![Arxiv:1910.13898V3 [Math.CV] 9 Jun 2020 Where Nti Ae R Oee Over Modeled Are Paper This in Oss Uhsrne Xeso Rprista Otltdi T in 1](http://data.docslib.org/img/b18b9e6e7b72dc5c9a4601833fd4c903-1.webp)
arXiv:1910.13898v3 [math.CV] 9 Jun 2020 where nti ae r oee over modeled are paper this in oss uhsrne xeso rprista otltdi t in 1. Hilb postulated Definition that than remark properties us let extension all stronger of much First possess countable. second be to supposed norm eune fcmlxnmeswt t aua clrpout( product scalar natural its with numbers complex of sequences xed oahlmrhcmap holomorphic a to extends ooopi mapping holomorphic ibr aiod(rsml ibr-atg when Hilbert-Hartogs simply (or manifold Hilbert otefiiedmninlcs esyta ope ibr manifold Hilbert complex a that say we case dimensional finite the to fradius of atg figure Hartogs ihvle in values with aua ubr esythat say We number. natural ..HletHrosmanifolds. Hilbert-Hartogs 1.1. .Lo pcso atg aiod r ibr-atg 5 Hilbert-Hartogs are manifolds Hartogs References of spaces Loop 3. Manifolds results Hilbert-Hartogs of 2. statement and Introduction 1. upre yteLbxCmiANR-11-LABX-0007-01. Cempi Labex the by Supported * phrases. and words Key 1991 Date || 0 ue1,2020. 11, June : ahmtc ujc Classification. Subject Mathematics rista ti otltdi hi ento.Fnly egv ito exa of list a give we Finally, definition. their in postulated b manifolds. a is Hilbert-Hartogs possess it manifolds that Hilbert-Hartogs that erties also prove We Hartogs. Abstract. z < r < || 2 > r OPSAE SHLETHROSMANIFOLDS HILBERT-HARTOGS AS SPACES LOOP = P 1 and 0 A . X k | z q epoeta eeaie opsae fHrosmnflsaeHilbe are manifolds Hartogs of spaces loop generalized that prove We xed oteui oyic∆ polydisc unit the to extends k cnaeHrosfiuein figure Hartogs -concave .Itouto n ttmn fresults of statement and Introduction 1. | 2 nadto u aiod,bt nt n nnt iesoa,are dimensional, infinite and finite both manifolds, our addition In . A H f ibr aiod nltcds,aayi otnain opspace. loop continuation, analytic disk, analytic manifold, Hilbert r q H 1 q 1 : ,r ( q ∞ H 2 r =(∆ := ) ( q ∆ = 1 r .AAKR .IVASHKOVICH S. ANAKKAR, M. ( X =(∆ := ) r f l ˜ ) 2 is h atrsad o h pc ftesur integrable square the of space the for stands latter The . ∆ : X → q ( q q r Hrosi vr ooopi apn fa of mapping holomorphic every if -Hartogs q 2 × +1 q ) Let rmr 21,Scnay-4G0 46T25. 46G20, - Secondary 32D15, - Primary × ∆( \ Contents fthe of X → ∆ B ¯ X r q ∞ )) ( r eafiiedmninlcmlxmanifold, complex dimensional finite a be ( 1 ∪ 1 r fteui oyikto polydisk unit the of sarn domain, ring a is ) l )) q 2 cnaeHrosfiuewt ausin values with figure Hartogs -concave A ∪ stefloigoe set open following the is 1 q − A q r, +1 q 1 1 q × − ee∆ Here . scerfo h otx)i every if context) the from clear is r, ∆ 1 × ⊂ B ∞ C q +1 te xeso prop- extension etter erdefinition. heir z,w q , ( r r saplds in polydisk a is ) r-atg manifolds ert-Hartogs 2 X = ) r > ibr manifolds Hilbert . X P 1 Analogously . k pe of mples sa is z k w ¯ q q k rt- -Hartogs -concave n the and (1.2) (1.1) q C X a 9 3 1 q 2 Section 1
Here B∞(r) stands for the ball in l2 of radius r > 0 centered at origin. When writing q 2 D × B, where D is a subset in C with coordinates z =(z1,...,zq) and B is a subset in l 2 with coordinates w =(w1,....) we always mean the obvious subset in l with coordinates 2 (z,w)=(z1,...,zq,w1,...), namely D × B = {(z,w) ∈ l : z ∈ D,w ∈B}. Theorem 1. Let X be a q-Hartogs Hilbert manifold. Then every holomorphic mapping ∞ ˜ q ∞ f : Hq (r) → X extends to a holomorphic mapping f : ∆ × B → X . As an immediate consequence we obtain the following statement. A domain D in a complex Hilbert manifold X with the boundary of class C2 is called q-pseudoconcave at its boundary point p if the complex Hessian of a local defining D function at p has at c least q negative eigenvalues when restricted to Tp ∂D. Here by saying that ρ defines D at p ∈ ∂D we mean that for some neighborhood U of p one has D ∩ U = {x : ρ(x) < 0} (and of course ∇ρ|∂D never vanishes). Corollary 1. Let D be a domain in a complex Hilbert manifold X which is q-pseudocon- cave at a boundary point p. Then every holomorphic map f : D→Y to a q-Hartogs Hilbert manifold Y extends holomorphically to a neighborhood of p. The proof of this corollary follows from Theorem 1 by appropriately placing the Hartogs ∞ figure Hq (r) near p ∈D.
1.2. Loop spaces as Hilbert-Hartogs manifolds. Our main goal in this paper is to show that one finds Hilbert-Hartogs manifolds more often than one could expect. In order to provide such examples we concentrate our attention in section 3 on loop spaces of complex manifolds. Fix a smooth real compact manifold S, with or without boundary, k,2 k,2 and a finite dimensional complex manifold X. Then the manifold WS,X := W (S,X) of Sobolev W k,2-maps carries a natural structure of a complex Hilbert manifold, see [L1] or, section 3. Here k > n = dim RS to ensure that mappings from this space are continuous. k,2 The manifold WS,X is usually called a generalized loops space of X. We prove the following statement. Theorem 2. A generalized loop space of a q-Hartogs complex manifold is a q-Hartogs Hilbert manifold. This provides us a lot of interesting examples of infinite dimensional Hilbert-Hartogs manifolds. For example let S = S1 be the circle and X a connected Riemann surface. Remark that X is 1-Hartogs if and only if X is different from the Riemann sphere P1 (see 1,2 1 1 X section 3 for details). Therefore all loop spaces WS ,X for =6 P are Hilbert-Hartogs. Another example is obtained by lettting G be a complex Lie group. Then by [ASY] G is Hartogs. Therefore the loop space, which is classically denoted as LG, is also Hartogs (1-Hartogs to be precise). From the results of [Iv1, Iv2, IS], which determine Hartogs manifolds of complex dimensions 2 and 3 we get the following statement. Corollary 2. Let X be a compact complex manifold of dimension 2 (resp. of dimension 3). Then: i) either X contains a 2-dimensional spherical shell (resp. a 3-dimensional shell), or a rational curve (resp. a uniruled complex surface), k,2 iii) or, every generalized loop space WS,X is 1-Hartogs (resp. 2-Hartogs). Hilbert-Hartogs Manifolds 3
X can contain both shells and rational curves. Both these objects are obstructions to holomorphic extension. But if X doesn’t contain neither shells no rational curves then X is Hartogs (1-Hartogs if dimX = 2 and 2-Hartogs if dimX = 3). It might be interesting + to think about X from this Corollary as being (an unknown) surface of class VII0 or as S6 (provided it exists as a complex manifold), see subsection 3.3 for more details and explanations. Acknowledgment. The Authors would like to express their gratitude to our colleague L´ea Blanc-Centi for useful discussions around the results of this paper. We would like also to thank the Referee of this paper who pointed to us a serious gap in its first version.
2. Hilbert-Hartogs Manifolds 2.1. 1-complete neighborhoods. In this subsection we recall and make a bit more precise the results from [AZ]. Let U be an open subset of l2. Let furthermore u ∈ C2(U,R). Definition 2.1. We say that u is strictly plurisubharmonic if for every z ∈ U ∞ 2 ∂ u 2 2 (z)vkv¯l > c(z)||v|| ∀v ∈ l , ∂zk∂z¯l k,lX=1 with c(z) strictly positive and continuous on U. Let X be a complex Hilbert manifold modeled over l2 and let V be an open subset in X . A function u : V → R is called strictly plurisubharmonic if for every coordinate chart 2 −1 ϕj : Vj → Uj ⊂ l the composition u ◦ ϕj is strictly plurisubharmonic in Uj. By a closed coordinate ball we shall mean a closed subset B¯ of X such that there exists a coordinate chart (Vj,ϕj,Uj) with ∞ 1) B¯ ⊂ Uj; ¯ −1 ¯∞ ¯∞ 2 2) B = ϕj (B ), where B is the closed unit ball in l . −1 ∞ In this situation B := ϕj (B ) will be called an open coordinate ball, or simply a coordinate ball. Here B∞ is the open unit ball in l2. Let K be a connected compact in X . Definition 2.2. By a 1-complete neighborhood of K we understand an open set V⊃K such that: i) V is contained in a finite union of open coordinate balls centered at points of K, i.e.,
N −1 ∞ −1 V ⊂ Vj with Vj = ϕj (B ) and ϕj (0) = kj ∈K; j[=1
ii) V possesses a strictly plurisubharmonic exhaustion function u : V→ [0,t0), i.e., −1 • for every t < t0 one has that u ([0,t)) ⊂V, −1 −1 • for all 0 6 t1 < t2 < t0 one has that u ([0,t1)) ⊂ u ([0,t2)), −1 −1 • V = u ([0,t0)) and u (0) is a single point in K. All closures are taken in the topology of X . We shall need the following result from [AZ]. By an analytic q-disk in a complex Hilbert manifold X we understand an image ϕ(D¯) of the closure of a bounded strictly pseudoconvex domain D ⊂ Cq under a holomorphic imbedding ϕ : D¯ → X , which is defined in a neighborhood of D¯. 4 Section 2
Theorem 2.1. Let ϕ : D¯ → X be an imbedded analytic q-disk in a second countable complex Hilbert manifold modeled over l2. Then ϕ(D¯) admits a fundamental system of 1-complete neighborhoods. The proof of 2.1 follows the main lines of the first part of Royden’s proof from [Ro]. After trivializing the tubular neighborhood of ϕ(D¯) to a sufficiently high degree one considers −1 2 for every 0 <λ< 1 a function ϕλ(u,v)= θ(u)+ λ ||v|| . Here u is a coordinate “along” ϕ(D) and θ is a strictly plurisubharmonic. exhaustion function of D¯ = {θ(u) 6 1}, v is a “normal” to ϕ(D) coordinate. The coordinates u and v are not holomorphic but only smooth ones obtained by patching the local holomorphic coordinates. Using a sufficient degree of triviality of the tubular neighborhood one proves that these ϕλ are strictly plurisubharmonic. And then one defines 1-complete neighborhoods of ϕ(D¯) as follows: Oλ = {ϕλ(u,v) < 1+ λ}. 2.2. Infinite dimensional Hartogs figures. Now we shall prove Theorem 1 from the 2 q Introduction. For a unit vector v ∈ l orthogonal to C set Lv := span{e1,...,eq,v}. Here q 2 2 ∞ we identify C with the subspace lq of l generated by e1,...,eq. Note that Lv ∩ Hq (r)= 1 ∞ Hq (r) and therefore given a holomorphic mapping f : Hq (r) → X its restriction to ∞ q ∞ Lv ∩ Hq (r) holomorphically extends to Lv ∩ (∆ × B ). We conclude that for every line q q ∞ < v >⊥ C the restriction f|Lv holomorphically extends to Lv ∩ (∆ × B ), giving us an extension f˜ of f to ∆q ×B∞. This extension is correctly defined because for unit vectors q q v =6 w orthogonal to C the spaces Lv and Lw intersect only along C ×{0}. q ∞ Let us prove the continuity of f˜. Consider a sequence (Zn)n>1 in ∆ × B . Write n n n q n ∞ it as Zn = (z ,w ) with z ∈ ∆ and w ∈ B . Suppose that Zn converges to Z0 = ∞ (z0,w0) ∈ ∆q × B . Take R such that 1 − r
+dist ((yn,ϕ0(yn),(x,ϕ0(x)) < δ + δ + ε< 3ε. ˜ 0 0 z0 zn ˜ n n ˜ Therefore f(z ,w )= ϕ0 R ,1 = limn→∞ ϕn R ,1 = limn→∞ f(z ,w ), i.e., f is contin- uous. Now let us prove that this extension is Gˆateaux differentiable. Take again some Z0 = 0 0 q ∞ q 2 (z ,w ) ∈ ∆ × B and fix some direction v = (v1,v2) in C × l . Let l = {Z0 + tv : t ∈ C} be the line through Z0 in the direction v. Set L := span{e1,...,eq,v2}. Remark that L contains Cq and l. Note also that ∞ 2 q 2 q 2 q+2 L ∩ Hq (r)= Hq (r) := ∆ × B (r) ∪ A1−r,1 × B ⊂ C (2.1) Loop spaces of Hartogs manifolds are Hilbert-Hartogs 5 ˜ is a Hartogs figure of bidimension (q,2). The restriction g := f|∆q×B2 is continuous and 2 q 2 holomorphic on Hq (r). Holomorphicity of g on the whole ∆ × B is a local question and follows from the classical Hartogs separate analyticity theorem. Therefore f˜ is holo- q 2 morphic on ∆ × B . In particular it is differentiable in the direction of v at Z0, i.e., is Gˆateaux differentiable. Every continuous Gˆateaux differentiable map is holomorphic, see Theorem 8.7 in [Mu], therefore Theorem 1 is proved. Remark 2.1. One can also consider the following version of an infinite dimensional Hartogs figure:
∞ ∞ ∞ ∞ ∞ ∞ H∞ (r)=(B × B (r)) ∪ B \ B¯ (1 − r) × B . (2.2) Analogously to the proof of theorem above one can prove the following. ∞ Proposition 2.1. Holomorphic maps from H∞ (r) to q-Hartogs Hilbert manifolds extend to B∞ × B∞ for any value of q > 1. ∞ 2 2 Indeed, let f : H∞ (r) → X . Choose vectors e1,...,eq−1 in l and let lq−1 denote their ⊥ 2 linear span. For any v ∈ lq−1 , one has: 2 ∞ ∞ (span{e1,...,eq−1,v} × l ) ∩ H∞ (r)= Hq (r). (2.3) ⊥ 2 ˜ ∞ ∞ Then, by Theorem 1 f extends along Lv for all v ∈ lq−1 to f : B × B → X . The proof of continuity and Gateaux differentiability is the same as above. 3. Loop spaces of Hartogs manifolds are Hilbert-Hartogs 3.1. Loop spaces of complex manifolds. Fix a compact, connected, n-dimensional smooth real manifold S with or without boundary and let us following [L1] describe the natural complex Hilbert structure on the Sobolev manifold W k,2(S,X) of W k,2-maps of S to a finite dimensional complex manifold X. To speak about Sobolev spaces it is convenient to suppose that X is smoothly imbedded to some RN by Whitney Imbedding Theorem, see ex. [Mi]. If X is not compact, we suppose that this imbedding is proper. For the following basic facts about Sobolev spaces we refer to [T]. i) f ∈ W k,2(Rn) ⇐⇒ (1+ ||ξ||)kfˆ∈ L2(Rn), where fˆ is the Fourier transform of f. Moreover this correspondence is an isometry by the Plancherel identity. One defines then for any positive real number s the Sobolev-Slobodetskii space W s(Rn)={f : (1+ ||ξ||)sfˆ∈ L2(Rn)}. > n s n α n ii) If s 2 + α with 0 <α< 1 then W (R ) ⊂ C (R ) and this inclusion is a compact n, n 1 n operator. In particular W 2(R ) ⊂ C 2 (R ). n s n k n iii) If s> 2 + k for a positive integer k, then W (R ) ⊂ C (R ). n 2n 6 s n n−2s n iv) If 0 s< 2 then W (R ) ⊂ L (R ). From (ii) one easily derives that if f,g ∈ W n,2(Rn) then fg ∈ W n,2(Rn). This enables one to define correctly a W k,2-vector bundle over an n-dimensional real manifold provided k > n. By that we mean that the transition functions of the bundle are in W k,2. Condition k > n will be always assumed from now on. Take now f ∈ W k,2(S,X). Note that by (ii) such f is H¨older continuous. Consider the pullback f ∗TX as a complex Sobolev bundle ∗ over S. A neighborhood Vf of the zero section of f TX → S is an open set of the complex 6 Section 3
k,2 ∗ Hilbert space W (S,f TX) of Sobolev sections of the pullback bundle. This Vf can be naturally identified with a neighborhood of f in W k,2(S,X), thus providing a structure of complex Hilbert manifold on W k,2(S,X); see [F] and Lemma 2.1 in [L1] for more details on this construction. Another way to understand this complex structure on W k,2(S,X) is to describe what are analytic disks in W k,2(S,X). Lemma 3.1. Let D and X be finite dimensional complex manifolds and let S be an n- dimensional compact real manifold with boundary. A mapping F : D × S → X represents a holomorphic map from D to W k,2(S,X) (denoted by the same letter F ) if and only if the following holds: i) for every s ∈ S the map F (·,s): D → X is holomorphic; ii) for every z ∈ D one has F (z,·) ∈ W k,2(S,X) and this correspondence D ∋ z → F (z,·) ∈ W k,2(S,X) is continuous with respect to the Sobolev topology on W k,2(S,X) (and the standard topology on D). For the proof we refer to [L1]. Now let us prove the Theorem 2 from the Introduction.
1 k,2 3.2. Proof of Theorem 2. Let F : Hq (r) → W (S,X) be a holomorphic map. We 1 represent it as a map F : Hq (r) × S → X enjoying properties (i) and (ii) of Lemma 3.1 above. From the fact that X is q-Hartogs we get that for every fixed s ∈ S the 1 q+1 mapping F (·,s): Hq (r) → X extends holomorphically to F (·,s) : ∆ → X and we get a mapping F : ∆q+1 ×S → X. We need to prove that for every fixed z ∈ ∆q+1 one has that F (z,·) ∈ W k,2(S,X) and that this correspondence is continuous. q+1 q+1 Fix some z0 ∈ ∆ and some s0 ∈ S. Take R< 1 such that z0 ∈ ∆R . Let gs0 be the map to the graph ΓF (∗,s0) of F (∗,s0), i.e., gs0 is defined by q+1 q+1 gs0 : ∆ ∋ z −→ (z,F (z,s0)) ∈ ∆ × X. (3.1) q+1 m By Royden’s Lemma for R< 1 there exists a holomorphic embedding G : ∆R × ∆R → q+1 q+1 m ∆ × X such that G(∗,0) = gs0 , where m = dim(X). Then V = G(∆R × ∆R ) contains q+1 1 k,2 ΓF (∗,s0) over ∆R . Since on Hq (r) the map z 7→ g∗(z) ∈ W (S,X) is continuous, there 1 exists an ǫ> 0 such that for s ∈ B(s0,ǫ) the graph Γ(∗,s) over Hq (r) is contained in V . q+1 Therefore by the Hartogs theorem for holomorphic functions the graph Γ(∗,s) over ∆R q+1 is contained in V as well. By the maximum principle one has for z ∈ ∆R and s ∈ B(s0,ǫ) that −1 −1 ||G (g (z))|| 6 max ∈ q × ||G (g (z))|| 6 s z ∂∆R ∂∆R s −1 6 Cmax ∈ q × ||G (g∗(z))|| k,2 m+q+1 . (3.2) z ∂∆R ∂∆R W (B(s0,ǫ),C ) Here C is the constant from the Sobolev Imbedding Theorem. The mapping z 7→ G−1(z,F (z,∗)) q 1 is continuous in a neighborhood of ∂∆R × ∂∆R ⊂ Hq (r). Set −1 k,2 m ψ(z)(s)= π2 ◦ G (z,F (z,s)) ∈ W (B(s0,ǫ),C ), q+1 m m where π2 : ∆R × ∆R → ∆R is the natural projection and consider its Fourier transform ψˆ. From the inequality (3.2) we see that there exists a constant, namely k M = max ∈ q × ||(1+ ||∗||) ψˆ (∗)|| 2 m , s0 z ∂∆R ∂∆R (z) L (B(s0,ǫ),C ) Loop spaces of Hartogs manifolds are Hilbert-Hartogs 7 such that for all z ∈ ∆q+1 one has k ˆ 2 m ||(1+ ||∗||) ψ(z)(∗)||L (B(s0,ǫ),C ) 6 Ms0 . (3.3)
Since S is compact one can cover it by a finite number of balls {B(s0,ǫ)}s0∈J and by taking the maximum M = maxs0∈J Ms0 one obtains the inequality k ˆ ||(1+ ||∗||) ψ(z)(∗)||L2(S,Cm) 6 M. (3.4)
q+1 k,2 n Therefore for all z ∈ ∆R the mapping ψ(z) is in W (S,C ) and consequently the map F (z,∗) is in W k,2(S,X). Now let us see that the correspondence z 7→ F (z,∗) is continuous k ˆ in the variable z in Sobolev topology. Indeed, the map z 7→ (1 + ||∗||) ψ(z)(∗) is a holomorphic Hilbert space valued mapping that satisfies the maximum modulus principle, i.e., in particular it will continuously depend on z. The theorem is proved. The following statement gives us one more example of pairs of open sets U $ Uˆ in a Hilbert manifold such that holomorphic mappings extend from U to Uˆ; the previous one ∞ q ∞ ˆ k,2 q n was Hq (r) $ ∆ ×B of Theorem 1. It shows that U := W (S,∆ ×∆ ) is the “envelope k,2 n n q n q n of holomorphy” of U := W (S,Hq (r)). Here Hq (r) := (∆ ×∆ (r))∪(A1−r,1 ×∆ ) stands for the q-concave Hartogs figure in Cq+n. Theorem 3.1. Let X be a q-Hartogs Hilbert manifold. Then every holomorphic map k,2 n ˜ k,2 q n F : W (S,Hq (r)) → X extends to a holomorphic map F : W (S,∆ × ∆ ) → X .
Proof. Let f =(f q,f n): S → ∆q × ∆n be an element of W k,2(S,∆q × ∆n). Consider the analytic q-disk in W k,2(S,∆q × ∆n) defined by q n ϕ :(z,s) ∈ ∆ × S → hf q(s)(z),f (s) , (3.5) q where ha is the following automorphism of ∆ interchanging a =(a1,...,aq) and 0
a1 − z1 aq − zq ha(z)= ,..., . 1 − a¯1z1 1 − a¯qzq Clearly Φ = ϕ(∆¯ q) is an analytic disk in W k,2(S,∆q × ∆n) possessing the following prop- erties: • ϕ(0,s) is our loop f. q q n k,2 n • For z ∈ ∂∆ one has that ϕ(z,·)(S) ⊂ A1−r,1 ×∆ , therefore ∂Φ ⊂ W (S,Hq (r)). Consider the (q + 1)-disk in W k,2(S,∆q × ∆n) defined by n ϕt(z,s) := ϕ(z,t,s)= hf q (s)(z),tf (s) , |t| < 1. Then q • ϕ0 ⊂ ∆ ×{0}. • ϕ1 = ϕ. k,2 n • For all t ∈ ∆ one has that ∂Φt ⊂ W (S,Hq (r)). Therefore for ε> 0 small enough the Hartogs figure 1 Hq (ε) := {||z|| < 1,|t| <ε or 1 − ε< ||z|| < 1+ ε,|t| < 1} k,2 n is mapped by ϕ to W (S,Hq (r)). 8 Section 3
∞ We can now thicken it to an infinite dimensional Hartogs figure Hq (ε) by multiplying ∞ ∞ it with B (ε), where B (ε) is a ball in W k,2(S,Cq+n). Decreasing ε> 0 if necessary, we can achieve that the map
ϕ˜ :(z,s,f2) → ϕ(z,t,s)+ f2(s) ∞ k,2 n n ∞ will send Hq (ε) to W (S,Hq (r)). Applying Theorem 1 we extendϕ ˜ to ∆ × B and n q n therefore F is extended to a neighborhood of ϕ. Finally, since Hq (r)) and ∆ × ∆ are contractible the manifold W k,2(S,∆q × ∆n) is simply connected for any S. This ensures that our extensions give a single valued holomorphic extension of F . Corollary 3.1. If X is a q-Hartogs complex manifold then every holomorphic mapping k,2 n k,2 ˆ k,2 q F : W (S,Hq (r)) → W (S,X) extends to a holomorphic mapping F : W (S,∆ × ∆n) → W k,2(S,X). This readily follows from Theorem 3.1 applied to q-Hartogs by Theorem 2 Hilbert manifold X = W k,2(S,X). 3.3. Loop spaces of compact complex manifolds are “almost Hartogs”. In this subsection we shall explain how to determine Hartogs manifolds among compact complex manifolds and their loop spaces. Let us start from dimension one. Proposition 3.1. Every connected Riemann surface except for the Riemann sphere is 1-Hartogs. Proof. Let X be a connected Riemann surface. If X is not compact then it is Stein, see [Na]. But every Stein manifold X is 1-Hartogs for the following reason. Since X can N 1 N be realized as a closed submanifold in C every holomorphic map F : H1 (r) → X ⊂ C is given by holomorphic functions which are extendable to ∆2 by the classical Hartogs theorem. This gives the extension of F . Now let X be compact. If X =6 P1 then its universal cover is either the disk ∆ or 1 2 the complex line C. Since H1 (r) and ∆ are both simply connected the extension of 1 ˜ 1 ˜ F : H1 (r) → X is equivalent to the extension of its lift F : H1 (r) → X ≡ ∆ or C. And the latter is just a holomorphic function. 1 2 Finally, if X = P then C \{0} ∋ (z1,z2) → [z1 : z2] gives an example of a non-extendable map. Therefore Theorem 2 implies the following Corollary 3.2. Let X be a connected Riemann surface different from the Riemann sphere. 1,2 Then the loop space WS1,X is a 1-Hartogs Hilbert manifold.
In [Iv2] we introduced the class Gq of q-disk convex complex manifolds having a strictly c ∞ positive dd -closed (q,q)-form. The sequence {Gq}q=1 is rather exhaustive: Gq contains all compact manifolds of dimension q + 1, see subsection 1.5 in [Iv2] for more details. We think that the following statement should be true:
Conjecture. Let X be a complex manifold from the class Gq. Then i) either X contains a (q +1)-dimensional spherical shell, or X contains an uniruled compact subvariety of dimension q; ii) or, X is q-Hartogs. Loop spaces of Hartogs manifolds are Hilbert-Hartogs 9
Remark 3.1. a) This was proved in [Iv2] for q = 1 (in fact this particular statement was proved already in [Iv1]), and in [IS] for q = 2. In the latter paper we proved almost the assertion stated above (for all values of q), but for holomorphic mappings with zero- dimensional fibers, see Proposition 12 there. The main difficulty in proving the general case of this conjecture is that the product ∆q × X doesn’t admit a positive ddc-closed (q,q)-form in general if even X does. This produces difficulties in estimating the areas of graphs of holomorphic maps from ∆q to X. These estimates play crucial role in the proofs in [Iv1, Iv2] and [IS]. 2. For q = 1 the item (i) means just that X contains a shell or a rational curve. For q =2 we need to add few explanations to [IS]. We proved there that a meromorphic map from 1 3 H2 (r) to such X meromorphically extends to ∆ \S, where S is a complete pluripolar set of Hausdoff dimension zero. If S =6 ∅ then X contains a spherical shell of dimension 3. 1 Otherwise S is empty. If our map f was in addition holomorphic on H2 (r) then the set ˜ If˜ of points of indeterminacy of the extension f can only be discrete and then it is clear ˜ that for every a ∈ If˜ its full image f[a] contains an uniruled analytic set of dimension two. 3. Proof of Corollary 2 from the Introduction. If X is not as in (i) or (ii) then X is 1-Hartogs if dimX = 2 (resp. 2-Hartogs if dimX = 3). Theorem 2 gives that the corresponding loop spaces are Hilbert-Hartogs as claimed. 4. Let us make few remarks about a possible interest in Corollary 2. The main problem in classifying surfaces of Kodaira class VII0 with second Betti number > 0 is to prove that they contain rational curves. Concerning spherical shells, if such X contains such a shell it is a deformation of a blown-up Hopf surface. Therefore the 2-dimensional part of Corollary 2 specifies that if X has no rational curves and no shells (i.e., is “totally unknown”) then one can say a good thing about it: it is 1-Hartogs itself as well as all its loop spaces. An analogous thing can be said about a complex structure on S6 (if it exists). Remark 3.2. Finally let us discuss compact manifolds of dimension q. Note that they belong to Gq. Indeed, take any volume form on such X. It has bidegree (q,q) and is vacously ddc-closed. It was proved by F. Chazal in [Ch] that every compact manifolds of dimension q is meromorphically q-Hartogs in the sence that meromorphic maps with 1 q+1 values in such X extend from Hq (r) to ∆ . Suppose there exists a holomorphic map 1 q+1 f : Hq (r) → X which extend to ∆ only meromorphically. By the same reasoning as in the previous remark we get a uniruled subpace of X of dimension q−1. On the other hand if every f extends holomorphically then this X is q-Hartogs and so are its generalized loop spaces.
References [ASY] Adachi K., Suzuki M., Yosida M.: Continuation of holomorphic mappings with values in a complex Lie group. Pacific J. Math. 47, 1-4 (1973). [AZ] Anakkar M., Zagorodnuk A.: On Hilbert-Hartogs Manifolds. arXiv: 1112.2504 (2019), to appear in Compl. Var. & Ell. Eq. DOI:10.1080/17476933.2019.1687458 . [Bh] Behnke H.: Zur Theorie der analytischen Funktionen mehrerer komplexen Ver¨anderlichen. Der Kontinuit¨atssats und die Regul¨arkonvexit¨at. Math. Ann. 113, 392-397 (1937). [Bu] Bungart L.: On analytic fibre bundles I. Holomorphic fibre bundles with infinite dimen- sional fibers. Topology 7, no: 1, 55-68, (1968). 10 Section 3
[Ch] Chazal F.: Un th´eor`eme de prolongement d’application m´eromorphes. Math. Ann. 320, no: 2, 285-297 (2001). [F] Forstneric F.: Manifolds of holomorphic mappings from strongly pseudoconvex domains. Asian. J. Math. 11, no. 1, 113-126 (2007). [Iv1] Ivashkovich S.: Spherical shells as obstructions for the extension of holomorphic mappings. J. Geom. Anal. 2, 231-371 (1992). [Iv2] Ivashkovich S.: Extension properties of meromorphic mappings with values in non-K¨ahler manifolds. Annals of Math. 160, 795-837 (2004). [IS] Ivashkovich S., Shiffman B.: Compact singularities of meromorphic mappings between complex 3-dimensional manifolds. Math. Res. Let. 7, 695-708 (2000). [L1] Lempert L.: Holomorphic functions on (generalized) Loop spaces. Math. Proc. R. Ir. Acad. 104A, no:1, 35-46 (2004). [L2] Lempert L.: Analytic continuation in mapping spaces. Pure Appl. Math. Q. 6, no: 4, 1051-1080 (2010). [Mi] Milnor J.: Lectures on the h-cobordism theorem. Princeton University Press (1965). [Mu] Mujica J.: Complex Analysis in Banach Spaces. Math. Studies 120, North-Holland (1985). [Na] Narasimhan R.: Analysis on Real and Complex Manifolds. North-Holland, (1968). [Ro] Royden H.: The extendsion of regular holomorphic maps. Proc. AMS 43, no: 2, 306-310 (1974). [T] Taylor M.: Partial differential equations I. Basic theory. Second edition. Applied Math- ematical Sciences, 115. Springer, New York, xxii+654 pp. ISBN: 978-1-4419-7054-1, 35-01 (2011).
* Universite´ de Lille-1, UFR de Mathematiques,´ 59655 Villeneuve d’Ascq, France. E-mail address: [email protected] , [email protected]