---
title: Flux Homomorphism in Symplectic Topology
url: https://www.emergentmind.com/topics/flux-homomorphism
type: topic
---

# Flux Homomorphism in Symplectic Topology

The flux homomorphism is a cohomological invariant attached to symplectic isotopies. For a symplectic path it is represented by the closed \(1\)-form obtained by integrating \(\iota_{X_t}\omega\) along the isotopy, and for a Lagrangian isotopy it measures the symplectic areas of the cylinders swept by loops in the Lagrangian. In symplectic topology it functions as the basic obstruction separating symplectic from Hamiltonian isotopies, and it also enters the study of hyperbolic surfaces and Anti-de Sitter geometry, \(C^0\)-rigidity questions, Novikov-Floer theory, and Lagrangian deformation theory [1309.1325, 1804.02044, 1712.02413].

## 1. Classical definition and exactness properties

Let \((M,\omega)\) be a closed symplectic manifold, let \(G=\mathrm{Symp}_0(M,\omega)\), and let \(\{\phi_t\}_{t\in[0,1]}\subset G\) be a symplectic isotopy with \(\phi_0=\mathrm{id}\). If \(X_t\) is defined by
\[
\frac{d\phi_t}{dt}=X_t\circ \phi_t,
\]
then \(\mathcal L_{X_t}\omega=d(\iota_{X_t}\omega)=0\), so \(\iota_{X_t}\omega\) is closed. The classical flux homomorphism is therefore
\[
\mathrm{Flux}\bigl([\{\phi_t\}]\bigr)=\left[\int_0^1 \iota_{X_t}\omega\,dt\right]\in H^1(M;\mathbb R),
\]
defined on the universal cover \(\widetilde{\mathrm{Symp}}_0(M,\omega)\). For loops in \(G\), one obtains a group homomorphism \(\mathrm{Flux}:\pi_1(G)\to H^1(M;\mathbb R)\), and its image \(\Gamma=\mathrm{Flux}(\pi_1(G))\) is the flux group. Banyaga’s theorem identifies \(\mathrm{Ham}(M,\omega)\) as exactly the kernel of \(\mathrm{Flux}\) [1309.1325].

On a closed connected oriented surface \((S,\omega)\) of genus at least two, the situation simplifies because \(\pi_1(\mathrm{Symp}_0(S,\omega))=0\). In that case the flux descends to
\[
\mathrm{Flux}_\omega:\mathrm{Symp}_0(S,\omega)\longrightarrow H^1(S;\mathbb R),
\qquad
\mathrm{Flux}_\omega(\phi)=\left[\int_0^1 i_{X_t}\omega\,dt\right],
\]
independently of the chosen isotopy from \(\mathrm{id}\) to \(\phi\). In the formulation used for closed hyperbolic surfaces, this yields the short exact sequence
\[
1\to \mathrm{Ham}(S,\Omega)\to \mathrm{Symp}_0(S,\Omega)\xrightarrow{\mathrm{Flux}} H^1(S;\mathbb R)\to 1,
\]
so the vanishing of flux is equivalent to Hamiltonianity [2102.12161, 1712.02413].

A closely related geometric interpretation arises for Lagrangian isotopies. If \(\{L_t\}\) is a Lagrangian isotopy and \(\alpha_0\subset L_0\) represents \(a\in H_1(L_0;\mathbb R)\), then the trace of \(\alpha_0\) under the isotopy defines a \(2\)-chain \(C_\alpha\), and
\[
\mathrm{Flux}(\{L_t\})(a)=\int_{C_\alpha}\omega.
\]
This realizes flux as the symplectic area swept by the isotopy [1804.02044].

## 2. Hyperbolic surfaces, the mod \(2\pi\) refinement, and Anti-de Sitter geometry

For two hyperbolic metrics \(h,h'\) on a closed surface \(S\), with area forms \(\Omega_h,\Omega_{h'}\), any symplectomorphism
\[
\psi:(S,\Omega_h)\to (S,\Omega_{h'})
\]
determines a bundle isomorphism
\[
b\in \Gamma(\mathrm{Isom}(TS,\psi^*h',h)),\qquad \det b=1,
\]
such that \(\psi^*h'=h(b\cdot,b\cdot)\). If \(\nabla\) and \(\nabla'\) are the Levi-Civita connections of \(h\) and \(\psi^*h'\), and if one chooses local orthonormal frames with connection \(1\)-forms \(\omega\) and \(\omega'\), then
\[
\eta_{\psi,b}=\omega'-\omega
\]
is closed. Changing \(b\) by an \(h\)-rotation changes \([\eta_{\psi,b}]\) by an element of \(H^1(S;2\pi\mathbb Z)\), so one obtains a well-defined class
\[
C_{h,h'}(\psi):=[\eta_{\psi,b}]\in H^1(S;\mathbb R)/H^1(S;2\pi\mathbb Z).
\]
When \(h'=h\), the map \(C_{h,h}\) is a Lie-group homomorphism on \(\mathrm{Symp}_0(S,\Omega_h)\), and it agrees with the classical flux modulo \(2\pi\):
\[
\pi\circ \mathrm{Flux}=C_{h,h},
\]
where \(\pi:H^1(S;\mathbb R)\to H^1(S;\mathbb R)/H^1(S;2\pi\mathbb Z)\) is the quotient map [1712.02413].

This construction is used in the AdS\(^3\) setting. If
\[
\rho=(\rho_l,\rho_r):\pi_1(S)\to \mathrm{PSL}_2(\mathbb R)\times \mathrm{PSL}_2(\mathbb R)
\]
is a pair of Fuchsian representations and \(\Sigma\hookrightarrow \mathrm{AdS}^3\) is a smooth spacelike \(\rho\)-invariant surface of negative curvature, then its Gauss map identifies \(\Sigma\) with a Lagrangian submanifold \(\Lambda_\Sigma\subset \mathbb H^2\times \mathbb H^2\). Under the curvature hypothesis \(K<0\), both projections to \(\mathbb H^2\) are diffeomorphisms, giving an equivariant diffeomorphism \(\widetilde\phi_\Sigma:\mathbb H^2\to \mathbb H^2\) descending to
\[
\phi_\Sigma:(S,\Omega_{h_l})\to (S,\Omega_{h_r}),
\]
a symplectomorphism isotopic to the identity. The paper proves that \(\phi_\Sigma\) admits a section \(b\) with
\[
\det b=1,\qquad d^{\nabla_l}b=0,\qquad b\neq -\mathrm{Id},
\]
and from the alternate formula for \(\eta_{\psi,b}\) it follows that \(\eta_{\phi_\Sigma,b}=0\), hence
\[
C_{h_l,h_r}(\phi_\Sigma)=0.
\]
The same vanishing holds for the unique minimal Lagrangian diffeomorphism from \((S,h_l)\) to \((S,h_r)\) isotopic to the identity. By additivity of \(C_{h,h'}\) under composition and by a deformation argument through invariant spacelike surfaces, one obtains the decomposition
\[
\phi_\Sigma=\phi\circ \psi,
\]
where \(\phi\) is the unique minimal Lagrangian diffeomorphism and \(\psi\in \mathrm{Symp}_0(S,\Omega_{h_l})\) is Hamiltonian. In this framework the flux modulo \(2\pi\) is the obstruction measuring the deviation from minimal Lagrangian behavior [1712.02413].

## 3. \(C^0\)-closure, continuity, and restrictions on commuting symplectomorphisms

The flux homomorphism is central to the \(C^0\)-flux conjecture. For a closed symplectic manifold \((M,\omega)\), the question “Is \(\mathrm{Ham}(M,\omega)\) \(C^0\)-closed in \(G=\mathrm{Symp}_0(M,\omega)\)?” is equivalent to asking whether the flux group \(\Gamma\) is closed in \(H^1(M;\mathbb R)\). Using the notation \(H^0\subset G\) for the \(C^0\)-closure of \(\mathrm{Ham}(M,\omega)\), \(\Gamma^0=\mathrm{Flux}(H^0)\), and \(\Gamma_{\mathrm{top}}=\mathrm{Flux}(\pi_1(\mathrm{Map}_0(M)))\), Buhovsky proves
\[
\Gamma^0\subset \Gamma_{\mathrm{top}}+\overline{\Gamma_{\mathrm{top}}^e}.
\]
Hence, if \(\Gamma_{\mathrm{top}}=\Gamma\), then \(\Gamma^0=\Gamma\) and the \(C^0\)-flux conjecture holds. The same paper proves continuity of flux in the \(C^0\)-topology on isotopies: there exist constants \(c>0\) and \(C>0\) such that
\[
\max_{t\in[0,1]} d(\phi_t,\mathrm{id})<c
\quad\Longrightarrow\quad
|\mathrm{Flux}(\{\phi_t\})|\le C\, d(\phi,\mathrm{id}),
\]
which implies \(C^0\)-rigidity of Hamiltonian paths [1309.1325].

On closed surfaces of genus \(l\ge 2\), flux also constrains the algebraic structure of \(\mathrm{Symp}_0(S,\omega)\). If \(f,g\in \mathrm{Symp}_0(S,\omega)\) commute, then
\[
\mathrm{Flux}_\omega(f)\smile \mathrm{Flux}_\omega(g)=0\in H^2(S;\mathbb R)\cong \mathbb R.
\]
Equivalently, the natural intersection pairing \(b\) on \(H^1(S;\mathbb R)\) satisfies
\[
b\bigl(\mathrm{Flux}(f),\mathrm{Flux}(g)\bigr)=0.
\]
Therefore the real span of the fluxes of an abelian subgroup is isotropic in the \(2l\)-dimensional symplectic vector space \(H^1(S;\mathbb R)\), and its dimension is at most \(l\). The proof combines a discrete-extension theorem for quasimorphisms with a non-extendability theorem for Py’s Calabi quasimorphism \(\mu_P\) [2102.12161].

A further dynamical application appears in Floer-Novikov theory. For a nondegenerate symplectic isotopy \(\phi=\{\phi_t\}\) on a closed monotone \((M,\omega)\), the flux class \([\alpha_\phi]\in H^1(M;\mathbb R)\) determines the integration cover \(\widetilde M\to M\), the Novikov ring \(\Lambda_{[\alpha]}\), and the action filtration. If \([\alpha_\phi]\) is “small,” namely \(\lambda<1/2A\) in the notation of the paper, then energy-depth estimates force properness of evaluation maps on the relevant Floer moduli spaces. Under this hypothesis the evaluation homomorphism
\[
\mathrm{ev}:\mathrm{FloerLoop}(\phi,J)\longrightarrow \pi_1(\widetilde M,[\alpha])
\]
is surjective, so the Floer moduli spaces associated to closed trajectories generate the Novikov fundamental group [2111.06151].

## 4. Boundary phenomena, non-orientability, and Euler classes

For surfaces with boundary, the target of flux may become relative cohomology or even \(\mathbb R\). On the closed unit disk \(D\subset \mathbb R^2\) with \(\omega=dx\wedge dy\), let
\[
G_{\mathrm{rel}}=\{g\in \mathrm{Diff}(D)\mid g^*\omega=\omega,\ g(0)=0,\ g|_{\partial D}=\mathrm{id}\}.
\]
With
\[
n=\frac{x\,dy-y\,dx}{2},\qquad dn=\omega,
\]
and the singular \(1\)-chain \(y(t)=(t,0)\), one defines
\[
\mathrm{Flux}_{\mathbb R}(g):=-\int_y (g^*n-n).
\]
This is a surjective group homomorphism \(G_{\mathrm{rel}}\to \mathbb R\), invariant under conjugation by the larger group \(G=\{g\in \mathrm{Diff}(D)\mid g^*\omega=\omega,\ g(0)=0\}\). Writing \(K=\ker \mathrm{Flux}_{\mathbb R}\), one obtains
\[
1\to K\to G_{\mathrm{rel}}\xrightarrow{\mathrm{Flux}_{\mathbb R}} \mathbb R\to 0,
\]
and the quotient \(E=G/K\) is a central \(\mathbb R\)-extension
\[
0\to \mathbb R\to E\to \mathrm{Diff}^+(S^1)\to 1.
\]
Its Euler class is the real-valued Euler class \(e_{\mathbb R}\), and the same class is represented by the Ismagilov-Losik-Michor \(2\)-cocycle after restriction to the relevant symplectomorphism group [1905.08029].

A twisted variant exists for non-orientable compact surfaces with one boundary component. Let \(N\) be such a surface, let \(L=L_N\) be its orientation line bundle, and let \(\omega\in \Omega^2(N;L)\) be an everywhere-positive twisted \(2\)-form. Since \(\omega\) is exact, one may choose \(\eta\in \Omega^1(N;L)\) with \(d\eta=\omega\). For
\[
G(N)=\mathrm{Diff}_\omega(N,\mathrm{near}\ \partial N)_0,
\]
the flux homomorphism is
\[
\mathrm{Flux}:G(N)\to H^1(N,\partial N;L),
\qquad
\mathrm{Flux}(g)=[\eta-g^*\eta].
\]
For a closed ordinary \(1\)-form \(\lambda\), the associated scalar flux is
\[
\mathrm{Flux}_\lambda(g)=\int_N (\eta-g^*\eta)\wedge \lambda.
\]
If \(\{g_t\}\) is an isotopy from \(\mathrm{id}\) to \(g\), then
\[
\mathrm{Flux}(g)=\left[\int_0^1 i_{X_t}\omega\,dt\right]\in H^1(N,\partial N;L),
\qquad
\mathrm{Flux}_\lambda(g)=\int_0^1\int_N i_{X_t}\omega\wedge \lambda\,dt.
\]
The boundary restriction map \(p:G(N)\to \mathrm{Diff}_0(\partial N)\) fits into
\[
1\to \mathrm{Diff}_\omega(N,\partial N)_0\to G(N)\to \mathrm{Diff}_0(S^1)\to 1,
\]
and the transgression of \(\mathrm{Flux}_\lambda\) is a nonzero multiple of the Euler class. The kernel \(K=\ker \mathrm{Flux}\) is simple, which implies that no Calabi-type homomorphism exists on \(K\) in this non-orientable setting [2508.12874].

## 5. Lagrangian flux, star-isotopies, and Floer-theoretic geometry

Solomon introduces a Lagrangian analogue of the flux homomorphism for a path
\[
\Lambda:[0,1]\to \mathrm{Lag}(X,L,d),
\qquad
t\mapsto \Lambda_t,
\]
of compactly supported oriented Lagrangian submanifolds. If \(f_t:L\to X\) is a lift with \(f_t(L)=\Lambda_t\) and \(\ell:S^1\to \Lambda_0\) is a loop, define
\[
\bar\ell(u,t)=f_t\bigl(f_0^{-1}(\ell(u))\bigr).
\]
The Lagrangian flux is
\[
\mathrm{flux}(\Lambda,\ell)=\int_{S^1\times[0,1]} \bar\ell^*\omega.
\]
This depends only on the end-point-preserving homotopy class of the path and on \([\ell]\in H_1(\Lambda_0;\mathbb Z)\), giving a homomorphism from end-point-preserving classes of Lagrangian paths to \(H^1(\Lambda_0;\mathbb R)\). If \(\Lambda_*\) is a basepoint and
\[
G_*=\{\mathrm{flux}([\Lambda])\mid \Lambda_0=\Lambda_1=\Lambda_*\}\subset H^1(\Lambda_*;\mathbb R),
\]
then \(G_*\) is a subgroup. If \(G_*\) is discrete, the \(\mathrm{Ham}(X,\omega)\)-orbit of \(\Lambda_*\) is \(C^1\)-closed in \(\mathrm{Lag}\). Under the additional hypothesis that \(H^1(X;\mathbb R)\to H^1(\Lambda_*;\mathbb R)\) is surjective, the kernel of \(\mathrm{flux}\) is exactly the set of classes homotopic rel endpoints to exact Lagrangian paths, and one gets a bijection
\[
(\text{path component of }\Lambda_*)/\mathrm{Ham}(X,\omega)
\cong
H^1(\Lambda_*;\mathbb R)/G_*.
\]
In the graph construction \(X=M\times M\), \(\omega=-\pi_1^*\omega_M+\pi_2^*\omega_M\), and \(\Lambda_t=\mathrm{Graph}(\phi_t)\), the Lagrangian flux recovers the classical flux [1209.4737].

The geometric paper on symplectic flux and Lagrangian torus fibrations studies isotopies whose flux grows linearly. A Lagrangian isotopy \(\{L_t\}\) is a star-isotopy if
\[
\mathrm{Flux}(\{L_s\}_{s\in[0,t]})=t\,\mathrm{Flux}(\{L_s\}_{s\in[0,1]})
\quad\forall t\in[0,1].
\]
For such isotopies the numerical invariant \(\Psi(L_t)\), defined via the Fukaya \(A_\infty\)-algebra, is continuous and concave on \([0,1]\). If \(L\subset X\) is monotone with monotonicity constant \(c>0\), \(U\subset X\) is a Liouville or Weinstein neighborhood of \(L\), and \(\beta\in H_2(X,L)\) has nonzero Maslov-index \(2\) disk count, then any Lagrangian isotopy of \(L\) in \(U\) with total flux \(f\in H^1(L;\mathbb R)\) satisfies
\[
f\cdot \beta>-2c.
\]
Equivalently, the shape \(Sh_L(U)\) lies in the half-space \(\{f:2c+f\cdot \beta>0\}\). In the Fano Gelfand-Cetlin or toric-type case, if \(P_L\) is the Newton polytope of the Landau-Ginzburg potential of a monotone torus fiber \(L\), then the allowed star-flux classes are exactly
\[
Sh^*_L(X)=2c\,(P_L)^\vee.
\]
These statements tie flux to holomorphic disk potentials and to mirror-symmetry structures [1804.02044].

## 6. Abstract analogues and distinct uses of the term

An abstract analogue of flux appears in the Fukaya-category framework. There, a class in \(\mathit{HH}^1(A,A)\) plays the role of a deformation field for families of objects, and periodicity is formulated in terms of perfect families over an affine curve following that field. In this setting the construction recovers the classical flux for loops of symplectic automorphisms. For \(M=T^2\), the relevant periodic class corresponds under mirror symmetry to \(\theta\otimes[dq]\in H^0(\bar S,\Omega^1)\otimes H^1(T^2)\). For symplectic mapping tori and their blowups, the resulting Hochschild-theoretic invariant distinguishes trivial from nontrivial mapping-torus constructions. This suggests a categorical enlargement of flux valued in \(\mathit{HH}^1(\mathit{Fuk}(M))\), probing phenomena not visible to \(\pi_1(\mathrm{Symp})\) or \(H^1(M)\) [1108.0394].

The term also occurs in a distinct C\(^*\)-algebraic setting in loop quantum gravity. For a graph \(\Gamma\) and a finite set of surfaces \(\mathcal S\), one defines a flux group \(G_{\mathcal S,\Gamma}\subset G^{2N}\) generated by path-flux assignments. This group acts on the analytic holonomy algebra \(C_0(\mathcal L_\Gamma)\cong C_0(G^N)\) by left translations,
\[
\alpha:G_{\mathcal S,\Gamma}\to \mathrm{Aut}(C_0(G^N)),
\]
and \(\alpha\) is a point-norm continuous group homomorphism. The resulting C\(^*\)-dynamical system gives rise to the holonomy-flux cross-product algebra
\[
C_0(\mathcal L_\Gamma)\rtimes_\alpha G_{\mathcal S,\Gamma}.
\]
This is a separate usage of “flux homomorphism,” but it preserves the common structural theme of encoding surface data by a group action on an algebra of observables [1108.4579].

Source: https://www.emergentmind.com/topics/flux-homomorphism