Purely three-dimensional Chern-Simons theory with supergroup gauge group: formulation and one-loop level shift

In "Branes and Supergroups" (Commun. Math. Phys. 2015), Mikhaylov and Witten derived, from a four-dimensional brane construction, an analytically-continued Chern-Simons theory with supergroup gauge group $U(m|n)$ or $OSp(m|2n)$, but explicitly left open the purely three-dimensional theory. Because the invariant quadratic form (the supertrace) on a Lie superalgebra is not positive-definite on its bosonic part, the three-dimensional path integral has no canonical integration cycle, the sign of the level $k$ is ill-defined, and the standard bosonic one-loop shift $k\to k+h^{\vee}\,\mathrm{sign}\,k$ has no agreed superanalog: analyzing $SU(m|n)$ sector by sector, the bosonic summands force different effective levels, and no value of the fermionic-sector contribution reconciles them with the naive shift $k\to k+(n-m)\,\mathrm{sign}\,k$. The authors tentatively propose that the odd (fermionic) sector produces no shift, admit this may conflict with the known $OSp(1|2n)$ quantum-group three-manifold invariants and with the four-dimensional picture in which a single effective level $K$ governs all directions, and conclude the appendix with the invitation: "Perhaps the reader can explain a better point of view." The problem is open: the cited works treat only the abelian and $\mathfrak{gl}(1|1)$-type cases, and no work defines the purely three-dimensional non-abelian supergroup Chern-Simons path integral or determines its one-loop level renormalization.

Background

Chern-Simons theory is a three-dimensional topological gauge theory: for a compact simple gauge group $G$ and an integer level $k$, the action is $I=ik\,CS(A)$ with $CS(A)=\frac{1}{4\pi}\int_W\mathrm{Tr}(A\wedge dA+\tfrac{2}{3}A\wedge A\wedge A)$ for a connection $A$ on a closed oriented three-manifold $W$; the partition function and the expectation values of Wilson lines (traces of holonomy along knots, taken in representations of $G$) are invariants of framed links and three-manifolds. Two structural facts are needed here. First, on a three-manifold with boundary, the theory is related to a two-dimensional Wess-Zumino-Witten (WZW) model, a sigma model with target $G$ whose holomorphic current algebra at level $k$ makes the theory exactly solvable. Second, the semiclassical expansion around a flat connection $A_0$ (a connection of zero field strength) contains oscillatory Gaussian integrals, each contributing a phase $\exp(i\pi\,\mathrm{sign}\,\lambda/4)$; the regularized product of these phases is the Atiyah-Patodi-Singer $\eta$-invariant, whose effect is the well-known one-loop shift $k\to k+h^{\vee}\,\mathrm{sign}\,k$ of the effective level, where $h^{\vee}$ is the dual Coxeter number of $G$ (a Lie-algebra constant, e.g. $N$ for $SU(N)$). Equivalently, the shift arises by Wick-rotating the integration cycle of the path integral to a cycle on which the integral converges; knot invariants depend on $q=\exp(2\pi i/K)$ with $K=k+h^{\vee}\,\mathrm{sign}\,k$.

A Lie supergroup is the $\mathbb{Z}_2$-graded analog of a Lie group: its Lie superalgebra $\mathfrak{sg}=\mathfrak{g}_{\bar0}\oplus\mathfrak{g}_{\bar1}$ has an even (bosonic) and an odd (fermionic) part and carries an invariant bilinear form given by the supertrace $\mathrm{Str}$, the graded trace on supermatrices. Supergroups such as $U(m|n)$ (the unitary supergroup, whose even part is $U(m)\times U(n)$) and $OSp(m|2n)$ (the orthosymplectic supergroup, whose even part is $SO(m)\times Sp(2n)$) arise as effective gauge symmetries of string-theory brane systems containing negative branes, i.e. branes carrying negative flux charge, reached by analytic continuation in the brane number. In the construction of Mikhaylov and Witten's "Branes and Supergroups", D3-branes (three-dimensional extended objects on which open strings end) stretching between two fivebranes (five-dimensional extended objects), with a theta-angle (a topological, $2\pi$-periodic coupling of gauge theory), lead, by $S$-duality (the strong-weak coupling duality of string theory) and descent to the boundary, to a three-dimensional Chern-Simons theory with supergroup gauge group $SG=U(m|n)$ or $OSp(m|2n)$; this generalizes Witten's earlier "Fivebranes and Knots" construction, in which the one-sided version with an ordinary gauge group reproduces Chern-Simons theory and, after a further duality, Khovanov homology, a knot invariant taking values in graded vector spaces that refines (categorifies) the Jones polynomial. Crucially, the four-dimensional super Yang-Mills description defines an analytically continued version of the supergroup theory, in the sense of Witten's "Analytic Continuation of Chern-Simons Theory": the path integral is made convergent by rotating the integration cycle, and on that cycle there is no one-loop shift of the effective coupling $K$; any three-dimensional shift must be absorbed into a relation $K=k+h_{sg}\,\mathrm{sign}\,k$, where $h_{sg}$ is the dual Coxeter number of the superalgebra (e.g. $n-m$ for $SU(m|n)$, $m\neq n$).

The unresolved difficulty is that the supertrace form is typically not positive-definite on the bosonic part $\mathfrak{g}_{\bar0}$: for $SU(m|n)$ it is positive on $\mathfrak{su}(n)$ and negative on $\mathfrak{su}(m)$; $OSp(1|2n)$ is the exception, where the form on the bosonic part is definite. Consequently, (i) there is no convergent WZW sigma model with target $SG$ and no canonical way to relate the three-dimensional theory to a boundary current algebra (the sign of $k$ is not even well defined); and (ii) the standard one-loop computation does not give the naive shift. Decomposing the one-loop Gaussian integrals for $SU(m|n)$ into $\mathfrak{g}_{\bar0}$-valued and $\mathfrak{g}_{\bar1}$-valued fields, the bosonic summands $\mathfrak{su}(n)$, $\mathfrak{su}(m)$ and $\mathfrak{u}(1)$ individually force effective levels $k+n\,\mathrm{sign}\,k$, $k+m\,\mathrm{sign}\,k$ and $k$, and no value of the fermionic contribution $\alpha$ — not even the authors' tentative proposal $\alpha=0$, motivated by the exact analyticity of the fermionic Gaussian integral $\int d\psi_1 d\psi_2\,e^{i\lambda\psi_1\psi_2}=i\lambda$ — can reconcile these with a single supergroup shift $k\to k+(n-m)\,\mathrm{sign}\,k$. For $OSp(1|2n)$, where the form is definite, Reshetikhin-Turaev-type three-manifold invariants — invariants of three-manifolds built combinatorially from the representation theory of a quantum group, here the quantum supergroups $U_q(\mathfrak{osp}(1|2))$ and $U_q(\mathfrak{osp}(1|2n))$, which are q-deformations of the universal enveloping algebras of the Lie superalgebras $\mathfrak{osp}(1|2)$ and $\mathfrak{osp}(1|2n)$ — do exist, and it is not known whether the $\alpha=0$ proposal is compatible with them.

Problem Statement

As posed in Appendix E of Mikhaylov-Witten, the problem is to formulate and understand purely three-dimensional Chern-Simons theory with a supergroup gauge group, without relying on the four-dimensional analytically-continued description. In the authors' words: "Does there exist an abstract 3d Chern-Simons theory of a supergroup $SG$, in which the coupling $k$ has to be an integer for topological reasons and whose path integrals give invariants of arbitrary framed three-manifolds (possibly containing framed links)?" For $OSp(1|2n)$ the answer is apparently yes; for a general supergroup (e.g. $SU(m|n)$ with $m\neq n$, or $OSp(m|2n)$ with $m\geq2$) at most a partial theory, with many but not all amplitudes well-defined, is expected. The problem is to: (1) define the theory — specify the integration cycle (or an equivalent regulator) for the path integral when the invariant form is indefinite on $\mathfrak{g}_{\bar0}$, and characterize which observables are finite, noting that amplitudes depending on inverse powers of the volume of $SG$ diverge; (2) determine the one-loop renormalization of the level in the semiclassical expansion around flat connections valued in a maximal bosonic subgroup — the superanalog of the shift $k\to k+h^{\vee}\,\mathrm{sign}\,k$, in particular whether the integration over odd ($\mathfrak{g}_{\bar1}$-valued) fields shifts the level; and (3) resolve the resulting three-way tension: the even-sector shifts (for $SU(m|n)$: $k+n\,\mathrm{sign}\,k$ on $\mathfrak{su}(n)$, $k+m\,\mathrm{sign}\,k$ on $\mathfrak{su}(m)$, $k$ on $\mathfrak{u}(1)$) cannot be combined into a single supergroup shift for any fermionic contribution, the tentative zero-shift proposal is unverified against the existing $U_q(\mathfrak{osp}(1|2n))$ three-manifold invariants, and the four-dimensional analytically-continued theory is described by one common effective level $K$ for all bosonic directions. The target is a construction — a well-defined, at least partial, three-dimensional theory together with its semiclassical expansion — addressing (1)-(3) jointly; the problem also encompasses whether one of the specific candidate formulations or shifts — the naive shift $k\to k+(n-m)\,\mathrm{sign}\,k$, the sector-wise even shifts, or the tentative zero-odd-shift proposal — is incompatible with the requirements stated in (1)-(3) in a way that the discussion above does not already establish, or whether no purely three-dimensional formulation compatible with conditions (1)-(3) exists.

References

1. V. Mikhaylov and E. Witten, "Branes And Supergroups", Commun. Math. Phys. 340 (2015), arXiv:1410.1175, DOI: 10.1007/s00220-015-2449-y, https://arxiv.org/abs/1410.1175
2. E. Witten, "Analytic Continuation Of Chern-Simons Theory", arXiv:1001.2933 (2010), https://arxiv.org/abs/1001.2933
3. E. Witten, "Fivebranes and Knots", arXiv:1101.3216 (2011), https://arxiv.org/abs/1101.3216
4. R. B. Zhang, "Three-Manifold Invariants Arising From U_q(osp(1|2))", Modern Physics Letters A 9 (1994) 1453-1465, DOI: 10.1142/S0217732394001283, https://doi.org/10.1142/S0217732394001283
5. S. Blumen, "Topological invariants of three-manifolds from U_q(osp(1|2n))", arXiv:math/0209218 (2002), https://arxiv.org/abs/math/0209218
6. V. Mikhaylov, "Analytic Torsion, 3d Mirror Symmetry And Supergroup Chern-Simons Theories", arXiv:1505.03130 (2015), https://arxiv.org/abs/1505.03130
7. N. Aghaei, A. M. Gainutdinov, M. Pawelkiewicz, V. Schomerus, "Combinatorial Quantisation of GL(1|1) Chern-Simons Theory I: The Torus", arXiv:1811.09123 (2018), https://arxiv.org/abs/1811.09123
8. N. Geer and M. B. Young, "Three dimensional topological quantum field theory from U_q(gl(1|1)) and U(1|1) Chern-Simons theory", arXiv:2210.04286 (2022), https://arxiv.org/abs/2210.04286
9. N. Garner, N. Geer, M. B. Young, "B-twisted Gaiotto-Witten theory and topological quantum field theory", arXiv:2401.16192 (2024), https://arxiv.org/abs/2401.16192
10. N. Garner and W. Niu, "Line Operators in U(1|1) Chern-Simons Theory", Commun. Math. Phys. 407 (2026), DOI: 10.1007/s00220-025-05546-5, https://doi.org/10.1007/s00220-025-05546-5
11. C. A.-M. Anghel, N. Geer, B. Patureau-Mirand, "Relative (pre)-modular categories from special linear Lie superalgebras", arXiv:2010.13759 (2020), https://arxiv.org/abs/2010.13759
12. F. Ferrari and P. Putrov, "Supergroups, q-series and 3-manifolds", arXiv:2009.14196 (2020), https://arxiv.org/abs/2009.14196
13. J. Chae, "A supergroup series for knot complements", J. Geom. Phys. 227 (2026) 105872, arXiv:2508.10279, DOI: 10.1016/j.geomphys.2026.105872, https://arxiv.org/abs/2508.10279
14. J. Chae, "Quantum invariants of 3-manifolds and links: a review", arXiv:2509.02939 (2025), https://arxiv.org/abs/2509.02939
15. T. Kimura, "Aspects of Supergroup Gauge Theory", Int. J. Mod. Phys. A 38 (2023) 2330001, arXiv:2301.05927, DOI: 10.1142/S0217751X23300016, https://arxiv.org/abs/2301.05927
16. F. Nieri, "Defects at the Intersection: the Supergroup Side", arXiv:2112.07603 (2021), https://arxiv.org/abs/2112.07603
17. R. Dijkgraaf, B. Heidenreich, P. Jefferson, C. Vafa, "Negative branes, supergroups and the signature of spacetime", JHEP 02 (2018) 050, DOI: 10.1007/JHEP02(2018)050, https://doi.org/10.1007/JHEP02(2018)050
18. K. Budzik, "Supergroup Invariants and the Brane/Negative Brane Expansion", arXiv:2509.20451 (2025), https://arxiv.org/abs/2509.20451
19. F. Costantino, M. Harper, A. Robertson, M. B. Young, "Non-semisimple topological field theory and $\widehat{Z}$-invariants from $\mathfrak{osp}(1|2)$", Lett. Math. Phys. 116 (2026) 8, DOI: 10.1007/s11005-025-02039-3, https://doi.org/10.1007/s11005-025-02039-3