---
title: Covariant Phase-Space Method
url: https://www.emergentmind.com/topics/covariant-phase-space-method-1b4ba6cb-1ee8-4964-8998-ae3f58cd94c3
type: topic
---

# Covariant Phase-Space Method

The covariant phase-space method is a manifestly covariant construction of the symplectic structure and conserved charges of a Lagrangian field theory from the space of its solutions. In its standard form, one starts from a spacetime manifold \(\mathcal M\), a Lagrangian \(d\)-form \(L[\Phi]\), and the decomposition
\[
\delta L = E_I[\Phi]\,\delta\phi^I + d\,\Theta[\delta\Phi;\Phi],
\]
where \(E_I=0\) are the equations of motion and \(\Theta\) is the presymplectic potential. The presymplectic current is the field-space exterior derivative of \(\Theta\),
\[
\omega[\delta_1\Phi,\delta_2\Phi;\Phi]
=\delta_1\Theta[\delta_2\Phi;\Phi]-\delta_2\Theta[\delta_1\Phi;\Phi],
\]
and its integral over a Cauchy slice \(\Sigma\) defines the presymplectic form
\[
\Omega[\delta_1\Phi,\delta_2\Phi;\Phi]=\int_\Sigma \omega.
\]
On shell, \(d\omega\approx 0\), so \(\Omega\) is independent of the choice of \(\Sigma\) after quotienting gauge directions. In recent treatments, the formalism is described as acting simultaneously on spacetime and on the infinite-dimensional solution phase space, with parallel geometric structures on both manifolds [2301.12418] [2603.14439].

## 1. Definition and geometric setting

The covariant phase space is the space of all solutions of the Euler–Lagrange equations. In the notation of the gravity-with-boundaries literature,
\[
C=\{\phi\mid E_I(\phi)=0\},
\]
while in the geometric solution-space formulation it is the Solution Phase Space (SPS) \(\mathcal P\), whose points are solutions \(\Phi\equiv\{\phi^I(x)\}\) modulo proper gauge transformations [2301.12418] [2603.14439].

A central structural point is that the method uses two exterior calculi. On spacetime one has \(d\), interior products \(i_\xi\), and Lie derivatives \(\mathcal L_\xi\); on field space one has \(\delta\), interior products \(\mathrm I_{\hat\xi}\), and Lie derivatives \(\mathrm L_{\hat\xi}\). The “fundamental link” is that a spacetime symmetry \(\xi\) induces a field-space vector \(\hat\xi\) via
\[
\hat\xi^I(\Phi)=\delta_\xi\phi^I=\mathcal L_\xi\phi^I.
\]
This parallelism is not merely notational. It underlies the construction of surface charges, the analysis of integrability, and later developments involving connections, torsion, and curvature on solution space [2603.14439].

In the local jet-bundle picture used for theories with boundaries, all bulk and boundary forms live in a bicomplex graded by spacetime degree and field-variation degree. The familiar first-variation formula then becomes the starting point of a bicomplex descent. The covariant phase-space two-form is obtained by pulling the presymplectic current back to the solution space and integrating it over a Cauchy slice. This is the sense in which the method replaces a noncovariant \(3+1\) split by a construction directly on the solution space [2008.01842].

## 2. Boundaries, corners, and the relative bicomplex

For manifolds with timelike boundaries, the boundary-free formalism is insufficient. One must add boundary terms to obtain a well-posed variational principle, and one must track boundary and corner contributions in the presymplectic structure. The relative bicomplex framework does this by replacing an ordinary Lagrangian form with a Lagrangian pair
\[
(L,\ell)\in\Omega^{(n,0)}_{\mathrm{loc}}((M,\partial M)\times F),
\]
or equivalently \(L_{\mathrm{rel}}=(L_{\mathrm{bulk}},L_{\mathrm{bdy}})\), together with a relative differential \(\underline d\) combining the bulk exterior derivative with pullbacks to the boundary [2008.01842] [2301.12418].

In this framework the generalized first-variation formula takes the form
\[
\underline d_V(L,\ell)=(E_I,b_I)\wedge \delta\phi^I+\underline d_H(\Theta^L,\theta^\ell),
\]
where \(E_I\) are the bulk Euler–Lagrange forms, \(b_I\) are the boundary “equations,” and \((\Theta^L,\theta^\ell)\) is the relative presymplectic potential. The relative presymplectic current is then
\[
(\Omega,\bar\omega)=\underline d_V(\Theta^L,\theta^\ell),
\]
and the presymplectic form on a Cauchy pair \((\Sigma,\partial\Sigma)\) is
\[
\Omega_S^\Sigma=\int_{(\Sigma,\partial\Sigma)}(\Omega,\bar\omega)
=\int_\Sigma d_V\Theta^L-\int_{\partial\Sigma} d_V\theta^\ell.
\]
In the formulation of [2301.12418], the same structure appears as
\[
\Omega_{\mathrm{rel}}=\int_\Sigma \omega_{\mathrm{bulk}}+\int_{\partial\Sigma}\omega_{\mathrm{bdy}}.
\]

Three general statements organize the boundary theory. First, any local Lagrangian pair admits a well-defined relative presymplectic potential. Second, the relative presymplectic current is \(\underline d\)-closed on shell, so \(\Omega_{\mathrm{rel}}[\Sigma]\) is independent of \(\Sigma\). Third, infinitesimal gauge symmetries generate conserved charges that depend only on boundary data [2301.12418].

A distinctive claim of the relative bicomplex literature is that it gives a formal equivalence between the relative version of a theory with boundary and the non-relative version of the same theory with no boundary. This is used to transfer bicomplex identities, Cartan formulae, and exactness theorems from the boundaryless setting to the theory with boundary and corners. This also underlies the claim that the framework is “clean and ambiguity-free” for the solution space and its symplectic structure [2008.01842] [2103.06362].

## 3. Symmetries, Noether currents, and Hamiltonian charges

For a vertical symmetry \(X\) satisfying
\[
L_X(L,\ell)=\underline d_H(S_X,s_X),
\]
the relative Noether current pair is
\[
(J_X,\jmath_X):=(S_X,s_X)-\iota_X(\Theta^L,\theta^\ell).
\]
On shell, \(d_H(J_X,\jmath_X)=0\), and the associated charge on a Cauchy pair is
\[
H_X^\Sigma=\int_{(\Sigma,\partial\Sigma)}(J_X,\jmath_X).
\]
For diffeomorphisms generated by \(\xi\), one obtains on shell a Noether current \(J_\xi=dQ_\xi\), so the Hamiltonian generator is a codimension-two boundary integral,
\[
Q_\xi=\int_{\partial\Sigma}Q_\xi,
\]
a statement emphasized in the relative-bicomplex treatment of gravity with timelike boundaries [2301.12418].

The boundary literature also gives an algorithmic derivation of the Hamiltonian in the presence of spatial boundaries. Starting from an action
\[
S[\phi]=\int_M L[\phi]+\int_{\partial M}\ell[\phi],
\]
one demands that, at solutions of the equations of motion, the action is stationary modulo future/past boundary terms under arbitrary variations obeying the spatial boundary conditions. From this, both the symplectic structure and the Hamiltonian for any diffeomorphism that preserves the theory are constructed, and the term denoted \(B\) in the original literature emerges naturally. The same procedure produces an additional boundary term \(C\), which can already be nonvanishing in general relativity with sufficiently permissive boundary conditions [1906.08616].

A related issue is ambiguity in the presymplectic potential under shifts \(\Theta\to\Theta+d\alpha\). For subregions, this produces explicit boundary ambiguities in the naive prescription \(\Omega=\int_\Sigma\omega\). A path-integral derivation resolves this by replacing the hypersurface integral with a contour integral around a partial Cauchy surface,
\[
\Omega(\delta_1\phi,\delta_2\phi)=\int_{\partial\mathcal U}\omega(\delta_1\phi,\delta_2\phi),
\]
where \(\partial(\partial\mathcal U)=\varnothing\). This makes the symplectic form unambiguous and implies that all gauge transformations, “large” or “small,” lie in the kernel of \(\Omega\) for the subregion construction [1901.09857].

The same boundary analysis is also the setting in which the Poisson bracket on covariant phase space is shown to coincide directly with the Peierls bracket, without non-covariant intermediate steps [1906.08616].

## 4. Gravitational realizations: metric, tetrad, and Palatini formulations

In gravity, the method is often presented through the comparison of metric and tetrad formulations on manifolds with boundary. For metric Einstein–Hilbert plus Gibbons–Hawking–York,
\[
L_{\mathrm{bulk}}=(R-2\Lambda)\,\mathrm{vol}_g,\qquad
L_{\mathrm{bdy}}=-2K\,\mathrm{vol}_\gamma.
\]
The presymplectic potential is
\[
\theta_{\mathrm{bulk}}
=(g^{\mu[\alpha}g^{\beta]\nu}\nabla_\nu\delta g_{\alpha\beta})\,\mathrm{vol}_\mu,
\qquad
\theta_{\mathrm{bdy}}=0.
\]
For the tetrad formulation with fields \((e^I,\omega^{IJ})\),
\[
L_{\mathrm{bulk}}
=\tfrac12\,\epsilon_{IJKL}(F^{IJ}-\Lambda\,e^I e^J)e^K e^L,
\]
\[
L_{\mathrm{bdy}}
=-\epsilon_{IJKL}N^I D N^J e^K e^L,
\]
and one finds
\[
\theta_{\mathrm{bulk}}
=\tfrac12\,\epsilon_{IJKL}\,e^I e^J\,\delta\omega^{KL},
\qquad
\theta_{\mathrm{bdy}}
=\epsilon_{IJKL}\,e^I e^J N^K \delta N^L.
\]
A direct pull-back argument gives, off shell,
\[
(\theta_{\mathrm{bulk,tetrad}},\theta_{\mathrm{bdy,tetrad}})
=\Phi^*(\theta_{\mathrm{bulk,metric}},\theta_{\mathrm{bdy,metric}})+d(\cdots),
\]
and hence on shell
\[
\Omega_{\mathrm{tetrad}}=\Phi^*\Omega_{\mathrm{metric}}.
\]
The same pattern holds for Palatini versus Einstein–Hilbert variables, yielding on shell
\[
\Omega_{\mathrm{Palatini}}=\pi^*\Omega_{\mathrm{EH}}.
\]
In the tetrad description, the only extra degeneracies are those along internal Lorentz rotations in \(\ker\Phi\) [2301.12418] [2103.06362].

This equivalence result addresses a long-standing comparison problem between metric and tetrad covariant phase spaces with boundaries. In the cohomological formulation, the two presymplectic structures differ only by relative exact terms, which vanish after integration over a Cauchy pair by the relative Stokes theorem [2103.06362].

## 5. Integrability, slicing dependence, and flux

Surface-charge integrability is one of the most technically sensitive parts of the method. In the standard expression,
\[
\slashed\delta Q(\xi)=-\,\mathrm I_{\hat\xi}\Omega
=\oint\Bigl[\delta(i_{\hat\xi}\Theta)-\mathrm L_{\hat\xi}\Theta\Bigr],
\]
\(\slashed\delta Q(\xi)\) need not be exact. The Wald–Zoupas criterion requires \(\delta\,\slashed\delta Q(\xi)=0\), or equivalently \(\delta\beta_a=0\) for a chosen slicing. A central result of the recent geometric SPS analysis is that this criterion is inherently slicing-dependent. The Frobenius theorem on SPS then yields a slicing-independent condition:
\[
\delta\beta_a\wedge \beta_1\wedge\cdots\wedge\beta_P=0,
\]
necessary and sufficient for the existence of an integrable slicing [2603.14439].

This geometric reformulation introduces an SPS connection, torsion, and curvature. In Cartan form,
\[
\delta\beta_a=\mathscr A_a{}^b\wedge\beta_b+\mathscr F_a,\qquad
\mathscr R_a{}^b=\delta\mathscr A_a{}^b-\mathscr A_a{}^c\wedge\mathscr A_c{}^b.
\]
Within this language, “fake” flux is encoded in the pure-gauge connection \(\mathscr A_a{}^b\), while “genuine” flux is measured by non-vanishing torsion \(\mathscr F_a\neq 0\). The latter is slicing-independent and directly proportional to the physical gravitational News tensor or bulk symplectic flux at the boundary. The same framework gives a boundary Liouville theorem when \(\mathscr F=0\) and classifies theories according to whether torsion and curvature vanish [2603.14439].

A different resolution of non-integrability appears when the boundary itself is allowed to fluctuate. Using a step-function reformulation on a boundary-less manifold, one obtains a total presymplectic potential
\[
\Theta^\mu\equiv \Theta^\mu_{\mathrm{LW}}+\chi^\mu \mathcal L
\]
and a symplectic form with both bulk and corner pieces. In this setup, all surface charges become integrable once \(\chi\neq 0\), and the charge takes the form
\[
Q(\xi)=-\int_S x_{\mu\nu}Q_N^{\mu\nu}(\xi)+Q_{\mathcal R}(\xi).
\]
The same formalism yields a charge algebra with a possible central term, and a flux law \(Q_2-Q_1=-\mathfrak F_\xi\), with nonzero \(\mathfrak F_\xi\) encoding radiation memory [2407.03259].

These developments suggest that some charge non-integrability previously attributed directly to radiation can instead arise from solution-space slicing choices, while other instances correspond to genuine boundary flux. The distinction is explicit only in the newer SPS geometry [2603.14439].

## 6. Representative extensions and applications

The method has become a standard tool for boundary-sensitive problems in gravity and gauge theory. In asymptotically locally AdS spacetimes with non-smooth conformal boundaries, a covariant phase-space treatment with relaxed boundary conditions and codimension-two corner terms matches holographic Noether charges to Wald Hamiltonians. For accelerating, non-rotating dyonic AdS\(_4\) black holes, this yields the first law with the “thermodynamic length” terms associated with cosmic-string tensions [2306.16187].

In BRST language, the formalism has been extended to a trigraded BRST covariant phase space. Starting from a BRST-invariant gauge-fixed Lagrangian, one obtains a BRST-invariant presymplectic form whose graded canonical brackets reduce to the classical covariant phase-space brackets when ghosts, antighosts, and antifields are set to zero. In that framework, the global BRST Noether charges are equivalent to the usual classical corner charges of large gauge transformations, and a unified Ward identity separates into the BRST–BV quantum master equation in the bulk and a boundary identity for large gauge transformations [2405.18898].

A semi-classical generalization replaces classical matter by a quantum state and defines a semi-classical symplectic two-form as the sum of the gravitational symplectic form and the Berry curvature of the matter state,
\[
F=\Omega_{\mathrm{grav}}+f.
\]
This two-form is slice-independent and satisfies a quantum generalization of the Hollands–Iyer–Wald identity. For gauge-invariant subregions, the quantum contribution is replaced by the Berry curvature of special purifications involving the Connes cocycle [2510.19939].

The method has also been used to analyze boundary-supported phase spaces in string theory. For tensile open strings in a constant Kalb–Ramond background, the presymplectic current splits into a bulk kinetic term plus an exact boundary term, and imposing mixed boundary conditions yields a boundary symplectic form whose inverse reproduces the Seiberg–Witten noncommutativity parameter. In the tensionless limit with constant \(B\)-field, the bulk contribution vanishes and the symplectic form localizes entirely on the endpoints, so the physical phase space becomes purely boundary-supported [2604.13163].

Across these applications, the recurrent theme is that the covariant phase-space method is not merely a recipe for \(\Omega=\int_\Sigma\omega\). In its contemporary form, it is a boundary-aware, cohomological, and increasingly geometric framework for presymplectic structures, Noether currents, Hamiltonians, fluxes, and charge algebras in diffeomorphism-invariant theories [2008.01842].

Source: https://www.emergentmind.com/topics/covariant-phase-space-method-1b4ba6cb-1ee8-4964-8998-ae3f58cd94c3