betaFeedback

WorksMathematics

Strictly pseudoconvex CRCR structures on \sfS^5 that can be turned inside out

Abstract

The existence of a complex structure on \sfS^{6} implies that of strictlty pseudoconvex structures on \sfS^{5} which can be turned inside out by a homotopy.

Introduction

A perhaps locally defined, smooth, real codimension one, hypersurface MM in a complex manifold XX inherits a CR structure from the ambient complex structure on XX. If that CR structure on MM is strictly pseudoconvex, then MM has two different sides that are distinguished from one another by geometric considerations. One is the pseudoconvex side, and the other is the pseudoconcave side. This distinction between the two sides is invariant with respect to local biholomorphic diffeomorphisms of the ambient complex structure on XX, and to local CR diffeomorphisms of MM. This distinction between the two sides can be understood in several different but equivalent ways:

Let xx be a point on MM. Then, as is well known, there is a local biholomorphic diffeomorphism of a neighborhood of xx in XX, which transforms MM into an elementary strictly convex hypersurface. The pseudoconvex side is then just the convex side, and the pseudoconcave side is just the other concave side. That makes it easy to see that there is no local holomorphic disk, of complex dimension equal to the CR dimension of MM, whose center is at xx, but otherwise stays on the pseudoconvex side. On the other hand, there is a local holomorphic disk, of complex dimension equal to the CR dimension of MM, whose center is at xx, but otherwise stays on the pseudoconcave side of MM. This can also be understood in terms of the scalar Levi form of MM at xx (see [1]), whose signature is a local biholomorphic, and local CR diffeomorphism invariant.

For example, let MM be the standard unit sphere S5S^5, as it sits in C3\mathbb{C}^3. Then, with the induced CR structure coming from C3\mathbb{C}^3, S5S^5 is strictly pseudoconvex, the pseudoconvex side is the side of the ball B6B^6 which the S5S^5 bounds, and the pseudoconcave side is the exterior side. At every point xx on S5S^5, there is a local complex 2-dimensional disk whose center is at xx, but otherwise stays on the exterior side. There are no such disks on the interior side. This is the usual standard strictly pseudoconvex CRCR structure one has on S5S^5.

Now suppose that, by not changing the orientation of the S5S^5 itself, one could smoothly deform the CRCR structure on the fixed S5S^5, through a one parameter family of CRCR structures on S5S^5, so as to switch the touching disks from the side on which they initially appear to the other side. That would seem to be impossible for this standard CRCR structure on S5S^5.

However, in this note we show that one can find strictly pseudoconvex CRCR structures on the fixed S5S^5, for which the touching disks do, in fact, switch sides in just that way. We refer to this by saying that these CRCR structures on S5S^5 can be ”turned inside out”. We conjecture that for spheres SkS^k, this phenomenon of having a strictly pseudoconvex CRCR structure on SkS^k, that can be turned inside out, occurs only when k=5k=5.

We also show that there are weakly pseudoconvex CRCR structures on S5S^5 whose Levi form is degenerate at each point.

All of these CRCR structures on S5S^5 are globally CRCR embeddable in S6S^6, where S6S^6 has some complex structure.

A homotopy of CR structures on S5

Fix a complex structure JJ on S6S^6 (see [2, 3]). We identify S6S^6 with the unit sphere:

S6≃{x∈R7∣∥x∥2=∑h=06xh2=1}S^6 \simeq\left\{x\in\mathbb{R}^7\mid\lVert x\rVert^2=\sum_{h=0}^{6}x_h^2=1\right\}

of R7\mathbb{R}^7 and set

Mt:={x∈S6∣x0=t}.M_t:=\{x\in S^6\mid x_0=t\}.

For −1<t<1-1<t<1, this is a smooth real 55-dimensional hypersurface of S6S^6 and inherits a CRCR structure from (S6,J)(S^6,J). Let τ\tau be the restriction of dx0\mathrm{d}x_0 to S6S^6. If −1<t<1-1<t<1 and x∈Mtx\in M_t, then τx\tau_x is a nonzero conormal to MtM_t at xx and J∗τxJ^*\tau_x generates the fiber Hx0MtH_x^0M_t of the characteristic bundle of MtM_t at xx.

We recall that, for a point xx of MtM_t, the scalar Levi form (see [1]) at ξ∈Hx0Mt\xi\in H_x^0M_t is the Hermitian symmetric form

Lξ(Z1,Z2)=(2i)−1ξ([Z~1,Z~‾2]),for Z1,Z2∈Tx1,0Mt.\mathcal{L}_{\xi}(Z_1,Z_2)=(2i)^{-1}\xi([\widetilde{Z}_1,\overline{\widetilde{Z}}_2]),\qquad\text{for }Z_1,Z_2\in T_x^{1,0}M_t.

where Z~1,Z~2∈Γ∞(Mt,T1,0Mt)\widetilde{Z}_1,\widetilde{Z}_2\in\Gamma^\infty(M_t,T^{1,0}M_t) and Z~i(x)=Zi\widetilde{Z}_i(x)=Z_i, which, for ξ=J∗τx\xi=J^*\tau_x yields

LJ∗τx(Z1,Z2)=Z1Z‾2x0,for Z1,Z2∈Tx1,0Mt.\mathcal{L}_{J^*\tau_x}(Z_1,Z_2)=Z_1\overline{Z}_2x_0,\qquad\text{for }Z_1,Z_2\in T_x^{1,0}M_t.

Take x1,…,x6x_1,\ldots,x_6 as real smooth coordinates on S6S^6 on neighbourhoods of the south pole p−:=(−1,0,…,0)p_-:=(-1,0,\ldots,0) and of the north pole p+=(+1,0,…,0)p_+=(+1,0,\ldots,0). Then

{x0=−1−x12−⋯−x62≃−1+12∑h=16xh2near p−,x0=+1−x12−⋯−x62≃+1−12∑h=16xh2near p+,\begin{cases} x_0=-\sqrt{1-x_1^2-\cdots-x_6^2}\simeq-1+\frac{1}{2}\sum_{h=1}^{6}x_h^2 & \text{near }p_-,\\ x_0=+\sqrt{1-x_1^2-\cdots-x_6^2}\simeq+1-\frac{1}{2}\sum_{h=1}^{6}x_h^2 & \text{near }p_+, \end{cases}

shows that the real valued function x0x_0 has positive definite Hessian at its critical point p−p_- and negative definite Hessian at its critical point p+p_+. Taking holomorphic coordinates in (S6,J)(S^6,J) near p+p_+, by the invariance of the signature of the Hessian at critical points, we get

Lemma 1. The real valued function x0∣S6x_0|_{S^6} is strictly pseudoconvex on a neighbourhood of p−p_- and strictly pseudoconcave on a neighbourhood of p+p_+. In particular, we can find ϵ>0\epsilon>0 such that MtM_t is a smooth strictly pseudoconvex CR hypersurface for real tt with 0<(1−t2)<ϵ0<(1-t^2)<\epsilon.

Theorem 1. S5S^5 can be endowed with a strictly pseudoconvex CR structure of hypersurface type which can be ”turned inside out”, via a homotopy of CR structures on S5S^5, all of which are of hypersurface type.

Proof. Let y=(y1,…,y6)\mathbf{y}=(y_1,\ldots,y_6) be Euclidean coordinates on R6\mathbb{R}^6. For real θ\theta, with 0<θ<π0<\theta<\pi, we denote by Tθ1,0S5T^{1,0}_\theta S^5 the CR structure on S5S^5 which is the pullback of the CR structure T1,0Mcos⁡(θ)=(T1,0S6)∩(CTMcos⁡(θ))T^{1,0}M_{\cos(\theta)}=(T^{1,0}S^6)\cap(\mathbb{C}T M_{\cos(\theta)}) of the real hypersurface Mcos⁡(θ)M_{\cos(\theta)} of S6S^6 by the parametrisation

ϕθ:S5∋y↦(cos⁡(θ),sin⁡(θ)⋅y)∈Mcos⁡(θ)⊂S6.\phi_\theta:S^5\ni y\mapsto(\cos(\theta),\sin(\theta)\cdot y)\in M_{\cos(\theta)}\subset S^6.

For 0<θ1<θ2<π0<\theta_1<\theta_2<\pi, the map

[θ1,θ2]×S5∋(θ,y)↦(cos⁡(θ),sin⁡(θ)⋅y)∈S6[\theta_1,\theta_2]\times S^5\ni(\theta,y)\mapsto(\cos(\theta),\sin(\theta)\cdot y)\in S^6

yields a homotopy between the CR structures Tθ11,0S5T^{1,0}_{\theta_1}S^5 and Tθ21,0S5T^{1,0}_{\theta_2}S^5. We note that ξθ=ϕθ∗(J∗τ)\xi_{\theta}=\phi_\theta^*(J^*\tau) is a nonzero one form of the characteristic bundle Hθ0S5\mathcal{H}^{0}_{\theta}S^5 of (S5,Tθ1,0S5)(S^5,T^{1,0}_{\theta}S^5). If we take θ1,θ2\theta_1,\theta_2 with 0<θ1<ϵ0<\theta_1<\epsilon and 0<(π−θ2)<ϵ0<(\pi-\theta_2)<\epsilon for 0<ϵ≪10<\epsilon\ll1, then by the previous lemma the Levi form LθL_{\theta} on (S5,Tθ1,0S5)(S^5,T^{1,0}_{\theta}S^5) is negative definite at every point for θ=θ1\theta=\theta_1 and positive definite at every point for θ=θ2\theta=\theta_2.

The open set {x∈S6∣x0<cos⁡(θ2)}\{x\in S^6\mid x_0<\cos(\theta_2)\} is a strictly pseudoconvex domain bounded by Mcos⁡(θ2)M_{\cos(\theta_2)}, while {x∈S6∣x0<cos⁡(θ1)}\{x\in S^6\mid x_0<\cos(\theta_1)\} is a strictly pseudoconcave domain bounded by Mcos⁡(θ1)M_{\cos(\theta_1)}.

Thus, in going from θ1\theta_1 to θ2\theta_2, the touching complex 2-dimensional local holomorphic disks have changed sides; i.e., the CR structure on S5S^5 has been turned inside out.

Remark 1. By the construction, these CR structures on S5S^5 are all globally CRCR-embeddable in a complex S6S^6.

Theorem 2. A complex S6S^6 contains a smooth real hypersurface, diffeomorphic to S5S^5, having a weakly pseudoconvex CR structure and whose scalar Levi forms degenerate at each point.

Proof. The CR structures {T1,0Mx0}\{T^{1,0}M_{x_0}\}, for x0∈(−1,1)x_0\in(-1,1), make a smooth complex vector subbundle T\mathcal{T} of rank two of T1,0(S6∖{p±})T^{1,0}(S^6\setminus\{p_\pm\}) and, likewise, the characteristic bundles H0Mx0\mathcal{H}^{0}M_{x_0}, for x0∈(−1,1)x_0\in(-1,1), make a rank one smooth subbundle H\mathcal{H} of the real bundle T∗(S6∖{p±})T^*(S^6\setminus\{p_\pm\}). The form ξ=J∗τ\xi=J^*\tau is a global section of H\mathcal{H} which is nonzero at all points of S6∖{p±}S^6\setminus\{p_\pm\}. The Hermitian symmetric forms on Tx\mathcal{T}_x make a rank four smooth real vector bundle L\mathcal{L} on S6∖{p±}S^6\setminus\{p_\pm\} and we can view the Levi forms at points of the real hypersurfaces MtM_t as a global smooth section x↦Lξxx \mapsto\mathcal{L}_{\xi_x} of L\mathcal{L} on S6∖{p±}S^6\setminus\{p_\pm\}.

Since every MtM_t is compact, the sets

A={t∈(−1,1)∣Lξx>0, ∀x∈Mt},B={t∈(0,1)∣Lξx has at least one negative proper value, ∀x∈Mt},\begin{aligned} A &= \{t\in(-1,1)\mid\mathcal{L}_{\xi_x}>0,\ \forall x\in M_t\},\\ B &= \{t\in(0,1)\mid\mathcal{L}_{\xi_x}\text{ has at least one negative proper value},\ \forall x\in M_t\}, \end{aligned}

are two disjoint open subsets of (−1,1)(-1,1). Thus C=(−1,1)∖(A∪B)≠∅C=(-1,1)\setminus(A\cup B)\ne\varnothing and, for all t∈Ct\in C, the hypersurface MtM_t is weakly pseudoconvex and has a Levi form which is degenerate at every point. □

Remark 2. This weakly pseudoconvex CR structure on S5S^5 cannot be the boundary of any region in any Stein manifold, but it is the boundary of a region of the complex S6S^6. This follows e.g. from [4]. In fact, a compact CR submanifold of a Stein manifold has a characteristic covector at which the scalar Levi form is definite.

Conjecture: Among spheres SkS^k, of real dimension kk, the only one that admits a strictly pseudoconvex CR-hypersurface structure, that can be turned inside out, is S5S^5.

Remark 3. In [6] it is shown that S5S^5 admits a Levi flat CR structure. On the other hand, a theorem of A. Heflinger (see [5]) implies that no S5S^5 may carry a Levi-flat real analytic CR structure. This in particular implies that, if the embedding of the complex S6S^6 in R7\mathbb{R}^7 is real analytic, then none of our MtM_t, for −1<t<1-1<t<1, can be Levi flat. A more general question would be: Can there be a C∞C^\infty smooth Levi flat real hypersurface, which is not simply connected, in some complex S6S^6?

C.D. Hill: Department of Mathematics, Stony Brook University, Stony Brook, N.Y. 11794 (USA)

Email address: [email protected]

Mauro Nacinovich: Dipartimento di Matematica, II Università di Roma “Tor Vergata”, Via della Ricerca Scientifica, 00133 Roma (Italy)

Email address: [email protected]

References

  1. [1]J. Brinkschulte, C.D. Hill, J. Leiterer, and M. Nacinovich, Aspects of the levi form, Boll. Unione Mat. Ital. 13 (2020), 71–89.arxiv.org/abs/1905.11362
  2. [2]Ph. Engel, Complex structures on s⁶, arXiv (2026), no. 2609.38442 [math.AG], 1–25.
  3. [3]G. Etesi, The complex structure on the six dimensional sphere, arXiv (2024), no. 1509.02300 [math.DG], 1–19.
  4. [4]C.D. Hill and M. Nacinovich, A necessary condition for global Stein immersion of compact CR-manifolds, Riv. Mat. Univ. Parma 1 (1993), 175–182.
  5. [5]S. Jekel, On two theorems of A. Hefflinger concerning foliations, Topology 15 (1976), 267–271.DOI
  6. [6]L. Meersseman and A. Verjovsky, A smooth foliation of the 5-sphere by complex surfaces, Annals of Math. 156 (2002), 915–930.DOI

Paper details

vlnc:W_aphag85p
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents