Lorentz-Covariant M2-Brane Matrix Model

• Lorentz-covariant M2-brane matrix models are formulated to preserve full 11-dimensional Lorentz invariance while capturing nonperturbative membrane dynamics through Nambu bracket techniques.
• The RVPD approach circumvents discretization obstacles by restricting volume-preserving diffeomorphisms, yielding consistent BPS sectors and vacuum configurations in matrix formulations.
• Higher gauge theory and crossed module reformulations provide a complementary framework that bridges covariant matrix models with ABJM-type descriptions and holographic benchmarks.

Searching arXiv for recent and foundational papers on Lorentz-covariant M2-brane matrix models, higher-gauge reformulations, and quantum M2-brane consistency. Lorentz-covariant M2-brane matrix models are attempts to formulate membrane dynamics, or a nonperturbative description of multiple M2-branes, in terms of matrix degrees of freedom without adopting the light-cone gauge that underlies BFSS and BMN. In the literature, the term covers several inequivalent programs: Lorentzian 3-algebra or BF membrane proposals, higher-gauge-theoretic reformulations in terms of crossed modules and Lie 2-groups, and more recent matrix regularizations of the supermembrane based on Restricted Volume-Preserving Deformations (RVPD) and Nambu brackets. Taken together, these works suggest that the central problem is not merely finding matrix variables, but reconciling manifest $11$-dimensional Lorentz invariance with the constraint algebra of membrane worldvolume symmetries, the Fundamental Identity of Nambu structures, $\kappa$-symmetry, and the expected spectrum of BPS and vacuum configurations (Katagiri, 8 Apr 2025, Katagiri, 10 Aug 2025, 0807.1074).

1. Conceptual setting and structural problem

The membrane starting point is the Nambu-bracket formulation of the bosonic M2-brane action,

$$S = \int d^3 \sigma\, \frac{1}{2} \{X^I, X^J, X^K\}^2 ,$$

with embedding coordinates $X^I(\sigma^1,\sigma^2,\sigma^3)$, $I=0,\dots,10$, and

$$\{X^I, X^J, X^K\} = \epsilon^{ijk} \frac{\partial X^I}{\partial \sigma^i} \frac{\partial X^J}{\partial \sigma^j} \frac{\partial X^K}{\partial \sigma^k}.$$

This action is invariant under volume-preserving diffeomorphisms (VPD),

$\delta_{\mathrm{VPD}} X^I = \{Q_1,Q_2,X^I\},$

for arbitrary functions $Q_1,Q_2$ (Katagiri, 8 Apr 2025).

The obstacle to a Lorentz-covariant matrix model is that naive discretization of the Nambu bracket typically breaks the Leibniz rule and the Fundamental Identity. That obstruction is closely tied to the membrane constraint algebra. Hoppe identified, for relativistic extended objects of arbitrary dimension, special diffeomorphism algebras generalizing the Witt-Virasoro algebra and a nontrivial dynamical symmetry crucial for quantization, integrability, and M(atrix) theory. In this formulation, the modes of the constraints obey

$$\{\eta_\alpha,\eta_\beta\}=f_{\alpha\beta}{}^\gamma \eta_\gamma,$$

while internal Lorentz generators satisfy

$$\{\mathbb{M}_{i-},\mathbb{M}_{j-}\}=\mathbb{M}^2 \mathbb{M}_{ij}, \qquad \mathbb{M}^2=H_\alpha H_\alpha = 2\eta H - P^2 .$$

These structures indicate that matrix regularization is constrained by more than kinematics: it must reproduce the higher-dimensional diffeomorphism algebra that replaces the string Virasoro algebra (Hoppe, 2010).

A complementary contrast is provided by the BMN matrix model in the large-$\kappa$0 classical limit. There the membrane description arises after light-cone gauge fixing and retains only residual SDiff$\kappa$1-type structures; manifest Lorentz invariance is not present. The corresponding membrane Hamiltonian in the maximally supersymmetric plane-wave background is

$\kappa$2

and the explicit covariant matrix formulation remains an open challenge in that framework (Axenides et al., 2017).

2. Early Lorentz-covariant proposals and vacuum-structure tests

An early line of work treated Lorentz-covariant M2-brane matrix models as Lorentzian three-algebra or BF membrane theories. These models were attractive because they were $\kappa$3 and formally covariant, but their interpretation as the worldvolume theory of multiple M2-branes was tested by mass deformation. The relevant expectation from M-theory and AdS/CFT was a discrete set of vacua in one-to-one correspondence with partitions of $\kappa$4 (0807.1074).

For the maximally supersymmetric mass deformation, the Lorentzian three-algebra vacuum equations were

$\kappa$5

In the Lorentzian three-algebra theory, these equations admitted only the trivial solution,

$\kappa$6

The resulting unique classical vacuum conflicted with the predicted discrete vacuum structure. On that basis, the Lorentzian three-algebra BF membrane model was argued likely not to describe multiple M2-branes in flat spacetime (0807.1074).

The same analysis positioned ABJM as the more successful, though still nontrivial, comparator. In the mass-deformed ABJM theory,

$\kappa$7

and the bosonic potential could be written as

$\kappa$8

For $\kappa$9, the vacuum equations reduce to

$$S = \int d^3 \sigma\, \frac{1}{2} \{X^I, X^J, X^K\}^2 ,$$0

These equations yield irreducible and block-diagonal reducible vacua, with fuzzy three-sphere features such as

$$S = \int d^3 \sigma\, \frac{1}{2} \{X^I, X^J, X^K\}^2 ,$$1

However, the classical vacuum count is larger than the partition counting expected from M2-brane physics, because each block can be a $$S = \int d^3 \sigma\, \frac{1}{2} \{X^I, X^J, X^K\}^2 ,$$2-block or $$S = \int d^3 \sigma\, \frac{1}{2} \{X^I, X^J, X^K\}^2 ,$$3-block. A recurring misconception is therefore that Lorentz covariance by itself guarantees the correct membrane interpretation; the mass-deformation test showed that covariance is insufficient unless the vacuum structure is also correct (0807.1074).

3. RVPD and the bosonic Lorentz-covariant matrix regularization

A more recent proposal constructs a Lorentz-covariant matrix model for bosonic M2-branes by starting from the Nambu-bracket action and introducing a gauge-fixing condition that restricts VPD to a residual subclass called Restricted Volume-Preserving Deformations. The defining gauge condition is

$$S = \int d^3 \sigma\, \frac{1}{2} \{X^I, X^J, X^K\}^2 ,$$4

with $$S = \int d^3 \sigma\, \frac{1}{2} \{X^I, X^J, X^K\}^2 ,$$5 a fixed Lorentz vector, so that

$$S = \int d^3 \sigma\, \frac{1}{2} \{X^I, X^J, X^K\}^2 ,$$6

Because $$S = \int d^3 \sigma\, \frac{1}{2} \{X^I, X^J, X^K\}^2 ,$$7 transforms as a Lorentz vector, the construction is presented as retaining full $$S = \int d^3 \sigma\, \frac{1}{2} \{X^I, X^J, X^K\}^2 ,$$8-dimensional Lorentz invariance (Katagiri, 8 Apr 2025).

The Nambu bracket is decomposed using the Poisson bracket on $$S = \int d^3 \sigma\, \frac{1}{2} \{X^I, X^J, X^K\}^2 ,$$9,

$X^I(\sigma^1,\sigma^2,\sigma^3)$0

together with

$X^I(\sigma^1,\sigma^2,\sigma^3)$1

and

$X^I(\sigma^1,\sigma^2,\sigma^3)$2

The resulting identity is

$X^I(\sigma^1,\sigma^2,\sigma^3)$3

RVPD is then defined by restrictions on the VPD charges, in particular

$X^I(\sigma^1,\sigma^2,\sigma^3)$4

These restrictions are designed to remove the terms that otherwise spoil matrix regularization (Katagiri, 8 Apr 2025).

The matrix substitution rules are

$X^I(\sigma^1,\sigma^2,\sigma^3)$5

and the regularized RVPD transformation becomes

$X^I(\sigma^1,\sigma^2,\sigma^3)$6

The bosonic matrix action is written as

$X^I(\sigma^1,\sigma^2,\sigma^3)$7

The central claim is that the RVPD restriction bypasses the long-standing obstruction associated with the Leibniz rule and the Fundamental Identity by retaining only a closed, tractable residual symmetry (Katagiri, 8 Apr 2025).

The model admits explicit configurations. A particle-like solution is

$X^I(\sigma^1,\sigma^2,\sigma^3)$8

while a noncommutative membrane solution is

$X^I(\sigma^1,\sigma^2,\sigma^3)$9

Higher-dimensional noncommutative configurations are obtained by adding further noncommuting pairs (Katagiri, 8 Apr 2025).

4. Supersymmetric extension, restricted $I=0,\dots,10$0-symmetry, and BPS structure

The supersymmetric extension starts from the Bergshoeff-Sezgin-Townsend supermembrane action

$I=0,\dots,10$1

with

$I=0,\dots,10$2

and

$I=0,\dots,10$3

After gauge-fixing $I=0,\dots,10$4, the full action can be written in Nambu-bracket form,

$I=0,\dots,10$5

$I=0,\dots,10$6

where $I=0,\dots,10$7 (Katagiri, 10 Aug 2025).

Under the RVPD gauge structure, ordinary $I=0,\dots,10$8-symmetry is reduced to a restricted form $I=0,\dots,10$9 satisfying

$$\{X^I, X^J, X^K\} = \epsilon^{ijk} \frac{\partial X^I}{\partial \sigma^i} \frac{\partial X^J}{\partial \sigma^j} \frac{\partial X^K}{\partial \sigma^k}.$$0

The restricted transformations are

$$\{X^I, X^J, X^K\} = \epsilon^{ijk} \frac{\partial X^I}{\partial \sigma^i} \frac{\partial X^J}{\partial \sigma^j} \frac{\partial X^K}{\partial \sigma^k}.$$1

Successive $$\{X^I, X^J, X^K\} = \epsilon^{ijk} \frac{\partial X^I}{\partial \sigma^i} \frac{\partial X^J}{\partial \sigma^j} \frac{\partial X^K}{\partial \sigma^k}.$$2-transformations close into an RVPD transformation, so the residual fermionic symmetry and RVPD form a closed algebra. This closure is one of the technical claims that distinguishes the RVPD approach from earlier covariant proposals (Katagiri, 10 Aug 2025).

The BPS classification is organized by

$$\{X^I, X^J, X^K\} = \epsilon^{ijk} \frac{\partial X^I}{\partial \sigma^i} \frac{\partial X^J}{\partial \sigma^j} \frac{\partial X^K}{\partial \sigma^k}.$$3

with $$\{X^I, X^J, X^K\} = \epsilon^{ijk} \frac{\partial X^I}{\partial \sigma^i} \frac{\partial X^J}{\partial \sigma^j} \frac{\partial X^K}{\partial \sigma^k}.$$4 chosen so that $$\{X^I, X^J, X^K\} = \epsilon^{ijk} \frac{\partial X^I}{\partial \sigma^i} \frac{\partial X^J}{\partial \sigma^j} \frac{\partial X^K}{\partial \sigma^k}.$$5 on the background. The preserved supersymmetry fractions reported for the model are as follows (Katagiri, 10 Aug 2025).

Configuration	Preserved supercharges	BPS fraction
Particle-like (commuting)	16	$$\{X^I, X^J, X^K\} = \epsilon^{ijk} \frac{\partial X^I}{\partial \sigma^i} \frac{\partial X^J}{\partial \sigma^j} \frac{\partial X^K}{\partial \sigma^k}.$$6-BPS
Noncommutative membrane (2d)	8	$$\{X^I, X^J, X^K\} = \epsilon^{ijk} \frac{\partial X^I}{\partial \sigma^i} \frac{\partial X^J}{\partial \sigma^j} \frac{\partial X^K}{\partial \sigma^k}.$$7-BPS
4-dimensional membrane	4	$$\{X^I, X^J, X^K\} = \epsilon^{ijk} \frac{\partial X^I}{\partial \sigma^i} \frac{\partial X^J}{\partial \sigma^j} \frac{\partial X^K}{\partial \sigma^k}.$$8-BPS
6-dimensional membrane	2	$$\{X^I, X^J, X^K\} = \epsilon^{ijk} \frac{\partial X^I}{\partial \sigma^i} \frac{\partial X^J}{\partial \sigma^j} \frac{\partial X^K}{\partial \sigma^k}.$$9-BPS
8-dimensional membrane	1	$\delta_{\mathrm{VPD}} X^I = \{Q_1,Q_2,X^I\},$0-BPS
10-dimensional membrane	0	non-BPS

Representative backgrounds include

$\delta_{\mathrm{VPD}} X^I = \{Q_1,Q_2,X^I\},$1

for the $\delta_{\mathrm{VPD}} X^I = \{Q_1,Q_2,X^I\},$2-BPS noncommutative membrane, with projection $\delta_{\mathrm{VPD}} X^I = \{Q_1,Q_2,X^I\},$3, and higher-dimensional analogues obtained by adding further independent noncommuting pairs (Katagiri, 10 Aug 2025).

A subsequent one-loop analysis around BPS backgrounds formulates a BRST complex for RVPD and restricted $\delta_{\mathrm{VPD}} X^I = \{Q_1,Q_2,X^I\},$4-symmetry. The key claim is that the closure of restricted $\delta_{\mathrm{VPD}} X^I = \{Q_1,Q_2,X^I\},$5-symmetry with RVPD causes the BRST complex to terminate without higher ghosts. Around BPS backgrounds with $\delta_{\mathrm{VPD}} X^I = \{Q_1,Q_2,X^I\},$6, the quadratic Euclidean fluctuation actions are

$\delta_{\mathrm{VPD}} X^I = \{Q_1,Q_2,X^I\},$7

$\delta_{\mathrm{VPD}} X^I = \{Q_1,Q_2,X^I\},$8

with $\delta_{\mathrm{VPD}} X^I = \{Q_1,Q_2,X^I\},$9. The resulting Main Theorem states that $Q_1,Q_2$0D, $Q_1,Q_2$1D, $Q_1,Q_2$2D, and $Q_1,Q_2$3D noncommutative membranes are one-loop stable, while the $Q_1,Q_2$4D configuration develops a tachyonic mode (Katagiri, 28 Sep 2025).

5. Higher gauge theory, crossed modules, and reductions to ABJM-type models

A distinct but related approach does not begin from matrix regularization of the Nambu bracket. Instead, it reinterprets the gauge structures appearing in M2-brane models in terms of differential crossed modules and Lie 2-groups. In this language, 3-Lie algebras are special cases of differential crossed modules $Q_1,Q_2$5 with trivial $Q_1,Q_2$6 and abelian $Q_1,Q_2$7, while more general crossed modules allow nontrivial $Q_1,Q_2$8 and nonabelian $Q_1,Q_2$9 (Palmer et al., 2012).

The defining identities are

$$\{\eta_\alpha,\eta_\beta\}=f_{\alpha\beta}{}^\gamma \eta_\gamma,$$0

For metric crossed modules, the 3-bracket is reconstructed by the Faulkner-type formula

$$\{\eta_\alpha,\eta_\beta\}=f_{\alpha\beta}{}^\gamma \eta_\gamma,$$1

In this framework, BLG-like M2-brane models and candidate M5-brane equations are placed on a common higher-gauge-theoretic footing. The reformulation yields Lorentz-covariant maximally supersymmetric equations for the $$\{\eta_\alpha,\eta_\beta\}=f_{\alpha\beta}{}^\gamma \eta_\gamma,$$2 tensor multiplet on any metric differential crossed module, with the fake-curvature constraint

$$\{\eta_\alpha,\eta_\beta\}=f_{\alpha\beta}{}^\gamma \eta_\gamma,$$3

and the self-duality conditions

$$\{\eta_\alpha,\eta_\beta\}=f_{\alpha\beta}{}^\gamma \eta_\gamma,$$4

A crucial limitation is also explicit in this formulation: fully covariant coupling of matter fields remains to be worked out, and full invariance under “fat” gauge transformations is not achieved (Palmer et al., 2012).

A six-dimensional realization of this higher-gauge program uses the string Lie 2-algebra as gauge structure. Its bosonic action contains a PST term that enforces self-duality while preserving Lorentz covariance: $$\{\eta_\alpha,\eta_\beta\}=f_{\alpha\beta}{}^\gamma \eta_\gamma,$$5 The higher curvatures are

$$\{\eta_\alpha,\eta_\beta\}=f_{\alpha\beta}{}^\gamma \eta_\gamma,$$6

$$\{\eta_\alpha,\eta_\beta\}=f_{\alpha\beta}{}^\gamma \eta_\gamma,$$7

together with higher forms $$\{\eta_\alpha,\eta_\beta\}=f_{\alpha\beta}{}^\gamma \eta_\gamma,$$8 and $$\{\eta_\alpha,\eta_\beta\}=f_{\alpha\beta}{}^\gamma \eta_\gamma,$$9. Upon compactification on $$\{\mathbb{M}_{i-},\mathbb{M}_{j-}\}=\mathbb{M}^2 \mathbb{M}_{ij}, \qquad \mathbb{M}^2=H_\alpha H_\alpha = 2\eta H - P^2 .$$0, with a gerbe background satisfying

$$\{\mathbb{M}_{i-},\mathbb{M}_{j-}\}=\mathbb{M}^2 \mathbb{M}_{ij}, \qquad \mathbb{M}^2=H_\alpha H_\alpha = 2\eta H - P^2 .$$1

the reduction yields a three-dimensional M2-brane model that is a deformation of ABJM, and the Chern-Simons level $$\{\mathbb{M}_{i-},\mathbb{M}_{j-}\}=\mathbb{M}^2 \mathbb{M}_{ij}, \qquad \mathbb{M}^2=H_\alpha H_\alpha = 2\eta H - P^2 .$$2 is interpreted as the Dixmier-Douady class of the gerbe (Saemann et al., 2017).

This higher-gauge route differs from RVPD regularization: it prioritizes categorified gauge symmetry and dimensional reduction rather than a direct finite-matrix transcription of the membrane bracket. A plausible implication is that Lorentz covariance in M2-brane modeling may be achievable either by restricting membrane reparametrization symmetry, as in RVPD, or by enlarging the gauge concept to Lie 2-group data, as in crossed-module and string-2-algebra constructions (Palmer et al., 2012, Saemann et al., 2017).

6. Quantum M2-brane benchmarks, comparative models, and open questions

A useful benchmark for any covariant membrane formalism is whether genuine quantum M2-brane calculations can be carried out consistently. In the large-$$\{\mathbb{M}_{i-},\mathbb{M}_{j-}\}=\mathbb{M}^2 \mathbb{M}_{ij}, \qquad \mathbb{M}^2=H_\alpha H_\alpha = 2\eta H - P^2 .$$3, fixed-$$\{\mathbb{M}_{i-},\mathbb{M}_{j-}\}=\mathbb{M}^2 \mathbb{M}_{ij}, \qquad \mathbb{M}^2=H_\alpha H_\alpha = 2\eta H - P^2 .$$4 regime of ABJM, the $$\{\mathbb{M}_{i-},\mathbb{M}_{j-}\}=\mathbb{M}^2 \mathbb{M}_{ij}, \qquad \mathbb{M}^2=H_\alpha H_\alpha = 2\eta H - P^2 .$$5-BPS circular Wilson loop has the localization result

$$\{\mathbb{M}_{i-},\mathbb{M}_{j-}\}=\mathbb{M}^2 \mathbb{M}_{ij}, \qquad \mathbb{M}^2=H_\alpha H_\alpha = 2\eta H - P^2 .$$6

while the dual wrapped M2-brane on $$\{\mathbb{M}_{i-},\mathbb{M}_{j-}\}=\mathbb{M}^2 \mathbb{M}_{ij}, \qquad \mathbb{M}^2=H_\alpha H_\alpha = 2\eta H - P^2 .$$7 has classical action

$$\{\mathbb{M}_{i-},\mathbb{M}_{j-}\}=\mathbb{M}^2 \mathbb{M}_{ij}, \qquad \mathbb{M}^2=H_\alpha H_\alpha = 2\eta H - P^2 .$$8

The one-loop determinant, including all Kaluza-Klein modes, yields

$$\{\mathbb{M}_{i-},\mathbb{M}_{j-}\}=\mathbb{M}^2 \mathbb{M}_{ij}, \qquad \mathbb{M}^2=H_\alpha H_\alpha = 2\eta H - P^2 .$$9

for $\kappa$00, exactly reproducing the field-theory prefactor. This was presented as the first exact matching of the overall numerical prefactor in a Wilson-loop expectation value against the dual holographic result, and as evidence that Lorentz-covariant quantum M2-brane theory is consistent in a highly supersymmetric setting (Giombi et al., 2023).

The broader comparative landscape remains mixed. Covariantized Matrix theory for D-particles already showed that manifest Lorentz covariance can be achieved in a matrix theory by introducing higher gauge symmetries associated with a discretized Nambu 3-bracket,

$\kappa$01

while reducing to BFSS in the light-front gauge (Yoneya, 2016). This supports the idea that covariant matrix theories require gauge structures larger than ordinary $\kappa$02. By contrast, the Lorentzian three-algebra BF membrane program failed a basic vacuum-counting test, and the higher-gauge crossed-module program still leaves open the covariant matter coupling and full higher-gauge invariance problems (0807.1074, Palmer et al., 2012).

Taken together, these results indicate several unresolved issues. One concerns higher-loop control: in the holographic wrapped-M2 calculation, subleading terms in the $\kappa$03 expansion were proposed to correspond to higher-loop corrections on the M2-brane worldvolume, and whether divergences cancel at those orders remains open (Giombi et al., 2023). A second concerns uniqueness: the existing literature contains multiple inequivalent Lorentz-covariant constructions rather than a single accepted model. A third concerns physical interpretation: the RVPD program offers a direct matrix regularization with explicit BPS sectors and one-loop stability results, but its relation to the established ABJM description of multiple M2-branes is still a programmatic connection rather than an exact equivalence. A final issue is extension to higher branes: both the RVPD papers and the higher-gauge literature describe M5-brane generalization as a natural next step, but not as a completed construction (Katagiri, 10 Aug 2025, Saemann et al., 2017).

In that sense, the Lorentz-covariant M2-brane matrix model is best understood as an active research domain defined by a common objective—maintaining Lorentz covariance in a nonperturbative membrane description—rather than by a settled formalism. The most technically explicit recent realization is the RVPD-based Nambu-bracket matrix model and its supersymmetric extension, while higher-gauge and holographic results provide complementary structural and quantum-consistency benchmarks (Katagiri, 8 Apr 2025, Katagiri, 10 Aug 2025, Giombi et al., 2023).