Clifford symplectic pencils and the multi-time Arnold conjecture beyond flat targets

The classical Arnold conjecture lower-bounds the critical points of the Hamiltonian action functional on the loop space of a symplectic manifold by the cup-length (plus one) or, in the non-degenerate case, the sum of Betti numbers. Hohloch-Noetzel-Salamon (2009) and Ginzburg-Hein (2012, 2013) extended the conjecture to multi-dimensional time: the time circle is replaced by a closed r-manifold M with a divergence-free frame, and the symplectic target by a Clifford symplectic pencil, i.e. r pairwise anti-commuting complex structures compatible with one metric whose associated 2-forms are closed and symplectic. All proven lower bounds confine the target to flat quotients of Clifford representations. The problem collects the three interlocking questions Ginzburg-Hein explicitly left open, with the common objective of determining the natural generality of the multi-time Arnold conjecture and proving it there: (i) whether non-flat manifolds carrying such pencils of rank r at least 4 exist at all (none is known, and rigidity classifications cover only parallel or homogeneous structures); (ii) the characterization of admissible time manifolds and of regular divergence-free frames; and (iii) proving (or disproving) the Arnold bounds beyond flat targets, for which Ginzburg-Hein suggest, without prescribing, a Floer-theoretic route, codimension-2 bubbling currently obstructing compactness. All three components are open; recent Fueter-equation programs (2015-2026) supply only partial compactness results.

Background

The Arnold conjecture in Hamiltonian dynamics predicts that a Hamiltonian diffeomorphism of a closed symplectic manifold $W$ has at least $\mathrm{CL}(W)+1$ fixed points, and at least $\mathrm{SB}(W)$ when all fixed points are non-degenerate, where $\mathrm{CL}(W)$ denotes the cup-length of $W$ (the maximal number of positive-degree cohomology classes with non-zero cup product) and $\mathrm{SB}(W)$ the sum of the Betti numbers. Equivalently, the Hamiltonian action functional on the loop space of $W$ has at least that many critical points. The case of the standard torus was Conley and Zehnder's original theorem, proved by a finite-dimensional reduction; the general case required Hamiltonian Floer theory, in which the $L^2$-gradient lines of the action functional are solutions of a Cauchy-Riemann equation.

The hyperkahler analogue replaces both sides of this picture. A hyperkahler manifold is a Riemannian manifold $(W,g)$ carrying three covariantly constant complex structures satisfying the quaternionic relations and compatible with $g$; equivalently its holonomy is contained in the compact symplectic group $\mathrm{Sp}(n)$, and the three associated Kahler forms are closed. Hohloch-Noetzel-Salamon replaced the time circle $S^1$ by a three-dimensional time manifold $M$ (a torus $T^3$ or $\mathrm{SU}(2)$, in a general hypercontact framework, where a hypercontact structure is a triple of contact forms tied together by the quaternionic relations) and the symplectic form by the hyperkahler triple, defining a hypersymplectic action functional $A_H$ on the space of maps $f:M\to W$. Its $L^2$-gradient is a Dirac-type operator $\not\partial=\sum_l J_l\nabla_{v_l}$, a first-order operator built from the complex structures and divergence-free vector fields $v_l$ on $M$, and the gradient-flow equation is a Cauchy-Riemann-Fueter equation, the triholomorphic analogue of the Cauchy-Riemann equation. They proved the non-degenerate Arnold bound, but only for flat targets, and their own account identifies codimension-2 bubbling (Hopf circles) as the obstruction to general compact hyperkahler targets.

Ginzburg-Hein generalized the linear algebra. A Clifford pencil of rank $r$ on a vector space $V$, or on a manifold, is a collection of $r$ complex structures $J_1,\dots,J_r$, orthogonal with respect to one inner product $g$ and pairwise anti-commuting, $J_lJ_k=-J_kJ_l$, such that each $\omega_l=g(J_l\,\cdot,\cdot)$ is symplectic; any non-trivial linear combination of the $\omega_l$ is then again symplectic, whence the name pencil. Equivalently, $V$ is an orthogonal representation of the real Clifford algebra $\mathrm{Cl}(0,r)$, the algebra generated by $r$ units squaring to $-1$; such a representation exists exactly when the unit sphere of $V$ carries $r$ linearly independent vector fields, so rank $r\ge 4$ forces the dimension of $V$ to be divisible by $8$. For $r=2$ one recovers the quaternionic situation, and, as Ginzburg-Hein note, any manifold carrying a Clifford pencil with $r\ge 2$ is automatically hyperkahler, with third complex structure $J_1J_2$; ranks above $3$ are genuinely new only through the additional structures. The known targets are flat: compact quotients $W=V/\Gamma$ of a Clifford representation by a group preserving all structures. For these, and for time manifolds equal to the torus $T^r$ (or $\mathrm{SU}(2)$ at $r=3$), Ginzburg-Hein proved both the degenerate bound, at least $\mathrm{CL}(W)+1$ critical points of $A_H$ for every Hamiltonian, and the non-degenerate bound, at least $\mathrm{SB}(W)$, by a Conley-Zehnder-style finite-dimensional reduction via generating functions. That method uses flatness of the target essentially, through a Fourier-mode decomposition of the Dirac operator. Their companion paper extends the theorem to arbitrary rank and to time manifolds carrying a frame of $r$ divergence-free vector fields that is regular, meaning the equation $\not\partial_v f=0$ has only constant solutions $f:M\to V$; on a manifold that admits one regular frame, the regular frames form an open dense subset of the divergence-free frames. Ellipticity of $\not\partial$ forces the frame to span $TM$, and beyond the flat-target, toral or Lie-group cases nothing systematic is known about which time manifolds support the construction.

Three interlocking issues are left open and define the problem. First, no non-flat manifold carrying a Clifford pencil of rank $r\ge 4$ is known. The parallel case is strongly constrained: Moroianu-Semmelmann classified parallel Clifford structures and showed that rank $\ge 4$ forces flatness in dimensions at least $9$, the rank-4 exception in dimension $8$ being products of two Ricci-flat Kahler surfaces, whose Clifford bundle is generally non-trivial and does not yield globally defined anti-commuting complex structures. The Ginzburg-Hein conditions, however, require only closed compatible forms, not parallelism, and no known classification or no-go theorem covers that category. Second, the correct admissibility conditions on the time manifold, the general-rank analogue of the hypercontact condition, are unidentified. Third, no Floer theory for Clifford pencils exists: extending even the rank-3 theory to non-flat hyperkahler targets is obstructed by bubbling, and compactness of the relevant Fueter-type moduli spaces is understood only partially.

Problem Statement

Let $r\ge 2$. The problem, as left open by Ginzburg-Hein, has three linked components whose common objective is to determine the natural generality in which the multi-time Arnold conjecture holds and to prove it there. Throughout, the time manifold is a closed $r$-manifold $M$ equipped with a volume form $\mu$ and $r$ divergence-free vector fields $v_1,\dots,v_r$, the target $W$ is a Riemannian manifold carrying a Clifford symplectic pencil of rank $r$ in the sense of component (1), and Hamiltonians are smooth functions $H:M\times W\to\mathbb{R}$. Following Ginzburg-Hein (arXiv:1105.0874, Sec. 2, Eq. (2.3); arXiv:1110.0529, Sec. 2.2.1, Eq. (2.2)), the action functional is defined on the space $E$ of smooth null-homotopic maps $f:M\to W$ in two steps: for a capping homotopy $F:[0,1]\times M\to W$ from $f$ to a constant map, the unperturbed functional is

$A(f)=-\sum_{l=1}^{r}\int_{[0,1]\times M}F^{*}\omega_l\wedge i_{v_l}\mu,$

and the total functional is

$A_H(f)=A(f)-\int_M H(f)\,\mu.$

Its $L^2$-gradient is $\not\partial f-\nabla H(f)$ with $\not\partial=\sum_l J_l\nabla_{v_l}$ (equivalently $\sum_l J_l\mathcal{L}_{v_l}$ on maps $f:M\to W$), so the critical points of $A_H$ are the solutions of $\not\partial f=\nabla H(f)$. As in the classical case, $A$, hence $A_H$, may be multivalued on a general target $W$, its value possibly depending on the choice of capping $F$; single-valuedness is a condition on the target $W$, so it enters the well-definedness question of component (2) as a joint condition on the pair $(M,W)$, not as a property of $M$ alone.

(1) Targets. Determine whether there exists a compact non-flat Riemannian manifold $W$ carrying a Clifford symplectic pencil of rank $r\ge 4$: orthogonal pairwise anti-commuting complex structures $J_1,\dots,J_r$ compatible with one metric $g$ such that every form $\omega_l=g(J_l\,\cdot,\cdot)$ is closed, hence symplectic. Since a pencil with $r\ge 2$ makes $W$ automatically hyperkahler, this asks in particular whether a non-flat hyperkahler manifold admits a further anti-commuting compatible almost-complex structure with closed fundamental form. Construct and classify all such compact $W$, or prove that none exist outside the flat quotients of Clifford representations; the corresponding existence and classification questions for non-compact $W$ are a separate case.

(2) Time manifolds. Identify the correct structure on a closed $r$-manifold $M$ — the general-rank analogue of the hypercontact condition of rank three — and the joint conditions with the target $W$ under which the action functional $A_H$ on maps $f:M\to W$ is well-defined, i.e. independent of the choice of capping homotopy $F$, and its $L^2$-gradient, the Dirac-type operator $\not\partial_v=\sum_l J_l\nabla_{v_l}$ built from a divergence-free frame, is elliptic; this includes a classification or structural characterization of regular divergence-free frames.

(3) Conjecture and framework. Prove, or disprove, the Arnold conjecture for Clifford symplectic pencils beyond flat targets: for a target $W$ carrying a Clifford pencil as in (1), including non-flat hyperkahler targets at $r=3$, and an admissible time manifold $M$ as in (2), show that for every Hamiltonian $H:M\times W\to\mathbb{R}$ the action functional $A_H$ has at least $\mathrm{CL}(W)+1$ critical points, where $\mathrm{CL}(W)$ is the cup-length of $W$, and at least $\mathrm{SB}(W)$, the sum of the Betti numbers of $W$, when $H$ is non-degenerate, i.e. when every critical point $f$ of $A_H$ (every solution of $\not\partial f=\nabla H(f)$) has one-to-one linearization of the $L^2$-gradient at $f$, equivalently Hessian $d_f^2A_H$ as an operator $T_fE\to T_fE$; non-degeneracy is known to be a generic (residual) condition on $H$ in $C^\infty(M\times W)$ only in the flat-target settings where the bounds are proved. Ginzburg-Hein suggest, as a separate open direction, that a version of Hamiltonian Floer theory for the Cauchy-Riemann-Fueter-type gradient equation can be developed beyond the flat setting; this suggestion does not prescribe the route by which the Arnold bounds of this component are to be established.

The three components are jointly needed for the common objective: (1) supplies the target class, (2) the time-manifold class, and (3) the method and the theorem. Partial answers, such as compactness theorems for restricted moduli problems or classifications under extra parallelism hypotheses, constitute progress but not completion.

References

1. V. L. Ginzburg, D. Hein, Hyperkähler Arnold conjecture and its generalizations, International Journal of Mathematics 23 (2012), no. 8, 1250077. DOI: 10.1142/S0129167X12500772, https://doi.org/10.1142/S0129167X12500772 (preprint arXiv:1105.0874, https://arxiv.org/abs/1105.0874)
2. V. L. Ginzburg, D. Hein, The Arnold conjecture for Clifford symplectic pencils, Israel Journal of Mathematics 196 (2013), 95-112. DOI: 10.1007/s11856-012-0146-5, https://doi.org/10.1007/s11856-012-0146-5 (preprint arXiv:1110.0529, https://arxiv.org/abs/1110.0529)
3. S. Hohloch, G. Noetzel, D. A. Salamon, Hypercontact structures and Floer homology, Geometry & Topology 13 (2009), 2543-2617. DOI: 10.2140/gt.2009.13.2543, https://doi.org/10.2140/gt.2009.13.2543
4. A. Moroianu, U. Semmelmann, Clifford structures on Riemannian manifolds, Advances in Mathematics 228 (2011), 940-967. DOI: 10.1016/j.aim.2011.06.006, https://doi.org/10.1016/j.aim.2011.06.006 (preprint arXiv:0912.4207, https://arxiv.org/abs/0912.4207)
5. A. Moroianu, M. Pilca, Higher rank homogeneous Clifford structures, Journal of the London Mathematical Society 87 (2013), 384-400. DOI: 10.1112/jlms/jds061, https://doi.org/10.1112/jlms/jds061 (preprint arXiv:1110.4260, https://arxiv.org/abs/1110.4260)
6. D. D. Joyce, Manifolds with many complex structures, The Quarterly Journal of Mathematics 46 (1995), 169-184. DOI: 10.1093/qmath/46.2.169, https://doi.org/10.1093/qmath/46.2.169
7. D. A. Salamon, The three-dimensional Fueter equation and divergence-free frames, Abhandlungen aus dem Mathematischen Seminar der Universität Hamburg 83 (2013), 1-28. DOI: 10.1007/s12188-013-0075-1, https://doi.org/10.1007/s12188-013-0075-1 (preprint arXiv:1202.4165, https://arxiv.org/abs/1202.4165)
8. C. C. Conley, E. Zehnder, The Birkhoff-Lewis fixed point theorem and a conjecture of V.I. Arnold, Inventiones Mathematicae 73 (1983), 33-49. DOI: 10.1007/BF01393824, https://doi.org/10.1007/BF01393824
9. S. Hohloch, Hyperkähler Floer theory as infinite dimensional Hamiltonian system, Proceedings of the American Mathematical Society 143 (2015), 3519-3524. DOI: 10.1090/S0002-9939-2015-12511-7, https://doi.org/10.1090/S0002-9939-2015-12511-7
10. T. Walpuski, A compactness theorem for Fueter sections, Commentarii Mathematici Helvetici 92 (2017), 751-776. DOI: 10.4171/CMH/423, https://doi.org/10.4171/CMH/423 (preprint arXiv:1507.03258, https://arxiv.org/abs/1507.03258)
11. A. Doan, S. Rezchikov, Holomorphic Floer Theory and the Fueter Equation, preprint (2022). arXiv:2210.12047, https://arxiv.org/abs/2210.12047
12. S. Habibi Esfahani, Towards a monopole Fueter Floer homology I: a compactness theorem, preprint (2023). arXiv:2305.09456, https://arxiv.org/abs/2305.09456
13. R. Brilleslijper, O. Fabert, An approach to Hamiltonian Floer theory for maps from surfaces, preprint (2024). arXiv:2404.12249, https://arxiv.org/abs/2404.12249
14. R. Brilleslijper, O. Fabert, Generalizing symplectic topology from 1 to 2 dimensions, preprint (2024). arXiv:2412.16223, https://arxiv.org/abs/2412.16223
15. R. Brilleslijper, O. Fabert, Floer sections in multisymplectic geometry, preprint (2025). arXiv:2512.05797, https://arxiv.org/abs/2512.05797
16. L. Asselle, R. Brilleslijper, Fueter equations and the search for a higher dimensional Hamiltonian Floer theory I: analytical foundations and compactness, preprint (2026). arXiv:2603.01832, https://arxiv.org/abs/2603.01832