Derived Critical Loci and Derived Gauge Quotients

Classical equations define more than a set of solutions when their differentials fail to be independent or when gauge transformations have stabilizers. A derived critical locus retains those failures as cohomological directions. Its degree-zero truncation recovers the familiar solution space, while neighboring degrees record infinitesimal automorphisms, physical deformations, and obstructions. The point is local and perturbative: it does not by itself construct a measure, a Hilbert space, or a nonperturbative quantum theory.

Required background. BV Quantization and Obstruction–Deformation Complexes supplies the cochain-complex interpretation of fields, ghosts, and obstructions. Elliptic Complexes and Factorization Observables supplies the elliptic deformation complexes used below. Helpful background. Existence, Construction, Reconstruction, and Continuum Claims separates a formal local model from a constructed QFT. Domains, Signatures, Supports, and Regularity explains why the compactness and smoothness assumptions matter. Chain Homotopy, Quasi-Isomorphisms, and Derived Vocabulary supplies the homological language.

Derived zero loci retain excess equations

Let $Y$ be a smooth space, $E\to Y$ a vector bundle, and $s\in\Gamma(Y,E)$ a section. The ordinary zero locus remembers the points at which $s=0$. Locally, the derived zero locus is represented by the Koszul commutative differential graded algebra

$$\left(\Gamma\!\left(U,\Lambda^{\bullet}E^{\vee}\right),\ \iota_s\right),$$

with generators in negative cohomological degrees. If $s$ is transverse to the zero section, the negative cohomology vanishes and the derived and ordinary loci agree. If the components of $s$ satisfy relations, negative cohomology records those syzygies. Thus derivation is not an ornamental thickening: it measures precisely the failure of transversality.

For an action functional $S$ on a smooth field space $\mathcal F$, its critical locus is the intersection of the graph of $dS$ with the zero section in $T^*\mathcal F$. Replacing this intersection by its homotopy fiber product gives

$$\operatorname{RCrit}(S) =\mathcal F\mathop{\times}^{h}_{T^*\mathcal F}\mathcal F.$$

In finite-dimensional derived algebraic geometry, the zero section and graph of $dS$ are Lagrangian, so this derived intersection is canonically $(-1)$-shifted symplectic Pantev–Toën–Vaquié–Vezzosi 2013, Corollary 2.11, pp. 39–40. For field theory this statement is a model, not an automatic theorem: one must choose a category of infinite-dimensional spaces, supports, completions, and a class of admissible functionals before the fiber product exists.

Gauge quotients are complexes, not orbit sets

At a classical solution $x$, the tangent object of a gauge theory has a cochain model

$$\mathfrak g_x \xrightarrow{\rho_x} T_x\mathcal F \xrightarrow{D_x\mathrm{EL}} \mathcal E_x \xrightarrow{N_x} \mathcal I_x,$$

placed in degrees $-1,0,1,2$. Here $\mathfrak g_x$ generates infinitesimal gauge transformations, $\mathcal E_x$ contains linearized equations, and $\mathcal I_x$ contains Noether identities. The identities $D_x\mathrm{EL}\,\rho_x=0$ and $N_xD_x\mathrm{EL}=0$ make this a complex. Its cohomology has a direct interpretation:

• $H^{-1}$ is the infinitesimal stabilizer of $x$;
• $H^0$ is the space of infinitesimal solutions modulo infinitesimal gauge transformations;
• $H^1$ contains primary obstructions to extending an infinitesimal solution;
• higher groups record identities among identities when the gauge symmetry is reducible.

The displayed four-term form is a convenient finite window. A reducible theory can extend farther in both directions, and a Lagrangian BV theory appends dual antifield terms. In a characteristic-zero derived-deformation setting, tangent quasi-isomorphisms are the weak equivalences underlying the corresponding formal models Pridham 2010, Theorem 4.55 and Corollary 4.57, PDF p. 47. Equality of degree-zero orbit sets does not provide that comparison.

First application: a reducible flat connection

Consider the setting of Gauge Orbits, Gauss Constraints, and Stabilizers. Let $P\to M$ be a principal bundle over a closed compact manifold, let $G$ be compact, and let $A$ be a smooth flat connection. The deformation problem for flat connections modulo gauge transformations is controlled by

$$\Omega^0(M,\operatorname{ad}P) \xrightarrow{d_A} \Omega^1(M,\operatorname{ad}P) \xrightarrow{d_A} \Omega^2(M,\operatorname{ad}P) \xrightarrow{d_A}\cdots .$$

After shifting so connection variations have degree zero, $H_A^0$ is the Lie algebra of the stabilizer, $H_A^1$ gives infinitesimal flat deformations modulo gauge, and $H_A^2$ is the primary obstruction space. Flatness ensures $d_A^2=[F_A,\cdot]=0$. Ellipticity on a closed manifold makes these groups finite-dimensional and Hodge theory supplies harmonic representatives. The nonlinear equation for a displacement $a\in\Omega^1$ is

$$d_Aa+\tfrac12[a,a]=0.$$

Projecting this equation to harmonic two-forms gives the leading Kuranishi obstruction. Near a reducible connection, $H_A^0\neq0$; the local moduli object therefore has automorphisms and cannot be a smooth orbit space. This flat sector sits inside the critical geometry of Yang–Mills: flat connections solve the Yang–Mills equations, but the full Yang–Mills BV complex also includes the linearized Euler–Lagrange operator and its dual terms.

An independent check is Euler-characteristic consistency. For an acyclic irreducible flat connection, $H_A^0=H_A^1=H_A^2=0$ in the relevant deformation window, so the derived enhancement is locally invisible, as transversality predicts. At a reducible connection the stabilizer dimension reappears in $H_A^0$, exactly where the quotient ceases to be locally free.

Failure test: truncating at the orbit space

Replace the derived quotient by the set of gauge orbits and examine the reducible connection above. The underlying point $[A]$ remains, but the orbit set has no place for $H_A^0$ and $H_A^2$. It therefore loses both the isotropy responsible for the singular quotient and the obstruction to extending a first-order deformation. Two gauge problems can have the same nearby orbit set while having non-quasi-isomorphic tangent complexes. Consequently, agreement of ordinary solution sets is not a converse to equivalence of derived critical loci.

Nor does the derived local model prove global existence. The construction is formal near $A$ and depends on the selected function spaces and completion. Gribov phenomena, disconnected bundle sectors, convergence of perturbation theory, reflection positivity, and a quantum state space require additional arguments.

Exercises

Let $s:\mathbb R\to\mathbb R^2$ be $s(x)=(x^2,x^3)$. Compute the cohomology at the origin of its Koszul complex and explain why the ordinary zero set misses information.

Solution

The local Koszul algebra is $\mathbb R[x]\otimes\Lambda(e_1,e_2)$ with $de_1=x^2$ and $de_2=x^3$. The degree $-1$ element $xe_1-e_2$ is closed because $x^3-x^3=0$. It is not generated by a degree $-2$ boundary with a unit coefficient, so negative cohomology survives at the origin. The ordinary zero set is only $\{0\}$ and cannot display this relation between the two equations.

For a flat connection, verify that an infinitesimal gauge transformation $a=d_A\epsilon$ solves the linearized flatness equation.

Solution

The linearized flatness operator is $d_A:\Omega^1\to\Omega^2$. Hence $d_Aa=d_A^2\epsilon=[F_A,\epsilon]=0$. This verifies that the image of the gauge differential lies in the kernel of the equation differential and explains why the quotient is cohomology rather than a set-theoretic subtraction.

References

• Pantev, Toën, Vaquié, and Vezzosi 2013, Shifted Symplectic Structures, Publications Mathématiques de l’IHÉS 117, 271–328. Open PDF
• Pridham 2010, Unifying Derived Deformation Theories, Advances in Mathematics 224(3), 772–826. Open PDF