On Duality Invariant Yang-Mills Theory
Carlo Alberto Cremonini, Erik Hundeshagen and Ivo Sachs · 2026
Read On Duality Invariant Yang-Mills Theory by Carlo Alberto Cremonini, Erik Hundeshagen and Ivo Sachs (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-05744.txt.
Introduction
Carlo Alberto Cremonini and 2 co-authors construct, in a 2026 paper, an interacting Yang-Mills theory that keeps duality invariance manifest. Duality swaps electric and magnetic descriptions of a field theory, and building an interacting deformation of Maxwell theory that preserves it openly has been a long-standing challenge.
Their building uses mutually local but interacting one-form and three-form fields. Unusually, the theory is written as a BRST-quantized gauge theory, a formalism used to handle gauge redundancy in quantum calculations. The underlying gauge-invariant Lagrangian before gauge fixing stays hidden. The gauge invariance itself rests on an associative symmetry rather than a standard Lie symmetry, marking a structurally new route to duality-invariant interactions. The result is a new kind of gauge theory, built to keep a key swap in plain sight.
Catalog record
| CREATOR | Carlo Alberto Cremonini, Erik Hundeshagen and Ivo Sachs |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 35 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-05744
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
On Duality Invariant Yang-Mills Theory Carlo Alberto Cremonini, Erik Hundeshagen and Ivo Sachs Abstract We provide an explicit construction of a manifestly duality invariant, interacting deformation of Maxwell theory in four dimensions in terms of mutually local, but interacting 1- and 3-forms. Interestingly, our theory is formulated directly as a BRST quantized gauge theory, while the underlying gauge invariant Lagrangian before gauge fixing is obscured. Furthermore, the underlying gauge invariance is based on an associative, rather than a Lie symmetry. I Introduction Since Maxwell’s equations are invariant under the exchange of electric and magnetic fields, it is natural to speculate whether electric-magnetic duality can be extended to an interacting theory. For a single Maxwell field, it is long known to be the case. Schrödinger noticed that the Born-Infeld action has this property. See 4 for an extensive discussion on this topic. On the other hand, Yang-Mills theory does not have a local and duality-invariant formulation. See e.g. 1 ; 9 and references therein. Below, we will present a manifestly duality invariant non-abelian interacting deformation of Maxwell theory, involving a finite number of auxiliary fields. By doing so, we avoid the no-go theorem in 1 in two aspects. First, the interacting theory is not a local functional of one vector potential. It is, however, a local functional of a 1-form and a 3-form potential combined with an auxiliary field. Second, while the free theory is manifestly invariant under Maxwell gauge transformations, the gauge invariance is not manifest in the interacting theory. Rather, the interacting theory is described as a deformation of the gauge fixed BRST-fication of the free theory. In other words, the deformation is formulated directly in BRST-quantization rather than a gauge theory, which is later BRST quantized. It is not clear to us whether our model has a formulation as a non-abelian gauge theory in the usual sense. A key new ingredient in our approach, which was not present in previous literature, is that we replace the wedge product by a Clifford product of forms and the exterior differential \mathrm{d} is replaced by the Kähler-Dirac operator K=\mathrm{d}+\mathrm{d}^{\dagger} . In a way, K is the natural choice since it makes Hodge duality for the kinetic term manifest. However, to include interactions, we need a multiplication that is compatible with Hodge duality. The Clifford product is obtained simply by substituting differentials \mathrm{d}x^{\mu} by a gamma matrix \gamma^{\mu} . In fact, this product turns out not to be invariant under Hodge duality, but its projection on top-forms (i.e. under the integral) is. We will find that this is sufficient to obtain a manifestly electric-magnetic invariant form of the action. An immediate consequence of introducing the Clifford product is that it induces not only Lie brackets of the gauge-generators, but also anti-commutators as well. This means that admissible gauge Lie algebras need to admit an associative structure as well. Such is the case for the gauge group U(n)^{\mathbb{C}} , for example. Another option is to consider supergroups instead of ordinary gauge groups. II Free Theory Let us begin with the familiar Maxwell action (the notations adopted throughout the text are explained in the last section, and in the whole text, the trace over the Lie algebra generators is understood, when applicable) \displaystyle S_{1}=\frac{1}{2}\int(\mathrm{d}A^{[1]},\mathrm{d}A^{[1]})\ , (1) and the "dual" formulation in terms of a 3-form \displaystyle S_{2}=\frac{1}{2}\int(\mathrm{d}^{\dagger}A^{[3]},\mathrm{d}^{\dagger}A^{[3]})\ , (2) where \int(\cdot,\cdot) is the Hodge inner product. Here we mean dual in the sense that if we write A^{[3]}=*\tilde{A}^{[1]} and interpret \tilde{A}^{[1]} as the potential naturally coupling to magnetic 4-current, then S_{2} naturally describes the dynamics of an electromagnetic field coupled to magnetic charges. To continue, we introduce an auxiliary 2-form to couple the two field strengths \mathrm{d}A^{[1]} and \mathrm{d}^{\dagger}A^{[3]} . Concretely, we set \displaystyle S=S_{1}+S_{2}+i\int \displaystyle(\mathrm{d}A^{[1]}+\mathrm{d}^{\dagger}A^{[3]},B^{[2]})\ . (3) Variation w.r.t. B^{[2]} implies \displaystyle\mathrm{d}A^{[1]}+\mathrm{d}^{\dagger}A^{[3]}=0\ , (4) thus identifying the two field strengths \mathrm{d}A^{[1]} and \mathrm{d}^{\dagger}A^{[3]} . Note that this is the generic duality symmetric coupling at linear order in the absence of boundaries. Before proceeding with the analysis of the equations of motion, let us consider the naive counting of the theory described by two Maxwell theories with the constraint identifying the two field-strengths. For A^{[3]}=*\tilde{A}^{[1]} one recovers a duality invariant action as in 6 and 5 for instance however, with an extra auxiliary field. The coupling ( 4 ) is implied in the quantization of the relativistic spinning particle 3 . In string theory, such a two-form arises as a massive field in the spectrum. We then expect that this identifies the degrees of freedom of the two Maxwell systems, such that of the 4+4 components of A^{[1]} and A^{[3]} , only 4 are independent. The gauge symmetry of the theory then allows us to identify two propagating degrees of freedom, as usual in Maxwell theory. If, furthermore, \tilde{A}^{[1]}=A^{[1]} , we recover the theory of a chiral 2-form, e.g. 2 with no local degrees of freedom. We should note that the variation w.r.t. A , on shell, furthermore implies \mathrm{d}B^{[2]}=\mathrm{d}^{\dagger}B^{[2]}=0 which suggests that B^{[2]} is the field strength of another Maxwell field. However, the components of B^{[2]} can be gauged away thanks to an on-shell 2-form symmetry, analogous to the residual gauge in Maxwell after imposing the Lorentz gauge-fixing condition. We could then include ghosts (and ghosts for ghosts as in 2 ) for these modes, but since B^{[2]} decouples, as we will see below, this won’t be necessary. To see that this intuition is correct, let us analyze the degrees of freedom in terms of constraints in phase space (see also 2 ). From ( 3 ), we read the momenta as \displaystyle\pi^{0}=0\ ,\ \pi^{i}=\partial^{i}A^{0}-\partial^{0}A^{i}+2B^{0i}\ ,\ \pi^{\mu\nu}=0\ ,\ \ldots (5) where \pi^{\mu\nu} are the six momenta conjugate to B_{\mu\nu} and " \ldots " denotes analogous expressions for the momenta of the other Maxwell system, which will be denoted by \tilde{\pi} . The introduction of the Lagrange multiplier B^{[2]} in the action leads to six more primary constraints, compared to the usual analysis for the two Maxwell theories with \phi_{1}=\pi^{0},\tilde{\phi}_{1}=\tilde{\pi}^{0} , given by the momenta \phi_{\mu\nu}=\pi_{\mu\nu} . The Hamiltonian on the constraint surface reads \displaystyle H_{0}= \displaystyle\frac{1}{2}\pi^{i}\pi_{i}+\frac{1}{2}\pi^{i}\partial_{i}A_{0}-\frac{1}{4}F_{ij}F^{ij}+ \displaystyle-6B_{0i}B^{0i}-B_{ij}\left(F^{ij}+\tilde{F}^{ij}\right)+\ldots\ , (6) and it is extended as \displaystyle H_{E}=H_{0}+g_{1}\phi_{1}+\tilde{g}_{1}\tilde{\phi}_{1}+g^{\mu\nu}\phi_{\mu\nu}\ . (7) These eight primary constraints generate, by consistency, eight more secondary constraints as \displaystyle\left\{H_{E},\pi^{0}\right\}=-\partial_{i}\pi^{i}\to\partial_{i}\pi^{i}\equiv\phi_{2}\ , (8a) \displaystyle\left\{H_{E},\tilde{\pi}^{0}\right\}=-\partial_{i}\tilde{\pi}^{i}\to\partial_{i}\tilde{\pi}^{i}\equiv\tilde{\phi}_{2}\ , (8b) \displaystyle\left\{H_{E},\pi_{0i}\right\}=-6B_{0i}\to B_{0i}\equiv\phi_{sec,i}\ , (8c) \displaystyle\left\{H_{E},\pi_{ij}\right\}=-\left(F_{ij}+\tilde{F}_{ij}\right)\to F_{ij}+\tilde{F}_{ij}\equiv\phi_{sec,ij}\ . (8d) The Hamiltonian is thus extended as \displaystyle H_{E}^{\prime}=H_{E}+g_{2}\phi_{2}+\tilde{g}_{2}\tilde{\phi}_{2}+g_{0i}\phi_{sec,i}+g_{ij}\phi_{sec,ij}\ . (9) First of all, notice that \phi_{sec,i} do not generate other constraints as \displaystyle\left\{H_{E}^{\prime},\phi_{sec,i}\right\}=g_{0i}\approx 0\ . (10) On the other hand, \phi_{sec,ij} generate more constraints as \displaystyle\left\{H_{E}^{\prime},\phi_{sec,ij}\right\}\approx-\partial_{i}\left(\pi_{j}+\tilde{\pi}_{j}\right)+\partial_{j}\left(\pi_{i}+\tilde{\pi}_{i}\right)\equiv\phi_{ter,ij}\ , (11) and \left\{H_{E}^{\prime},\phi_{ter,ij}\right\}\approx 0 . It is not hard to see that \phi_{1},\tilde{\phi}_{1},\phi_{2},\tilde{\phi}_{2} and \phi_{ij} are first-class constraints, while \phi_{0i},\phi_{sec,i},\phi_{sec,ij} and \phi_{ter,ij} are second-class constraints. One subtlety in counting the number of constraints is that the combination \phi_{2}+\tilde{\phi}_{2} , \phi_{sec,ij} and \phi_{ter,ij} contribute altogether to reduce the number of canonical variables by six, instead of eight, as one could think by counting them as a first-class constraint and six second-class constraint. However, a more detailed analysis, shows that these constraints lead to the identification of the electric and magnetic fields of the two theories, that is, they fix six canonical variables 1 1 1 This can be analogously seen by turning off