Rigid Backgrounds, Topological Twists, and Localization

Localization is not a single theorem applied to an infinite-dimensional integral. It is a chain of constructions: choose a supersymmetric background, obtain a globally defined odd symmetry, identify its cohomology, and declare the original field-space integration cycle before introducing the deformation. One must then combine the symmetry with gauge fixing, determine every saddle and zero mode, and transport or decompose that same cycle through the regulated finite-dimensional reduction. This chapter develops the full chain and makes clear which statements are exact theorems and which remain controlled path-integral arguments.

Helpful background. Fredholm and Dirac index theorems explain how unpaired modes replace a formal quotient of infinitely many eigenvalues. Heat kernels, zeta functions, and spectral determinants supply determinant regulators. BRST cohomology and physical observables supplies the gauge-theory differential used after a localizing supercharge is chosen.

Enter this chapter

The central diagnostic is simple:

A localization formula is only as well defined as its least specified ingredient.

In many Euclidean supersymmetric constructions, variables related by Lorentzian conjugation are first complexified and treated independently. A middle-dimensional integration cycle then supplies the reality or steepest-descent condition of the functional integral. Thus real nonnegativity of a bosonic deformation is a statement about a chosen real cycle; a complex cycle instead requires control of its real part and a separate identification of its critical set. Neither follows automatically from the mnemonic $t=-i\tau$ . Likewise, $Q^2$ need not vanish: it may be an isometry, an R rotation, a flavor transformation, and a field-dependent gauge transformation. Every deformation, insertion, boundary condition, and regulator must be invariant under that full even symmetry.

The pages are ordered so that each construction supplies the input of the next.

Stage	Question answered	Output needed downstream
Supercurrent compatibility	Which current multiplet can couple to the desired off-shell supergravity formulation?	Admissible nondynamical sources and obstruction classes
Rigid curved backgrounds	Which frozen bosonic backgrounds make every background-fermion variation vanish?	Generalized Killing-spinor equations and off-shell algebra
Global spin-R bundles	Does the local spinor patch to a global supercharge?	Bundle, flux, zero-locus, and boundary data
Topological and holomorphic twists	Can Lorentz and R symmetry be recombined to produce a scalar or holomorphic differential?	Twisted fields, $Q$ -complex, and protected observables
Finite-dimensional localization	What does the Atiyah–Bott–Berline–Vergne theorem actually prove?	The precise fixed-locus model and its hypotheses
$Q$ -exact path-integral deformation	Why should a selected correlator be deformation independent?	Ward identity and a list of possible failure terms
Gauge fixing and the deformation complex	How are supersymmetry and gauge redundancy combined?	A graded elliptic or transversely elliptic complex
Loci and one-loop determinants	Which saddles, collective coordinates, and unpaired modes remain?	Sector sum, classical weights, and regulated determinant
Contours and regularization	How is the predeclared integration cycle transported, and which spectral prescription defines the answer?	Thimble coefficients, determinant phase, and scheme dependence
Boundaries, gluing, and JK residues	How do cutting, boundary anomalies, and residue chambers fit together?	A gluing pairing or chamber-qualified residue formula

The validity chain

It is useful to write a proposed computation as the following data rather than as the slogan “add a $Q$ -exact term”:

\mathfrak L= \bigl(M,\mathcal B,\Gamma,Q,Q^2,\mathcal O,V, \mathcal M_Q,\mathcal C_{\rm def},\operatorname{Reg},S_{\rm ct},\mathfrak b\bigr).

Here $M$ is the spacetime (possibly with boundary), $\mathcal B$ its spin-R and background bundles, $\Gamma$ the field-space cycle, $Q$ the odd symmetry, $\mathcal O$ the insertion, and $V$ the deformation functional. The remaining entries are the full BPS locus $\mathcal M_Q$ , the gauge-fixed fluctuation complex $\mathcal C_{\rm def}$ , the spectral regulator, allowed local counterterms, and boundary or gluing data $\mathfrak b$ . A claimed exact quantity should remain qualified until every relevant entry is fixed.

The order matters. The expression $S_t=S_E+tQV$ defines a supersymmetric deformation only if $Q^2V=0$ after all bosonic symmetries, gauge transformations, boundary conditions, and ghost transformations are included. Even then, the Ward-identity step needs a $Q$ -invariant measure and cycle. Reality and nonnegativity on a real cycle, or controlled real-part decay on a complex cycle, do not by themselves prove completeness of the fixed locus or uniform control as $t\to\infty$ ; those are separate inputs. The numbered localization-validity diagram keeps these obligations and their failure exits visible in one place.

Three distinct conclusions must not be conflated:

Cohomological independence: a Ward identity gives $d\langle\mathcal O\rangle_t/dt=0$ for finite $t$ .
Asymptotic localization: the $t\to\infty$ integral is controlled by the zeros of the bosonic deformation on the chosen cycle.
Evaluation: the resulting finite-dimensional integral, sum, or residue has been assigned a contour, regulator, phase, and normalization.

The first does not by itself prove the second, and neither fixes the third. This distinction is emphasized in the finite- and infinite-dimensional treatments of Schwarz and Zaboronsky 1997, pp. 463–476 and in the gauge-theory construction of Pestun 2012, §§3–4.

Reading a localized answer

Once the chain succeeds, a typical result has the schematic form

Z= \sum_{\alpha} \int_{\Gamma_\alpha\subset\mathcal M_\alpha} d\mu_{\alpha}^{\rm zero}\, e^{-S_{E,\alpha}^{\rm cl}} Z_{\alpha}^{\rm 1\text{-}loop} Z_{\alpha}^{\rm rem}.

This compact line hides most of the work:

$\alpha$ ranges over every allowed bundle, flux, instanton, defect, and boundary sector, not merely the smooth saddle found first;
$\mathcal M_\alpha$ is a quotient or stack-like saddle space, so stabilizers and residual gauge volume affect $d\mu_{\alpha}^{\rm zero}$ ;
$Z_{\alpha}^{\rm 1\text{-}loop}$ is a regulated ratio of primed determinants, with true zero modes removed and integrated through the measure instead;
$Z_{\alpha}^{\rm rem}$ denotes any lower-dimensional, instanton, boundary, or other contribution not captured by the Gaussian complex; and
$\Gamma_\alpha$ records the chosen real slice, thimble combination, or residue chamber. When a fixed original cycle is transported across a Stokes wall, its thimble basis and coefficients can jump oppositely while the integral stays unchanged. The value changes only if the prescribed homology class or lateral chamber changes—for example, through a pole crossing or a different contour prescription—even though the local saddle equations may be identical Witten 2011, §§2–3.

An equality between two displayed formulas is therefore meaningful only after their sector sets, cycles, determinant phases, counterterm schemes, and normalizations have been matched. Depending on the observable, allowed supersymmetric local counterterms may leave only derivatives, ratios, absolute values, or anomaly-controlled pieces scheme independent.

Review the chapter

Take a localization formula you know and answer, in order:

What is the globally defined bundle of the supersymmetry parameter?
What is $Q^2$ on physical fields, ghosts, and boundary fields?
Which integration cycle makes the bosonic deformation convergent?
Are all topological sectors and fixed components included?
Which modes are removed from $\det'$ and how are they integrated instead?
Which regulator fixes the determinant phase and scale?
Which supersymmetric local counterterms can change the reported observable?
If the space is cut, what polarization, anomaly inflow, and gauge quotient enter the pairing?
Which parameter chamber and normalization convention make the result comparable with another calculation?

If any answer is missing, the formula may still be a useful formal expression, but it is not yet a complete definition.

References

Pestun, Vasily. “Localization of Gauge Theory on a Four-Sphere and Supersymmetric Wilson Loops.” Communications in Mathematical Physics 313 (2012): 71–129. doi:10.1007/s00220-012-1485-0. Open preprint.
Schwarz, Albert, and Oleg Zaboronsky. “Supersymmetry and Localization.” Communications in Mathematical Physics 183 (1997): 463–476. doi:10.1007/BF02506415. Open preprint.
Witten, Edward. “Analytic Continuation of Chern–Simons Theory.” In Chern–Simons Gauge Theory: 20 Years After, AMS/IP Studies in Advanced Mathematics 50, 347–446. Providence, RI: American Mathematical Society, 2011. Open preprint.