---
title: 'Pfaffian Fibrations: Intrinsic PDE Geometry'
url: https://www.emergentmind.com/topics/pfaffian-fibrations
type: topic
---

# Pfaffian Fibrations: Intrinsic PDE Geometry

Pfaffian fibrations are an intrinsic geometric framework for encoding the essential data of a partial differential equation without privileging an ambient jet bundle. In the formulation developed in “From PDEs to Pfaffian fibrations” [1901.02084], the basic object is a fibration equipped with a Cartan-type \(1\)-form, or equivalently a distinguished distribution whose holonomic sections play the role of solutions. Subsequent work relating Pfaffian fibrations to relative algebroids shows that they also admit a canonical infinitesimal model with the same prolongations and local solutions, and that this model is compatible with natural notions of symmetry [2510.01517].

## 1. Definition and equivalent formulations

A Pfaffian fibration may be presented in either form language or distributional language. In the form-theoretic presentation, a Pfaffian fibration \((P,\theta)\) over a manifold \(M\) consists of a fibration \(\pi:P\to M\) together with a pointwise surjective differential form
\[
\theta\in\Omega^1(P,N),
\]
where \(N\to P\) is a vector bundle, such that \(\theta\) is \(\pi\)-regular and \(\pi\)-involutive [1901.02084]. The \(\pi\)-regularity condition is
\[
TP=\ker(\theta)+T^\pi P,
\]
so the kernel distribution is transverse to the \(\pi\)-fibres, while \(\pi\)-involutivity requires the symbol distribution
\[
g(\theta):=T^\pi P\cap \ker(\theta)
\]
to be involutive in the Frobenius sense [1901.02084].

The equivalent distributional presentation, emphasized in the later symmetry-theoretic treatment, describes a Pfaffian fibration as a triple
\[
(P,C,\pi),
\]
where \(\pi:P\to X\) is a submersion and \(C\subset TP\) is a distribution satisfying the transversality condition
\[
C+\ker T\pi=TP
\]
and the \(\pi\)-involutivity condition
\[
C^\pi:=C\cap \ker T\pi
\]
is involutive [2510.01517]. The passage between the two descriptions is immediate: if \(H=\ker\theta\), then \(H\) is a \(\pi\)-transversal distribution with involutive symbol, and conversely a distribution \(H\subset TP\) determines a surjective quotient form
\[
\theta_H:TP\to TP/H\cong N_H
\]
[1901.02084].

This dual description is structurally important. The form viewpoint makes contact with Cartan forms and Spencer-operator-type constructions, whereas the distribution viewpoint is adapted to prolongations, relative algebroids, and symmetry groupoids. A local section is holonomic precisely when it annihilates the Pfaffian form or, equivalently, when its tangent image lies in the Pfaffian distribution:
\[
b^*\theta=0
\qquad\text{or}\qquad
db(TM)\subseteq H
\]
[1901.02084]; equivalently, for \(C=\ker\theta_C\),
\[
\operatorname{im}(T\sigma)\subset C
\quad\Longleftrightarrow\quad
\sigma^*\theta_C=0
\]
[2510.01517].

## 2. Relation to partial differential equations and jet bundles

The motivating example is the Cartan geometry of jet bundles. For a submersion \(q:Q\to X\), the first jet bundle \(J^1q\to X\) with its Cartan distribution \(C^1\) is a Pfaffian fibration
\[
(J^1q,C^1,\pi),
\]
and any PDE \(\mathcal E\subset J^1q\) becomes a Pfaffian fibration by restriction:
\[
(\mathcal E,C^1,\pi)
\]
[2510.01517]. In the general \(k\)-th order setting, if \(P\subset J^kR\) is a PDE on a fibration \(R\to M\), then the restricted Cartan form \(\theta_{\mathrm{can}|_P}\) already captures the essential geometric structure, so the pair
\[
(P,\theta_{\mathrm{can}|_P})
\]
is the intrinsic object of interest [1901.02084].

The central claim of the intrinsic theory is that the ambient jet bundle is not the essential datum. A section \(s\) of \(R\to M\) solves the PDE exactly when its jet prolongation is tangent to \(P\), equivalently
\[
j^k s\in P,
\]
or, in Pfaffian terms,
\[
s^*\theta_{\mathrm{can}}=0
\]
[1901.02084]. This is the precise sense in which a PDE can be encoded by a fibration together with a Cartan-type \(1\)-form.

The standard slogan recorded in the foundational paper is that a PDE is encoded by a fibration together with a geometric device that detects solutions, and the jet bundle is background structure rather than essential structure [1901.02084]. This suggests that Pfaffian fibrations isolate precisely those aspects of PDE geometry needed for prolongation, integrability, and linearisation, while discarding extraneous choices of embedding.

## 3. Holonomic sections, curvature, prolongations, and integrability

The intrinsic solution concept is given by holonomic sections. A section \(b:M\to P\) is holonomic if
\[
b^*\theta=0,
\]
and a Pfaffian fibration is PDE-integrable if through every point \(p\in P\) there exists a local holonomic section passing through \(p\) [1901.02084]. In the distributional formulation, holonomicity is the condition that the tangent map of the section lands in the Pfaffian distribution [2510.01517].

Curvature controls the failure of the Pfaffian distribution to be integrable. For a Pfaffian distribution \(H\subset TP\), the curvature is
\[
K_H:H\times H\to N_H,
\qquad
K_H(U,V)=[U,V]\bmod H,
\]
or, in form language,
\[
K_\theta(U,V)=\theta([U,V]),
\qquad U,V\in\ker\theta
\]
[1901.02084]. In the later notation, the curvature map entering prolongation theory is
\[
\kappa_C(X,Y)=[\tilde X,\tilde Y]\bmod C
\]
[2510.01517].

Two first-order prolongation spaces are distinguished. The partial prolongation is
\[
J^1_C P=\{h\in J^1\pi\mid h^*\theta_C=0\},
\]
while the full prolongation is
\[
P^{(1)}=\{h\in J^1\pi\mid h^*\theta_C=0,\; h^*\kappa_C=0\}
\]
[2510.01517]. In the earlier notation, the partial prolongation \(J_H P\) is the space of partial integral elements,
\[
J_H P:=\{(p,\xi)\in J^1P\mid \xi(T_xM)\subset H_p\},
\]
and the classical prolongation is the space of integral elements,
\[
\operatorname{Prol}(P,H)=\{(p,\xi)\in J_HP\mid \xi^*K_H=0\}
\]
[1901.02084].

These prolongation spaces are not merely formal definitions. The paper proves that \(J_HP\) is smooth and that the projection
\[
\operatorname{pr}:J_HP\to P
\]
is an affine bundle modeled on
\[
\operatorname{Hom}(T^*M,g(H))
\]
[1901.02084]. Moreover, \(\operatorname{Prol}(P,H)\) is a smooth affine subbundle of \(J^1P\) if and only if the torsion vanishes and the first prolongation of the symbol bundle has constant rank; in that case its model vector bundle is \(g(H)^{(1)}\), and the projection to \(P\) is the largest normalised prolongation [1901.02084].

Formal integrability is defined by iterating prolongation. In the later terminology, a Pfaffian fibration is \(1\)-integrable if \(P^{(1)}\to P\) is a submersion, \(k\)-integrable if all prolongations up to order \(k\) exist, and formally integrable if it is \(k\)-integrable for all \(k\) [2510.01517]. The foundational paper gives a recursive obstruction theory in terms of higher torsions and symbol prolongations, and extends Goldschmidt’s criterion: if the symbol space \(g\) is \(2\)-acyclic,
\[
H^{\ell,2}(g)=0 \quad \forall \ell\ge 0,
\]
and \(g^{(1)}\) is smooth while the first prolongation is surjective, then the Pfaffian fibration is formally integrable [1901.02084].

## 4. Linear theory, relative connections, and linearisation

The linear theory of Pfaffian fibrations is modeled on linear PDEs and Spencer operators. If \(E\to M\) is a vector bundle, a linear Pfaffian fibration may be described either by a pointwise surjective linear form
\[
\theta\in\Omega^1(E,\pi^*F),
\]
or by a linear distribution \(H\subset TE\); there is a \(1\)-to-\(1\) correspondence between the two descriptions [1901.02084].

The form-theoretic description is equivalent to a relative connection
\[
D:\Gamma(E)\to \Omega^1(M,F),
\]
with symbol map \(o:E\to F\), satisfying the Leibniz rule
\[
D(fs)(X)=fD(s)(X)+X(f)\,o(s)
\]
[1901.02084]. In the classical jet-bundle case, this recovers the Spencer operator, so linear Pfaffian fibrations supply the intrinsic language behind standard linear PDE operators.

Linearisation along a holonomic section provides the nonlinear-to-linear passage. Given a Pfaffian fibration \((P,\theta)\) and a holonomic section \(b\), the linearisation is
\[
\big(\operatorname{Lin}_b(P,\theta),D_b\big),
\qquad
\operatorname{Lin}_b(P,\theta)=b^*T^\pi P,
\]
where \(D_b\) is a relative connection obtained by differentiating families of sections through \(b\) [1901.02084]. For a linear Pfaffian fibration, linearisation at the zero section recovers the original linear object, and for jet bundles the construction recovers the classical Spencer operator [1901.02084].

This identifies the linearised Pfaffian geometry of a solution with the familiar infinitesimal theory of PDEs. A plausible implication is that the Pfaffian-fibration formalism is designed not only to encode solution spaces intrinsically, but also to preserve the standard deformation-theoretic structures used in the analysis of formal integrability.

## 5. Relative algebroids and the formal equivalence theorem

A principal development in the later literature is the construction of a canonical relative algebroid underlying any Pfaffian fibration. Starting from a Pfaffian fibration \((P,C,\pi)\), the relevant geometric data are the distribution \(C\) and the foliation
\[
C^\pi=C\cap \ker T\pi
\]
[2510.01517]. The associated vector bundle is
\[
B:=\pi^*TX\to P,
\]
which carries a canonical flat \(C^\pi\)-connection, and from the geometry of \(C\) one obtains a relative derivation
\[
{}^C\in \Gamma\!\left(D^1_{(\pi^*TX,\,)}\right)
\]
[2510.01517]. The construction is described concretely by
\[
{}^C\alpha=\Pi\big(I(\alpha)\big)
\]
for flat forms \(\alpha\), where \(I\) and \(\Pi\) are the canonical maps induced by the splitting of the normal bundle [2510.01517]. The resulting statement is that a Pfaffian fibration \((P,C,\pi)\) gives rise to a relative algebroid
\[
(\pi^*TX,\overline\nabla,{}^C)
\]
over the foliation \(C^\pi\) [2510.01517].

The main structural result is that Pfaffian fibrations and their underlying relative algebroids have the same formal theory. The theorem states that the partial prolongation spaces of \((P,C,\pi)\) and \((\pi^*TX,\, ,{}^C)\) are canonically isomorphic, hence the Pfaffian fibration is \(1\)-integrable if and only if the relative algebroid is; that their tableau maps are canonically identified; that if \((P,C,\pi)\) is \(1\)-integrable then the relative algebroid underlying \((P^{(1)},C^1,\pi^{(1)})\) is canonically isomorphic to the prolongation of the original relative algebroid; and that germs of holonomic sections are in one-to-one correspondence with germs of realizations of the relative algebroid, modulo diffeomorphism [2510.01517].

The theorem’s content is explicit: the spaces of first-order admissible jets coincide, the symbol or tableau data coincide, the prolongation towers coincide when defined, and the local solutions coincide up to the expected equivalence on the algebroid side [2510.01517]. This shows that the relative algebroid is not merely an associated invariant. It captures the same formal PDE data as the original Pfaffian fibration.

Naturality is also built into the construction. A morphism of Pfaffian fibrations
\[
(\Phi,\varphi):(P_1,C_1,\pi_1)\to(P_2,C_2,\pi_2)
\]
satisfying
\[
T\Phi(C_1)\subseteq C_2
\]
induces a bundle map
\[
\Phi_*:\pi_1^*TX_1\to \pi_2^*TX_2
\]
which is a morphism of relative algebroids [2510.01517]. This establishes a functorial passage from Pfaffian fibrations to relative algebroids.

## 6. Symmetries, Pfaffian groupoids, and terminological scope

The symmetry theory distinguishes two notions. Internal symmetries are local diffeomorphisms preserving the Cartan distribution,
\[
(P,C):=\{\varphi\in (P)\mid T\varphi(C)\subseteq C\},
\]
whereas Pfaffian symmetries are local diffeomorphisms preserving both \(C\) and its vertical part,
\[
(P,C,\pi):=\left\{\varphi\in(P)\mid T\varphi(C)\subseteq C,\; T\varphi(C^\pi)\subseteq C^\pi\right\}
\]
[2510.01517]. The distinction is essential because a Pfaffian symmetry need not preserve \(\ker T\pi\), so it need not be a morphism of Pfaffian fibrations over the base [2510.01517].

Prolongation links the two notions. An internal symmetry \(\varphi\) prolongs to a local map
\[
\varphi^{(1)}:J^1_C P\dasharrow J^1_C P,
\]
and, if \(P^{(1)}\) is smooth, to
\[
\varphi^{(1)}:P^{(1)}\dasharrow P^{(1)}
\]
[2510.01517]. The main theorem on this point states that if \(\varphi\in(P,C)\), then \(\varphi^{(1)}\) is a Pfaffian symmetry of \(J^1_C P\); if \(P\) is \(1\)-integrable, then \(\varphi^{(1)}\) is also a Pfaffian symmetry of \(P^{(1)}\); any Pfaffian symmetry of \(J^1_C P\) is locally the prolongation of an internal symmetry of \((P,C,\pi)\); and if \(P\) is \(1\)-integrable and the tableau map is involutive, then every Pfaffian symmetry of \(P^{(1)}\) is locally the prolongation of an internal symmetry of \((P,C,\pi)\) [2510.01517].

The symmetry pseudogroups are organized into Pfaffian groupoids. The internal symmetry groupoid is
\[
G_C=\{j^1_x\varphi\in J^1(P)\mid T_x\varphi(C_x)\subseteq C_{\varphi(x)}\},
\]
and the Pfaffian symmetry groupoid is
\[
G_{C,\pi}=\{j^1_x\varphi\in J^1(P)\mid T_x\varphi(C_x)\subseteq C_{\varphi(x)},\; T_x\varphi(C^\pi_x)\subseteq C^\pi_{\varphi(x)}\}
\]
[2510.01517]. These inherit the Cartan distribution \(H^1\) from \(J^1(P)\), and the canonical representation
\[
G^{(1)}_{C,\pi}\curvearrowright D^1_{(\pi^*TX,\,)}
\]
leaves the derivation \({}^C\) invariant [2510.01517]. In particular, Pfaffian symmetries induce local isomorphisms of the underlying relative algebroid and preserve both the flat connection and the derivation [2510.01517].

The framework extends to actions of arbitrary Pfaffian groupoids \((G,H)\) on a Pfaffian fibration \((P,C,\pi)\) with moment map \(\mu:P\to M\). An action is by internal symmetries if
\[
Tm\left(H\lefttensor[_{Ts}]{\times}{_{T\mu} C\right)\subseteq C,
\]
and by Pfaffian symmetries if, in addition,
\[
Tm\left(H\lefttensor[_{Ts}]{\times}{_{T\mu} C^\pi\right)\subseteq C^\pi
\]
[2510.01517]. If \((G,H)\curvearrowright(P,C,\pi)\) is an action by Pfaffian symmetries and \((G,H)\) is \(1\)-integrable, then the induced action
\[
(G^{(2)},G^{(1)})\curvearrowright(\pi^*TX,\, ,{}^C)
\]
is by symmetries of relative algebroids [2510.01517]. The paper identifies applications in particular to partial differential equations with Lie pseudogroup symmetries [2510.01517].

A recurrent source of confusion is terminological. Several other works use the word “Pfaffian” in unrelated senses: for example, one paper explicitly states that it does not use the phrase “Pfaffian fibration” in the differential-geometric sense, even though it studies a fibration of odd orthogonal Grassmannians whose degeneracy classes are governed by Pfaffian formulas [1602.04448]; another emphasizes that its “Pfaffian structure” concerns random-matrix partition functions rather than a geometric fibration in the differential-geometric sense [2205.02527]. In the PDE literature, by contrast, a Pfaffian fibration means precisely the intrinsic geometric object defined by a fibration together with a Pfaffian distribution or form satisfying transversality and involutivity conditions [1901.02084].

The resulting picture is that Pfaffian fibrations form a bridge between nonlinear PDEs, prolongation theory, Spencer-type linearisation, relative algebroids, and Lie pseudogroup symmetry. This suggests that their role is not merely terminological but foundational: they provide the intrinsic object on which formal integrability, infinitesimal structure, and symmetry reduction can all be formulated in a common language.

Source: https://www.emergentmind.com/topics/pfaffian-fibrations