and obstruction without asserting a field-theory classification.
Decide whether a topological number labels a physical sector, anomaly, phase, or selection rule	Begin: at the page defining that number. Continue: to the corresponding Symmetry and Gauge Structure, Nonperturbative Dynamics, or Mathematical QFT treatment below.	Separate the mathematical obstruction from existence, dynamics, superselection, anomaly matching, and classification claims.

Name the equivalence relation first

The word topological does not identify one universal quotient. Each page compares different objects, permits different changes, and produces an invariant with a different logical strength.

Objects, equivalence relations, and limits of the resulting invariants

Object and allowed comparison	Resulting class or invariant	What equality of the invariant does not generally prove
Maps related by an admissible homotopy	Homotopy class, winding, or degree	That arbitrary maps are homotopic, or that a field configuration exists dynamically or is stable
Chains modulo boundaries and cochains modulo coboundaries, with fixed coefficients	Homology or cohomology class and exact-sequence obstruction	Equality of homotopy type, or that changing coefficients preserves torsion information
Closed real forms modulo exact real forms	de Rham class, period, duality pairing, or intersection number	Equality of integral classes when they can differ by torsion
A fixed bundle with varying connection, or bundles compared by pullback and isomorphism	Characteristic class or characteristic number	That characteristic classes completely classify bundles, or that a real curvature representative contains flat holonomy and torsion data
Fredholm operators under admissible deformation or compact perturbation	Kernel-minus-cokernel index; for a chiral Dirac block, a net chiral count	The two kernel dimensions separately, absence of zero modes when the index vanishes, or an anomaly coefficient
Closed structured manifolds that bound one compact structured manifold a dimension higher	Structured bordism class and characteristic-number obstructions to filling	That vanishing tests construct a filling, or that a bare bordism group classifies anomalies, phases, or field theories

Homotopy, covering spaces, chains, induced maps, degree, and cohomology are developed in Hatcher 2002, §§ 1.1, 1.3, 2.1–2.2, and 3.1, pp. 25–31, 56–76, 105–118, 128–144, and 189–201, author-hosted PDF. The route from those objects through de Rham representatives, Chern–Weil theory, and index orientation is developed in Nakahara 2003, Chapters 3–4, §§ 6.2–6.4, 11.1–11.4, 12.1–12.2, and 12.6, pp. 98–151, 230–240, 419–441, 453–460, and 468–472, official publisher record. Milnor 1962, pp. 16–23, stable journal record supplies the distinct bordism equivalence relation and characteristic-number tests for boundaries.

How the six pages fit together

The displayed page order is pedagogical, not one compulsory chain.

1. Deformations of maps. Homotopy asks whether one map can be moved to another while preserving the stated basepoint, boundary, regularity, or gauge conditions. Winding and degree give obstructions.
2. Algebraic quotients. Chains and cochains turn boundaries and failed fillings into homology, cohomology, and exact-sequence data with explicit coefficients.
3. Smooth real representatives. de Rham theory realizes real cohomology through closed forms modulo exact forms, and integration pairs those classes with cycles.
4. Bundle invariants. Chern–Weil theory turns curvature into closed forms whose cohomology classes do not depend on the chosen connection.
5. The analytic-to-topological bridge. Index theory relates a stable Fredholm defect, or a net chiral zero-mode count, to symbol and characteristic data.
6. A different equivalence relation. Bordism asks whether the manifold and every declared structure extend across a compact filling. It is a parallel advanced entrance, not a consequence of the index theorem.

There are six required preparation links:

• de Rham cohomology requires differential forms, integration, orientation, and Stokes;
• characteristic classes require bundle connections and de Rham cohomology; and
• index theory requires spectral analysis, characteristic classes, and spin structures with Dirac operators.

Four of those links arrive from other chapters. The two internal required links are de Rham cohomology before characteristic classes and characteristic classes before index theory.

There are three recommended links: chains and cochains before de Rham cohomology, and homotopy plus characteristic classes before bordism. In particular, the bordism primer does not require index theory, and de Rham cohomology does not require the chain page.

Shared conventions and failure checks

Coefficients are part of every homology or cohomology statement. The symbols $H^\bullet(M;\mathbb Z)$, $H^\bullet(M;\mathbb R)$, and $H^\bullet(M;\mathbb Z_N)$ are not interchangeable. Passing to real de Rham cohomology erases torsion.

The chain boundary lowers degree, the coboundary raises degree, and the evaluation convention is

$$\begin{aligned} \partial &: C_k\longrightarrow C_{k-1}, & \delta &: C^k\longrightarrow C^{k+1}, \\ \partial^2&=0, & \delta^2&=0, \\ \langle\delta\alpha,c\rangle &=\langle\alpha,\partial c\rangle. \end{aligned}$$

Homology is covariant under $f_*$, while cohomology is contravariant under $f^*$. Smooth forms obey $\mathrm d^2=0$. Wedge order is retained: $\alpha\wedge\beta=(-1)^{pq}\beta\wedge\alpha$ for degrees $p$ and $q$. Poincaré duality in the basic chapter treatment assumes a compact, oriented, boundaryless manifold; it is not the metric-dependent Hodge duality.

Orientations are stated where a sign becomes consequential. Degree names the orientations of source and target. Intersections use the displayed order. Bordism uses outward-normal-first boundary orientation, so a bordism from $M_0$ to $M_1$ satisfies

$$\partial W\cong(-M_0)\sqcup M_1.$$

Gauge formulas inherit the site’s Hermitian-generator convention,

$$D=\mathrm d-i g_{\mathrm{YM}}A.$$

When a mathematical source uses anti-Hermitian connection and curvature matrices, the translation is performed before importing a characteristic form:

$$\mathbf A=-i g_{\mathrm{YM}}A, \qquad \mathbf F=-i g_{\mathrm{YM}}F.$$

The trace representation, coupling placement, orientation, and integral normalization remain explicit. Gauge invariance of a form, connection independence of its class, and integrality of a characteristic number are three separate statements.

The adjective closed is also typed by its noun: a closed form has zero exterior derivative, a closed manifold is compact and boundaryless, and a closed operator has a closed graph. For the Euclidean index bridge, analytic Clifford multiplication and the index-theorem grading are

$$\mathbf c=i\gamma, \qquad \Gamma_{\mathrm{ind}}=(-1)^m\Gamma_E \quad\text{in dimension }2m.$$

The bordism page states whether it uses stable tangent or stable normal structures.

Exact page guide

Homotopy, Degree, Winding, and Covering Spaces

Homotopy, Degree, Winding, and Covering Spaces asks when maps or field configurations can be deformed into one another and which integers obstruct the deformation. It has no required or recommended preparation.

Use it to declare the admissible homotopy, compute circle winding through the universal cover, use degree with orientations visible, and understand transition-function winding. Unequal invariants obstruct a homotopy; equal degree does not generally classify arbitrary maps. Continue to Large Gauge Transformations and Topological Sectors for the physical status of winding sectors and large transformations.

Chains, Homology, Cohomology, and Exact Sequences

Chains, Homology, Cohomology, and Exact Sequences asks how cycles, boundaries, cocycles, and exact sequences encode topological invariants. It has no required or recommended preparation.

Use it when coefficients, torsion, relative cycles, induced maps, pairings, or connecting homomorphisms matter. It is recommended before de Rham cohomology but remains an independent algebraic entrance. Continue to BF Couplings and Discrete Topological Data for a developed physical use of finite cohomology–homology pairings.

de Rham Cohomology, Periods, Duality, and Intersection

de Rham Cohomology, Periods, Duality, and Intersection asks how closed forms, periods, and dual cycles turn local differential data into global invariants. It requires Differential Forms, Integration, Orientation, and Stokes Theorem; chains and cochains are recommended.

Use it to distinguish closed from exact, real from integral periods, and de Rham from torsion data. State the compactness, orientation, support, and boundary hypotheses before using Poincaré duality or intersection pairings. Continue to Higher-Form Currents, Charges, Backgrounds, and Ward Identities for the physical current and charge interpretation.

Characteristic Classes and Chern–Weil Theory

Characteristic Classes and Chern–Weil Theory asks how invariant polynomials of curvature produce connection-independent topological classes. It requires Bundle Connections, Curvature, Gauge Transformations, and Bianchi Identities and de Rham cohomology.

Use it to separate a curvature representative from an integral class, to translate anti-Hermitian and Hermitian gauge conventions, and to fix the trace and orientation of a characteristic number. Real Chern–Weil forms do not detect torsion or Stiefel–Whitney data. Continue to Theta Terms, Periodicity, and Vacuum Sectors for the physical weighting and vacuum interpretation.

Fredholm and Dirac Index Theorems and Zero-Mode Counting

Fredholm and Dirac Index Theorems and Zero-Mode Counting asks how analytic zero-mode data can be related to topological characteristic data. It requires Spectra, Resolvents, Spectral Measures, and Functional Calculus, characteristic classes, and Spin Structures and Dirac Operators.

Use it for a closed elliptic Fredholm problem and a stable kernel-minus-cokernel count. The twisted chiral Dirac index is a net chirality, not the two kernel dimensions separately. Boundary, noncompact, Lorentzian, and unspecified-domain problems require different theorems. Continue to Perturbative Chiral and Gauge Anomalies or Fermion Zero Modes, Index Data, and Selection Rules only after preserving that distinction.

Bordism and Tangential Structures: a Primer

Bordism and Tangential Structures: a Primer asks what new equivalence relation bordism imposes and which background structures must be carried along. It has no required preparation; homotopy and characteristic classes are recommended comparisons.

Use it to parse a symbol such as $\Omega_d^\theta(X)$ by naming the dimension, smooth category, tangential structure, background map, and whether the structure is stated on stable tangent or stable normal data. A characteristic number can obstruct a filling, but its vanishing need not construct one. Continue to Bordism Categories and Symmetric Monoidal TQFTs for the categorical structure and field-theory assignments deliberately omitted here.

One line bundle across the chapter

Let $S^2$ carry its standard orientation, cover it by northern and southern hemispheres, and let $L_n\to S^2$ be a Hermitian line bundle with structure group $U(1)$ whose equatorial transition function is

$$g_{NS}(e^{i\varphi})=e^{in\varphi}, \qquad n\in\mathbb Z.$$

With the equator oriented by increasing $\varphi$ and the convention $v_N=g_{NS}v_S$, its winding is

$$\nu(g_{NS}) = \frac1{2\pi i} \int_{S^1}g_{NS}^{-1}\,\mathrm d g_{NS} =n.$$

A compatible coupling-absorbed real connection has global curvature

$$f = \frac n2\sin\theta\, \mathrm d\theta\wedge\mathrm d\varphi,$$

and therefore

$$\frac1{2\pi}\int_{S^2}f=n.$$

Equip $S^2$ with a Riemannian metric and its unique spin structure, equip $L_n$ with a Hermitian metric, and interpret the compatible $U(1)$ connection as a unitary connection. Under these choices, the chapter’s central comparison is

$$\begin{aligned} \nu(g_{NS}) &= \left\langle c_1(L_n),[S^2]\right\rangle = \frac1{2\pi}\int_{S^2}f, \\ &= \operatorname{ind}\mathcal D_{R,L_n}^{+} =n. \end{aligned}$$

The last equality uses the closed Riemannian spin surface, the unitary connection, the twisted elliptic Dirac operator, and $\Gamma_{\mathrm{ind}}$. In two dimensions, $\Gamma_{\mathrm{ind}}=-\Gamma_E$, so using the earlier $\Gamma_E$ labels would reverse the displayed index sign; the index page gives the full translation. Reversing the sphere or equator orientation or inverting the transition convention likewise changes the corresponding signs and must be translated consistently. The winding, flux, Chern, and two-dimensional index versions are developed in Nakahara 2003, Chapter 4, § 10.5.2, Example 11.2, and § 12.6, pp. 121–151, 400–401, 432–433, and 468–472, official publisher record.

The pair $(S^2,L_n)$ viewed through all six pages

Page viewpoint	Question asked of the line bundle	Invariant and limitation
Homotopy	Can the equatorial transition map be deformed to the constant map while preserving the patching data?	Its winding is $n$. Nonzero winding obstructs that deformation but does not prove a monopole solution exists or is stable.
Chains and cochains	Which integral cycle and cohomology class are paired?	The fundamental class $[S^2]$ pairs with $c_1(L_n)$ to give $n$; coefficients and orientation are part of the pairing.
de Rham theory	Which smooth form represents the real class and which period is invariant?	$[f/(2\pi)]_{\mathrm{dR}}$ has period $n$. Closedness preserves periods on homologous cycles, while the integral first Chern class and its normalization supply integrality.
Characteristic classes	Why does changing the connection not change the integer?	$[f/(2\pi)]_{\mathrm{dR}}=c_1(L_n)\otimes1$; a connection change alters the form by an exact term, while the integral class remains fixed.
Index theory	What does the twisted chiral Dirac operator count?	$\operatorname{ind}\mathcal D_{R,L_n}^{+}=n$ in the stated grading. It fixes $n_+-n_-$, not $n_+$ and $n_-$ separately.
Bordism	Does the line bundle extend across a three-dimensional filling?	The oriented sphere bounds $D^3$, but the structured-surface calculation shows that $(S^2,L_n)$ is not null-bordant over $BU(1)$ for $n\ne0$, because its first Chern number would have to vanish on a structured boundary.

The bordism conclusion treats the line bundle as a background map $S^2\to BU(1)$; it does not require a connection to extend. Adding that connection would define a stronger geometric problem. The distinction between the underlying sphere and the structured pair is the point of the bordism entrance, not a classification of a QFT anomaly or phase.

Chapter synthesis

Three distinctions organize the chapter.

An invariant is typed by an equivalence relation. Winding compares maps under homotopy, homology compares cycles modulo boundaries, de Rham cohomology compares closed real forms modulo exact forms, the Fredholm index survives operator deformations, and bordism compares structured manifolds through a filling. The same integer can appear in several descriptions without making those descriptions synonymous.

A representative is not the whole global class. A curvature form can represent the real image of an integral characteristic class, but real forms lose torsion and do not record every flat-holonomy datum. Gauge invariance, connection independence, and integrality require separate arguments.

An obstruction is not automatically a classification or a physical claim. Unequal invariants can rule out an equivalence. Equal invariants need not construct a homotopy, bundle isomorphism, filling, or field configuration. An index is a stable difference, and neither an index nor a bordism class by itself proves an anomaly, phase, selection rule, or superselection sector.

A compact working rule is:

Before using a topological label, state the objects, allowed equivalence, coefficients and structures, orientation and normalization, theorem hypotheses, and the strongest conclusion that the invariant actually licenses.

Review the chapter

A successful response meets the stated criterion; use the linked page to repair the first missing step.

Choose the equivalence relation. For a circle map, an integral cycle, a closed real form, a connection on a fixed bundle, a chiral Fredholm operator, and a structured closed manifold, state what changes are allowed and what quotient or stable number results. A successful response does not use homotopy, homology, cohomology, index, and bordism as synonyms. Repair with the comparison table and the corresponding page guide.

Track coefficients and torsion. Give an example of information that can be present over $\mathbb Z$ or $\mathbb Z_N$ but vanish after passage to real de Rham cohomology. A successful response names the coefficient map and does not ask a real curvature form to detect torsion. Repair with Chains, Homology, Cohomology, and Exact Sequences and the de Rham page.

Reconstruct the line-bundle chain. Starting from $g_{NS}(e^{i\varphi})=e^{in\varphi}$, recover its winding, the normalized curvature period, the first Chern pairing, and the two-dimensional twisted Dirac index. Translate $\mathbf A=-i g_{\mathrm{YM}}A$ and $\mathbf F=-i g_{\mathrm{YM}}F$, and translate $\Gamma_{\mathrm{ind}}=(-1)^m\Gamma_E$, before comparing signs. A successful response states the orientation, transition convention, $U(1)$ structure group, integral Chern normalization, and index grading; preserves the Chern number under that notation translation; and explains why the index is only a net count. Repair with the four relevant pages in the exact guide.

Build the index preparation route. Explain why spin geometry, characteristic classes, and spectral Fredholm theory are three distinct required inputs. A successful response identifies the chiral block, its domain and target, the closed elliptic setting, and the characteristic density without treating the full self-adjoint Dirac operator as a nonzero index problem. Repair with Fredholm and Dirac Index Theorems and Zero-Mode Counting.

Compare a boundary with a structured boundary. Using the structured-surface calculation, explain why $S^2$ bounds as an oriented manifold but $(S^2,L_n)$ with $n\ne0$ does not bound over $BU(1)$. A successful response says which map or bundle must extend and why the first Chern number obstructs that extension. It does not claim that one vanishing number constructs every filling. Repair with Bordism and Tangential Structures: a Primer.

Transfer without overclaim. Choose one of winding, flux, a Chern number, an index, or a bordism class and identify a QFT page that gives it physical meaning. A successful response lists the extra dynamical, operator, regularization, state, or classification assumptions needed before the physical conclusion follows. Repair with the relevant continuation below.

Continue from here

Within Mathematical Methods:

• Differential Geometry, Bundles, Connections, and Curvature supplies the forms, bundles, connections, metrics, and spin geometry required by the de Rham, characteristic-class, and index routes.
• Functional and Spectral Analysis supplies closed operators, Fredholm theory, spectral data, and functional calculus for the analytic side of the index bridge.
• Categorical, Homological, and Local-to-Global Methods continues from complexes to chain homotopy, quasi-isomorphisms, sheaves, descent, and higher structural language.
• Mathematical Methods lets you choose another chapter route without treating this specialist chapter as a universal prerequisite.

For developed QFT applications:

• Large Gauge Transformations and Topological Sectors gives winding data its boundary-condition, gauge-group, and sector meaning.
• BF Couplings and Discrete Topological Data develops finite pairings and discrete coefficients in physical models.
• Higher-Form Currents, Charges, Backgrounds, and Ward Identities turns closed-current and dual-cycle language into physical charges and identities.
• Theta Terms, Periodicity, and Vacuum Sectors supplies the physical weighting and global qualifications of characteristic numbers.
• Background Responses and Invertible Phases develops when a quantized background functional defines an invertible response and what additional boundary and anomaly data are required.
• Perturbative Chiral and Gauge Anomalies develops the field-theoretic anomaly calculation without identifying its coefficients with a bare index by inspection.
• Fermion Zero Modes, Index Data, and Selection Rules adds normalizability, Grassmann saturation, and the physical selection-rule argument.
• Bordism Categories and Symmetric Monoidal TQFTs develops categorical bordisms, field-theory assignments, and gluing beyond this chapter’s primer.
• Dualizability and the Cobordism Hypothesis states the target-category, full-dualizability, extension-depth, and equivalence data needed for the classification theorem.

References

• Allen Hatcher, Algebraic Topology, Cambridge University Press (2002), Hatcher 2002, §§ 1.1 and 1.3, pp. 25–31 and 56–76; §§ 2.1–2.2, pp. 105–118 and 128–144; and § 3.1, pp. 189–201, author-hosted PDF. Homotopy, covering spaces, degree, chains, homology, cohomology, relative groups, and exact sequences.
• John Milnor, “A Survey of Cobordism Theory,” L’Enseignement Mathématique 8 (1962), 16–23, stable journal record. Bordism as an equivalence relation, structured variants, and characteristic-number tests for boundaries.
• Mikio Nakahara, Geometry, Topology and Physics, second edition, Institute of Physics Publishing (2003), Nakahara 2003, Chapters 3–4, pp. 98–151; §§ 6.2–6.4, pp. 230–240; §§ 11.1–11.4, pp. 419–441; and §§ 12.1–12.2 and 12.6, pp. 453–460 and 468–472, official publisher record. Chains, homotopy, de Rham cohomology, characteristic classes, Chern–Weil theory, and index orientation.