---
title: Manifestly Covariant Canonical Operator Formalism
url: https://www.emergentmind.com/topics/manifestly-covariant-canonical-operator-formalism
type: topic
---

# Manifestly Covariant Canonical Operator Formalism

Searching arXiv for papers on manifestly covariant canonical/operator formalisms and closely related frameworks.
Search results from arXiv:
- "Manifestly Covariant Canonical Formalism of Quadratic Gravity" [2505.09149]
- "Covariant Canonical Quantization" [1907.00645]
- "Poisson Bracket and Symplectic Structure of Covariant Canonical Formalism of Fields" [1703.06718]
- "Generators of local gauge transformations in the covariant canonical formalism of fields" [1909.06779]
- "Linking Covariant and Canonical General Relativity via Local Observers" [1206.0658]
- "Canonical analysis of Holst action without second-class constraints" [1911.09690]
- "Quantum Conformal Gravity" [2311.09582]
- "Covariant operator formalism for higher derivative systems: Vector spin-0 dual model as a prelude to generalized QED4" [2404.11805]
- "Duality and Self-duality of the Spin-1 Model in the Covariant Operator Formalism" [1906.07742]
- "A manifestly Lorentz covariant, interacting and non-commutative Dirac equation" [1512.04253]
Manifestly covariant canonical operator formalism designates a family of canonical and operator-based constructions in which spacetime covariance is kept explicit rather than being traded for an initial \(3+1\) split. In these constructions, canonical variables, brackets, constraints, and operator equations are reformulated so that Lorentz covariance, gauge covariance, or diffeomorphism covariance remains manifest, while canonical machinery is recovered through graded symplectic forms, covariant phase spaces, extended Hilbert spaces, BRST charges, Lorentz-covariant connection variables, or modified operator products. The resulting literature is not a single uniform formalism but a cluster of related programs that share the aim of reconciling canonical structure with manifest covariance [1703.06718].

## 1. Defining scope and recurrent structural ideas

A central motivation is the observation that standard Hamiltonian formulations privilege a time variable and may therefore obscure spacetime symmetries. One response is to retain all dynamical objects as spacetime fields and encode canonical structure covariantly. In the covariant canonical formalism of fields, the basic variables are differential forms and the canonical equations are written directly on spacetime, without gauge fixing or Dirac brackets, while remaining Lorentz-, gauge-, or diffeomorphism-covariant [1703.06718]. In the observer-field approach to general relativity, the same aim is pursued differently: all fields remain spacetime fields, but an observer field distinguishes “spatial” and “temporal” components in a way that is still covariant under \(\mathrm{Diff}(M)\) and local \(SO(3,1)\) transformations [1206.0658].

A second recurrent idea is that the primitive commutator need not be an equal-time one. In covariant canonical quantization of a scalar field, the fundamental operator relation is the covariant commutator
\[
[\Phi(x),\Phi(y)] = i\,\Delta(x-y),
\]
with the Pauli–Jordan function \(\Delta\), and equal-time commutators appear only as a derived or optional specialization [1907.00645]. In BRST/Kugo–Ojima–Nakanishi operator formalisms, by contrast, equal-time canonical brackets are retained, but the total Hilbert space and subsidiary conditions are arranged so that covariance survives gauge fixing and indefinite metric sectors are controlled cohomologically [1906.07742].

A third theme is that manifest covariance does not imply the disappearance of canonical variables; it changes their presentation. In Lorentz-covariant canonical gravity, one can work with covariant connection–momentum pairs and first-class constraints only, imposing the time gauge only at the end if one wishes to recover Ashtekar–Barbero variables [1911.09690]. In non-commutative Dirac theory, manifest covariance is restored not by abandoning operator methods but by modifying the operator product itself so that Lorentz generators act as derivations on field products [1512.04253]. This suggests that the expression “manifestly covariant canonical operator formalism” is best understood as a methodological family rather than a unique axiomatic system.

## 2. Differential-form phase space, graded brackets, and symplectic structure

One major branch of the subject replaces ordinary canonical coordinates by differential forms on an oriented \(n\)-manifold \(M\). The algebra \(A(M)=\bigoplus_{k=0}^n A^k(M)\) of differential forms is treated as a \(\mathbb{Z}_2\)-graded commutative algebra, and the covariant phase space is identified with the ringed space \((M,\mathcal{O}_M)\), where \(\mathcal{O}_M(U)=A(U)\) for open \(U\subset M\) [1703.06718]. Canonical coordinates are themselves forms,
\[
\phi^a,\quad p_a,\quad \psi^\alpha,\quad \tau_\alpha,
\]
with parities determined by the spacetime dimension.

Starting from a Lagrangian \(n\)-form \(L\), one defines the conjugate forms by
\[
p_a := \frac{\partial L}{\partial(d\phi^a)},\qquad \tau_\alpha := \frac{\partial L}{\partial(d\psi^\alpha)},
\]
and the Hamiltonian \(n\)-form
\[
H = d\phi^a\wedge p_a + d\psi^\alpha\wedge \tau_\alpha - L.
\]
The covariant canonical equations then take the form
\[
d\phi^a = \frac{\delta H}{\delta p_a},\qquad dp_a = -\frac{\delta H}{\delta \phi^a},\qquad
d\psi^\alpha = \frac{\delta H}{\delta \tau_\alpha},\qquad d\tau_\alpha = -\frac{\delta H}{\delta \psi^\alpha},
\]
or, equivalently, \(dF=-\{H,F\}\) for \(F\in A(M)\) [1703.06718].

The symplectic structure is encoded in the closed, non-degenerate 2-form
\[
\Omega = -\,d\phi^a\wedge d p_a - d\psi^\alpha\wedge d\tau_\alpha.
\]
Using \(df=i_{X_f}\Omega\) to define Hamiltonian vector fields, one obtains the graded Poisson bracket
\[
\{f,g\}:= -\,i_{X_f}i_{X_g}\Omega = -\,L_{X_f}g.
\]
Its parity is \(|\{f,g\}|=|f|+|g|+n+1 \pmod 2\), so the bracket is odd for even \(n\) and even for odd \(n\) [1703.06718]. The standard one-dimensional Poisson bracket is recovered by restricting to 0-forms and \(n=1\).

The gauge-theoretic extension of this formalism identifies local gauge generators directly in the covariant bracket language. For fields \(\psi^A\in\Omega^p(M)\) with conjugates \(\pi_A\), the Hamiltonian form is
\[
H = d\psi^A\wedge \pi_A - L(\psi,d\psi),
\]
and the canonical equations become
\[
d\psi^A = -\,\frac{\partial H}{\partial \pi_A},\qquad d\pi_A = -\,\frac{\partial H}{\partial \psi^A}.
\]
If \(\epsilon^r(x)\) are local gauge parameters, the full generator is
\[
G[\epsilon] = \epsilon^r G_r + d\epsilon^r \wedge F_r,
\]
with
\[
\{G_r,G_s\} = f^t{}_{rs}\,G_t.
\]
For gauge fields and gravity, \(G_r=-\{F_r,H\}\); for matter fields, \(F_r=0\) [1909.06779]. This is one of the clearest formulations of local gauge symmetry in a manifestly covariant canonical language.

The operator-theoretic continuation is explicit: one promotes \(f\mapsto \hat f\) and replaces the graded Poisson bracket by the super-commutator,
\[
\{f,g\}\ \mapsto\ \frac{1}{i\hbar}[\hat f,\hat g]_s,\qquad
[\hat A,\hat B]_s = \hat A\hat B - (-1)^{|\hat A||\hat B|}\hat B\hat A,
\]
with ordering constrained by total degree [1703.06718]. In this sense, the differential-form formalism is already a covariant canonical operator framework in embryo.

## 3. Extended Hilbert spaces and covariant canonical quantization

A distinct realization of manifest covariance is provided by covariant canonical quantization of a real scalar field on an extended Hilbert space
\[
H_{\mathrm{real}} \subset H_{\mathrm{phys}} \subset H_{\mathrm{total}}.
\]
Here \(H_{\mathrm{total}}\) contains off-shell modes, \(H_{\mathrm{phys}}\) is obtained by imposing the field-equation constraint, and \(H_{\mathrm{real}}\) is the usual on-shell Fock space of conventional canonical quantization [1907.00645]. The field operator is expanded as
\[
\Phi(x) = \int d^4k\,(2\pi)^{-4}\,\tilde\Phi(k)\,e^{-ik\cdot x},
\]
without imposing the Klein–Gordon equation at the outset.

The formalism takes the covariant commutator as primitive:
\[
[\Phi(x),\Phi(y)] = i\,\Delta(x-y),
\]
with \(\Delta\) the Pauli–Jordan function. In momentum space,
\[
[\tilde\Phi(k),\tilde\Phi(k')] = (2\pi)^4\delta^4(k+k')\,\delta(k^2-m^2)\,\mathrm{sign}(k_0).
\]
Equal-time canonical commutators involving \(\Pi^\mu(x)\) can be recovered, but they are not the foundational postulate; all spacetime components are treated symmetrically [1907.00645].

Creation and annihilation operators are defined covariantly by
\[
\hat a(k):=\tilde\Phi(k),\qquad \hat a^\dagger(k):=\tilde\Phi(-k),
\]
so that
\[
[\hat a(k),\hat a^\dagger(k')] = (2\pi)^4\delta^4(k-k')\,\delta(k^2-m^2)\,\mathrm{sign}(k_0).
\]
The covariant number-operator density is
\[
\hat N(k) := (2\pi)^{-5}\hat a^\dagger(k)\hat a(k),
\]
and its covariance follows from the Lorentz invariance of \(d^4k\,\delta(k^2-m^2)\,\mathrm{sign}(k_0)\) [1907.00645].

A characteristic feature of this construction is the use of two symmetric vacua, \(|0_+\rangle\) and \(|0_-\rangle\), rather than a single vacuum selected by an explicit time-ordering prescription. These vacua satisfy complementary annihilation conditions on positive- and negative-energy sectors. The Feynman propagator is then reconstructed as
\[
\Delta_F(x-y)
= -i\langle 0_+|T(\Phi(x)\Phi(y))|0_+\rangle
= -i\Big[\Theta(x^0-y^0)\langle 0_+|\Phi(x)\Phi(y)|0_+\rangle
+\Theta(y^0-x^0)\langle 0_-|\Phi(x)\Phi(y)|0_-\rangle\Big],
\]
which reproduces the standard momentum-space propagator \((k^2-m^2+i\epsilon)^{-1}\) without explicitly treating time-ordering as primary [1907.00645].

The same framework rederives LSZ reduction through a projection limit. In/out operators are defined by asymptotic evolution with the covariant “Hamiltonian” enforcing \(k^2=m^2\), and the difference between in and out creation operators is written as a spacetime integral involving \((\Box+m^2)\Phi(x)\). The resulting S-matrix formula reduces exactly to the usual LSZ expression once external legs are placed on shell by \(\delta(k^2-m^2)\) factors [1907.00645].

A further result concerns vacuum energy. The total energy operator derived from the canonical energy–momentum tensor remains covariant at the level of spacetime integration, but the familiar zero-point divergence appears only after one performs a spacetime split and selects a single physical vacuum \(|0_+\rangle\). In this formulation, the divergence is therefore not primitive but arises a posteriori under the same specialization that reproduces ordinary canonical quantization [1907.00645].

## 4. Lorentz-covariant canonical gravity

In canonical gravity, manifest covariance is often lost at the step where spacetime is foliated. One line of work avoids that loss by introducing a local observer field \(y^I(x)\in H^3\subset\mathbb{R}^{3,1}\), satisfying \(y^Iy_I=-1\), together with a spacetime vector field \(u^\mu\) defined by
\[
y^I=e^I{}_\mu\,u^\mu.
\]
Under diffeomorphisms, \(u^\mu\) transforms as a spacetime vector; under local Lorentz transformations, \(e^I\) and \(y^I\) rotate while \(u^\mu\) remains the same spacetime vector. The pair \((e^I,u^I)\) is therefore covariant under both \(\mathrm{Diff}(M)\) and local \(SO(3,1)\) [1206.0658].

Relative to this observer field, the coframe and connection split as
\[
e^I = E^I + \hat u\,y^I,\qquad
\omega^{IJ} = \Omega^{IJ} + \hat u\,\Xi^{IJ},
\]
with \(E^I(u)=0\), \(y_I E^I=0\), \(\Omega^{IJ}(u)=0\), and \(y_I\Omega^{IJ}=0\). Starting from the Palatini–Holst Lagrangian,
\[
L[e,\omega] = \frac{1}{2\kappa}\Bigl(\epsilon_{IJKL}\,e^I\wedge e^J\wedge R^{KL}[\omega]
+\frac{2}{\gamma}e_I\wedge e_J\wedge R^{IJ}[\omega]\Bigr),
\]
one obtains the presymplectic current
\[
\Theta(\delta)=\frac{1}{2\kappa}\Bigl(\epsilon_{IJKL}\,e^I\wedge e^J\wedge \delta\omega^{KL}
+\frac{2}{\gamma}e_I\wedge e_J\wedge \delta\omega^{IJ}\Bigr)
\]
and the presymplectic 2-form
\[
\Omega(\delta_1,\delta_2)=\int_\Sigma \bigl[\delta_1\Theta(\delta_2)-\delta_2\Theta(\delta_1)\bigr].
\]
When the observer field is normal to a foliation and one imposes time gauge \(y^I=(1,0,0,0)\), the Ashtekar–Barbero connection
\[
A_a^i=\Gamma_a^i+\gamma K_a^i
\]
emerges, together with the Gauss, vector, and scalar constraints and the canonical bracket
\[
\{A_a^i(x),E^b{}_j(y)\}=\kappa\gamma\,\delta_a^b\,\delta^i_j\,\delta^3(x,y).
\]
The covariant presymplectic form reduces to
\[
\Omega_{\mathrm{can}}=\frac{1}{\kappa\gamma}\int_\Sigma \delta E_i^a\wedge \delta A_a^i,
\]
so the usual Hamiltonian theory is recovered as a specialization rather than a starting point [1206.0658].

A complementary Lorentz-covariant route begins directly from the Holst action with cosmological constant and avoids second-class constraints by separating dynamical and nondynamical components of the connection from the outset. The spatial connection \(\omega_{aIJ}\) is decomposed into twelve dynamical fields \(C_{aI}\) and six symmetric auxiliary fields \(\lambda_{ab}\), with the latter entering algebraically and therefore being integrated out without invoking the standard second-class-constraint machinery [1911.09690]. The resulting Hamiltonian action is purely first-class:
\[
S=\kappa\int dt\,d^3x\,
\Bigl(
2\,\widetilde\Pi^{aI}\dot C_{aI}
-\lambda_{IJ}\widetilde{\mathcal G}^{IJ}
-2N^a\widetilde{\mathcal D}_a
-N\,\widetilde{\widetilde{\mathcal H}}
\Bigr).
\]

This formalism admits a manifestly covariant connection–momentum pair
\[
A_a{}^{IJ}:=2\,P^{IJ}{}_{KL}\,C_a{}^K n^L,\qquad
\Pi^a{}_{IJ}:=\frac12\epsilon_{IJKL}\widetilde\Pi^{aK}n^L+\frac{1}{\gamma}\widetilde\Pi^a{}_{[I}n_{J]},
\]
with bracket
\[
\{A_a^{IJ}(x),\Pi^b{}_{KL}(y)\}
=\delta_a^b\,\delta^{[I}_K\delta^{J]}_L\,\delta^3(x,y).
\]
The Gauss, vector, and scalar constraints then take fully Lorentz-covariant form, and the Dirac algebra of hypersurface deformations is recovered [1911.09690].

The same analysis exhibits a two-parameter family of canonical transformations
\[
C_{aI}\longrightarrow X_{aI}
=
C_{aI}-W_a{}^b{}_{IJK}\Bigl(\alpha\,\Gamma_b{}^{JK}
+\frac{\beta}{\gamma}\,\ast\Gamma_b{}^{JK}\Bigr).
\]
In time gauge, these variables either collapse to the \(SO(3)\) ADM formulation or yield the Ashtekar–Barbero connection with a rescaled Immirzi parameter \(\gamma/(1-\beta)\), depending on \(\beta\) [1911.09690]. A common misconception is therefore that manifest Lorentz covariance and canonical gravity are mutually exclusive; these constructions show instead that the canonical description can be postponed, reorganized, or recovered from a covariant starting point.

## 5. BRST, indefinite metric, and covariant operator quantization of gauge systems

In gauge theories, manifestly covariant operator formalisms are often built in an indefinite-metric Hilbert space and controlled by BRST cohomology. For the \(2+1\)-dimensional spin-1 self-dual model, the Kugo–Ojima–Nakanishi formalism quantizes the theory in the Heisenberg picture with a Nakanishi–Lautrup \(B\)-field and Faddeev–Popov ghosts. The canonical momenta satisfy
\[
\pi^\mu(x)=\frac{\partial\mathcal{L}}{\partial(\partial_0A_\mu)}
=\frac{m}{2}\epsilon^{0\mu\nu}A_\nu(x)+\eta^{\mu 0}B(x),
\]
with nonvanishing equal-time brackets
\[
[A_\mu(x),\pi^\nu(y)]_{x^0=y^0}=i\,\delta_\mu{}^\nu\,\delta^2(\mathbf{x}-\mathbf{y}),
\]
along with the corresponding ghost anticommutators. After eliminating momenta, one may write, for example,
\[
[A_i(x),A_j(y)]_{x^0=y^0}=-\,i\,\epsilon_{ij}\,\delta^2(\mathbf{x}-\mathbf{y}),\qquad
[A_0(x),B(y)]_{x^0=y^0}=i\,\delta^2(\mathbf{x}-\mathbf{y}),
\]
and
\[
[B(x),B(y)]_{x^0=y^0}=-\,i\,m\,\delta^2(\mathbf{x}-\mathbf{y}) .
\]
The nilpotent BRST transformation
\[
sA_\mu=\partial_\mu c,\qquad sc=0,\qquad s\bar c=iB,\qquad sB=0
\]
leads to a conserved charge \(Q_{\mathrm{BRST}}\) with \(Q_{\mathrm{BRST}}^2=0\), and the quartet mechanism removes \((A_0,B,c,\bar c)\) from the physical spectrum [1906.07742].

The physical one-particle excitation is created by
\[
U_\mu(x)=A_\mu(x)-\frac{1}{m^2}\partial_\mu B(x),
\]
which obeys
\[
(\Box+m^2)U_\mu=0,\qquad \partial^\mu U_\mu=0.
\]
Its propagator agrees with that of Maxwell–Chern–Simons theory, and it satisfies the self-duality relation
\[
\frac{1}{m}\epsilon_\mu{}^{\alpha\beta}\partial_\alpha U_\beta = U_\mu.
\]
The duality to the gauge-invariant Maxwell–Chern–Simons description is therefore established at the operator level through the physical subspace rather than by a purely classical field redefinition [1906.07742].

The same operator technology can be extended to higher-derivative systems by enlarging phase space à la Ostrogradski and replacing Poisson brackets with Dirac brackets before quantization. For the vector spin-0 dual model in \(2+1\) dimensions, the higher-derivative Lagrangian is formulated in terms of \((B_\mu,\dot B_\mu,\phi_\mu,\Omega)\), with primary second-class constraints enforced through an inverse constraint matrix \(\mathcal{G}^{-1}_{IJ}\). After the replacement \(\{\ ,\ \}_D\to \frac{1}{i}[\ ,\ ]\), one obtains manifestly covariant unequal-time commutators such as
\[
[B_\mu(x),B_\nu(y)]
=
\frac{i}{m^4}\partial_\mu\partial_\nu[\Delta(x-y;m^2)-\Delta(x-y;0)]
+\frac{i}{m^2}\partial_\mu\partial_\nu E(x-y;0),
\]
and analogous expressions for commutators involving \(\phi_\mu\) and \(\Omega\) [2404.11805].

For generalized QED\(_4\) of Bopp–Podolsky type, the higher-derivative gauge field is likewise quantized through an extended phase space with second-class constraints. The resulting propagator contains the transverse combination
\[
D_{\mu\nu}(p)=\frac{i}{p^2}\theta_{\mu\nu}-\frac{i}{p^2-m^2}\theta_{\mu\nu},
\qquad
\theta_{\mu\nu}=\eta_{\mu\nu}-\frac{p_\mu p_\nu}{p^2},
\]
so the massive mode carries negative norm through the residue \(-1\). In the interacting regime, the positive-norm subspace is no longer time invariant because the interaction can create negative-norm states from an initially ghost-free one. The spectral density acquires a negative delta contribution,
\[
\rho(s)=\rho_{\mathrm{QED}}(s)-Z_m\,\delta(s-m^2),\qquad Z_m>0,
\]
and the ultraviolet improvement \(\sim 1/p^4\) is tied directly to these Lee–Wick-type poles [2404.11805]. The same work exhibits a toy higher-derivative interacting model with an extra discrete \(\mathcal{O}\) symmetry and subsidiary conditions
\[
B_\mu^{(+)}|{\rm phys}\rangle=0,\qquad
\tilde B_\mu^{(+)}|{\rm phys}\rangle=0,\qquad
\mathcal{O}|{\rm phys}\rangle=|{\rm phys}\rangle,
\]
for which a positive-norm, time-invariant subspace can be maintained [2404.11805].

## 6. Higher-derivative gravity, conformal gravity, and non-commutative operator extensions

In higher-derivative gravity, manifest covariance can be preserved together with a canonical operator algebra by working in a BRST-fixed first-order formalism. For quadratic gravity in four dimensions, the classical Lagrangian
\[
L_c=L_{EH}+C_{R^2}+C_{C^2}
\]
is rewritten using an auxiliary symmetric tensor \(K_{\mu\nu}\) and a Stückelberg vector \(A_\mu\), and the de Donder condition
\[
\partial_\nu(\sqrt{-g}\,g^{\mu\nu})=0
\]
is imposed together with \(\nabla^\mu K_{\mu\nu}=0\) [2505.09149]. The BRST-invariant gauge-fixed theory carries canonical pairs \((g_{\mu\nu},\pi^{\mu\nu})\), \((K_{\mu\nu},\Pi^{\mu\nu})\), and \((A_\mu,\pi_A^\mu)\), plus ghosts and Nakanishi–Lautrup fields, with canonical commutators such as
\[
[g_{\mu\nu}(t,\vec x),\pi^{\rho\sigma}(t,\vec x')]=
i\,\delta_{\mu\nu}^{\rho\sigma}\delta^3(\vec x-\vec x').
\]
A striking result is that, using identities implied by the de Donder gauge and the relation
\[
[A,\dot B]=\partial_t[A,B]-[\partial_tA,B],
\]
all commutators among \(g_{\mu\nu}\) and its time derivatives vanish identically:
\[
[\partial_t^m g_{\mu\nu}(t,\vec x),\partial_t^n g_{\rho\sigma}(t,\vec x')]=0
\qquad (m,n\ge 0).
\]
The physical content of the theory nonetheless contains a massless graviton, a massive scalar, and a massive spin-2 ghost; the ghost has negative norm and spoils unitarity of the S-matrix unless some nonperturbative confinement mechanism exists, a possibility mentioned but not realized canonically in the paper [2505.09149].

Quantum conformal gravity extends this BRST-covariant operator program to a Weyl-invariant scalar–tensor sector plus conformal gravity. The local symmetries comprise diffeomorphisms, Weyl rescalings, and Stückelberg shifts, with gauge conditions
\[
\partial_\mu(\sqrt{-g}\,\phi\,g^{\mu\nu})=0,\qquad
K-2\nabla\!\cdot\!A=0,\qquad
\nabla^\mu K_{\mu\nu}=0.
\]
The theory possesses two nilpotent BRST charges \(Q_{(1)}\) and \(Q_{(2)}\), satisfying
\[
Q_{(i)}^2=0,\qquad \{Q_{(1)},Q_{(2)}\}=0,
\]
and equal-time canonical brackets for \(g_{\mu\nu}\), \(\phi\), \(K_{\mu\nu}\), \(A_\mu\), and the ghost multiplets [2311.09582]. Physical states satisfy the Kugo–Ojima conditions
\[
Q_{(1)}|{\rm phys}\rangle=0,\qquad
Q_{(2)}|{\rm phys}\rangle=0.
\]
The on-shell cohomology contains two transverse polarizations of a massless graviton and five polarizations of a massive spin-2 ghost, while the dilaton, \(A_\mu\), and the ghost/NL sectors form BRST quartets and decouple. The gauge-fixed action further exhibits a Poincaré-like global \(\mathrm{IOSp}(8|8)\) symmetry, and the reduction from \(\mathrm{IOSp}(10|10)\) is attributed to the presence of the Stückelberg symmetry [2311.09582].

A different but related operator extension appears in non-commutative Dirac theory. There the spacetime coordinates satisfy
\[
[\hat x^\mu,\hat x^\nu]=i\,\theta^{\mu\nu},
\]
with fields realized as Hilbert–Schmidt operators on a configuration Hilbert space. Derivatives are inner commutators,
\[
\hat\partial_\mu\psi = -\,i\,(\theta^{-1})_{\mu\nu}[\hat x^\nu,\psi],
\]
but the ordinary operator product fails to respect Lorentz covariance under the twisted co-product. Manifest covariance is restored by replacing ordinary multiplication with
\[
\psi * \phi
=
\exp\!\Bigl(-\frac{i}{2}\theta^{\mu\nu}\hat\partial_\mu\hat\partial_\nu\Bigr)[\psi\,\phi],
\]
for which the Lorentz generators satisfy the Leibniz rule on \(*\)-products [1512.04253]. The free Dirac equation retains its standard dispersion relation,
\[
(i\gamma^\mu\hat\partial_\mu-m)\hat\psi=0,\qquad k^2=m^2,
\]
while gauge coupling to an external electromagnetic potential is incorporated covariantly through
\[
(i\gamma^\mu\hat\partial_\mu-q\,\gamma^\mu\hat A_\mu * -m)\hat\psi=0.
\]

At the operator level, the action is defined by the trace inner product on the configuration Hilbert space,
\[
S[\hat\psi]
=
{\rm Tr}_{H_c}\Bigl\{\hat\psi^\ddagger\gamma^0
\bigl(i\gamma^\mu\hat\partial_\mu-q\,\gamma^\mu\hat A_\mu * -m\bigr)\hat\psi\Bigr\},
\]
and coherent-state symbols then produce an effective non-local action on ordinary \(4\)-dimensional Minkowski spacetime [1512.04253]. In a constant magnetic field, the Landau-level spectrum remains the usual one, but the physical localization length and Aharonov–Bohm phase are shifted, with
\[
B_{\rm eff}=\frac{B}{1+q\theta B},
\]
leading to an upper bound \(\Delta x\,\Delta y\ge |\theta|\) as \(B\to\infty\). The formalism differs from standard Moyal-based NCFT in that the operator-level \(*\)-product is commutative, while non-commutative effects reappear as non-locality in the emergent spacetime action [1512.04253].

Taken together, these developments show that manifest covariance at the operator level does not by itself resolve the hard problems of quantization. In quadratic and conformal gravity it coexists with negative-norm massive spin-2 modes; in higher-derivative electrodynamics it exposes rather than removes Lee–Wick-type ghost poles; and in non-commutative fermion theory it shifts the issue from operator noncommutativity to non-locality. The common achievement is structural: canonical brackets, operator equations, and symmetry generators can be formulated covariantly. The common limitation is equally clear: covariance does not guarantee positivity, unitarity, or locality.

Source: https://www.emergentmind.com/topics/manifestly-covariant-canonical-operator-formalism