Introduction
Definition 1.1. Let X ⊂ P M X \subset\mathbb{P}^{M} X ⊂ P M be an irreducible nondegenerate variety of codimension at least 1. Let k ≥ 1 k \ge1 k ≥ 1 be an integer. We say that X X X is k − k- k − extendable if there exists a variety W ⊂ P M + k W \subset\mathbb{P}^{M+k} W ⊂ P M + k different from a cone, with dim W = dim X + k \dim W = \dim X + k dim W = dim X + k and having X X X as a section by a M − M- M − dimensional linear space such that W W W is smooth along X = W ∩ P M X = W \cap\mathbb{P}^{M} X = W ∩ P M . We say that X X X is precisely k − k- k − extendable if it is k − k- k − extendable but not ( k + 1 ) − (k + 1)- ( k + 1 ) − extendable. The variety W W W is called a k k k -extension of X X X . We say that X X X is extendable if it is 1 − 1- 1 − extendable.
Extendability of a projective variety is a natural and fundamental question in projective geometry. This classical question has been the topic of intense research for decades and has revealed deep connections between the geometry of an embedding, Gaussian map of curve sections, deformations of cones over the hyperplane sections etc. We refer to [28 ] for an excellent recent survey of this topic.
This article revisits the question of extendability of an embedded prime K 3 K3 K 3 surface and its applications to the boundedness, classification and Hilbert schemes of prime Fano threefolds of Picard rank one. In [13 ] , [12 ] and [14 ] , Ciliberto, Lopez and Miranda analyzed the extendability of general K 3 K3 K 3 surfaces by computing the coranks of Wahl maps of the canonical curve sections (see also [2 ] , [8 ] , [9 ] , [10 ] , [11 ] , [Ko1], [25 ] , [26 ] , [34 ] for related and more recent work).
2020 Mathematics Subject Classification. 14B10, 14C05, 14N05, 14J26, 14J28, 14J45.
Key words and phrases. extendability, K 3 K3 K 3 surfaces, canonical curves, Fano threefolds, Mukai varieties, projective degenerations, ribbons.
The questions on extendability has almost always involved the Gaussian map, either on the surface or going to the hyperplane curve section. As noted in [28 ] “the emerging philosophy is that if one can control the corank of the Gaussian maps of the curve section of X X X , then one can give information on the extendability of X X X .”
In [5 ] an approach to the question of extendability via degeneration to ribbons was introduced. It avoids Gaussian maps altogether, either on the surface or on the hyperplane curve section, and provides a general method to study extendability, by directly computing α ( X ) = h 0 ( N X / P M ( − 1 ) ) − M − 1 \alpha(X)=h^0(N_{X/\mathbb{P}^M}(-1))-M-1 α ( X ) = h 0 ( N X / P M ( − 1 )) − M − 1 and h 0 ( N X / P M ( − 2 ) ) h^0(N_{X/\mathbb{P}^M}(-2)) h 0 ( N X / P M ( − 2 )) on a ribbon degeneration of X X X . Once α ( X ) \alpha(X) α ( X ) and h 0 ( N X / P M ( − 2 ) ) h^0(N_{X/\mathbb{P}^M}(-2)) h 0 ( N X / P M ( − 2 )) are computed, we conclude using the following theorem due to Zak-L’vovsky.
Theorem 1.2. ([29 ] , [Zak91]) Let X ⊂ P M X \subset\mathbb{P}^M X ⊂ P M be a smooth irreducible nondegenerate variety of codimension at least 1 1 1 and suppose X X X is not a quadric. If α ( X ) ≤ 0 \alpha(X) \le0 α ( X ) ≤ 0 , then X X X is not extendable. Further, given an integer k ≥ 2 k \ge2 k ≥ 2 , suppose that either:
α ( X ) < M \alpha(X)<M α ( X ) < M or
H 0 ( N X / P M ( − 2 ) ) = 0 H^0(N_{X/\mathbb{P}^M}(-2))=0 H 0 ( N X / P M ( − 2 )) = 0 ,
If α ( X ) ≤ k − 1 \alpha(X)\le k-1 α ( X ) ≤ k − 1 , then X X X is not k k k -extendable.
In [6 ] this method was used to give a new proof of the optimal results of [14 ] on the extendability of general non-prime K 3 K3 K 3 surfaces and of the classification of non-prime Fano threefolds of Picard rank one. In this article we treat the prime case and thereby complete this program, i.e., we prove the optimal results on the extendability of general prime K 3 K3 K 3 surfaces, and the classification of prime Fano threefolds of Picard rank one.
By semicontinuity it is enough to give sharp upper bounds to α ( X ) \alpha(X) α ( X ) by computing it over an embedded degeneration of X X X . Let Y = F e Y=\mathbb{F}_e Y = F e , e ∈ { 0 , 1 } e\in\{0,1\} e ∈ { 0 , 1 } , be a Hirzebruch surface embedded in P g \mathbb{P}^g P g by the complete linear series of H = C 0 + b f H=C_0+bf H = C 0 + b f , b ≥ e + 2 b\ge e+2 b ≥ e + 2 , as a rational normal scroll of degree g − 1 g-1 g − 1 , where g = 2 b − e + 1 g=2b-e+1 g = 2 b − e + 1 . Every genus g ≥ 6 g\ge6 g ≥ 6 arises in this way. Let Y ~ ⊂ P g \widetilde{Y}\subset\mathbb{P}^g Y ⊂ P g be the K 3 K3 K 3 carpet on Y Y Y (see Definition 2.1 ), that is, the unique double structure on Y Y Y with trivial dualizing sheaf and h 1 ( O Y ~ ) = 0 h^1(\mathcal{O}_{\widetilde{Y}})=0 h 1 ( O Y ) = 0 [20 ] , and let H ~ = O Y ~ ( 1 ) \widetilde{H}=\mathcal{O}_{\widetilde{Y}}(1) H = O Y ( 1 ) . By [20 ] (see also [7 ] , [19 ] , [17 ] , [18 ] ), Y ~ \widetilde{Y} Y is smoothable in P g \mathbb{P}^g P g to prime K 3 K3 K 3 surfaces of genus g g g (see Theorem 2.4 ), so that α ( X ) ≤ α ( Y ~ ) : = h 0 ( N Y ~ / P g ( − H ~ ) ) − g − 1 \alpha(X)\le\alpha(\widetilde{Y}):=h^0(N_{\widetilde{Y}/\mathbb{P}^g}(-\widetilde{H}))-g-1 α ( X ) ≤ α ( Y ) := h 0 ( N Y / P g ( − H )) − g − 1 for a general prime K 3 K3 K 3 surface X X X of genus g g g , and similarly for h 0 ( N X / P g ( − 2 ) ) h^0(N_{X/\mathbb{P}^g}(-2)) h 0 ( N X / P g ( − 2 )) (see (44) ). It is worth noting that by [4 ] , such a Y ~ \widetilde{Y} Y lies in the closure of a union of two scrolls meeting transversally along an anti-canonical elliptic curve, a degeneration that was used in [13 ] . As in [5 ] and [6 ] , the computability of α ( Y ~ ) \alpha(\widetilde{Y}) α ( Y ) is facilitated by the fact that Y ~ ⊂ P g \widetilde{Y}\subset\mathbb{P}^g Y ⊂ P g is an embedded ribbon which is a local complete intersection scheme whose reduced part is the scroll Y ⊂ P g Y\subset\mathbb{P}^g Y ⊂ P g . Three exact sequences (see Remark 3.1) reduce the computation of the cohomology of twists of the normal bundle N Y ~ / P g N_{\widetilde{Y}/\mathbb{P}^g} N Y / P g to the computation of the cohomology of twists of N Y / P g N_{Y/\mathbb{P}^g} N Y / P g and to the analysis of the related coboundary maps. The new difficulty in the prime case is that one of these coboundary maps cannot be controlled by cohomological means alone. We give a deformation theoretic interpretation of this map, which we explain in § 1.1 ; this is the key new ingredient of this article.
We now state our results. Let H g \mathcal{H}_g H g denote the Hilbert scheme of prime K 3 K3 K 3 surfaces of genus g g g (see Definition 2.3 and § 5.1 ), i.e., of smooth K 3 K3 K 3 surfaces X ⊂ P g X\subset\mathbb{P}^g X ⊂ P g embedded by the complete linear series of a primitive very ample line bundle B B B with B 2 = 2 g − 2 B^2=2g-2 B 2 = 2 g − 2 . It is irreducible, and a point of H g \mathcal{H}_g H g is called general if it lies in a suitable dense open subset. Our main theorem computes α ( Y ~ ) \alpha(\widetilde{Y}) α ( Y ) exactly, for K 3 K3 K 3 carpets on rational normal scrolls.
Theorem 1.3. (see Theorem 3.2, Proposition 4.2 , Proposition 4.3 ) Let Y = F e ⊂ P M Y=\mathbb{F}_e\subset\mathbb{P}^M Y = F e ⊂ P M , e ≥ 0 e\ge0 e ≥ 0 , be embedded by the complete linear series of H = C 0 + b f H=C_0+bf H = C 0 + b f with b ≥ e + 2 b\ge e+2 b ≥ e + 2 , so that M = 2 b − e + 1 M=2b-e+1 M = 2 b − e + 1 , and let Y ~ ⊂ P M \widetilde{Y}\subset\mathbb{P}^M Y ⊂ P M be the K 3 K3 K 3 carpet on Y Y Y .
(1) If b ≥ e + 3 b \ge e+3 b ≥ e + 3 or ( e , b ) = ( 1 , 3 ) (e,b)=(1,3) ( e , b ) = ( 1 , 3 ) , then
α ( Y ~ ) = h 0 ( O Y ( − H − 2 K Y ) ) = h 0 ( O Y ( 3 C 0 + ( 2 e + 4 − b ) f ) ) . \alpha(\widetilde{Y})=h^0(\mathcal{O}_Y(-H-2K_Y))=h^0(\mathcal{O}_Y(3C_0+(2e+4-b)f)). α ( Y ) = h 0 ( O Y ( − H − 2 K Y )) = h 0 ( O Y ( 3 C 0 + ( 2 e + 4 − b ) f )) .
In particular α ( Y ~ ) = 0 \alpha(\widetilde{Y})=0 α ( Y ) = 0 if and only if b ≥ 2 e + 5 b\ge2e+5 b ≥ 2 e + 5 .
(2) h 0 ( N Y ~ / P M ( − 2 H ~ ) ) ≤ h 0 ( O Y ( − 2 H − 2 K Y ) ) h^0(N_{\widetilde{Y}/\mathbb{P}M}(-2\widetilde{H}))\le h^0(\mathcal{O}_Y(-2H-2K_Y)) h 0 ( N Y / P M ( − 2 H )) ≤ h 0 ( O Y ( − 2 H − 2 K Y )) . In particular H 0 ( N Y ~ / P M ( − 2 H ~ ) ) = 0 H^0(N_{\widetilde{Y}/\mathbb{P}M}(-2\widetilde{H}))=0 H 0 ( N Y / P M ( − 2 H )) = 0 if b ≥ e + 3 b\ge e+3 b ≥ e + 3 , and h 0 ( N Y ~ / P M ( − 2 H ~ ) ) ≤ 1 h^0(N_{\widetilde{Y}/\mathbb{P}M}(-2\widetilde{H}))\le1 h 0 ( N Y / P M ( − 2 H )) ≤ 1 if ( e , b ) = ( 1 , 3 ) (e,b)=(1,3) ( e , b ) = ( 1 , 3 ) .
For general prime K3 surfaces we obtain:
Theorem 1.4. (see Theorem 4.1, Corollary 4.5, Corollary 4.8 ) Let X ⊂ P g X\subset\mathbb{P}^g X ⊂ P g be a general prime K3 surface of genus g ≥ 6 g\ge6 g ≥ 6 .
(1) If g = 12 g=12 g = 12 , then α ( X ) = 1 \alpha(X)=1 α ( X ) = 1 .
(2) If g = 10 g=10 g = 10 , then α ( X ) = 3 \alpha(X)=3 α ( X ) = 3 .
(3) If g = 9 g=9 g = 9 , then α ( X ) = 4 \alpha(X)=4 α ( X ) = 4 .
(4) If g = 8 g=8 g = 8 , then α ( X ) = 6 \alpha(X)=6 α ( X ) = 6 .
(5) If g = 7 g=7 g = 7 , then α ( X ) = 8 \alpha(X)=8 α ( X ) = 8 .
(6) If g = 6 g=6 g = 6 , then α ( X ) = 10 \alpha(X)=10 α ( X ) = 10 .
(7) If g = 11 g=11 g = 11 or g ≥ 13 g\ge13 g ≥ 13 , then α ( X ) = 0 \alpha(X)=0 α ( X ) = 0 and consequently X ⊂ P g X\subset\mathbb{P}^g X ⊂ P g is not extendable.
(8) H 0 ( N X / P g ( − 2 ) ) = 0 H^0(N_{X/\mathbb{P}^g}(-2))=0 H 0 ( N X / P g ( − 2 )) = 0 if g ≥ 7 g\ge7 g ≥ 7 , and h 0 ( N X / P g ( − 2 ) ) = 1 h^0(N_{X/\mathbb{P}^g}(-2))=1 h 0 ( N X / P g ( − 2 )) = 1 if g = 6 g=6 g = 6 .
Theorem 1.4 recovers the results of [13 ] on general prime K3 surfaces of every genus g ≥ 6 g\ge6 g ≥ 6 (for g ≤ 5 g\le5 g ≤ 5 a general prime K3 surface is a complete intersection and the results are classical), while the non-prime case was treated in [6 ] .
One of the key technical results proved in [13 ] and [14 ] , that crucially depends on the corank of the Wahl map, concerns the dimension of the component of the Hilbert scheme at points representing cones over canonical curves. Using Theorem 1.4 we prove this result directly for the cone over a K3 surface.
Corollary 1.5. (see Corollary 4.5) Let X ⊂ P g X\subset\mathbb{P}^g X ⊂ P g be a general prime K3 surface of genus g ≥ 6 g\ge6 g ≥ 6 and let C ( X ) ⊂ P g + 1 C(X)\subset\mathbb{P}^{g+1} C ( X ) ⊂ P g + 1 be the cone over X X X . Then the tangent space to the Hilbert scheme of P g + 1 \mathbb{P}^{g+1} P g + 1 at [ C ( X ) ] [C(X)] [ C ( X )] has dimension
dim T [ C ( X ) ] = 18 + ( g + 1 ) 2 + ( g + 1 ) + α ( X ) + h 0 ( N X / P g ( − 2 ) ) . \dim T_{[C(X)]}=18+(g+1)^2+(g+1)+\alpha(X)+h^0(N_{X/\mathbb{P}^g}(-2)). dim T [ C ( X )] = 18 + ( g + 1 ) 2 + ( g + 1 ) + α ( X ) + h 0 ( N X / P g ( − 2 )) .
The following theorem, which is a consequence of our results and a general principle outlined in [13 ] and [14 ] , gives a new proof of the classification of prime Fano threefolds of Picard rank one due to Fano, Iskovskikh and Mukai, and of the irreducibility of their Hilbert schemes.
Theorem 1.6. (see Theorem 4.7) Let g ≥ 6 g\ge6 g ≥ 6 and let V g \mathcal{V}_g V g denote the locus, in the Hilbert scheme of P g + 1 \mathbb{P}^{g+1} P g + 1 , of smooth Fano threefolds V ⊂ P g + 1 V\subset\mathbb{P}^{g+1} V ⊂ P g + 1 of Picard rank one, index one and genus g g g , embedded by the complete linear series ∣ − K V ∣ |-K_V| ∣ − K V ∣ (see Definition 4.4).
(1) V g = ∅ \mathcal{V}_g=\varnothing V g = ∅ if g = 11 g=11 g = 11 or g ≥ 13 g\ge13 g ≥ 13 .
(2) For 6 ≤ g ≤ 10 6\le g\le10 6 ≤ g ≤ 10 and g = 12 g=12 g = 12 , V g \mathcal{V}_g V g is irreducible, and the families of Fano and Iskovskikh form an open dense subset of V g \mathcal{V}_g V g .
The key new ingredient in the proof
We keep the notation of Theorem 1.3 and assume b ≥ e + 3 b\ge e+3 b ≥ e + 3 ; the case ( e , b ) = ( 1 , 3 ) (e,b)=(1,3) ( e , b ) = ( 1 , 3 ) requires a small additional argument (see Proposition 4.2). We write N Y : = N Y / P M N_Y:=N_{Y/\mathbb{P}^M} N Y := N Y / P M and N ~ Y : = N Y ~ / P M \widetilde{N}_Y:=N_{\widetilde{Y}/\mathbb{P}^M} N Y := N Y / P M . Let H ⊂ P M H\subset\mathbb{P}^M H ⊂ P M be a hyperplane meeting Y Y Y transversally along a smooth curve C : = Y ∩ H C:=Y\cap H C := Y ∩ H which is a rational normal curve of degree g − 1 g-1 g − 1 in H ≅ P g − 1 H\cong\mathbb{P}^{g-1} H ≅ P g − 1 (note that O Y ( H ) = O Y ( C ) \mathcal{O}_Y(H)=\mathcal{O}_Y(C) O Y ( H ) = O Y ( C ) ). Let C ~ : = Y ~ ∩ H \widetilde{C}:=\widetilde{Y}\cap H C := Y ∩ H , which is a canonical ribbon on C C C of arithmetic genus g g g (see [3 ] , [20 ] ). Put N C : = N C / H N_C:=N_{C/H} N C := N C / H and L C : = K Y ∣ C ≅ O P 1 ( − g − 1 ) L_C:=K_Y|_C\cong\mathcal{O}_{\mathbb{P}^1}(-g-1) L C := K Y ∣ C ≅ O P 1 ( − g − 1 ) , the conormal bundle of C ~ \widetilde{C} C .
The cohomological procedure of [5 ] , [6 ] , applied to the three exact sequences (2) –(4) , the normal sequence of Y ⊂ P M Y \subset\mathbb{P}^{M} Y ⊂ P M and the Euler sequence, gives (see (34) )
α ( Y ~ ) = h 0 ( O Y ( − H − 2 K Y ) ) + dim coker F , \alpha(\widetilde{Y}) = h^{0}\left(\mathcal{O}_{Y}(-H-2K_{Y})\right) + \dim\operatorname{coker} F, α ( Y ) = h 0 ( O Y ( − H − 2 K Y ) ) + dim coker F ,
where F : H 0 ( N Y ~ ⊗ O Y ( − H ) ) ⟶ H 1 ( N Y ~ ⊗ K Y ( − H ) ) F\colon H^{0}\left(N_{\widetilde{Y}}\otimes\mathcal{O}_{Y}(-H)\right)\longrightarrow H^{1}\left(N_{\widetilde{Y}}\otimes K_{Y}(-H)\right) F : H 0 ( N Y ⊗ O Y ( − H ) ) ⟶ H 1 ( N Y ⊗ K Y ( − H ) ) is the coboundary map of the exact sequence 0 ⟶ N Y ~ ⊗ K Y ⟶ N Y ~ ⟶ N Y ~ ⊗ O Y ⟶ 0 0\longrightarrow N_{\widetilde{Y}}\otimes K_{Y}\longrightarrow N_{\widetilde{Y}}\longrightarrow N_{\widetilde{Y}}\otimes\mathcal{O}_{Y}\longrightarrow0 0 ⟶ N Y ⊗ K Y ⟶ N Y ⟶ N Y ⊗ O Y ⟶ 0 , relating the normal bundle of the K 3 K3 K 3 carpet to its restriction to the scroll, twisted by O Y ~ ( − H ~ ) \mathcal{O}_{\widetilde{Y}}(-\widetilde{H}) O Y ( − H ) . Now Theorem 1.3 , (1) is equivalent to the surjectivity of F F F . The surjectivity of F F F is proved in the following steps.
Reduction to a map a a a between cohomology groups of twists of N Y N_{Y} N Y (see the proof of Theorem 3.2, Lemma 3.3 and Proposition 3.4 ). Using the sequences (3) and (4) , the group H 0 ( N Y ( − H ) ) H^{0}(N_{Y}(-H)) H 0 ( N Y ( − H )) injects into the source of F F F and H 1 ( N Y ⊗ K Y ( − H ) ) H^{1}(N_{Y}\otimes K_{Y}(-H)) H 1 ( N Y ⊗ K Y ( − H )) injects into its target and it is shown that F F F is surjective if and only if the induced map
a : H 0 ( N Y ( − H ) ) ⟶ H 1 ( N Y ⊗ K Y ( − H ) ) a\colon H^{0}(N_{Y}(-H))\longrightarrow H^{1}(N_{Y}\otimes K_{Y}(-H)) a : H 0 ( N Y ( − H )) ⟶ H 1 ( N Y ⊗ K Y ( − H ))
is surjective. Note that H 0 ( N Y ( − H ) ) = H 0 ( N Y ( − C ) ) H^{0}(N_{Y}(-H))=H^{0}(N_{Y}(-C)) H 0 ( N Y ( − H )) = H 0 ( N Y ( − C )) is the tangent space T [ Y ] S C T_{[Y]}\mathcal{S}_{C} T [ Y ] S C at [ Y ] [Y] [ Y ] to the space S C \mathcal{S}_{C} S C of smooth rational normal scrolls Y ′ ⊂ P M Y' \subset\mathbb{P}^{M} Y ′ ⊂ P M with Y ′ ∩ H = C Y'\cap H=C Y ′ ∩ H = C , which is smooth of dimension 2 ( g − 1 ) 2(g-1) 2 ( g − 1 ) at [ Y ] [Y] [ Y ] (see Lemma 3.7 ). S C \mathcal{S}_{C} S C is the fiber over C ⊂ H C\subset H C ⊂ H of the flag Hilbert scheme parametrizing pairs ( Y ′ , C ′ ) (Y',C') ( Y ′ , C ′ ) , where Y ′ Y' Y ′ is a rational normal scroll and C ′ C' C ′ is a hyperplane section of it, which is a rational normal curve.
The isomorphism δ ˉ \bar{\delta} δ ˉ (see Lemma 3.5, Lemma 3.6 and Lemma 3.8 ). We then obtain an isomorphism
δ ˉ : T [ C ~ ] R C ⟶ ∼ H 1 ( N Y ⊗ K Y ( − H ) ) \bar{\delta}\colon T_{[\widetilde{C}]}\mathcal{R}_{C}\overset{\sim}{\longrightarrow}H^{1}(N_{Y}\otimes K_{Y}(-H)) δ ˉ : T [ C ] R C ⟶ ∼ H 1 ( N Y ⊗ K Y ( − H ))
where R C : = P ( H 0 ( N C ⊗ L C ) ) ≅ P g − 3 \mathcal{R}_{C}:=\mathbb{P}\left(H^{0}(N_{C}\otimes L_{C})\right)\cong\mathbb{P}^{g-3} R C := P ( H 0 ( N C ⊗ L C ) ) ≅ P g − 3 , is the space of canonical ribbons on C C C . Thus we have to show that
δ ˉ − 1 ∘ a : T [ Y ] S C ⟶ T [ C ~ ] R C \bar{\delta}^{-1}\circ a\colon T_{[Y]}\mathcal{S}_{C}\longrightarrow T_{[\widetilde{C}]}\mathcal{R}_{C} δ ˉ − 1 ∘ a : T [ Y ] S C ⟶ T [ C ] R C
is surjective.
The map Ψ \Psi Ψ and the deformation theoretic interpretation of δ ˉ − 1 ∘ a \bar{\delta}^{-1}\circ a δ ˉ − 1 ∘ a (see Theorem 3.9 , Lemma 3.10 , Lemma 3.11 , Lemma 3.12 and Proposition 3.13). For Y ′ ∈ S C Y'\in\mathcal{S}_{C} Y ′ ∈ S C , let Y ~ ′ \widetilde{Y}' Y ′ be the unique K 3 K3 K 3 carpet on Y ′ Y' Y ′ . Then Y ~ ′ ∩ H \widetilde{Y}'\cap H Y ′ ∩ H is a canonical ribbon on C C C , and we show that the G H G_{H} G H invariant map (where G H G_{H} G H denotes the automorphisms of P g \mathbb{P}^{g} P g inducing identity on H H H )
Ψ : S C ⟶ R C , Y ′ ⟼ [ Y ~ ′ ∩ H ] \Psi\colon\mathcal{S}_{C}\longrightarrow\mathcal{R}_{C},\qquad Y'\longmapsto[\widetilde{Y}'\cap H] Ψ : S C ⟶ R C , Y ′ ⟼ [ Y ′ ∩ H ]
satisfies d Ψ = δ ˉ − 1 ∘ a d\Psi=\bar{\delta}^{-1}\circ a d Ψ = δ ˉ − 1 ∘ a . Thus it is enough to show that d Ψ d\Psi d Ψ is surjective.
Local sections of Ψ \Psi Ψ and surjectivity of d Ψ d\Psi d Ψ (see Lemma 3.14, Theorem 3.15 , Lemma 3.16 , Proposition 3.17 and Proposition 3.18 , Theorem A.3 ). We finally finish the proof by showing that around every point C ~ \widetilde{C} C of R C \mathcal{R}_{C} R C , there exists an open neighbourhood U U U of C ~ \widetilde{C} C and a section Γ U \Gamma_{U} Γ U of Ψ \Psi Ψ defined on U U U . Due to the G H G_{H} G H invariance of Ψ \Psi Ψ , this shows the surjectivity of d Ψ d\Psi d Ψ . For this the crucial ingredient is Theorem A.3 which we state below:
Let C ~ \widetilde{C} C be a ribbon on a smooth curve C C C of genus g C g_{C} g C with conormal bundle L C L_{C} L C . Then V : = Ω C ~ ∣ C V:=\left.\Omega_{\widetilde{C}}\right|_{C} V := Ω C C is a rank two vector bundle on C C C sitting in the restricted cotangent sequence 0 ⟶ L C ⟶ V ⟶ K C ⟶ 0 0\longrightarrow L_{C}\longrightarrow V\longrightarrow K_{C}\longrightarrow0 0 ⟶ L C ⟶ V ⟶ K C ⟶ 0 , and this sequence determines C ~ \widetilde{C} C (see [3 ] ). Let Y : = P ( V ) Y:=\mathbb{P}(V) Y := P ( V ) with projection p : Y ⟶ C p\colon Y\longrightarrow C p : Y ⟶ C and let i : C ↪ Y i\colon C\hookrightarrow Y i : C ↪ Y be the section defined by the surjection V ⟶ K C V\longrightarrow K_{C} V ⟶ K C . Then K Y ∣ C = L C K_{Y}|_{C}=L_{C} K Y ∣ C = L C . A K K K -trivial carpet on Y Y Y is a ribbon on Y Y Y with conormal bundle K Y K_{Y} K Y . Such carpets are parametrized by Ext 1 ( Ω Y , K Y ) \operatorname{Ext}^{1}(\Omega_{Y},K_{Y}) Ext 1 ( Ω Y , K Y ) (see [3 ] ).
Theorem 1.7. (see Theorem A.3 ) In the above notation, assume that C ~ \widetilde{C} C is non-split.
There is a unique K K K -trivial carpet Y ~ \widetilde{Y} Y on Y Y Y such that the embedding i : C ↪ Y i\colon C\hookrightarrow Y i : C ↪ Y extends to a closed embedding C ~ ↪ Y ~ \widetilde{C}\hookrightarrow\widetilde{Y} C ↪ Y inducing the identity K Y ∣ C = L C K_{Y}|_{C}=L_{C} K Y ∣ C = L C on conormal bundles. Moreover Y ~ \widetilde{Y} Y is non-split.
(2) Let g C = 0 g_C = 0 g C = 0 and let C ~ ⊂ P g − 1 \widetilde{C} \subset\mathbb{P}^{g-1} C ⊂ P g − 1 be a canonically embedded ribbon of arithmetic genus g g g , so that C ⊂ P g − 1 C \subset\mathbb{P}^{g-1} C ⊂ P g − 1 is a rational normal curve. Then O Y ( C ) \mathcal{O}_Y(C) O Y ( C ) is very ample. Embed Y Y Y in P g \mathbb{P}^{g} P g by its complete linear series, as a rational normal scroll, and identify P g − 1 \mathbb{P}^{g-1} P g − 1 with a hyperplane H ⊂ P g H \subset\mathbb{P}^{g} H ⊂ P g so that Y ∩ H = C Y \cap H = C Y ∩ H = C . Then the K 3 K3 K 3 carpet Y ~ ⊂ P g \widetilde{Y} \subset\mathbb{P}^{g} Y ⊂ P g on Y Y Y satisfies
Y ~ ∩ H = C ~ . \widetilde{Y} \cap H = \widetilde{C}. Y ∩ H = C .
The genus zero case of this result was stated in [3 ] . What is needed in step (4) is the precise form given in Theorem 1.7 (2): the canonical ribbon C ~ \widetilde{C} C is the hyperplane section of the K 3 K3 K 3 carpet on the rational normal scroll P ( Ω C ~ ∣ C ) \mathbb{P}(\Omega_{\widetilde{C}}|_C) P ( Ω C ∣ C ) canonically attached to C ~ \widetilde{C} C by its restricted cotangent sequence, and the embedding C ~ ↪ Y ~ \widetilde{C} \hookrightarrow\widetilde{Y} C ↪ Y extends the given section C ↪ Y C \hookrightarrow Y C ↪ Y .
Conventions. Throughout this article we work over an algebraically closed field k k k of characteristic zero. For a closed subscheme Z ⊂ P M Z \subset\mathbb{P}^{M} Z ⊂ P M we write N Z / P M = H o m ( I Z , O Z ) N_{Z/\mathbb{P}^{M}} = \mathcal{H}om(\mathcal{I}_Z,\mathcal{O}_Z) N Z / P M = H o m ( I Z , O Z ) for its normal sheaf.
Acknowledgements. We thank Professor Purnaprajna Bangere for many stimulating discussions on this topic.
AI Disclosure. Throughout, we have used Claude (Anthropic) Fable 5.1 for LaTeX typesetting, proofreading the arguments and for improving the presentation of the article. Lemma 3.3 has been worked out by Fable 5.1 replacing an earlier erroneous version where the author had claimed that h ∘ f = g h \circ f = g h ∘ f = g . It also suggested a correct and much shorter alternative argument for the injectivity of s s s in Step 2(b) of the proof of Proposition 4.3 , via the Yoneda product with the class ∂ ( s ) \partial(s) ∂ ( s ) of Lemma 3.5 and the Leray spectral sequence. The proof given here although longer, is more transparent to the author and follows [6 ] instead. It also has substantial contributions in helping the author write Proposition 3.13 in its present form. Finally Claude Fable 5.1 has claimed to have a different and more computational syzygy theoretic proof of Proposition 3.4 (which is the surjectivity of the coboundary map mentioned in the abstract) using equations and syzygies of K 3 K3 K 3 carpets described by Eisenbud and Schreyer in [15 ] , whose correctness has not yet been verified by the author and is a potential topic for further research. All mathematical statements and proofs in this article have been verified by the author, who takes full responsibility for them.
Preliminaries on degeneration of K3 surfaces into K3 double structures on Hirzebruch surfaces
We recall some definitions and a theorem on some special non-reduced degenerations of K 3 K3 K 3 surfaces in a projective space.
Definition 2.1. Let Y Y Y be a reduced connected projective variety.
A scheme Y ~ \widetilde{Y} Y is called a ribbon on Y Y Y with conormal bundle L L L if Y ~ r e d = Y \widetilde{Y}_{\mathrm{red}} = Y Y red = Y , I Y / Y ~ 2 = ( 0 ) \mathcal{I}_{Y/\widetilde{Y}}^{2} = (0) I Y / Y 2 = ( 0 ) and I Y / Y ~ ≅ L \mathcal{I}_{Y/\widetilde{Y}} \cong L I Y / Y ≅ L as an O Y \mathcal{O}_Y O Y module where I Y / Y ~ \mathcal{I}_{Y/\widetilde{Y}} I Y / Y is the ideal sheaf of Y Y Y inside Y ~ \widetilde{Y} Y and L L L is a line bundle on Y Y Y .
If Y Y Y is a smooth surface, then a ribbon Y ~ \widetilde{Y} Y on Y Y Y is called a K 3 K3 K 3 carpet if the dualizing sheaf K Y ~ ≅ O Y ~ K_{\widetilde{Y}} \cong\mathcal{O}_{\widetilde{Y}} K Y ≅ O Y and H 1 ( O Y ~ ) = 0 H^{1}(\mathcal{O}_{\widetilde{Y}}) = 0 H 1 ( O Y ) = 0 .
The following lemma characterizes the conormal bundle of a K 3 K3 K 3 carpet Y ~ \widetilde{Y} Y on a smooth surface Y Y Y .
Lemma 2.2. (see [16 Proposition 1.5] ) A ribbon Y ~ \widetilde{Y} Y on a smooth regular surface Y Y Y with conormal bundle L L L is a K 3 K3 K 3 carpet if and only if L ≅ K Y L \cong K_Y L ≅ K Y .
We recall the definition of the Hilbert scheme of K 3 K3 K 3 surfaces.
Definition 2.3. Let H r , g \mathcal{H}_{r,g} H r , g denote the Hilbert scheme of K 3 K3 K 3 surfaces of index r r r and genus g g g , i.e., K 3 K3 K 3 surfaces X ⊂ P 1 + r 2 ( g − 1 ) = P M X \subset\mathbb{P}^{1+r^{2}(g-1)} = \mathbb{P}^{M} X ⊂ P 1 + r 2 ( g − 1 ) = P M embedded by the complete linear series of r B rB r B where B B B is a primitive
very ample line bundle with B 2 = 2 g − 2 B^2 = 2g - 2 B 2 = 2 g − 2 . If r = 1 r = 1 r = 1 , we call H r , g \mathcal{H}_{r,g} H r , g the Hilbert scheme of prime K 3 K3 K 3 surfaces while if r ≥ 2 r \ge2 r ≥ 2 , H r , g \mathcal{H}_{r,g} H r , g is called the Hilbert scheme of non-prime K 3 K3 K 3 surfaces. Recall that dim H r , g = h 0 ( N X / P M ) = 18 + ( M + 1 ) 2 \dim\mathcal{H}_{r,g} = h^0(N_{X/\mathbb{P}^M}) = 18 + (M + 1)^2 dim H r , g = h 0 ( N X / P M ) = 18 + ( M + 1 ) 2 .
Next we recall a degeneration theorem of K 3 K3 K 3 surfaces into K 3 K3 K 3 carpets on Hirzebruch surfaces.
Theorem 2.4. ([7 Theorems 3.1, 4.3, 4.4, 5.1] and [20 Proposition 1.7, Theorems 3.5, 4.1] ) Let Y = F e Y = \mathbb{F}_e Y = F e be a Hirzebruch surface. Let H = a C 0 + b f H = aC_0 + bf H = a C 0 + b f be a very ample line bundle on Y Y Y , where C 0 C_0 C 0 is the class of section, satisfying C 0 2 = − e C_0^2 = -e C 0 2 = − e and f f f is the class of a fiber satisfying f 2 = 0 f^2 = 0 f 2 = 0 (this happens if and only if b ≥ a e + 1 b \ge ae + 1 b ≥ a e + 1 ). Consider the embedding Y ↪ P N Y \hookrightarrow\mathbb{P}^N Y ↪ P N induced by the complete linear series ∣ H ∣ |H| ∣ H ∣ .
(1) Then there exists K 3 K3 K 3 carpets Y ~ \widetilde{Y} Y on Y Y Y along with a very ample line bundle H ~ \widetilde{H} H on Y ~ \widetilde{Y} Y such that H ~ ∣ Y = H \widetilde{H}|_Y = H H ∣ Y = H .
(2) If Y ~ ↪ P M \widetilde{Y} \hookrightarrow\mathbb{P}^M Y ↪ P M denote the embedding induced by the complete linear series ∣ H ~ ∣ |\widetilde{H}| ∣ H ∣ , then Y ~ \widetilde{Y} Y is smoothable in P M \mathbb{P}^M P M into smooth K 3 K3 K 3 surfaces embedded by the complete linear series of a very ample line bundle.
(3) For 0 ≤ e ≤ 2 0 \le e \le2 0 ≤ e ≤ 2 , the Hilbert point of Y ~ \widetilde{Y} Y is smooth and for 0 ≤ e ≤ 1 0 \le e \le1 0 ≤ e ≤ 1 , the general smooth surface of the unique Hilbert component containing Y ~ \widetilde{Y} Y is a
(a) prime K 3 K3 K 3 surface if gcd ( a , b ) = 1 \gcd(a,b) = 1 g cd( a , b ) = 1
(b) a non-prime K 3 K3 K 3 surface embedded by r B rB r B , with r = gcd ( a , b ) r = \gcd(a,b) r = g cd( a , b ) and B B B is a primitive very ample line bundle.
Proof. Part (1), (2) and the statement on smoothness of Hilbert point in part (3) of the Theorem follows from [7 Theorem 3.1(3), Theorem 4.3, Theorem 4.4, Theorem 5.1] and [20 Proposition 1.7, Theorem 3.5, Theorem 4.1] .
We prove part (3). Let Y = F e Y = \mathbb{F}_e Y = F e with 0 ≤ e ≤ 2 0 \le e \le2 0 ≤ e ≤ 2 . Consider the K 3 K3 K 3 double cover of π : X → Y \pi: X \to Y π : X → Y branched along a very general smooth D ∈ ∣ − 2 K Y ∣ D \in|-2K_Y| D ∈ ∣ − 2 K Y ∣ . Let H 0 = π ∗ ( O Y ( 1 ) ) = π ∗ ( a C 0 + b f ) H_0 = \pi^*(\mathcal{O}_Y(1)) = \pi^*(aC_0 + bf) H 0 = π ∗ ( O Y ( 1 )) = π ∗ ( a C 0 + b f ) and consider the composition
X → φ P M π ↓ Y → P M \begin{CD}
X @>{\varphi}>> \mathbb{P}^M \\
@V{\pi}VV @. \\
Y @>>> \mathbb{P}^M
\end{CD} X π ↓ ⏐ Y φ P M P M
Recall the smoothing fibers X t → φ t P M X_t \xrightarrow{\varphi_t} \mathbb{P}^M X t φ t P M are obtained as the image of an embedding φ t \varphi_t φ t which is a general deformation of φ \varphi φ along a one parameter family T T T and are hence embedded inside P M \mathbb{P}^M P M by the complete linear series of a line bundle H t H_t H t which is a deformation of H 0 H_0 H 0 . Since the multiplicity of a polarization remains constant on a smooth family, we have to find the multiplicity of H 0 H_0 H 0 on the double cover X X X . Now since D D D is very general, smooth and irreducible part (3) for Y = F e Y = \mathbb{F}_e Y = F e , 0 ≤ e ≤ 1 0 \le e \le1 0 ≤ e ≤ 1 follows from the fact that Pic ( X ) \operatorname{Pic}(X) Pic ( X ) is generated by π ∗ O Y ( C 0 ) \pi^*\mathcal{O}_Y(C_0) π ∗ O Y ( C 0 ) and π ∗ O Y ( f ) \pi^*\mathcal{O}_Y(f) π ∗ O Y ( f ) (see [1 Proposition 6.3] and [1 Proposition 6.5 (i), (iii)] ).
Extendability of K3 carpets on rational normal scrolls
Set-up and notation
3.1. Set-up and notation. Throughout this section Y = F e Y = \mathbb{F}_e Y = F e , e ≥ 0 e \ge0 e ≥ 0 , is a Hirzebruch surface with ruling p : Y → P 1 p : Y \to\mathbb{P}^1 p : Y → P 1 , fibre class f f f and negative section C 0 C_0 C 0 (C 0 2 = − e C_0^2 = -e C 0 2 = − e ), so that K Y = − 2 C 0 − ( e + 2 ) f K_Y = -2C_0 - (e + 2)f K Y = − 2 C 0 − ( e + 2 ) f . We fix an integer b ≥ e + 3 b \ge e + 3 b ≥ e + 3 and put
H : = C 0 + b f . H := C_0 + bf. H := C 0 + b f .
The complete linear series ∣ H ∣ |H| ∣ H ∣ embeds Y Y Y into P M \mathbb{P}^M P M as a rational normal scroll of type ( b − e , b ) (b-e,b) ( b − e , b ) , where
M + 1 = h 0 ( O Y ( H ) ) = 2 b − e + 2. M + 1 = h^0(\mathcal{O}_Y(H)) = 2b - e + 2. M + 1 = h 0 ( O Y ( H )) = 2 b − e + 2.
We write g : = M = 2 b − e + 1 g := M = 2b - e + 1 g := M = 2 b − e + 1 ; it is the arithmetic genus of a hyperplane section of the K3 carpet below.
Let Y ~ ⊂ P M \widetilde{Y} \subset\mathbb{P}^{M} Y ⊂ P M be the K3 carpet on Y Y Y , i.e. the unique double structure on Y Y Y inside P M \mathbb{P}^{M} P M with I Y / I Y ~ ≅ K Y I_Y/I_{\widetilde{Y}} \cong K_Y I Y / I Y ≅ K Y [20 Proposition 1.7] . Following [3 ] , Y ~ \widetilde{Y} Y is determined by a surjection
s : N Y ∨ = I Y / I Y 2 ⟶ K Y , I Y ~ / I Y 2 = ker s , (1) s: N_Y^{\vee}=I_Y/I_Y^2 \longrightarrow K_Y,\qquad I_{\widetilde{Y}}/I_Y^2=\ker s, \tag*{(1)} s : N Y ∨ = I Y / I Y 2 ⟶ K Y , I Y / I Y 2 = ker s , (1)
and by [20 Proposition 1.7] such a surjection exists and is unique up to a nonzero scalar (we recover this in Lemma 3.5 below). The scheme Y ~ \widetilde{Y} Y is arithmetically Cohen–Macaulay, indeed arithmetically Gorenstein [15 Theorem 3.2] . We put H ~ : = O Y ~ ( 1 ) \widetilde{H}:=\mathcal{O}_{\widetilde{Y}}(1) H := O Y ( 1 ) .
Abbreviations. We write N Y : = N Y / P M N_Y:=N_{Y/\mathbb{P}^{M}} N Y := N Y / P M and N Y ~ : = N Y ~ / P M N_{\widetilde{Y}}:=N_{\widetilde{Y}/\mathbb{P}^{M}} N Y := N Y / P M for the normal sheaves, and
H : = H o m ( I Y ~ / I Y 2 , O Y ) . \mathcal{H}:=\mathcal{H}om(I_{\widetilde{Y}}/I_Y^2,\mathcal{O}_Y). H := H o m ( I Y / I Y 2 , O Y ) .
Conventions on twists. For a coherent sheaf F \mathcal{F} F on Y ~ \widetilde{Y} Y and a divisor class D D D on Y Y Y we write F ( D ) : = F ⊗ O Y ~ O Y ( D ) \mathcal{F}(D):=\mathcal{F}\otimes_{\mathcal{O}_{\widetilde{Y}}}\mathcal{O}_Y(D) F ( D ) := F ⊗ O Y O Y ( D ) ; this is a sheaf on Y Y Y . Thus N Y ~ ( − H ) = N Y ~ ⊗ O Y ( − H ) N_{\widetilde{Y}}(-H)=N_{\widetilde{Y}}\otimes\mathcal{O}_Y(-H) N Y ( − H ) = N Y ⊗ O Y ( − H ) and N Y ~ ( − H + K Y ) = N Y ~ ⊗ O Y ( − H + K Y ) N_{\widetilde{Y}}(-H+K_Y)=N_{\widetilde{Y}}\otimes\mathcal{O}_Y(-H+K_Y) N Y ( − H + K Y ) = N Y ⊗ O Y ( − H + K Y ) are sheaves on Y Y Y , whereas N Y ~ ( − H ~ ) = N Y ~ ⊗ O Y ~ ( − H ~ ) N_{\widetilde{Y}}(-\widetilde{H})=N_{\widetilde{Y}}\otimes\mathcal{O}_{\widetilde{Y}}(-\widetilde{H}) N Y ( − H ) = N Y ⊗ O Y ( − H ) is a sheaf on Y ~ \widetilde{Y} Y . For a sheaf on Y Y Y the twist has its usual meaning. Following [5 ] we set
α ( Y ~ ) : = h 0 ( N Y ~ ( − H ~ ) ) − ( M + 1 ) , \alpha(\widetilde{Y}):=h^0(N_{\widetilde{Y}}(-\widetilde{H}))-(M+1), α ( Y ) := h 0 ( N Y ( − H )) − ( M + 1 ) ,
the Zak–L’vovsky invariant of Y ~ ⊂ P M \widetilde{Y}\subset\mathbb{P}^{M} Y ⊂ P M .
Remark 3.1. For the carpet Y ~ \widetilde{Y} Y we have the following exact sequences (see [16 Lemma 4.1] )
0 ⟶ N Y ~ ( K Y ) ⟶ N Y ~ ⟶ N Y ~ ⊗ O Y ⟶ 0 (2) 0\longrightarrow N_{\widetilde{Y}}(K_Y)\longrightarrow N_{\widetilde{Y}}\longrightarrow N_{\widetilde{Y}}\otimes\mathcal{O}_Y\longrightarrow0 \tag*{(2)} 0 ⟶ N Y ( K Y ) ⟶ N Y ⟶ N Y ⊗ O Y ⟶ 0 (2)
0 ⟶ H ⟶ N Y ~ ⊗ O Y ⟶ O Y ( − 2 K Y ) ⟶ 0 (3) 0\longrightarrow\mathcal{H}\longrightarrow N_{\widetilde{Y}}\otimes\mathcal{O}_Y\longrightarrow\mathcal{O}_Y(-2K_Y)\longrightarrow0 \tag*{(3)} 0 ⟶ H ⟶ N Y ⊗ O Y ⟶ O Y ( − 2 K Y ) ⟶ 0 (3)
0 ⟶ O Y ( − K Y ) ⟶ N Y ⟶ H ⟶ 0 (4) 0\longrightarrow\mathcal{O}_Y(-K_Y)\longrightarrow N_Y\longrightarrow\mathcal{H}\longrightarrow0 \tag*{(4)} 0 ⟶ O Y ( − K Y ) ⟶ N Y ⟶ H ⟶ 0 (4)
In (4) the first map is the transpose of s s s in (1) and the second is restriction of homomorphisms from I Y / I Y 2 I_Y/I_Y^2 I Y / I Y 2 to its subsheaf I Y ~ / I Y 2 I_{\widetilde{Y}}/I_Y^2 I Y / I Y 2 . We shall also use the normal sequence of Y ⊂ P M Y\subset\mathbb{P}^{M} Y ⊂ P M ,
0 ⟶ T Y ⟶ T P M ∣ Y ⟶ N Y ⟶ 0 (5) 0\longrightarrow T_Y\longrightarrow T_{\mathbb{P}^{M}}|_Y\longrightarrow N_Y\longrightarrow0 \tag*{(5)} 0 ⟶ T Y ⟶ T P M ∣ Y ⟶ N Y ⟶ 0 (5)
and the Euler exact sequence restricted to Y Y Y ,
0 ⟶ O Y ⟶ O Y ( H ) ⊕ M + 1 ⟶ T P M ∣ Y ⟶ 0. (6) 0\longrightarrow\mathcal{O}_Y\longrightarrow\mathcal{O}_Y(H)^{\oplus M+1}\longrightarrow T_{\mathbb{P}^{M}}|_Y\longrightarrow0. \tag*{(6)} 0 ⟶ O Y ⟶ O Y ( H ) ⊕ M + 1 ⟶ T P M ∣ Y ⟶ 0. (6)
Main theorem
As mentioned before, to give an upper bound to α ( X ) \alpha(X) α ( X ) for a K3 surface X ⊂ P M X\subset\mathbb{P}^{M} X ⊂ P M , we will degenerate X ⊂ P M X\subset\mathbb{P}^{M} X ⊂ P M to a K3 carpet Y ~ ⊂ P M \widetilde{Y}\subset\mathbb{P}^{M} Y ⊂ P M , using Theorem 2.4 and compute α ( Y ~ ) \alpha(\widetilde{Y}) α ( Y ) . The main tool for all our computations that follow consist of the three exact sequences (2) , (3) and (4) that decomposes the cohomology of the twisted normal bundle N Y ~ / P M ( − H ~ ) N_{\widetilde{Y}/\mathbb{P}^{M}}(-\widetilde{H}) N Y / P M ( − H ) of a ribbon Y ~ ⊂ P M \widetilde{Y}\subset\mathbb{P}^{M} Y ⊂ P M in terms of cohomology of various twists of the normal bundle N Y / P M N_{Y/\mathbb{P}^{M}} N Y / P M of the reduced part.
Theorem 3.2. Let Y ~ ⊂ P M \widetilde{Y}\subset\mathbb{P}^{M} Y ⊂ P M be the K3 carpet on the rational normal scroll Y = F e ⊂ P M Y=F_e\subset\mathbb{P}^{M} Y = F e ⊂ P M embedded by ∣ H ∣ |H| ∣ H ∣ , H = C 0 + b f H=C_0+bf H = C 0 + b f , and assume b ≥ e + 3 b\ge e+3 b ≥ e + 3 . Then
α ( Y ~ ) = h 0 ( O Y ( − H − 2 K Y ) ) = h 0 ( O Y ( 3 C 0 + ( 2 e + 4 − b ) f ) ) . \alpha(\widetilde{Y})=h^0(\mathcal{O}_Y(-H-2K_Y))=h^0(\mathcal{O}_Y(3C_0+(2e+4-b)f)). α ( Y ) = h 0 ( O Y ( − H − 2 K Y )) = h 0 ( O Y ( 3 C 0 + ( 2 e + 4 − b ) f )) .
In particular α ( Y ~ ) = 0 \alpha(\widetilde{Y})=0 α ( Y ) = 0 if and only if b ≥ 2 e + 5 b\ge2e+5 b ≥ 2 e + 5 .
Proof. Tensoring sequence (2) by − H ~ -\widetilde{H} − H and taking the cohomology, we have
0 ⟶ H 0 ( N Y ~ ( − H + K Y ) ) ⟶ H 0 ( N Y ~ ( − H ~ ) ) ⟶ H 0 ( N Y ~ ( − H ) ) ⟶ H 1 ( N Y ~ ( − H + K Y ) ) ⟶ H 1 ( N Y ~ ( − H ~ ) ) ⟶ H 1 ( N Y ~ ( − H ) ) ⟶ H 2 ( N Y ~ ( − H + K Y ) ) ⟶ H 2 ( N Y ~ ( − H ~ ) ) ⟶ H 2 ( N Y ~ ( − H ) ) ⟶ 0 (7) \begin{aligned}
0 &\longrightarrow H^0(N_{\widetilde{Y}}(-H+K_Y))\longrightarrow H^0(N_{\widetilde{Y}}(-\widetilde{H}))\longrightarrow H^0(N_{\widetilde{Y}}(-H))\longrightarrow H^1(N_{\widetilde{Y}}(-H+K_Y))\\
&\longrightarrow H^1(N_{\widetilde{Y}}(-\widetilde{H}))\longrightarrow H^1(N_{\widetilde{Y}}(-H))\longrightarrow H^2(N_{\widetilde{Y}}(-H+K_Y))\\
&\longrightarrow H^2(N_{\widetilde{Y}}(-\widetilde{H}))\longrightarrow H^2(N_{\widetilde{Y}}(-H))\longrightarrow0 \tag*{(7)}
\end{aligned} 0 ⟶ H 0 ( N Y ( − H + K Y )) ⟶ H 0 ( N Y ( − H )) ⟶ H 0 ( N Y ( − H )) ⟶ H 1 ( N Y ( − H + K Y )) ⟶ H 1 ( N Y ( − H )) ⟶ H 1 ( N Y ( − H )) ⟶ H 2 ( N Y ( − H + K Y )) ⟶ H 2 ( N Y ( − H )) ⟶ H 2 ( N Y ( − H )) ⟶ 0 (7)
We compute H i ( N Y ~ ( − H + K Y ) ) H^i(N_{\widetilde{Y}}(-H+K_Y)) H i ( N Y ( − H + K Y )) for i = 0 , 1 , 2 i=0,1,2 i = 0 , 1 , 2 . Tensoring sequence (3) by − H + K Y -H+K_Y − H + K Y and taking the cohomology, we have that
0 → H 0 ( H ( − H + K Y ) ) → H 0 ( N Y ~ ( − H + K Y ) ) → H 0 ( O Y ( − H − K Y ) ) → H 1 ( H ( − H + K Y ) ) → H 1 ( N Y ~ ( − H + K Y ) ) → H 1 ( O Y ( − H − K Y ) ) → H 2 ( H ( − H + K Y ) ) → H 2 ( N Y ~ ( − H + K Y ) ) → H 2 ( O Y ( − H − K Y ) ) → 0 (8) \begin{aligned}
0 &\to H^0(\mathcal{H}(-H+K_Y)) \to H^0(N_{\widetilde{Y}}(-H+K_Y)) \to H^0(\mathcal{O}_Y(-H-K_Y)) \to H^1(\mathcal{H}(-H+K_Y))\\
&\to H^1(N_{\widetilde{Y}}(-H+K_Y)) \to H^1(\mathcal{O}_Y(-H-K_Y)) \to H^2(\mathcal{H}(-H+K_Y)) \to H^2(N_{\widetilde{Y}}(-H+K_Y))\\
&\to H^2(\mathcal{O}_Y(-H-K_Y)) \to0 \tag*{(8)}
\end{aligned} 0 → H 0 ( H ( − H + K Y )) → H 0 ( N Y ( − H + K Y )) → H 0 ( O Y ( − H − K Y )) → H 1 ( H ( − H + K Y )) → H 1 ( N Y ( − H + K Y )) → H 1 ( O Y ( − H − K Y )) → H 2 ( H ( − H + K Y )) → H 2 ( N Y ( − H + K Y )) → H 2 ( O Y ( − H − K Y )) → 0 (8)
Note that since H = C 0 + b f H=C_0+bf H = C 0 + b f , we have that H 0 ( − H − K Y ) = H 0 ( C 0 + ( e + 2 − b ) f ) = 0 H^0(-H-K_Y)=H^0(C_0+(e+2-b)f)=0 H 0 ( − H − K Y ) = H 0 ( C 0 + ( e + 2 − b ) f ) = 0 if b ≥ e + 3 b\ge e+3 b ≥ e + 3 . Further, h 2 ( − H − K Y ) = h 0 ( H + 2 K Y ) = h 0 ( − 3 C 0 + ( b − 2 e − 4 ) f ) = 0 h^2(-H-K_Y)=h^0(H+2K_Y)=h^0(-3C_0+(b-2e-4)f)=0 h 2 ( − H − K Y ) = h 0 ( H + 2 K Y ) = h 0 ( − 3 C 0 + ( b − 2 e − 4 ) f ) = 0 .
Tensoring sequence (4) by − H + K Y -H+K_Y − H + K Y and taking the cohomology, we have that
0 → H 0 ( − H ) = ( 0 ) → H 0 ( N Y ( − H + K Y ) ) → H 0 ( H ( − H + K Y ) ) → H 1 ( − H ) = ( 0 ) → H 1 ( N Y ( − H + K Y ) ) → H 1 ( H ( − H + K Y ) ) → H 2 ( − H ) = ( 0 ) → H 2 ( N Y ( − H + K Y ) ) → H 2 ( H ( − H + K Y ) ) → 0 (9) \begin{aligned}
0 &\to H^0(-H)=(0)\to H^0(N_Y(-H+K_Y))\to H^0(\mathcal{H}(-H+K_Y))\to H^1(-H)=(0)\\
&\to H^1(N_Y(-H+K_Y))\to H^1(\mathcal{H}(-H+K_Y))\to H^2(-H)=(0)\\
&\to H^2(N_Y(-H+K_Y))\to H^2(\mathcal{H}(-H+K_Y))\to0 \tag*{(9)}
\end{aligned} 0 → H 0 ( − H ) = ( 0 ) → H 0 ( N Y ( − H + K Y )) → H 0 ( H ( − H + K Y )) → H 1 ( − H ) = ( 0 ) → H 1 ( N Y ( − H + K Y )) → H 1 ( H ( − H + K Y )) → H 2 ( − H ) = ( 0 ) → H 2 ( N Y ( − H + K Y )) → H 2 ( H ( − H + K Y )) → 0 (9)
Now using the Euler exact sequence (6) tensored with − H + K Y -H+K_Y − H + K Y , we have that
0 → H 0 ( − H + K Y ) → H 0 ( K Y ) ⊕ M + 1 = ( 0 ) → H 0 ( T P M ∣ Y ( − H + K Y ) ) → H 1 ( − H + K Y ) → H 1 ( K Y ) ⊕ M + 1 → H 1 ( T P M ∣ Y ( − H + K Y ) ) → H 2 ( − H + K Y ) → H 2 ( K Y ) ⊕ M + 1 → H 2 ( T P M ∣ Y ( − H + K Y ) ) → 0 (10) \begin{aligned}
0 &\to H^0(-H+K_Y)\to H^0(K_Y)^{\oplus M+1}=(0)\to H^0(T_{\mathbb{P}^M|Y}(-H+K_Y))\to H^1(-H+K_Y)\\
&\to H^1(K_Y)^{\oplus M+1}\to H^1(T_{\mathbb{P}^M|Y}(-H+K_Y))\to H^2(-H+K_Y)\\
&\to H^2(K_Y)^{\oplus M+1}\to H^2(T_{\mathbb{P}^M|Y}(-H+K_Y))\to0 \tag*{(10)}
\end{aligned} 0 → H 0 ( − H + K Y ) → H 0 ( K Y ) ⊕ M + 1 = ( 0 ) → H 0 ( T P M ∣ Y ( − H + K Y )) → H 1 ( − H + K Y ) → H 1 ( K Y ) ⊕ M + 1 → H 1 ( T P M ∣ Y ( − H + K Y )) → H 2 ( − H + K Y ) → H 2 ( K Y ) ⊕ M + 1 → H 2 ( T P M ∣ Y ( − H + K Y )) → 0 (10)
We have H 1 ( − H + K Y ) = H 1 ( H ) ∨ = 0 H^1(-H+K_Y)=H^1(H)^\vee=0 H 1 ( − H + K Y ) = H 1 ( H ) ∨ = 0 by Kodaira vanishing, and H 1 ( K Y ) = 0 H^1(K_Y)=0 H 1 ( K Y ) = 0 . Further the map H 2 ( − H + K Y ) → H 2 ( K Y ) ⊕ M + 1 = H 2 ( K Y ) ⊗ H 0 ( H ) H^2(-H+K_Y)\to H^2(K_Y)^{\oplus M+1}=H^2(K_Y)\otimes H^0(H) H 2 ( − H + K Y ) → H 2 ( K Y ) ⊕ M + 1 = H 2 ( K Y ) ⊗ H 0 ( H ) is an isomorphism since it is the dual of the isomorphism H 0 ( O Y ) ⊗ H 0 ( H ) → H 0 ( H ) H^0(\mathcal{O}_Y)\otimes H^0(H)\to H^0(H) H 0 ( O Y ) ⊗ H 0 ( H ) → H 0 ( H ) . Putting all this together in sequence (10) , we have
H i ( T P M ∣ Y ( − H + K Y ) ) = 0 , i = 0 , 1 , 2. (11) H^i(T_{\mathbb{P}^M|Y}(-H+K_Y))=0,\qquad i=0,1,2. \tag*{(11)} H i ( T P M ∣ Y ( − H + K Y )) = 0 , i = 0 , 1 , 2. (11)
On the other hand, from the sequence
0 → T Y / P 1 ( − H + K Y ) = − C 0 − ( b + 2 ) f → T Y ( − H + K Y ) → p ∗ T P 1 ( − H + K Y ) = − 3 C 0 − ( b + e ) f → 0 (12) 0\to T_{Y/\mathbb{P}^1}(-H+K_Y)=-C_0-(b+2)f\to T_Y(-H+K_Y)\to p^*T_{\mathbb{P}^1}(-H+K_Y)=-3C_0-(b+e)f\to0 \tag*{(12)} 0 → T Y / P 1 ( − H + K Y ) = − C 0 − ( b + 2 ) f → T Y ( − H + K Y ) → p ∗ T P 1 ( − H + K Y ) = − 3 C 0 − ( b + e ) f → 0 (12)
we have that, if b ≥ e + 1 b\ge e+1 b ≥ e + 1 ,
H i ( T Y ( − H + K Y ) ) = 0 , i = 0 , 1 , H 2 ( T Y ( − H + K Y ) ) = H 0 ( C 0 + ( b − 2 ) f ) ∨ . (13) H^i(T_Y(-H+K_Y))=0,\quad i=0,1,\qquad H^2(T_Y(-H+K_Y))=H^0(C_0+(b-2)f)^\vee. \tag*{(13)} H i ( T Y ( − H + K Y )) = 0 , i = 0 , 1 , H 2 ( T Y ( − H + K Y )) = H 0 ( C 0 + ( b − 2 ) f ) ∨ . (13)
Hence from the sequence (5) tensored with − H + K Y -H+K_Y − H + K Y we conclude that for b ≥ e + 1 b\ge e+1 b ≥ e + 1 ,
H 0 ( N Y ( − H + K Y ) ) = 0 , H 2 ( N Y ( − H + K Y ) ) = 0 , H 1 ( N Y ( − H + K Y ) ) = H 2 ( T Y ( − H + K Y ) ) = H 0 ( C 0 + ( b − 2 ) f ) ∨ . (14) \begin{aligned}
H^0(N_Y(-H+K_Y))&=0,\qquad H^2(N_Y(-H+K_Y))=0,\\
H^1(N_Y(-H+K_Y))&=H^2(T_Y(-H+K_Y))=H^0(C_0+(b-2)f)^\vee. \tag*{(14)}
\end{aligned} H 0 ( N Y ( − H + K Y )) H 1 ( N Y ( − H + K Y )) = 0 , H 2 ( N Y ( − H + K Y )) = 0 , = H 2 ( T Y ( − H + K Y )) = H 0 ( C 0 + ( b − 2 ) f ) ∨ . (14)
Plugging everything back into sequence (9) , we have that
H 0 ( H ( − H + K Y ) ) = 0 , H 2 ( H ( − H + K Y ) ) = 0 , H 1 ( H ( − H + K Y ) ) = H 2 ( T Y ( − H + K Y ) ) = H 0 ( C 0 + ( b − 2 ) f ) ∨ . (15) \begin{aligned}
H^0(\mathcal{H}(-H+K_Y))&=0,\qquad H^2(\mathcal{H}(-H+K_Y))=0,\\
H^1(\mathcal{H}(-H+K_Y))&=H^2(T_Y(-H+K_Y))=H^0(C_0+(b-2)f)^\vee. \tag*{(15)}
\end{aligned} H 0 ( H ( − H + K Y )) H 1 ( H ( − H + K Y )) = 0 , H 2 ( H ( − H + K Y )) = 0 , = H 2 ( T Y ( − H + K Y )) = H 0 ( C 0 + ( b − 2 ) f ) ∨ . (15)
Going back to sequence (8) , and noting that H 0 ( − H − K Y ) = 0 H^0(-H-K_Y)=0 H 0 ( − H − K Y ) = 0 for b ≥ e + 3 b\ge e+3 b ≥ e + 3 , we have that for b ≥ e + 3 b\ge e+3 b ≥ e + 3 ,
H 0 ( N Y ~ ( − H + K Y ) ) = 0 , H 2 ( N Y ~ ( − H + K Y ) ) = 0 , (16) H^0(N_{\widetilde{Y}}(-H+K_Y))=0,\qquad H^2(N_{\widetilde{Y}}(-H+K_Y))=0, \tag*{(16)} H 0 ( N Y ( − H + K Y )) = 0 , H 2 ( N Y ( − H + K Y )) = 0 , (16)
and an exact sequence
0 → H 2 ( T Y ( − H + K Y ) ) → H 1 ( N Y ~ ( − H + K Y ) ) → H 1 ( − H − K Y ) → 0 (17) 0\to H^2(T_Y(-H+K_Y))\to H^1(N_{\widetilde{Y}}(-H+K_Y))\to H^1(-H-K_Y)\to0 \tag*{(17)} 0 → H 2 ( T Y ( − H + K Y )) → H 1 ( N Y ( − H + K Y )) → H 1 ( − H − K Y ) → 0 (17)
yielding h 1 ( N Y ~ ( − H + K Y ) ) = h 2 ( T Y ( − H + K Y ) ) + h 1 ( − H − K Y ) h^1(N_{\widetilde{Y}}(-H+K_Y))=h^2(T_Y(-H+K_Y))+h^1(-H-K_Y) h 1 ( N Y ( − H + K Y )) = h 2 ( T Y ( − H + K Y )) + h 1 ( − H − K Y ) .
Now we compute H i ( N ~ Y ( − H ) ) H^i(\widetilde{N}_Y(-H)) H i ( N Y ( − H )) for i = 0 , 1 , 2 i = 0,1,2 i = 0 , 1 , 2 . Tensoring sequence (3) by − H -H − H and taking the cohomology, we have that
0 → H 0 ( H ( − H ) ) → H 0 ( N ~ Y ( − H ) ) → H 0 ( O Y ( − H − 2 K Y ) ) → H 1 ( H ( − H ) ) → H 1 ( N ~ Y ( − H ) ) → H 1 ( O Y ( − H − 2 K Y ) ) → H 2 ( H ( − H ) ) → H 2 ( N ~ Y ( − H ) ) → H 2 ( O Y ( − H − 2 K Y ) ) → 0 (18) \begin{aligned}
0 &\to H^0(\mathcal{H}(-H)) \to H^0(\widetilde{N}_Y(-H)) \to H^0(\mathcal{O}_Y(-H-2K_Y)) \to H^1(\mathcal{H}(-H)) \to H^1(\widetilde{N}_Y(-H)) \\
&\to H^1(\mathcal{O}_Y(-H-2K_Y)) \to H^2(\mathcal{H}(-H)) \to H^2(\widetilde{N}_Y(-H)) \to H^2(\mathcal{O}_Y(-H-2K_Y)) \to0 \tag*{(18)}
\end{aligned} 0 → H 0 ( H ( − H )) → H 0 ( N Y ( − H )) → H 0 ( O Y ( − H − 2 K Y )) → H 1 ( H ( − H )) → H 1 ( N Y ( − H )) → H 1 ( O Y ( − H − 2 K Y )) → H 2 ( H ( − H )) → H 2 ( N Y ( − H )) → H 2 ( O Y ( − H − 2 K Y )) → 0 (18)
Note that h 2 ( − H − 2 K Y ) = h 0 ( H + 3 K Y ) = h 0 ( − 5 C 0 + ( b − 3 e − 6 ) f ) = 0 h^2(-H-2K_Y)=h^0(H+3K_Y)=h^0(-5C_0+(b-3e-6)f)=0 h 2 ( − H − 2 K Y ) = h 0 ( H + 3 K Y ) = h 0 ( − 5 C 0 + ( b − 3 e − 6 ) f ) = 0 .
Tensoring sequence (4) by − H -H − H and taking the cohomology, we have that
0 → H 0 ( − H − K Y ) = ( 0 ) → H 0 ( N Y ( − H ) ) → H 0 ( H ( − H ) ) → H 1 ( − H − K Y ) → H 1 ( N Y ( − H ) ) → H 1 ( H ( − H ) ) → H 2 ( − H − K Y ) = ( 0 ) → H 2 ( N Y ( − H ) ) → H 2 ( H ( − H ) ) → 0 (19) \begin{aligned}
0 &\to H^0(-H-K_Y)=(0) \to H^0(N_Y(-H)) \to H^0(\mathcal{H}(-H)) \to H^1(-H-K_Y) \\
&\to H^1(N_Y(-H)) \to H^1(\mathcal{H}(-H)) \to H^2(-H-K_Y)=(0) \to H^2(N_Y(-H)) \to H^2(\mathcal{H}(-H)) \to0 \tag*{(19)}
\end{aligned} 0 → H 0 ( − H − K Y ) = ( 0 ) → H 0 ( N Y ( − H )) → H 0 ( H ( − H )) → H 1 ( − H − K Y ) → H 1 ( N Y ( − H )) → H 1 ( H ( − H )) → H 2 ( − H − K Y ) = ( 0 ) → H 2 ( N Y ( − H )) → H 2 ( H ( − H )) → 0 (19)
Now using the Euler exact sequence (6) tensored with − H -H − H , we have that
0 → H 0 ( − H ) → H 0 ( O Y ) ⊕ M + 1 → H 0 ( T P M ∣ Y ( − H ) ) → H 1 ( − H ) = ( 0 ) → H 1 ( O Y ) ⊕ M + 1 = ( 0 ) → H 1 ( T P M ∣ Y ( − H ) ) → H 2 ( − H ) → H 2 ( O Y ) ⊕ M + 1 → H 2 ( T P M ∣ Y ( − H ) ) → 0 (20) \begin{aligned}
0 &\to H^0(-H) \to H^0(\mathcal{O}_Y)^{\oplus M+1} \to H^0(T_{\mathbb{P}^M|Y}(-H)) \to H^1(-H)=(0) \to H^1(\mathcal{O}_Y)^{\oplus M+1}=(0) \\
&\to H^1(T_{\mathbb{P}^M|Y}(-H)) \to H^2(-H) \to H^2(\mathcal{O}_Y)^{\oplus M+1} \to H^2(T_{\mathbb{P}^M|Y}(-H)) \to0 \tag*{(20)}
\end{aligned} 0 → H 0 ( − H ) → H 0 ( O Y ) ⊕ M + 1 → H 0 ( T P M ∣ Y ( − H )) → H 1 ( − H ) = ( 0 ) → H 1 ( O Y ) ⊕ M + 1 = ( 0 ) → H 1 ( T P M ∣ Y ( − H )) → H 2 ( − H ) → H 2 ( O Y ) ⊕ M + 1 → H 2 ( T P M ∣ Y ( − H )) → 0 (20)
We have H 0 ( − H ) = 0 H^0(-H)=0 H 0 ( − H ) = 0 , H 1 ( − H ) = 0 H^1(-H)=0 H 1 ( − H ) = 0 by Kodaira vanishing theorem, H 2 ( − H ) = 0 H^2(-H)=0 H 2 ( − H ) = 0 by pushing forward to P 1 \mathbb{P}^1 P 1 since a = 1 a=1 a = 1 . Putting all this together in sequence (20) , we have
h 0 ( T P M ∣ Y ( − H ) ) = M + 1 , h 1 ( T P M ∣ Y ( − H ) ) = h 2 ( T P M ∣ Y ( − H ) ) = 0. (21) h^0(T_{\mathbb{P}^M|Y}(-H))=M+1,\qquad h^1(T_{\mathbb{P}^M|Y}(-H))=h^2(T_{\mathbb{P}^M|Y}(-H))=0. \tag*{(21)} h 0 ( T P M ∣ Y ( − H )) = M + 1 , h 1 ( T P M ∣ Y ( − H )) = h 2 ( T P M ∣ Y ( − H )) = 0. (21)
On the other hand, from the sequence
0 → T Y / P 1 ( − H ) = C 0 − ( b − e ) f → T Y ( − H ) → p ∗ T P 1 ( − H ) = − C 0 − ( b − 2 ) f → 0 (22) 0\to T_{Y/\mathbb{P}^1}(-H)=C_0-(b-e)f\to T_Y(-H)\to p^*T_{\mathbb{P}^1}(-H)=-C_0-(b-2)f\to0 \tag*{(22)} 0 → T Y / P 1 ( − H ) = C 0 − ( b − e ) f → T Y ( − H ) → p ∗ T P 1 ( − H ) = − C 0 − ( b − 2 ) f → 0 (22)
we have that if b ≥ e + 1 b\ge e+1 b ≥ e + 1
H i ( T Y ( − H ) ) = 0 , i = 0 , 2 , H 1 ( T Y ( − H ) ) = H 1 ( C 0 − ( b − e ) f ) . (23) H^i(T_Y(-H))=0,\ i=0,2,\qquad H^1(T_Y(-H))=H^1(C_0-(b-e)f). \tag*{(23)} H i ( T Y ( − H )) = 0 , i = 0 , 2 , H 1 ( T Y ( − H )) = H 1 ( C 0 − ( b − e ) f ) . (23)
Hence from the sequence (5) tensored with − H -H − H we conclude that for b ≥ e + 1 b\ge e+1 b ≥ e + 1 ,
H 1 ( N Y ( − H ) ) = H 2 ( N Y ( − H ) ) = 0 (24) H^1(N_Y(-H))=H^2(N_Y(-H))=0 \tag*{(24)} H 1 ( N Y ( − H )) = H 2 ( N Y ( − H )) = 0 (24)
and we have an exact sequence
0 → H 0 ( T P M ∣ Y ( − H ) ) = C M + 1 → H 0 ( N Y ( − H ) ) → H 1 ( T Y ( − H ) ) → 0. (25) 0\to H^0(T_{\mathbb{P}^M|Y}(-H))=\mathbb{C}^{M+1}\to H^0(N_Y(-H))\to H^1(T_Y(-H))\to0. \tag*{(25)} 0 → H 0 ( T P M ∣ Y ( − H )) = C M + 1 → H 0 ( N Y ( − H )) → H 1 ( T Y ( − H )) → 0. (25)
Plugging everything back into sequence (19) , we have that
H 1 ( H ( − H ) ) = H 2 ( H ( − H ) ) = 0 (26) H^1(\mathcal{H}(-H))=H^2(\mathcal{H}(-H))=0 \tag*{(26)} H 1 ( H ( − H )) = H 2 ( H ( − H )) = 0 (26)
and an exact sequence
0 → H 0 ( N Y ( − H ) ) → H 0 ( H ( − H ) ) → H 1 ( − H − K Y ) → 0. (27) 0\to H^0(N_Y(-H))\to H^0(\mathcal{H}(-H))\to H^1(-H-K_Y)\to0. \tag*{(27)} 0 → H 0 ( N Y ( − H )) → H 0 ( H ( − H )) → H 1 ( − H − K Y ) → 0. (27)
Going back to sequence (18) , we have that for b ≥ e + 3 b\ge e+3 b ≥ e + 3
0 → H 0 ( H ( − H ) ) → H 0 ( N ~ Y ( − H ) ) → H 0 ( − H − 2 K Y ) → 0 , H 1 ( N ~ Y ( − H ) ) = H 1 ( − H − 2 K Y ) , H 2 ( N ~ Y ( − H ) ) = 0. (28) \begin{aligned}
0&\to H^0(\mathcal{H}(-H))\to H^0(\widetilde{N}_Y(-H))\to H^0(-H-2K_Y)\to0,\\
H^1(\widetilde{N}_Y(-H))&=H^1(-H-2K_Y),\qquad H^2(\widetilde{N}_Y(-H))=0. \tag*{(28)}
\end{aligned} 0 H 1 ( N Y ( − H )) → H 0 ( H ( − H )) → H 0 ( N Y ( − H )) → H 0 ( − H − 2 K Y ) → 0 , = H 1 ( − H − 2 K Y ) , H 2 ( N Y ( − H )) = 0. (28)
Plugging (16) , (17) , (28) into the exact sequence (7) , we have
h 0 ( N ~ Y ( − H ~ ) ) = h 0 ( N ~ Y ( − H ) ) − h 1 ( N ~ Y ( − H + K Y ) ) + h 1 ( N ~ Y ( − H ~ ) ) − h 1 ( N ~ Y ( − H ) ) = M + 1 + h 1 ( T Y ( − H ) ) + h 1 ( − H − K Y ) + h 0 ( − H − 2 K Y ) (29) \begin{aligned}
h^0(\widetilde{N}_Y(-\widetilde{H}))={}&h^0(\widetilde{N}_Y(-H))-h^1(\widetilde{N}_Y(-H+K_Y))+h^1(\widetilde{N}_Y(-\widetilde{H}))-h^1(\widetilde{N}_Y(-H))\\
={}&M+1+h^1(T_Y(-H))+h^1(-H-K_Y)+h^0(-H-2K_Y) \tag*{(29)}
\end{aligned} h 0 ( N Y ( − H )) = = h 0 ( N Y ( − H )) − h 1 ( N Y ( − H + K Y )) + h 1 ( N Y ( − H )) − h 1 ( N Y ( − H )) M + 1 + h 1 ( T Y ( − H )) + h 1 ( − H − K Y ) + h 0 ( − H − 2 K Y ) (29)
− h 2 ( T Y ( − H + K Y ) ) − h 1 ( − H − K Y ) + h 1 ( N ~ Y ( − H ~ ) ) − h 1 ( − H − 2 K Y ) . (30) -h^2(T_Y(-H+K_Y))-h^1(-H-K_Y)+h^1(\widetilde{N}_Y(-\widetilde{H}))-h^1(-H-2K_Y). \tag*{(30)} − h 2 ( T Y ( − H + K Y )) − h 1 ( − H − K Y ) + h 1 ( N Y ( − H )) − h 1 ( − H − 2 K Y ) . (30)
Now note that by (23) , h 1 ( T Y ( − H ) ) = h 1 ( C 0 − ( b − e ) f ) = 2 b − e − 2 h^1(T_Y(-H))=h^1(C_0-(b-e)f)=2b-e-2 h 1 ( T Y ( − H )) = h 1 ( C 0 − ( b − e ) f ) = 2 b − e − 2 and h 2 ( T Y ( − H + K Y ) ) = h 0 ( C 0 + ( b − 2 ) f ) = 2 b − e − 2 h^2(T_Y(-H+K_Y))=h^0(C_0+(b-2)f)=2b-e-2 h 2 ( T Y ( − H + K Y )) = h 0 ( C 0 + ( b − 2 ) f ) = 2 b − e − 2 . Hence
h 0 ( N Y ~ ( − H ~ ) ) = M + 1 + h 0 ( − H − 2 K Y ) + h 1 ( N Y ~ ( − H ~ ) ) − h 1 ( − H − 2 K Y ) . (31) h^0(N_{\widetilde{Y}}(-\widetilde{H}))=M+1+h^0(-H-2K_Y)+h^1(N_{\widetilde{Y}}(-\widetilde{H}))-h^1(-H-2K_Y). \tag*{(31)} h 0 ( N Y ( − H )) = M + 1 + h 0 ( − H − 2 K Y ) + h 1 ( N Y ( − H )) − h 1 ( − H − 2 K Y ) . (31)
that is,
α ( Y ~ ) = h 0 ( − H − 2 K Y ) + h 1 ( N Y ~ ( − H ~ ) ) − h 1 ( − H − 2 K Y ) . (32) \alpha(\widetilde{Y})=h^0(-H-2K_Y)+h^1(N_{\widetilde{Y}}(-\widetilde{H}))-h^1(-H-2K_Y). \tag*{(32)} α ( Y ) = h 0 ( − H − 2 K Y ) + h 1 ( N Y ( − H )) − h 1 ( − H − 2 K Y ) . (32)
We end the proof by showing h 1 ( N Y ~ ( − H ~ ) ) = h 1 ( − H − 2 K Y ) h^1(N_{\widetilde{Y}}(-\widetilde{H}))=h^1(-H-2K_Y) h 1 ( N Y ( − H )) = h 1 ( − H − 2 K Y ) if b ≥ e + 3 b\geq e+3 b ≥ e + 3 . Note that from sequence (7) and equations (16) and (28) , we already have
0 → H 0 ( N Y ~ ( − H ~ ) ) → H 0 ( N Y ~ ( − H ) ) → F H 1 ( N Y ~ ( − H + K Y ) ) → H 1 ( N Y ~ ( − H ~ ) ) → H 1 ( − H − 2 K Y ) → 0 (33) 0\to H^0(N_{\widetilde{Y}}(-\widetilde{H}))\to H^0(N_{\widetilde{Y}}(-H))\xrightarrow{F}H^1(N_{\widetilde{Y}}(-H+K_Y))
\to H^1(N_{\widetilde{Y}}(-\widetilde{H}))\to H^1(-H-2K_Y)\to0 \tag*{(33)} 0 → H 0 ( N Y ( − H )) → H 0 ( N Y ( − H )) F H 1 ( N Y ( − H + K Y )) → H 1 ( N Y ( − H )) → H 1 ( − H − 2 K Y ) → 0 (33)
so that h 1 ( N Y ~ ( − H ~ ) ) = h 1 ( − H − 2 K Y ) + dim coker F h^1(N_{\widetilde{Y}}(-\widetilde{H}))=h^1(-H-2K_Y)+\dim\operatorname{coker}F h 1 ( N Y ( − H )) = h 1 ( − H − 2 K Y ) + dim coker F , and by (32)
α ( Y ~ ) = h 0 ( − H − 2 K Y ) + dim coker F . (34) \alpha(\widetilde{Y})=h^0(-H-2K_Y)+\dim\operatorname{coker}F. \tag*{(34)} α ( Y ) = h 0 ( − H − 2 K Y ) + dim coker F . (34)
Hence Theorem 3.2 is equivalent to the surjectivity of the connecting homomorphism F F F in (33) . We show that F F F is surjective. First note that we have the sequence (28)
0 → H 0 ( H ( − H ) ) → H 0 ( N Y ~ ( − H ) ) → H 0 ( − H − 2 K Y ) → 0. 0\to H^0(\mathcal{H}(-H))\to H^0(N_{\widetilde{Y}}(-H))\to H^0(-H-2K_Y)\to0. 0 → H 0 ( H ( − H )) → H 0 ( N Y ( − H )) → H 0 ( − H − 2 K Y ) → 0.
It is enough to show that the composed map
H 0 ( H ( − H ) ) → f H 1 ( N Y ~ ( − H + K Y ) ) H^0(\mathcal{H}(-H))\xrightarrow{f}H^1(N_{\widetilde{Y}}(-H+K_Y)) H 0 ( H ( − H )) f H 1 ( N Y ( − H + K Y ))
is surjective. Note that from (27) , we have
0 → H 0 ( N Y ( − H ) ) → H 0 ( H ( − H ) ) → g H 1 ( − H − K Y ) → 0 0\to H^0(N_Y(-H))\to H^0(\mathcal{H}(-H))\xrightarrow{g}H^1(-H-K_Y)\to0 0 → H 0 ( N Y ( − H )) → H 0 ( H ( − H )) g H 1 ( − H − K Y ) → 0
while from (14) , (15) and (17) , we have the exact sequence
0 → H 1 ( N Y ( − H + K Y ) ) = H 2 ( T Y ( − H + K Y ) ) → H 1 ( N Y ~ ( − H + K Y ) ) → h H 1 ( − H − K Y ) → 0. 0\to H^1(N_Y(-H+K_Y))=H^2(T_Y(-H+K_Y))\to H^1(N_{\widetilde{Y}}(-H+K_Y))\xrightarrow{h}H^1(-H-K_Y)\to0. 0 → H 1 ( N Y ( − H + K Y )) = H 2 ( T Y ( − H + K Y )) → H 1 ( N Y ( − H + K Y )) h H 1 ( − H − K Y ) → 0.
A Čech cocycle computation in Lemma 3.3 shows that h ∘ f = 2 g h\circ f=2g h ∘ f = 2 g . Therefore we have a commutative diagram of exact sequences
0 → H 0 ( N Y ( − H ) ) → H 0 ( H ( − H ) ) → g H 1 ( − H − K Y ) → 0 → ↓ a ↓ f ↓ 2 ⋅ id 0 → H 1 ( N Y ( − H + K Y ) ) → H 1 ( N Y ~ ( − H + K Y ) ) → h H 1 ( − H − K Y ) → 0 (35) \begin{aligned}
0&\to H^0(N_Y(-H))\to H^0(\mathcal{H}(-H))\xrightarrow{g}H^1(-H-K_Y)\to0\\
&\phantom{\to}\downarrow a\qquad\qquad\qquad\qquad\downarrow f\qquad\qquad\qquad\downarrow2\cdot\operatorname{id}\\
0&\to H^1(N_Y(-H+K_Y))\to H^1(N_{\widetilde{Y}}(-H+K_Y))\xrightarrow{h}H^1(-H-K_Y)\to0 \tag*{(35)}
\end{aligned} 0 0 → H 0 ( N Y ( − H )) → H 0 ( H ( − H )) g H 1 ( − H − K Y ) → 0 → ↓ a ↓ f ↓ 2 ⋅ id → H 1 ( N Y ( − H + K Y )) → H 1 ( N Y ( − H + K Y )) h H 1 ( − H − K Y ) → 0 (35)
where a a a is the map induced by f f f on the kernels. Since the right-hand vertical map is an isomorphism, the snake lemma gives coker a ≅ coker f \operatorname{coker}a\cong\operatorname{coker}f coker a ≅ coker f . Therefore, to show that f f f is surjective, it is enough to show that a a a is surjective. Now the proof is completed by Proposition 3.4. □
We now prove Lemma 3.3. We keep the notation of the proof of Theorem 3.2. Let H ( − H ) → N Y ~ ( − H ) \mathcal{H}(-H)\to N_{\widetilde{Y}}(-H) H ( − H ) → N Y ( − H ) be the first map of (3) tensored with ( − H ) (-H) ( − H ) , and let κ 0 \kappa_0 κ 0 denote the induced map on global sections, while F F F is the connecting map of (33) , f = F ∘ κ 0 f=F\circ\kappa_0 f = F ∘ κ 0 , g g g is the connecting map of (4) tensored with ( − H ) (-H) ( − H ) , and h : H 1 ( N Y ~ ( − H + K Y ) ) → H 1 ( − H − K Y ) h:H^1(N_{\widetilde{Y}}(-H+K_Y))\to H^1(-H-K_Y) h : H 1 ( N Y ( − H + K Y )) → H 1 ( − H − K Y ) . We identify Hom ( K Y , O Y ( − H ) ) ≅ Hom ( K Y ⊗ 2 , K Y ( − H ) ) \operatorname{Hom}(K_Y,\mathcal{O}_Y(-H))\cong\operatorname{Hom}(K_Y^{\otimes2},K_Y(-H)) Hom ( K Y , O Y ( − H )) ≅ Hom ( K Y ⊗ 2 , K Y ( − H )) via ϑ ↦ ϑ ⊗ id K Y \vartheta\mapsto\vartheta\otimes\operatorname{id}_{K_Y} ϑ ↦ ϑ ⊗ id K Y .
Lemma 3.3. h ∘ f = 2 g h\circ f=2g h ∘ f = 2 g . Consequently Im ( h ∘ f ) = Im g \operatorname{Im}(h\circ f)=\operatorname{Im}g Im ( h ∘ f ) = Im g , and the right square of diagram (35) commutes with 2 ⋅ id 2\cdot\operatorname{id} 2 ⋅ id as its right-hand vertical map.
Proof. We use Čech cocycles on an affine cover { U i } \{U_i\} { U i } of Y ~ \widetilde{Y} Y (equivalently of Y Y Y ), U i j = U i ∩ U j U_{ij}=U_i\cap U_j U ij = U i ∩ U j , and write L : = I Y / I Y ~ ≅ K Y L:=I_Y/I_{\widetilde{Y}}\cong K_Y L := I Y / I Y ≅ K Y . Fix φ ∈ H 0 ( H ( − H ) ) \varphi\in H^0(\mathcal{H}(-H)) φ ∈ H 0 ( H ( − H )) , i.e. φ : I Y ~ / I Y 2 → O Y ( − H ) \varphi:I_{\widetilde{Y}}/I_Y^2\to\mathcal{O}_Y(-H) φ : I Y / I Y 2 → O Y ( − H ) , and let φ ˉ = κ 0 ( φ ) = φ ∘ π \bar{\varphi}=\kappa_0(\varphi)=\varphi\circ\pi φ ˉ = κ 0 ( φ ) = φ ∘ π with π : I Y ~ / I Y I Y ~ → I Y ~ / I Y 2 \pi:I_{\widetilde{Y}}/I_YI_{\widetilde{Y}}\to I_{\widetilde{Y}}/I_Y^2 π : I Y / I Y I Y → I Y / I Y 2 . Note that φ ˉ ( I Y ~ 2 ) = 0 \bar{\varphi}(I_{\widetilde{Y}}^2)=0 φ ˉ ( I Y 2 ) = 0 .
Let us find the cocycle for f ( φ ) = F ( φ ˉ ) f(\varphi)=F(\bar{\varphi}) f ( φ ) = F ( φ ˉ ) . Choose O Y ~ \mathcal{O}_{\widetilde{Y}} O Y -linear lifts Φ i : I Y ~ / I Y 2 ∣ U i → O Y ~ ( − H ~ ) ∣ U i \Phi_i:I_{\widetilde{Y}}/I_Y^2|_{U_i}\to\mathcal{O}_{\widetilde{Y}}(-\widetilde{H})|_{U_i} Φ i : I Y / I Y 2 ∣ U i → O Y ( − H ) ∣ U i of φ ˉ \bar{\varphi} φ ˉ . Then Φ i j = Φ i − Φ j \Phi_{ij}=\Phi_i-\Phi_j Φ ij = Φ i − Φ j takes values in L ( − H ) L(-H) L ( − H ) , and f ( φ ) = [ { Φ i j } ] f(\varphi)=[\{\Phi_{ij}\}] f ( φ ) = [{ Φ ij }] . Since L ⊗ 2 ≅ I Y 2 / I Y I Y ~ L^{\otimes2}\cong I_Y^2/I_YI_{\widetilde{Y}} L ⊗ 2 ≅ I Y 2 / I Y I Y ,
h ( f ( φ ) ) = [ { Φ i j ∣ L ⊗ 2 } ] , Φ i j ∣ L ⊗ 2 ( x ˉ ⊗ y ˉ ) = Φ i ( x y ) − Φ j ( x y ) ( x , y ∈ I Y ) . h(f(\varphi))=[\{\Phi_{ij}|_{L^{\otimes2}}\}],\qquad\Phi_{ij}|_{L^{\otimes2}}(\bar{x}\otimes\bar{y})=\Phi_i(xy)-\Phi_j(xy)\quad(x,y\in I_Y). h ( f ( φ )) = [{ Φ ij ∣ L ⊗ 2 }] , Φ ij ∣ L ⊗ 2 ( x ˉ ⊗ y ˉ ) = Φ i ( x y ) − Φ j ( x y ) ( x , y ∈ I Y ) .
Now we find the cocycle for g ( φ ) g(\varphi) g ( φ ) . Choose O Y \mathcal{O}_Y O Y -linear lifts ψ i : I Y / I Y 2 ∣ U i → O Y ( − H ) \psi_i:I_Y/I_Y^2|_{U_i}\to\mathcal{O}_Y(-H) ψ i : I Y / I Y 2 ∣ U i → O Y ( − H ) with ψ i ∣ I Y ~ / I Y 2 = φ \psi_i|_{I_{\widetilde{Y}}/I_Y^2}=\varphi ψ i ∣ I Y / I Y 2 = φ ; then ψ i j = ψ i − ψ j \psi_{ij}=\psi_i-\psi_j ψ ij = ψ i − ψ j factors through L L L and g ( φ ) = [ { ψ i j } ] g(\varphi)=[\{\psi_{ij}\}] g ( φ ) = [{ ψ ij }] .
Now we compare the two cocycles. Put B i ( x , y ) : = Φ i ( x y ) ∈ L ( − H ) B_i(x,y):=\Phi_i(xy)\in L(-H) B i ( x , y ) := Φ i ( x y ) ∈ L ( − H ) for x , y ∈ I Y x,y\in I_Y x , y ∈ I Y (well defined since φ ˉ ( I Y 2 ) = 0 \bar{\varphi}(I_Y^2)=0 φ ˉ ( I Y 2 ) = 0 ). Then: (i) B i B_i B i is symmetric, O P M \mathcal{O}_{\mathbb{P}^M} O P M -bilinear, and vanishes if x x x or y y y lies in I Y 2 I_Y^2 I Y 2 (if x = ∑ k x k ′ x k ′ ′ x=\sum_k x'_kx''_k x = ∑ k x k ′ x k ′′ , then Φ i ( x k ′ x k ′ ′ y ) = x ˉ k ′ ′ ⋅ Φ i ( x k ′ y ) ∈ L ⋅ L ( − H ) = 0 \Phi_i(x'_kx''_ky)=\bar{x}''_k\cdot\Phi_i(x'_ky)\in L\cdot L(-H)=0 Φ i ( x k ′ x k ′′ y ) = x ˉ k ′′ ⋅ Φ i ( x k ′ y ) ∈ L ⋅ L ( − H ) = 0 ), so it descends to I Y / I Y 2 I_Y/I_Y^2 I Y / I Y 2 ; (ii) if x ∈ I Y ~ x\in I_{\widetilde{Y}} x ∈ I Y then B i ( x , y ) = y ˉ ⋅ Φ i ( x ) = φ ( x ˉ ) ⊗ y ˉ B_i(x,y)=\bar{y}\cdot\Phi_i(x)=\varphi(\bar{x})\otimes\bar{y} B i ( x , y ) = y ˉ ⋅ Φ i ( x ) = φ ( x ˉ ) ⊗ y ˉ , because multiplication by y ˉ ∈ L \bar{y}\in L y ˉ ∈ L on O Y ~ ( − H ~ ) \mathcal{O}_{\widetilde{Y}}(-\widetilde{H}) O Y ( − H ) factors through O Y ( − H ) \mathcal{O}_Y(-H) O Y ( − H ) . Similarly, if y ∈ I Y ~ y\in I_{\widetilde{Y}} y ∈ I Y then B i ( x , y ) = x ˉ ⋅ Φ i ( y ) = φ ( y ˉ ) ⊗ x ˉ B_i(x,y)=\bar{x}\cdot\Phi_i(y)=\varphi(\bar{y})\otimes\bar{x} B i ( x , y ) = x ˉ ⋅ Φ i ( y ) = φ ( y ˉ ) ⊗ x ˉ . Put B i ′ ( x , y ) : = ψ i ( x ) ⊗ y ˉ + ψ i ( y ) ⊗ x ˉ B'_i(x,y):=\psi_i(x)\otimes\bar{y}+\psi_i(y)\otimes\bar{x} B i ′ ( x , y ) := ψ i ( x ) ⊗ y ˉ + ψ i ( y ) ⊗ x ˉ . By (ii), B i = B i ′ B_i=B'_i B i = B i ′ whenever one argument lies in I Y ~ / I Y 2 I_{\widetilde{Y}}/I_Y^2 I Y / I Y 2 , so C i : = B i − B i ′ C_i:=B_i-B'_i C i := B i − B i ′ descends to a section of Hom ( L ⊗ 2 , L ( − H ) ) = L − 1 ( − H ) \operatorname{Hom}(L^{\otimes2},L(-H))=L^{-1}(-H) Hom ( L ⊗ 2 , L ( − H )) = L − 1 ( − H ) over U i U_i U i . Over U i j U_{ij} U ij , as forms on L × L L\times L L × L ,
B i − B j = ( C i − C j ) + ( B i ′ − B j ′ ) , ( B i ′ − B j ′ ) ( u , v ) = ψ i j ( u ) ⊗ v + ψ i j ( v ) ⊗ u = 2 ψ i j ( u ) ⊗ v , B_i-B_j=(C_i-C_j)+(B'_i-B'_j),\qquad(B'_i-B'_j)(u,v)=\psi_{ij}(u)\otimes v+\psi_{ij}(v)\otimes u=2\psi_{ij}(u)\otimes v, B i − B j = ( C i − C j ) + ( B i ′ − B j ′ ) , ( B i ′ − B j ′ ) ( u , v ) = ψ ij ( u ) ⊗ v + ψ ij ( v ) ⊗ u = 2 ψ ij ( u ) ⊗ v ,
the last step because L L L is a line bundle (ϑ ( u ) ⊗ v = ϑ ( v ) ⊗ u \vartheta(u)\otimes v=\vartheta(v)\otimes u ϑ ( u ) ⊗ v = ϑ ( v ) ⊗ u ). Hence { B i − B j } = δ { C i } + 2 { ψ i j ⊗ id L } \{B_i-B_j\}=\delta\{C_i\}+2\{\psi_{ij}\otimes\operatorname{id}_L\} { B i − B j } = δ { C i } + 2 { ψ ij ⊗ id L } , i.e. h ( f ( φ ) ) = 2 g ( φ ) h(f(\varphi))=2g(\varphi) h ( f ( φ )) = 2 g ( φ ) . □
Proposition 3.4. The map a : H 0 ( N Y ( − H ) ) → H 1 ( N Y ( − H + K Y ) ) a:H^0(N_Y(-H))\to H^1(N_Y(-H+K_Y)) a : H 0 ( N Y ( − H )) → H 1 ( N Y ( − H + K Y )) in diagram (35) is surjective.
Proof of Proposition 3.4
This subsection is devoted to the proof of Proposition 3.4. We outline the steps to do it for the ease of navigation:
Construct an isomorphism δ ˉ : H 0 ( N C ⊗ L C ) / k s C → ∼ H 1 ( N Y ( − H + K Y ) ) \bar{\delta}:H^0(N_C\otimes L_C)/k\,s_C\xrightarrow{\sim}H^1(N_Y(-H+K_Y)) δ ˉ : H 0 ( N C ⊗ L C ) / k s C ∼ H 1 ( N Y ( − H + K Y )) and conclude that it is enough to show that δ ˉ − 1 ∘ a \bar{\delta}^{-1}\circ a δ ˉ − 1 ∘ a is surjective (see Lemma 3.6).
Give a deformation theoretic interpretation of the map δ ˉ − 1 ∘ a \bar{\delta}^{-1}\circ a δ ˉ − 1 ∘ a by showing that δ ˉ − 1 ∘ a = d Ψ \bar{\delta}^{-1}\circ a=d\Psi δ ˉ − 1 ∘ a = d Ψ (see Proposition 3.13) where Ψ \Psi Ψ is the map defined in Theorem 3.9 , (3).
Construct local sections Γ \Gamma Γ of the map Ψ \Psi Ψ (see Theorem 3.15 ) and hence conclude that d Ψ d\Psi d Ψ is surjective (see Proposition 3.18 ).
The isomorphism δ ˉ \bar{\delta} δ ˉ
We now rewrite the target H 1 ( N Y ( − H + K Y ) ) H^1(N_Y(-H+K_Y)) H 1 ( N Y ( − H + K Y )) of a a a by composing it with an isomorphism. The composed map will have a deformation theoretic interpretation, which will enable us to show that it is surjective. Choose a hyperplane H ⊂ P M H\subset\mathbb{P}^M H ⊂ P M (we use the same letter) meeting Y Y Y transversally along a smooth curve
C : = Y ∩ H , C:=Y\cap H, C := Y ∩ H ,
a rational normal curve of degree M − 1 = g − 1 M-1=g-1 M − 1 = g − 1 in H ≅ P g − 1 H\cong\mathbb{P}^{g-1} H ≅ P g − 1 , and put C ~ : = Y ~ ∩ H \widetilde{C}:=\widetilde{Y}\cap H C := Y ∩ H , the hyperplane section of the carpet, a canonical ribbon of arithmetic genus g g g on C C C . Since O Y ( − H ) = O Y ( − C ) \mathcal{O}_Y(-H)=\mathcal{O}_Y(-C) O Y ( − H ) = O Y ( − C ) , all the twists by − H -H − H above are twists by − C -C − C . Put
N C : = N C / H , L C : = K Y ∣ C ≅ O P 1 ( − g − 1 ) , s C : = s ∣ C ∈ H 0 ( N C ⊗ L C ) , N_C:=N_{C/H},\qquad L_C:=K_Y|_C\cong\mathcal{O}_{\mathbb{P}^1}(-g-1),\qquad s_C:=s|_C\in H^0(N_C\otimes L_C), N C := N C / H , L C := K Y ∣ C ≅ O P 1 ( − g − 1 ) , s C := s ∣ C ∈ H 0 ( N C ⊗ L C ) ,
where K Y ⋅ C = − g − 1 K_Y\cdot C=-g-1 K Y ⋅ C = − g − 1 and N Y ∣ C ≅ N C N_Y|_C\cong N_C N Y ∣ C ≅ N C by transversality.
Lemma 3.5. Let Y ′ → P 1 Y'\to\mathbb{P}^1 Y ′ → P 1 be any smooth rational normal scroll in P M \mathbb{P}^M P M of degree M − 1 M-1 M − 1 (a P 1 \mathbb{P}^1 P 1 -bundle over P 1 \mathbb{P}^1 P 1 embedded by a line bundle of fibre degree 1 1 1 ), with ruling p p p . Then
p ∗ ( N Y ′ / P M ⊗ K Y ′ ) ≅ O P 1 , R 1 p ∗ ( N Y ′ / P M ⊗ K Y ′ ) = 0. p_*(N_{Y'/\mathbb{P}^M}\otimes K_{Y'})\cong\mathcal{O}_{\mathbb{P}^1},\qquad R^1p_*(N_{Y'/\mathbb{P}^M}\otimes K_{Y'})=0. p ∗ ( N Y ′ / P M ⊗ K Y ′ ) ≅ O P 1 , R 1 p ∗ ( N Y ′ / P M ⊗ K Y ′ ) = 0.
Consequently h 0 ( N Y ′ / P M ⊗ K Y ′ ) = 1 h^0(N_{Y'/\mathbb{P}^M}\otimes K_{Y'})=1 h 0 ( N Y ′ / P M ⊗ K Y ′ ) = 1 and h 1 ( N Y ′ / P M ⊗ K Y ′ ) = 0 h^1(N_{Y'/\mathbb{P}^M}\otimes K_{Y'})=0 h 1 ( N Y ′ / P M ⊗ K Y ′ ) = 0 . Moreover, if s ′ s' s ′ spans H 0 ( N Y ′ / P M ⊗ K Y ′ ) H^0(N_{Y'/\mathbb{P}^M}\otimes K_{Y'}) H 0 ( N Y ′ / P M ⊗ K Y ′ ) and ∂ : H 0 ( N Y ′ / P M ⊗ K Y ′ ) → H 1 ( T Y ′ ⊗ K Y ′ ) \partial:H^0(N_{Y'/\mathbb{P}^M}\otimes K_{Y'})\to H^1(T_{Y'}\otimes K_{Y'}) ∂ : H 0 ( N Y ′ / P M ⊗ K Y ′ ) → H 1 ( T Y ′ ⊗ K Y ′ ) is the connecting map of the normal sequence tensored with K Y ′ K_{Y'} K Y ′ , then the image of ∂ ( s ′ ) \partial(s') ∂ ( s ′ ) under the Leray map
H 1 ( T Y ′ ⊗ K Y ′ ) ⟶ H 0 ( R 1 p ∗ ( T Y ′ ⊗ K Y ′ ) ) = H 0 ( O P 1 ) = k H^1(T_{Y'}\otimes K_{Y'})\longrightarrow H^0(R^1p_*(T_{Y'}\otimes K_{Y'}))=H^0(\mathcal{O}_{\mathbb{P}^1})=k H 1 ( T Y ′ ⊗ K Y ′ ) ⟶ H 0 ( R 1 p ∗ ( T Y ′ ⊗ K Y ′ )) = H 0 ( O P 1 ) = k
is nonzero.
Proof. Write Y ′ = Y Y'=Y Y ′ = Y , H H H for the hyperplane class. From 0 → T Y / P 1 → T Y → p ∗ T P 1 → 0 0\to T_{Y/\mathbb{P}^1}\to T_Y\to p^*T_{\mathbb{P}^1}\to0 0 → T Y / P 1 → T Y → p ∗ T P 1 → 0 tensored with K Y K_Y K Y , using T Y / P 1 ⊗ K Y = p ∗ K P 1 T_{Y/\mathbb{P}^1}\otimes K_Y=p^*K_{\mathbb{P}^1} T Y / P 1 ⊗ K Y = p ∗ K P 1 , p ∗ K Y = 0 p_*K_Y=0 p ∗ K Y = 0 , R 1 p ∗ K Y = K P 1 R^1p_*K_Y=K_{\mathbb{P}^1} R 1 p ∗ K Y = K P 1 (relative duality), we get p ∗ ( T Y ⊗ K Y ) = K P 1 = O ( − 2 ) p_*(T_Y\otimes K_Y)=K_{\mathbb{P}^1}=\mathcal{O}(-2) p ∗ ( T Y ⊗ K Y ) = K P 1 = O ( − 2 ) and R 1 p ∗ ( T Y ⊗ K Y ) ≅ T P 1 ⊗ R 1 p ∗ K Y = O P 1 R^1p_*(T_Y\otimes K_Y)\cong T_{\mathbb{P}^1}\otimes R^1p_*K_Y=\mathcal{O}_{\mathbb{P}^1} R 1 p ∗ ( T Y ⊗ K Y ) ≅ T P 1 ⊗ R 1 p ∗ K Y = O P 1 . From the Euler sequence (6) tensored with K Y K_Y K Y , using R p ∗ ( K Y ( H ) ) = 0 Rp_*(K_Y(H))=0 R p ∗ ( K Y ( H )) = 0 (the fibre degree of K Y ( H ) K_Y(H) K Y ( H ) is − 1 -1 − 1 ), we get p ∗ ( T P M ∣ Y ⊗ K Y ) ≅ R 1 p ∗ K Y = O ( − 2 ) p_*(T_{\mathbb{P}^M}|_Y\otimes K_Y)\cong R^1p_*K_Y=\mathcal{O}(-2) p ∗ ( T P M ∣ Y ⊗ K Y ) ≅ R 1 p ∗ K Y = O ( − 2 ) and R 1 p ∗ ( T P M ∣ Y ⊗ K Y ) = 0 R^1p_*(T_{\mathbb{P}^M}|_Y\otimes K_Y)=0 R 1 p ∗ ( T P M ∣ Y ⊗ K Y ) = 0 . Pushing forward (5) tensored with K Y K_Y K Y gives the exact sequence
0 → p ∗ ( T Y ⊗ K Y ) → p ∗ ( T P M ∣ Y ⊗ K Y ) → p ∗ ( N Y ⊗ K Y ) → R 1 p ∗ ( T Y ⊗ K Y ) → R 1 p ∗ ( T P M ∣ Y ⊗ K Y ) = 0. 0\to p_*(T_Y\otimes K_Y)\to p_*(T_{\mathbb{P}^M}|_Y\otimes K_Y)\to p_*(N_Y\otimes K_Y)\to R^1p_*(T_Y\otimes K_Y)\to R^1p_*(T_{\mathbb{P}^M}|_Y\otimes K_Y)=0. 0 → p ∗ ( T Y ⊗ K Y ) → p ∗ ( T P M ∣ Y ⊗ K Y ) → p ∗ ( N Y ⊗ K Y ) → R 1 p ∗ ( T Y ⊗ K Y ) → R 1 p ∗ ( T P M ∣ Y ⊗ K Y ) = 0.
The first map is an injection O ( − 2 ) → O ( − 2 ) \mathcal{O}(-2)\to\mathcal{O}(-2) O ( − 2 ) → O ( − 2 ) , hence an isomorphism. Therefore p ∗ ( N Y ⊗ K Y ) → ∼ R 1 p ∗ ( T Y ⊗ K Y ) ≅ O P 1 p_*(N_Y\otimes K_Y)\xrightarrow{\sim}R^1p_*(T_Y\otimes K_Y)\cong\mathcal{O}_{\mathbb{P}^1} p ∗ ( N Y ⊗ K Y ) ∼ R 1 p ∗ ( T Y ⊗ K Y ) ≅ O P 1 , and R 1 p ∗ ( N Y ⊗ K Y ) = 0 R^1p_*(N_Y\otimes K_Y)=0 R 1 p ∗ ( N Y ⊗ K Y ) = 0 because R 1 p ∗ ( T P M ∣ Y ⊗ K Y ) = 0 = R 2 p ∗ R^1p_*(T_{\mathbb{P}^M}|_Y\otimes K_Y)=0=R^2p_* R 1 p ∗ ( T P M ∣ Y ⊗ K Y ) = 0 = R 2 p ∗ . Taking global sections gives h 0 = 1 h^0=1 h 0 = 1 , h 1 = 0 h^1=0 h 1 = 0 . Finally, the composite of ∂ \partial ∂ with the Leray map is H 0 H^0 H 0 of the sheaf-level connecting map p ∗ ( N Y ⊗ K Y ) → R 1 p ∗ ( T Y ⊗ K Y ) p_*(N_Y\otimes K_Y)\to R^1p_*(T_Y\otimes K_Y) p ∗ ( N Y ⊗ K Y ) → R 1 p ∗ ( T Y ⊗ K Y ) , which we have just shown to be an isomorphism.
Lemma 3.6. Let δ : H 0 ( N C ⊗ L C ) → H 1 ( N Y ( − H + K Y ) ) \delta:H^0(N_C\otimes L_C)\to H^1(N_Y(-H+K_Y)) δ : H 0 ( N C ⊗ L C ) → H 1 ( N Y ( − H + K Y )) be the connecting map of the restriction sequence
0 → N Y ⊗ K Y ( − C ) → N Y ⊗ K Y → ( N Y ⊗ K Y ) ∣ C = N C ⊗ L C → 0. 0\to N_Y\otimes K_Y(-C)\to N_Y\otimes K_Y\to(N_Y\otimes K_Y)|_C=N_C\otimes L_C\to0. 0 → N Y ⊗ K Y ( − C ) → N Y ⊗ K Y → ( N Y ⊗ K Y ) ∣ C = N C ⊗ L C → 0. Then δ \delta δ is surjective with kernel k ⋅ s C k\cdot s_C k ⋅ s C . Hence it induces an isomorphism
δ ˉ : H 0 ( N C ⊗ L C ) / k s C → ∼ H 1 ( N Y ( − H + K Y ) ) . \bar{\delta}:H^0(N_C\otimes L_C)/k\,s_C\xrightarrow{\sim}H^1(N_Y(-H+K_Y)). δ ˉ : H 0 ( N C ⊗ L C ) / k s C ∼ H 1 ( N Y ( − H + K Y )) .
Proof. By Lemma 3.5, H 0 ( N Y ⊗ K Y ) = k s H^0(N_Y\otimes K_Y)=k\,s H 0 ( N Y ⊗ K Y ) = k s and H 1 ( N Y ⊗ K Y ) = 0 H^1(N_Y\otimes K_Y)=0 H 1 ( N Y ⊗ K Y ) = 0 ; the long exact sequence of the restriction sequence gives the claim. (As a check on dimensions: N C ≅ O P 1 ( g + 1 ) ⊕ g − 2 N_C\cong\mathcal{O}_{\mathbb{P}^1}(g+1)^{\oplus g-2} N C ≅ O P 1 ( g + 1 ) ⊕ g − 2 for a rational normal curve of degree g − 1 g-1 g − 1 , so h 0 ( N C ⊗ L C ) = g − 2 h^0(N_C\otimes L_C)=g-2 h 0 ( N C ⊗ L C ) = g − 2 , while h 1 ( N Y ( − H + K Y ) ) = 2 b − e − 2 = g − 3 h^1(N_Y(-H+K_Y))=2b-e-2=g-3 h 1 ( N Y ( − H + K Y )) = 2 b − e − 2 = g − 3 by (14) .)
Proof of Proposition 3.4. Composing a a a with δ ˉ − 1 \bar{\delta}^{-1} δ ˉ − 1 we conclude that it is enough to show that the map
δ ˉ − 1 ∘ a : H 0 ( N Y ( − H ) ) ⟶ H 0 ( N C ⊗ K Y ∣ C ) / k s C (36) \bar{\delta}^{-1}\circ a:H^0(N_Y(-H))\longrightarrow H^0(N_C\otimes K_Y|_C)/k\,s_C \tag*{(36)} δ ˉ − 1 ∘ a : H 0 ( N Y ( − H )) ⟶ H 0 ( N C ⊗ K Y ∣ C ) / k s C (36)
is surjective. Now the proof follows from Proposition 3.19.
In this subsection, we will interpret the map δ ˉ − 1 ∘ a \bar{\delta}^{-1}\circ a δ ˉ − 1 ∘ a as the differential d Ψ d\Psi d Ψ of a map Ψ \Psi Ψ that we will define below in Theorem 3.9 . In this and the following subsections D D D denotes the scheme Spec k [ ε ] / ( ε 2 ) \operatorname{Spec} k[\varepsilon]/(\varepsilon^2) Spec k [ ε ] / ( ε 2 ) , and for a closed subscheme Z ⊂ P M Z\subset\mathbb{P}^M Z ⊂ P M we write Z D : = Z × D Z_D:=Z\times D Z D := Z × D for the constant family, a closed subscheme of P D M \mathbb{P}^M_D P D M . We use freely the identification of first-order embedded deformations of a closed subscheme Z ⊂ P M Z\subset\mathbb{P}^M Z ⊂ P M with sections of N Z / P M = H o m ( I Z , O Z ) N_{Z/\mathbb{P}^M}=\mathcal{H}om(I_Z,\mathcal{O}_Z) N Z / P M = H o m ( I Z , O Z ) [33 §3.2] . A section η \eta η corresponds the closed subscheme Z ε ⊂ P D M Z_\varepsilon\subset\mathbb{P}^M_D Z ε ⊂ P D M with ideal
I Z ε = { f + ε g : f ∈ I Z , g ∈ O P M , g ‾ = η ( f ) m o d I Z } ⊂ O P M [ ε ] / ( ε 2 ) , I_{Z_\varepsilon}=\{f+\varepsilon g:f\in I_Z,\ g\in\mathcal{O}_{\mathbb{P}^M},\ \overline{g}=\eta(f)\bmod I_Z\}\subset\mathcal{O}_{\mathbb{P}^M[\varepsilon]/(\varepsilon^2)}, I Z ε = { f + ε g : f ∈ I Z , g ∈ O P M , g = η ( f ) mod I Z } ⊂ O P M [ ε ] / ( ε 2 ) ,
the preimage of the graph of η \eta η . Locally I Z ε I_{Z_\varepsilon} I Z ε is generated by the elements f + ε η ~ ( f ) f+\varepsilon\widetilde{\eta}(f) f + ε η ( f ) , f ∈ I Z f\in I_Z f ∈ I Z , where η ~ ( f ) \widetilde{\eta}(f) η ( f ) is any local section of O P M \mathcal{O}_{\mathbb{P}^M} O P M lifting η ( f ) ∈ O Z = O P M / I Z \eta(f)\in\mathcal{O}_Z=\mathcal{O}_{\mathbb{P}^M}/I_Z η ( f ) ∈ O Z = O P M / I Z ; the ideal does not depend on the choice of lifts, since ε g = ε ( g + ε η ~ ( g ) ) \varepsilon g=\varepsilon(g+\varepsilon\widetilde{\eta}(g)) ε g = ε ( g + ε η ( g )) for g ∈ I Z g\in I_Z g ∈ I Z .
Lemma 3.7. Let F F F be the flag Hilbert scheme of pairs ( C ′ ⊂ Y ′ ) (C' \subset Y') ( C ′ ⊂ Y ′ ) of closed subschemes of P M \mathbb{P}^{M} P M with Hilbert polynomials
P C ′ ( n ) = ( 2 b − e ) n + 1 , P Y ′ ( n ) = ( 2 b − e ) ( n + 1 2 ) + n + 1 , P_{C'}(n)=(2b-e)n+1,\qquad P_{Y'}(n)=(2b-e)\binom{n+1}{2}+n+1, P C ′ ( n ) = ( 2 b − e ) n + 1 , P Y ′ ( n ) = ( 2 b − e ) ( 2 n + 1 ) + n + 1 , the Hilbert polynomials of C C C and Y Y Y (both of degree 2 b − e = M − 1 2b-e=M-1 2 b − e = M − 1 ), and let S C S_C S C be the fibre over the point [ C ] [C] [ C ] of the Hilbert scheme of rational normal curves of the open subscheme of F F F where Y ′ Y' Y ′ is a smooth rational normal scroll. Thus S C S_C S C is the locus of smooth rational normal scrolls Y ′ ⊂ P M Y'\subset\mathbb{P}^{M} Y ′ ⊂ P M containing C C C ; for such Y ′ Y' Y ′ , Y ′ ∩ H Y'\cap H Y ′ ∩ H is an effective divisor of class H ∣ Y ′ H|_{Y'} H ∣ Y ′ containing C C C with deg C = H 2 \deg C=H^2 deg C = H 2 , so Y ′ ∩ H = C Y'\cap H=C Y ′ ∩ H = C , and H H H is the only hyperplane with this property, since C C C spans H H H .
(1) T [ Y ] S C = H 0 ( N Y ( − H ) ) T_{[Y]}S_C=H^0(N_Y(-H)) T [ Y ] S C = H 0 ( N Y ( − H )) . More precisely, a first-order deformation Y ε Y_\varepsilon Y ε of Y Y Y with class x ∈ H 0 ( N Y ) x\in H^0(N_Y) x ∈ H 0 ( N Y ) contains C D C_D C D if and only if x ∈ H 0 ( N Y ( − C ) ) = H 0 ( N Y ( − H ) ) x\in H^0(N_Y(-C))=H^0(N_Y(-H)) x ∈ H 0 ( N Y ( − C )) = H 0 ( N Y ( − H )) .
(2) S C S_C S C is smooth at [ Y ] [Y] [ Y ] (of dimension h 0 ( N Y ( − H ) ) = 4 b − 2 e h^0(N_Y(-H))=4b-2e h 0 ( N Y ( − H )) = 4 b − 2 e ), hence reduced near [ Y ] [Y] [ Y ] .
(3) Let G H : = { γ ∈ PGL M + 1 ( k ) : γ ∣ H = id H } G_H:=\{\gamma\in\operatorname{PGL}_{M+1}(k):\gamma|_H=\operatorname{id}_H\} G H := { γ ∈ PGL M + 1 ( k ) : γ ∣ H = id H } ; its Lie algebra is H 0 ( T P M ( − H ) ) ≅ k M + 1 H^0(T_{\mathbb{P}^{M}}(-H))\cong k^{M+1} H 0 ( T P M ( − H )) ≅ k M + 1 . G H G_H G H acts on S C S_C S C , and the differential at 1 1 1 of the orbit map γ ↦ γ Y \gamma\mapsto\gamma Y γ ↦ γ Y is the map
c : H 0 ( T P M ∣ Y ( − H ) ) = C M + 1 ⟶ H 0 ( N Y ( − H ) ) c:H^0(T_{\mathbb{P}^{M}}|_Y(-H))=\mathbb{C}^{M+1}\longrightarrow H^0(N_Y(-H)) c : H 0 ( T P M ∣ Y ( − H )) = C M + 1 ⟶ H 0 ( N Y ( − H )) of (25) , which is injective. Thus Im c = T [ Y ] ( G H ⋅ Y ) \operatorname{Im}c=T_{[Y]}(G_H\cdot Y) Im c = T [ Y ] ( G H ⋅ Y ) , and the quotient H 0 ( N Y ( − H ) ) / Im c ≅ H 1 ( T Y ( − H ) ) H^0(N_Y(-H))/\operatorname{Im}c\cong H^1(T_Y(-H)) H 0 ( N Y ( − H )) / Im c ≅ H 1 ( T Y ( − H )) has dimension 2 b − e − 2 = g − 3 2b-e-2=g-3 2 b − e − 2 = g − 3 .
Proof. Each of the statements (1)–(3) are standard for the flag Hilbert scheme (see [27 ] , [33 §4.5] and [13 ] ); For the dimension computations of the cohomology groups appearing in (1)–(3) see (21) , (23) , (24) and (25) . □
We recall some well-known facts about the space of canonical ribbons on a rational curve C C C and the tangent space to that space at a given ribbon C ~ \widetilde{C} C .
Lemma 3.8. Let R C : = P ( H 0 ( N C ⊗ L C ) ) ≅ P g − 3 \mathcal{R}_C:=\mathbb{P}(H^0(N_C\otimes L_C))\cong\mathbb{P}^{g-3} R C := P ( H 0 ( N C ⊗ L C )) ≅ P g − 3 , and recall that s C = s ∣ C : N C ∨ ≅ N Y ∨ ∣ C → K Y ∣ C = L C s_C=s|_C:N_C^\vee\cong N_Y^\vee|_C\to K_Y|_C=L_C s C = s ∣ C : N C ∨ ≅ N Y ∨ ∣ C → K Y ∣ C = L C is the restriction to C C C of the surjection s s s defining the carpet, an element of H 0 ( N C ⊗ L C ) H^0(N_C\otimes L_C) H 0 ( N C ⊗ L C ) .
(1) Every nonzero t ∈ H 0 ( N C ⊗ L C ) = Hom ( N C ∨ , L C ) t\in H^0(N_C\otimes L_C)=\operatorname{Hom}(N_C^\vee,L_C) t ∈ H 0 ( N C ⊗ L C ) = Hom ( N C ∨ , L C ) is surjective and defines a double structure C ~ t ⊂ H \widetilde{C}_t\subset H C t ⊂ H on C C C by I C ~ t : = ker ( I C → N C ∨ → t L C ) I_{\widetilde{C}_t}:=\ker(I_C\to N_C^\vee\xrightarrow{t}L_C) I C t := ker ( I C → N C ∨ t L C ) ; it satisfies I C 2 ⊂ I C ~ t I_C^2\subset I_{\widetilde{C}_t} I C 2 ⊂ I C t and I C / I C ~ t ≅ L C I_C/I_{\widetilde{C}_t}\cong L_C I C / I C t ≅ L C , and depends only on [ t ] ∈ R C [t]\in\mathcal{R}_C [ t ] ∈ R C . Conversely every double structure on C C C in H H H with conormal bundle L C L_C L C is of this form.
(2) T [ C ~ ] R C = H 0 ( N C ⊗ L C ) / k s C T_{[\widetilde{C}]}\mathcal{R}_C=H^0(N_C\otimes L_C)/k s_C T [ C ] R C = H 0 ( N C ⊗ L C ) / k s C . Concretely, a first-order deformation of C ~ \widetilde{C} C inside H H H which is a double structure on C D C_D C D with invertible conormal is ker ( s C + ε t ) \ker(s_C+\varepsilon t) ker ( s C + εt ) for some t ∈ H 0 ( N C ⊗ L C ) t\in H^0(N_C\otimes L_C) t ∈ H 0 ( N C ⊗ L C ) , and s C + ε t s_C+\varepsilon t s C + εt , s C + ε t ′ s_C+\varepsilon t' s C + ε t ′ define the same subscheme if and only if t − t ′ ∈ k s C t-t'\in k s_C t − t ′ ∈ k s C .
Proof. (1) For part (1) see [3 §1] .
(2) An invertible sheaf on C D C_D C D restricting to L C L_C L C is L C ⊗ O D L_C\otimes\mathcal{O}_D L C ⊗ O D (as H 1 ( O C ) = 0 H^1(\mathcal{O}_C)=0 H 1 ( O C ) = 0 ), with automorphisms O D × \mathcal{O}_D^\times O D × ; so such a deformation is the kernel of a surjection s C + ε t : N C ∨ ⊗ O D → L C ⊗ O D s_C+\varepsilon t:N_C^\vee\otimes\mathcal{O}_D\to L_C\otimes\mathcal{O}_D s C + εt : N C ∨ ⊗ O D → L C ⊗ O D , well defined up to multiplication by a unit 1 + ε λ 1+\varepsilon\lambda 1 + ε λ , i.e. up to t ↦ t + λ s C t\mapsto t+\lambda s_C t ↦ t + λ s C . □
Below we define the crucial map
Ψ : S C ⟶ R C \Psi:S_C\longrightarrow\mathcal{R}_C Ψ : S C ⟶ R C
whose differential will turn out to be the map
δ ˉ − 1 ∘ a : H 0 ( N Y ( − H ) ) ⟶ H 0 ( N C ⊗ K Y ∣ C ) / k s C \bar{\delta}^{-1}\circ a:H^0(N_Y(-H))\longrightarrow H^0(N_C\otimes K_Y|_C)/k s_C δ ˉ − 1 ∘ a : H 0 ( N Y ( − H )) ⟶ H 0 ( N C ⊗ K Y ∣ C ) / k s C
in (36) .
Put V : = H 0 ( N C ⊗ L C ) V:=H^0(N_C\otimes L_C) V := H 0 ( N C ⊗ L C ) , so R C = P ( V ) \mathcal{R}_C=\mathbb{P}(V) R C = P ( V ) , and let pr 2 : C × R C → R C \operatorname{pr}_2:C\times\mathcal{R}_C\to\mathcal{R}_C pr 2 : C × R C → R C . Since pr 2 ∗ H o m ( N C ∨ ⊠ O R C , L C ⊠ O R C ( 1 ) ) = V ⊗ O R C ( 1 ) \operatorname{pr}_{2*}\mathcal{H}om(N_C^\vee\boxtimes\mathcal{O}_{\mathcal{R}_C},L_C\boxtimes\mathcal{O}_{\mathcal{R}_C}(1))=V\otimes\mathcal{O}_{\mathcal{R}_C}(1) pr 2 ∗ H o m ( N C ∨ ⊠ O R C , L C ⊠ O R C ( 1 )) = V ⊗ O R C ( 1 ) , the tautological line subbundle O R C ( − 1 ) → V ⊗ O R C \mathcal{O}_{\mathcal{R}_C}(-1)\to V\otimes\mathcal{O}_{\mathcal{R}_C} O R C ( − 1 ) → V ⊗ O R C , viewed as a nowhere vanishing section of V ⊗ O R C ( 1 ) V\otimes\mathcal{O}_{\mathcal{R}_C}(1) V ⊗ O R C ( 1 ) , is a homomorphism
τ : N C ∨ ⊠ O R C ⟶ L C ⊠ O R C ( 1 ) (37) \tau:N_C^\vee\boxtimes\mathcal{O}_{\mathcal{R}_C}\longrightarrow L_C\boxtimes\mathcal{O}_{\mathcal{R}_C}(1) \tag*{(37)} τ : N C ∨ ⊠ O R C ⟶ L C ⊠ O R C ( 1 ) (37)
whose restriction to C × { [ t ] } C \times\{[t]\} C × {[ t ]} is a nonzero multiple of t t t ; hence τ \tau τ is surjective by Lemma 3.8 (1). Let C ~ R C ⊂ H × R C \widetilde{C}_{R_C} \subset H \times R_C C R C ⊂ H × R C be the closed subscheme with
I C ~ R C : = ker ( I C × R C ⟶ I C × R C / I C × R C 2 = N C ∨ ⊠ O R C → τ L C ⊠ O R C ( 1 ) ) . I_{\widetilde{C}_{R_C}} := \ker\left(I_{C \times R_C} \longrightarrow I_{C \times R_C}/I^2_{C \times R_C} = N_C^\vee\boxtimes\mathcal{O}_{R_C} \xrightarrow{\tau} L_C \boxtimes\mathcal{O}_{R_C}(1)\right). I C R C := ker ( I C × R C ⟶ I C × R C / I C × R C 2 = N C ∨ ⊠ O R C τ L C ⊠ O R C ( 1 ) ) .
From 0 → L C ⊠ O R C ( 1 ) → O C ~ R C → O C × R C → 0 0 \to L_C \boxtimes\mathcal{O}_{R_C}(1) \to\mathcal{O}_{\widetilde{C}_{R_C}} \to\mathcal{O}_{C \times R_C} \to0 0 → L C ⊠ O R C ( 1 ) → O C R C → O C × R C → 0 , C ~ R C \widetilde{C}_{R_C} C R C is flat over R C R_C R C , it is a double structure on C × R C C \times R_C C × R C with conormal bundle L C ⊠ O R C ( 1 ) L_C \boxtimes\mathcal{O}_{R_C}(1) L C ⊠ O R C ( 1 ) , and its fibre over [ t ] [t] [ t ] is C ~ t \widetilde{C}_t C t .
Theorem 3.9. Let Y ⊂ P M × S C \mathcal{Y} \subset\mathbb{P}^M \times S_C Y ⊂ P M × S C be the universal scroll and pr : Y → S C \operatorname{pr}: \mathcal{Y} \to S_C pr : Y → S C the projection; thus Y ∩ ( H × S C ) = C × S C \mathcal{Y} \cap(H \times S_C) = C \times S_C Y ∩ ( H × S C ) = C × S C .
(1) There is a unique closed subscheme Y ~ ⊂ P M × S C \widetilde{\mathcal{Y}} \subset\mathbb{P}^M \times S_C Y ⊂ P M × S C , flat over S C S_C S C , with Y ~ r e d = Y \widetilde{\mathcal{Y}}_{\mathrm{red}} = \mathcal{Y} Y red = Y and all of whose fibres are K3 carpets; its fibre over [ Y ′ ] [Y'] [ Y ′ ] is the K3 carpet Y ~ ′ \widetilde{Y}' Y ′ on Y ′ Y' Y ′ . Moreover N : = pr ∗ ( N Y ⊗ ω Y / S C ) \mathcal{N} := \operatorname{pr}_*(N_{\mathcal{Y}} \otimes\omega_{\mathcal{Y}/S_C}) N := pr ∗ ( N Y ⊗ ω Y / S C ) is a line bundle on S C S_C S C , I Y ⊂ I Y ~ I_{\mathcal{Y}} \subset I_{\widetilde{\mathcal{Y}}} I Y ⊂ I Y , and I Y / I Y ~ ≅ ω Y / S C ⊗ pr ∗ N − 1 I_{\mathcal{Y}}/I_{\widetilde{\mathcal{Y}}} \cong\omega_{\mathcal{Y}/S_C} \otimes\operatorname{pr}^*\mathcal{N}^{-1} I Y / I Y ≅ ω Y / S C ⊗ pr ∗ N − 1 .
(2) C ~ : = Y ~ ∩ ( H × S C ) \widetilde{C} := \widetilde{\mathcal{Y}} \cap(H \times S_C) C := Y ∩ ( H × S C ) is flat over S C S_C S C and is a double structure on C × S C C \times S_C C × S C in H × S C H \times S_C H × S C with conormal bundle I C × S C / I C ~ ≅ L C ⊠ N − 1 I_{C \times S_C}/I_{\widetilde{C}} \cong L_C \boxtimes\mathcal{N}^{-1} I C × S C / I C ≅ L C ⊠ N − 1 ; its fibre over [ Y ′ ] [Y'] [ Y ′ ] is the ribbon Y ~ ′ ∩ H \widetilde{Y}' \cap H Y ′ ∩ H .
(3) There is a unique morphism Ψ : S C → R C \Psi: S_C \to R_C Ψ : S C → R C with ( id H × Ψ ) − 1 ( C ~ R C ) = C ~ (\operatorname{id}_H \times\Psi)^{-1}(\widetilde{C}_{R_C}) = \widetilde{C} ( id H × Ψ ) − 1 ( C R C ) = C . It satisfies Ψ ∗ O R C ( − 1 ) ≅ N \Psi^*\mathcal{O}_{R_C}(-1) \cong\mathcal{N} Ψ ∗ O R C ( − 1 ) ≅ N and, on closed points, Ψ ( [ Y ′ ] ) = [ Y ~ ′ ∩ H ] \Psi([Y']) = [\widetilde{Y}' \cap H] Ψ ([ Y ′ ]) = [ Y ′ ∩ H ] .
Proof. (1) By Lemma 3.7 (2), S C S_C S C is smooth at every [ Y ′ ] ∈ S C [Y'] \in S_C [ Y ′ ] ∈ S C . Hence [20 Prop. 3.7] applies to the flat family pr : Y → S C \operatorname{pr}: \mathcal{Y} \to S_C pr : Y → S C of smooth rational normal scrolls in P M × S C = P ( V ) \mathbb{P}^M \times S_C = \mathbb{P}(\mathcal{V}) P M × S C = P ( V ) and gives the existence and uniqueness of Y ~ \widetilde{\mathcal{Y}} Y . Its fibre over [ Y ′ ] [Y'] [ Y ′ ] is a K3 carpet on Y ′ Y' Y ′ , hence equals Y ~ ′ \widetilde{Y}' Y ′ [20 Prop. 1.7] . The fact that N \mathcal{N} N is a line bundle again follows from [20 Prop. 1.7] . The last two claims follow from the fact that 1 ∈ H 0 ( O S C ) = H 0 ( N Y ⊗ ω Y / S C ⊗ pr ∗ N − 1 ) 1 \in H^0(\mathcal{O}_{S_C}) = H^0(N_{\mathcal{Y}} \otimes\omega_{\mathcal{Y}/S_C} \otimes\operatorname{pr}^*\mathcal{N}^{-1}) 1 ∈ H 0 ( O S C ) = H 0 ( N Y ⊗ ω Y / S C ⊗ pr ∗ N − 1 ) is a nowhere vanishing section, and hence gives a surjection N Y ∨ → ω Y / S C ⊗ pr ∗ N − 1 N^\vee_{\mathcal{Y}} \to\omega_{\mathcal{Y}/S_C} \otimes\operatorname{pr}^*\mathcal{N}^{-1} N Y ∨ → ω Y / S C ⊗ pr ∗ N − 1 , and I Y ~ I_{\widetilde{\mathcal{Y}}} I Y is the kernel of I Y → N Y ∨ → ω Y / S C ⊗ pr ∗ N − 1 I_{\mathcal{Y}} \to N^\vee_{\mathcal{Y}} \to\omega_{\mathcal{Y}/S_C} \otimes\operatorname{pr}^*\mathcal{N}^{-1} I Y → N Y ∨ → ω Y / S C ⊗ pr ∗ N − 1 .
(2) Let h h h be the equation of H H H . For every [ Y ′ ] ∈ S C [Y'] \in S_C [ Y ′ ] ∈ S C , h h h is a nonzerodivisor on O Y ′ \mathcal{O}_{Y'} O Y ′ and on O Y ~ ′ \mathcal{O}_{\widetilde{Y}'} O Y ′ and Y ′ Y' Y ′ is smooth, Y ~ ′ \widetilde{Y}' Y ′ is Cohen–Macaulay of pure dimension 2 2 2 with support Y ′ ⊄ H Y' \not\subset H Y ′ ⊂ H . As Y \mathcal{Y} Y and Y ~ \widetilde{\mathcal{Y}} Y are flat over S C S_C S C , the local criterion of flatness [30 Cor. to Thm. 22.5] shows that h h h is a nonzerodivisor on O Y \mathcal{O}_{\mathcal{Y}} O Y and on O Y ~ \mathcal{O}_{\widetilde{\mathcal{Y}}} O Y , and that O C ~ = O Y ~ / h O Y ~ \mathcal{O}_{\widetilde{C}} = \mathcal{O}_{\widetilde{\mathcal{Y}}}/h\mathcal{O}_{\widetilde{\mathcal{Y}}} O C = O Y / h O Y is flat over S C S_C S C . We have I C × S C = ( I Y + ( h ) ) I_{C \times S_C} = (I_{\mathcal{Y}} + (h)) I C × S C = ( I Y + ( h )) and I C ~ = ( I Y ~ + ( h ) ) I_{\widetilde{C}} = (I_{\widetilde{\mathcal{Y}}} + (h)) I C = ( I Y + ( h )) . Further, I Y ∩ ( h ) = h I Y I_{\mathcal{Y}} \cap(h) = hI_{\mathcal{Y}} I Y ∩ ( h ) = h I Y . So we have I C × S C 2 ⊂ I Y ~ 2 + ( h ) ⊂ I C ~ I^2_{C \times S_C} \subset I^2_{\widetilde{\mathcal{Y}}} + (h) \subset I_{\widetilde{C}} I C × S C 2 ⊂ I Y 2 + ( h ) ⊂ I C and
I C × S C / I C ~ = ( I Y + ( h ) ) / ( I Y ~ + ( h ) ) = I Y / ( I Y ~ + h I Y ) = ( I Y / I Y ~ ) ∣ C × S C ≅ ω Y / S C ∣ C × S C ⊗ pr ∗ N − 1 , I_{C \times S_C}/I_{\widetilde{C}} = (I_{\mathcal{Y}} + (h))/(I_{\widetilde{\mathcal{Y}}} + (h)) = I_{\mathcal{Y}}/(I_{\widetilde{\mathcal{Y}}} + hI_{\mathcal{Y}}) = (I_{\mathcal{Y}}/I_{\widetilde{\mathcal{Y}}})|_{C \times S_C} \cong\omega_{\mathcal{Y}/S_C}|_{C \times S_C} \otimes\operatorname{pr}^*\mathcal{N}^{-1}, I C × S C / I C = ( I Y + ( h )) / ( I Y + ( h )) = I Y / ( I Y + h I Y ) = ( I Y / I Y ) ∣ C × S C ≅ ω Y / S C ∣ C × S C ⊗ pr ∗ N − 1 ,
Finally, C × S C ⊂ Y C \times S_C \subset\mathcal{Y} C × S C ⊂ Y is the divisor of h ∈ H 0 ( O Y ( 1 ) ) h \in H^0(\mathcal{O}_{\mathcal{Y}}(1)) h ∈ H 0 ( O Y ( 1 )) , so by adjunction ω Y / S C ∣ C × S C ≅ ω C × S C / S C ⊗ O Y ( − 1 ) ∣ C × S C = ( K C ⊗ O C ( − 1 ) ) ⊠ O S C = L C ⊠ O S C \omega_{\mathcal{Y}/S_C}|_{C \times S_C} \cong\omega_{C \times S_C/S_C} \otimes\mathcal{O}_{\mathcal{Y}}(-1)|_{C \times S_C} = (K_C \otimes\mathcal{O}_C(-1)) \boxtimes\mathcal{O}_{S_C} = L_C \boxtimes\mathcal{O}_{S_C} ω Y / S C ∣ C × S C ≅ ω C × S C / S C ⊗ O Y ( − 1 ) ∣ C × S C = ( K C ⊗ O C ( − 1 )) ⊠ O S C = L C ⊠ O S C . The fibre of C ~ \widetilde{C} C over [ Y ′ ] [Y'] [ Y ′ ] is Y ~ ′ ∩ H \widetilde{Y}' \cap H Y ′ ∩ H because base change commutes with fibre products.
(3) By (2), C ~ \widetilde{C} C is defined by the surjection
q : N C ∨ ⊠ O S C = I C × S C / I C × S C 2 ⟶ I C × S C / I C ~ ≅ L C ⊠ N − 1 , q: N_C^\vee\boxtimes\mathcal{O}_{S_C} = I_{C \times S_C}/I^2_{C \times S_C} \longrightarrow I_{C \times S_C}/I_{\widetilde{C}} \cong L_C \boxtimes\mathcal{N}^{-1}, q : N C ∨ ⊠ O S C = I C × S C / I C × S C 2 ⟶ I C × S C / I C ≅ L C ⊠ N − 1 ,
that is, I C ~ = ker ( I C × S C → N C ∨ ⊠ O S C → q L C ⊠ N − 1 ) I_{\widetilde{C}} = \ker\left(I_{C \times S_C} \to N_C^\vee\boxtimes\mathcal{O}_{S_C} \xrightarrow{q} L_C \boxtimes\mathcal{N}^{-1}\right) I C = ker ( I C × S C → N C ∨ ⊠ O S C q L C ⊠ N − 1 ) . By the projection formula q ∈ H 0 ( S C , V ⊗ N − 1 ) q \in H^0(S_C,\mathcal{V} \otimes\mathcal{N}^{-1}) q ∈ H 0 ( S C , V ⊗ N − 1 ) , and q q q vanishes nowhere on S C S_C S C , its restriction to each C × { [ Y ′ ] } C \times\{[Y']\} C × {[ Y ′ ]} being a surjection N C ∨ → L C N_C^\vee\to L_C N C ∨ → L C . Thus q q q is a line subbundle N C ⊗ L C ↪ V ⊗ O S C N_C \otimes L_C \hookrightarrow\mathcal{V} \otimes\mathcal{O}_{S_C} N C ⊗ L C ↪ V ⊗ O S C , i.e. a morphism Ψ : S C → P ( V ) = R C \Psi: S_C \to\mathbb{P}(\mathcal{V}) = R_C Ψ : S C → P ( V ) = R C with Ψ ∗ O R C ( − 1 ) = N \Psi^*\mathcal{O}_{R_C}(-1) = \mathcal{N} Ψ ∗ O R C ( − 1 ) = N and ( id C × Ψ ) ∗ q = q (\operatorname{id}_C \times\Psi)^*q = q ( id C × Ψ ) ∗ q = q where τ \tau τ is as in (37). Since L C ⊠ O R C ( 1 ) L_C \boxtimes\mathcal{O}_{R_C}(1) L C ⊠ O R C ( 1 ) is a line bundle, we have that the pullback along id H × Ψ \operatorname{id}_H \times\Psi id H × Ψ of
0 ⟶ I C ~ R C ⟶ I C × R C → τ L C ⊠ O R C ( 1 ) ⟶ 0 0 \longrightarrow I_{\widetilde{C}_{R_C}} \longrightarrow I_{C \times R_C} \xrightarrow{\tau} L_C \boxtimes\mathcal{O}_{R_C}(1) \longrightarrow0 0 ⟶ I C R C ⟶ I C × R C τ L C ⊠ O R C ( 1 ) ⟶ 0
is the exact sequence
0 ⟶ ( id H × Ψ ) ∗ I C ~ R C ⟶ I C × S C → q L C ⊠ N − 1 ⟶ 0. 0 \longrightarrow(\operatorname{id}_H \times\Psi)^*I_{\widetilde{C}_{R_C}} \longrightarrow I_{C \times S_C} \xrightarrow{q} L_C \boxtimes\mathcal{N}^{-1} \longrightarrow0. 0 ⟶ ( id H × Ψ ) ∗ I C R C ⟶ I C × S C q L C ⊠ N − 1 ⟶ 0.
and hence ( id H × Ψ ) − 1 ( C ~ R C ) = C ~ (\operatorname{id}_{H} \times\Psi)^{-1}(\widetilde{C}_{R_C}) = \widetilde{C} ( id H × Ψ ) − 1 ( C R C ) = C . To see the uniqueness, note that ( id H × Ψ ) − 1 ( C ~ R C ) (\operatorname{id}_{H} \times\Psi)^{-1}(\widetilde{C}_{R_C}) ( id H × Ψ ) − 1 ( C R C ) determines the surjection N C ∨ ⊠ O S C → I C × S C / I C ~ N_{C}^{\vee} \boxtimes\mathcal{O}_{S_C} \to I_{C \times S_C}/I_{\widetilde{C}} N C ∨ ⊠ O S C → I C × S C / I C up to an automorphism of the target, hence the line subbundle Ψ ∗ O R C ( − 1 ) ⊂ V ⊗ O S C \Psi^{*}\mathcal{O}_{R_C}(-1) \subset V \otimes\mathcal{O}_{S_C} Ψ ∗ O R C ( − 1 ) ⊂ V ⊗ O S C , hence Ψ \Psi Ψ . On closed points, Ψ ( [ Y ′ ] ) \Psi([Y']) Ψ ([ Y ′ ]) is the class of the fibre of C ~ \widetilde{C} C over [ Y ′ ] [Y'] [ Y ′ ] , i.e. [ Y ~ ′ ∩ H ] [\widetilde{Y}' \cap H] [ Y ′ ∩ H ] by (2). □
In the following three lemmas, we describe the map d Ψ : H 0 ( N Y ( − H ) ) → H 0 ( N C ⊗ L C ) / k s C d\Psi: H^{0}(N_Y(-H)) \to H^{0}(N_C \otimes L_C)/k s_C d Ψ : H 0 ( N Y ( − H )) → H 0 ( N C ⊗ L C ) / k s C , which is the differential of the map Ψ \Psi Ψ . An element of H 0 ( N Y ( − H ) ) H^{0}(N_Y(-H)) H 0 ( N Y ( − H )) consists of a flag C D = Y ε ∩ H D ⊂ Y ε ⊂ P D M C_D = Y_{\varepsilon} \cap H_D \subset Y_{\varepsilon} \subset\mathbb{P}^{M}_{D} C D = Y ε ∩ H D ⊂ Y ε ⊂ P D M where Y ε ⊂ P D M Y_{\varepsilon} \subset\mathbb{P}^{M}_{D} Y ε ⊂ P D M is an embedded first order deformation Y ⊂ P M Y \subset\mathbb{P}^{M} Y ⊂ P M . In the following lemma we prove a slight generalization of [20 Prop. 3.7] , to show that given a first order deformation Y ε ⊂ P D M Y_{\varepsilon} \subset\mathbb{P}^{M}_{D} Y ε ⊂ P D M of the scroll Y ⊂ P M Y \subset\mathbb{P}^{M} Y ⊂ P M there exists a unique embedded double structure Y ε ⊂ Y ~ ε ⊂ P D M Y_{\varepsilon} \subset\widetilde{Y}_{\varepsilon} \subset\mathbb{P}^{M}_{D} Y ε ⊂ Y ε ⊂ P D M that restricts on the central fiber to the unique embedded K 3 K3 K 3 carpet structure Y ⊂ Y ~ ⊂ P M Y \subset\widetilde{Y} \subset\mathbb{P}^{M} Y ⊂ Y ⊂ P M .
Lemma 3.10. Let Y ε ⊂ P D M Y_{\varepsilon} \subset\mathbb{P}^{M}_{D} Y ε ⊂ P D M be a first-order deformation of Y Y Y . There is a unique closed subscheme Y ~ ε ⊂ P D M \widetilde{Y}_{\varepsilon} \subset\mathbb{P}^{M}_{D} Y ε ⊂ P D M , flat over D D D , with Y ~ ε × D { 0 } = Y ~ \widetilde{Y}_{\varepsilon} \times_{D} \{0\} = \widetilde{Y} Y ε × D { 0 } = Y , Y ε ⊂ Y ~ ε Y_{\varepsilon} \subset\widetilde{Y}_{\varepsilon} Y ε ⊂ Y ε , I Y ε ⊂ I Y ~ ε I_{Y_{\varepsilon}} \subset I_{\widetilde{Y}_{\varepsilon}} I Y ε ⊂ I Y ε and I Y ε / I Y ~ ε I_{Y_{\varepsilon}}/I_{\widetilde{Y}_{\varepsilon}} I Y ε / I Y ε invertible on Y ε Y_{\varepsilon} Y ε . Further in this case, I Y ε / I Y ~ ε ≅ ω Y ε / D I_{Y_{\varepsilon}}/I_{\widetilde{Y}_{\varepsilon}} \cong\omega_{Y_{\varepsilon}/D} I Y ε / I Y ε ≅ ω Y ε / D . We write Φ ( Y ε ) : = Y ~ ε \Phi(Y_{\varepsilon}) := \widetilde{Y}_{\varepsilon} Φ ( Y ε ) := Y ε and d Φ : H 0 ( N Y ) → H 0 ( N Y ~ ) d\Phi: H^{0}(N_Y) \to H^{0}(N_{\widetilde{Y}}) d Φ : H 0 ( N Y ) → H 0 ( N Y ) for the resulting map. Moreover:
(1) the image of d Φ ( x ) d\Phi(x) d Φ ( x ) under H 0 ( N Y ~ ) = Hom ( I Y ~ , O Y ~ ) → H 0 ( N Y ~ ⊗ O Y ) = Hom ( I Y ~ , O Y ) H^{0}(N_{\widetilde{Y}}) = \operatorname{Hom}(I_{\widetilde{Y}}, \mathcal{O}_{\widetilde{Y}}) \to H^{0}(N_{\widetilde{Y}} \otimes\mathcal{O}_{Y}) = \operatorname{Hom}(I_{\widetilde{Y}}, \mathcal{O}_{Y}) H 0 ( N Y ) = Hom ( I Y , O Y ) → H 0 ( N Y ⊗ O Y ) = Hom ( I Y , O Y ) is ι ( x ) : = x ∣ I Y ~ \iota(x) := x|_{I_{\widetilde{Y}}} ι ( x ) := x ∣ I Y , the image of x x x under the composite H 0 ( N Y ) = Hom ( I Y , O Y ) → H 0 ( H ) → H 0 ( N Y ~ ⊗ O Y ) = Hom ( I Y ~ , O Y ) H^{0}(N_Y) = \operatorname{Hom}(I_Y, \mathcal{O}_Y) \to H^{0}(\mathcal{H}) \to H^{0}(N_{\widetilde{Y}} \otimes\mathcal{O}_{Y}) = \operatorname{Hom}(I_{\widetilde{Y}}, \mathcal{O}_{Y}) H 0 ( N Y ) = Hom ( I Y , O Y ) → H 0 ( H ) → H 0 ( N Y ⊗ O Y ) = Hom ( I Y , O Y ) of the maps in (4) and (3) ;
(2) for γ ε ∈ PGL M + 1 ( D ) \gamma_{\varepsilon} \in\operatorname{PGL}_{M+1}(D) γ ε ∈ PGL M + 1 ( D ) one has Φ ( γ ε Y ) = γ ε Y ~ \Phi(\gamma_{\varepsilon}Y) = \gamma_{\varepsilon}\widetilde{Y} Φ ( γ ε Y ) = γ ε Y .
Proof. Such a Y ~ ε \widetilde{Y}_{\varepsilon} Y ε is the kernel of a surjection s ε : N Y ε / P D M ∨ → L ε s_{\varepsilon}: N^{\vee}_{Y_{\varepsilon}/\mathbb{P}^{M}_{D}} \to L_{\varepsilon} s ε : N Y ε / P D M ∨ → L ε onto an invertible sheaf with L ε ∣ Y ≅ K Y L_{\varepsilon}|_{Y} \cong K_Y L ε ∣ Y ≅ K Y . Since H 1 ( O Y ) = 0 H^{1}(\mathcal{O}_{Y}) = 0 H 1 ( O Y ) = 0 , we have that Pic ( Y ε ) → Pic ( Y ) \operatorname{Pic}(Y_{\varepsilon}) \to\operatorname{Pic}(Y) Pic ( Y ε ) → Pic ( Y ) is injective. Hence L ε ≅ ω Y ε / D L_{\varepsilon} \cong\omega_{Y_{\varepsilon}/D} L ε ≅ ω Y ε / D , and s ε s_{\varepsilon} s ε is determined by Y ~ ε \widetilde{Y}_{\varepsilon} Y ε up to H 0 ( O Y ε ) = ( k [ ε ] / ε 2 ) × H^{0}(\mathcal{O}_{Y_{\varepsilon}}) = (k[\varepsilon]/\varepsilon^{2})^{\times} H 0 ( O Y ε ) = ( k [ ε ] / ε 2 ) × . Conversely every such surjection lifting s s s defines such a Y ~ ε \widetilde{Y}_{\varepsilon} Y ε . Put F : = N Y ε / P D M ⊗ ω Y ε / D \mathcal{F} := N_{Y_{\varepsilon}/\mathbb{P}^{M}_{D}} \otimes\omega_{Y_{\varepsilon}/D} F := N Y ε / P D M ⊗ ω Y ε / D , a locally free sheaf on Y ε Y_{\varepsilon} Y ε . Tensoring the exact sequence
0 → ε O Y → O Y ε → O Y → 0 0 \to\varepsilon\mathcal{O}_{Y} \to\mathcal{O}_{Y_{\varepsilon}} \to\mathcal{O}_{Y} \to0 0 → ε O Y → O Y ε → O Y → 0
by F \mathcal{F} F , we have
0 → ε F → F → N Y ⊗ K Y → 0 0 \to\varepsilon\mathcal{F} \to\mathcal{F} \to N_Y \otimes K_Y \to0 0 → ε F → F → N Y ⊗ K Y → 0
and ε F ≅ N Y ⊗ K Y \varepsilon\mathcal{F} \cong N_Y \otimes K_Y ε F ≅ N Y ⊗ K Y . By Lemma 3.5, H 0 ( N Y ⊗ K Y ) = k s H^{0}(N_Y \otimes K_Y) = k s H 0 ( N Y ⊗ K Y ) = k s , H 1 ( N Y ⊗ K Y ) = 0 H^{1}(N_Y \otimes K_Y) = 0 H 1 ( N Y ⊗ K Y ) = 0 . Therefore, H 0 ( F ) → H 0 ( N Y ⊗ K Y ) = k s H^{0}(\mathcal{F}) \to H^{0}(N_Y \otimes K_Y) = k s H 0 ( F ) → H 0 ( N Y ⊗ K Y ) = k s is surjective with kernel ε H 0 ( F ) = ε k s \varepsilon H^{0}(\mathcal{F}) = \varepsilon k s ε H 0 ( F ) = ε k s . So lifts of s s s exist, are surjective by Nakayama, and form the set { ( 1 + ε λ ) s ε } \{(1+\varepsilon\lambda)s_{\varepsilon}\} {( 1 + ε λ ) s ε } . But every such surjection has the same kernel and therefore we have a unique double structure Y ~ ε \widetilde{Y}_{\varepsilon} Y ε .
(1) With η : = d Φ ( x ) \eta:= d\Phi(x) η := d Φ ( x ) , I Y ~ ε ⊂ I Y ε I_{\widetilde{Y}_{\varepsilon}} \subset I_{Y_{\varepsilon}} I Y ε ⊂ I Y ε means that for f ∈ I Y ~ f \in I_{\widetilde{Y}} f ∈ I Y and any g g g with g ≡ η ( f ) m o d I Y ~ g \equiv\eta(f) \mod I_{\widetilde{Y}} g ≡ η ( f ) mod I Y one has g ≡ x ( f ) m o d I Y g \equiv x(f) \mod I_Y g ≡ x ( f ) mod I Y , i.e. η ( f ) ≡ x ( f ) m o d I Y \eta(f) \equiv x(f) \mod I_Y η ( f ) ≡ x ( f ) mod I Y for all f ∈ I Y ~ f \in I_{\widetilde{Y}} f ∈ I Y . (2) follows from uniqueness. □
Lemma 3.11. Let x ∈ H 0 ( N Y ( − H ) ) = T [ Y ] S C x \in H^{0}(N_Y(-H)) = T_{[Y]}S_C x ∈ H 0 ( N Y ( − H )) = T [ Y ] S C , let Y ε ⊃ C D Y_{\varepsilon} \supset C_D Y ε ⊃ C D be the corresponding deformation (Lemma 3.7 (1)), Y ~ ε : = Φ ( Y ε ) \widetilde{Y}_{\varepsilon} := \Phi(Y_{\varepsilon}) Y ε := Φ ( Y ε ) , and
C ~ ε : = Y ~ ε ∩ H D ⊂ H D \widetilde{C}_{\varepsilon} := \widetilde{Y}_{\varepsilon} \cap H_D \subset H_D C ε := Y ε ∩ H D ⊂ H D (scheme-theoretic intersection). Then:
(i) C ~ ε \widetilde{C}_{\varepsilon} C ε is flat over D D D with closed fibre C ~ \widetilde{C} C .
(ii) C ~ ε \widetilde{C}_{\varepsilon} C ε is a double structure on C D C_D C D with conormal bundle L C ⊗ O D L_C \otimes\mathcal{O}_D L C ⊗ O D .
Hence in particular, by Lemma 3.8 (2), C ~ ε = ker ( s C + ε t ) \widetilde{C}_{\varepsilon} = \ker(s_C+\varepsilon t) C ε = ker ( s C + εt ) for a t ∈ H 0 ( N C ⊗ L C ) t \in H^{0}(N_C \otimes L_C) t ∈ H 0 ( N C ⊗ L C ) well defined modulo k s C k s_C k s C .
Proof. To see (i) note that h h h is a nonzerodivisor on O Y ~ \mathcal{O}_{\widetilde{Y}} O Y , since Y ~ \widetilde{Y} Y is Cohen–Macaulay of pure dimension 2 2 2 and h h h is a nonzerodivisor on O Y \mathcal{O}_{Y} O Y . Now the conclusion follows from local criterion of flatness.
To see (ii), note that since Y ~ ε \widetilde{Y}_{\varepsilon} Y ε is double structure over Y ε Y_{\varepsilon} Y ε with conormal bundle ω Y ε / D \omega_{Y_{\varepsilon}/D} ω Y ε / D by Lemma 3.10 , C ~ ε = Y ~ ε ∩ H D \widetilde{C}_{\varepsilon}=\widetilde{Y}_{\varepsilon}\cap H_D C ε = Y ε ∩ H D is a double structure on Y ε ∩ H D = C D Y_{\varepsilon}\cap H_D=C_D Y ε ∩ H D = C D with conormal bundle ω Y ε / D ∣ C D = L C ⊗ O D \omega_{Y_{\varepsilon}/D}|_{C_D}=L_C\otimes\mathcal{O}_D ω Y ε / D ∣ C D = L C ⊗ O D .
Lemma 3.12. The differential
d Ψ : T [ Y ] S C = H 0 ( N Y ( − H ) ) ⟶ H 0 ( N C ⊗ L C ) / k s C = T [ C ~ ] R C d\Psi:T_{[Y]}\mathcal{S}_C=H^0(N_Y(-H))\longrightarrow H^0(N_C\otimes L_C)/k\,s_C=T_{[\widetilde{C}]}\mathcal{R}_C d Ψ : T [ Y ] S C = H 0 ( N Y ( − H )) ⟶ H 0 ( N C ⊗ L C ) / k s C = T [ C ] R C of the map Ψ \Psi Ψ of Theorem 3.9 is given by
d Ψ ( Y ε ⊃ C D ) = C ~ ε = Y ~ ε ∩ H D . d\Psi(Y_{\varepsilon}\supset C_D)=\widetilde{C}_{\varepsilon}=\widetilde{Y}_{\varepsilon}\cap H_D. d Ψ ( Y ε ⊃ C D ) = C ε = Y ε ∩ H D . In other words, if x ∈ H 0 ( N Y ( − H ) ) x\in H^0(N_Y(-H)) x ∈ H 0 ( N Y ( − H )) and C ~ ε = ker ( s C + ε t ) \widetilde{C}_{\varepsilon}=\ker(s_C+\varepsilon t) C ε = ker ( s C + εt ) for a t ∈ H 0 ( N C ⊗ L C ) t\in H^0(N_C\otimes L_C) t ∈ H 0 ( N C ⊗ L C ) , then d Ψ ( x ) = [ t ] d\Psi(x)=[t] d Ψ ( x ) = [ t ] , where [ t ] [t] [ t ] denotes the class of t t t modulo k s C k\,s_C k s C .
Proof. Regard Y ε ⊃ C D Y_{\varepsilon}\supset C_D Y ε ⊃ C D as a morphism D → S C D\to\mathcal{S}_C D → S C . Then d Ψ ( Y ε ⊃ C D ) d\Psi(Y_{\varepsilon}\supset C_D) d Ψ ( Y ε ⊃ C D ) is given by the composition
D ⟶ S C → Ψ R C . D\longrightarrow\mathcal{S}_C\xrightarrow{\Psi}\mathcal{R}_C. D ⟶ S C Ψ R C .
By Theorem 3.9 (1) and flatness, Y ~ × S C D \widetilde{Y}\times_{\mathcal{S}_C}D Y × S C D is flat over D D D and is a double structure on Y ε = Y × S C D Y_{\varepsilon}=Y\times_{\mathcal{S}_C}D Y ε = Y × S C D such that its restriction to the central fiber is the K3 double structure Y ~ \widetilde{Y} Y on Y Y Y . So Y ~ × S C D = Y ~ ε \widetilde{Y}\times_{\mathcal{S}_C}D=\widetilde{Y}_{\varepsilon} Y × S C D = Y ε by the uniqueness in Lemma 3.10 . Hence C ~ ε = C ~ × S C D \widetilde{C}_{\varepsilon}=\widetilde{C}\times_{\mathcal{S}_C}D C ε = C × S C D is the pullback of C ~ R C \widetilde{C}_{\mathcal{R}_C} C R C along D → S C → Ψ R C D\to\mathcal{S}_C\xrightarrow{\Psi}\mathcal{R}_C D → S C Ψ R C by Theorem 3.9 (3). Therefore d Ψ ( Y ε ⊃ C D ) = C ~ ε = Y ~ ε ∩ H D d\Psi(Y_{\varepsilon}\supset C_D)=\widetilde{C}_{\varepsilon}=\widetilde{Y}_{\varepsilon}\cap H_D d Ψ ( Y ε ⊃ C D ) = C ε = Y ε ∩ H D .
Proposition 3.13
Proposition 3.13. δ ‾ − 1 ∘ a = d Ψ \overline{\delta}^{-1}\circ a=d\Psi δ − 1 ∘ a = d Ψ . Equivalently, the map (36) is d Ψ d\Psi d Ψ .
Proof. Throughout we regard H 1 ( N Y ( − H + K Y ) ) H^1(N_Y(-H+K_Y)) H 1 ( N Y ( − H + K Y )) as a subspace of H 1 ( N Y ~ ( − H + K Y ) ) H^1(N_{\widetilde{Y}}(-H+K_Y)) H 1 ( N Y ( − H + K Y )) via the injective map j j j of the bottom row of (35) . The map j j j is induced by the sheaf map N Y → N Y ~ ⊗ O Y N_Y\to N_{\widetilde{Y}}\otimes\mathcal{O}_Y N Y → N Y ⊗ O Y , which restricts a homomorphism from I Y I_Y I Y to I Y ~ I_{\widetilde{Y}} I Y (the composite of (4) and (3)), and so is the injection ι : H 0 ( N Y ( − H ) ) ↪ H 0 ( N Y ~ ( − H ) ) \iota:H^0(N_Y(-H))\hookrightarrow H^0(N_{\widetilde{Y}}(-H)) ι : H 0 ( N Y ( − H )) ↪ H 0 ( N Y ( − H )) of (27), (28). The maps in play are
0 → H 0 ( N Y ( − H ) ) → ι H 0 ( N Y ~ ( − H ) ) d Ψ ↓ ↓ a H 0 ( N C ⊗ L C ) → δ H 1 ( N Y ~ ( − H + K Y ) ) ↓ ↑ δ ‾ H 0 ( N C ⊗ L C ) / k s C = T [ C ~ ] R C (38) \begin{CD}
0 @>>> H^0(N_Y(-H)) @>{\iota}>> H^0(N_{\widetilde{Y}}(-H)) \\
@. @V{d\Psi}VV @VV{a}V \\
@. H^0(N_C\otimes L_C) @>{\delta}>> H^1(N_{\widetilde{Y}}(-H+K_Y)) \\
@. @VVV @AA{\overline{\delta}}A \\
@. H^0(N_C\otimes L_C)/k\,s_C=T_{[\widetilde{C}]}\mathcal{R}_C \tag*{(38)}
\end{CD} 0 H 0 ( N Y ( − H )) d Ψ ↓ ⏐ H 0 ( N C ⊗ L C ) ↓ ⏐ H 0 ( N C ⊗ L C ) / k s C = T [ C ] R C ι δ H 0 ( N Y ( − H )) ↓ ⏐ a H 1 ( N Y ( − H + K Y )) ⏐ ↑ δ (38)
Here F F F is the connecting map of (33) , i.e. of (2) twisted by − H ~ -\widetilde{H} − H , and by definition of a a a (diagram (35) ) a = F ∘ ι a=F\circ\iota a = F ∘ ι . The row through δ \delta δ is the cohomology of the restriction sequence
0 ⟶ N Y ⊗ K Y ( − C ) ⟶ N Y ⊗ K Y ⟶ N C ⊗ L C ⟶ 0 0\longrightarrow N_Y\otimes K_Y(-C)\longrightarrow N_Y\otimes K_Y\longrightarrow N_C\otimes L_C\longrightarrow0 0 ⟶ N Y ⊗ K Y ( − C ) ⟶ N Y ⊗ K Y ⟶ N C ⊗ L C ⟶ 0
of Lemma 3.6. δ \delta δ is its connecting map, its kernel k s C k\,s_C k s C is the image of H 0 ( N Y ⊗ K Y ) = k s H^0(N_Y\otimes K_Y)=k\,s H 0 ( N Y ⊗ K Y ) = k s , and δ ‾ \overline{\delta} δ is the induced isomorphism. The dashed arrow is explained below; it is well defined modulo k s C = ker δ k\,s_C=\ker\delta k s C = ker δ .
Pick x ∈ H 0 ( N Y ( − H ) ) x\in H^0(N_Y(-H)) x ∈ H 0 ( N Y ( − H )) and set ξ : = ι ( x ) \xi:=\iota(x) ξ := ι ( x ) . We need to show δ ‾ ( d Ψ ( x ) ) = F ( ξ ) \overline{\delta}(d\Psi(x))=F(\xi) δ ( d Ψ ( x )) = F ( ξ ) .
The vector x x x is a first-order deformation Y ε ⊂ P D M Y_{\varepsilon}\subset\mathbb{P}^M_D Y ε ⊂ P D M of Y Y Y containing C D C_D C D , and Y ε ∩ H D = C D Y_{\varepsilon}\cap H_D=C_D Y ε ∩ H D = C D (Lemma 3.7 (1), Lemma 3.11 (ii)). By Lemma 3.10 (the version of [20 Prop. 3.7] over $D) there is a K3 carpet Y ~ ε = Φ ( Y ε ) ⊂ P D M \widetilde{Y}_{\varepsilon}=\Phi(Y_{\varepsilon})\subset\mathbb{P}^M_D Y ε = Φ ( Y ε ) ⊂ P D M on Y ε Y_{\varepsilon} Y ε ; since Y ~ ε × D Spec k = Y ~ \widetilde{Y}_{\varepsilon}\times_D\operatorname{Spec}k=\widetilde{Y} Y ε × D Spec k = Y , it is an embedded first-order deformation of Y ~ ⊂ P M \widetilde{Y} \subset\mathbb{P}^{M} Y ⊂ P M . Call its class η ∈ H 0 ( N Y ~ ) \eta\in H^{0}(N_{\widetilde{Y}}) η ∈ H 0 ( N Y ) . Then C ~ ε : = Y ~ ε ∩ H D \widetilde{C}_{\varepsilon}:=\widetilde{Y}_{\varepsilon}\cap H_{D} C ε := Y ε ∩ H D is a ribbon on C D C_{D} C D with conormal bundle L C ⊗ O D L_{C}\otimes\mathcal{O}_{D} L C ⊗ O D , hence a D D D -point of R C \mathcal{R}_{C} R C at [ C ~ ] [\widetilde{C}] [ C ] , and by definition (Lemma 3.11)
d Ψ ( x ) = [ C ~ ε ] ∈ T [ C ~ ] R C = H 0 ( N C ⊗ L C ) / k s C . d\Psi(x)=[\widetilde{C}_{\varepsilon}]\in T_{[\widetilde{C}]}\mathcal{R}_{C}=H^{0}(N_{C}\otimes L_{C})/k\,s_{C}. d Ψ ( x ) = [ C ε ] ∈ T [ C ] R C = H 0 ( N C ⊗ L C ) / k s C .
Put N ~ C : = N C ~ / H = H o m ( I C ~ , O C ~ ) \widetilde{N}_{C}:=N_{\widetilde{C}/H}=\mathcal{H}om(I_{\widetilde{C}},\mathcal{O}_{\widetilde{C}}) N C := N C / H = H o m ( I C , O C ) . Composing η : I Y ~ → O Y ~ \eta:I_{\widetilde{Y}}\to\mathcal{O}_{\widetilde{Y}} η : I Y → O Y with O Y ~ → O C ~ \mathcal{O}_{\widetilde{Y}}\to\mathcal{O}_{\widetilde{C}} O Y → O C and using I Y ~ ⊗ O H = I C ~ / H I_{\widetilde{Y}}\otimes\mathcal{O}_{H}=I_{\widetilde{C}/H} I Y ⊗ O H = I C / H gives η ∣ C ~ ∈ H 0 ( N ~ C ) \eta|_{\widetilde{C}}\in H^{0}(\widetilde{N}_{C}) η ∣ C ∈ H 0 ( N C ) . The class of C ~ ε \widetilde{C}_{\varepsilon} C ε as a first-order deformation of C ~ \widetilde{C} C inside H H H (reduce the generators f + ε η ~ ( f ) f+\varepsilon\widetilde{\eta}(f) f + ε η ( f ) of I Y ~ I_{\widetilde{Y}} I Y modulo h h h ). For the ribbon C ~ ⊂ H \widetilde{C}\subset H C ⊂ H , whose conormal bundle is L C = I C / I C ~ L_{C}=I_{C}/I_{\widetilde{C}} L C = I C / I C , we have the analogues of (2)–(4) [16 Lemma 4.1] :
0 → N ~ C ⊗ L C → N ~ C → N ~ C ⊗ O C → 0 , 0 → H C → N ~ C ⊗ O C → L C − 2 → 0 , 0 → L C − 1 → s C V N C → H C → 0 , where H C : = H o m ( I C ~ / I C 2 , O C ) . (39) \begin{aligned}
0&\to\widetilde{N}_{C}\otimes L_{C}\to\widetilde{N}_{C}\to\widetilde{N}_{C}\otimes\mathcal{O}_{C}\to0, \qquad
0\to\mathcal{H}_{C}\to\widetilde{N}_{C}\otimes\mathcal{O}_{C}\to L_{C}^{-2}\to0,\\
0&\to L_{C}^{-1}\xrightarrow{s_{C}^{V}}N_{C}\to\mathcal{H}_{C}\to0,\qquad
\text{where }\mathcal{H}_{C}:=\mathcal{H}om(I_{\widetilde{C}}/I_{C}^{2},\mathcal{O}_{C}). \tag*{(39)}
\end{aligned} 0 0 → N C ⊗ L C → N C → N C ⊗ O C → 0 , 0 → H C → N C ⊗ O C → L C − 2 → 0 , → L C − 1 s C V N C → H C → 0 , where H C := H o m ( I C / I C 2 , O C ) . (39)
Claim.
(i) η ∣ C ~ ∈ H 0 ( N ~ C ⊗ L C ) ⊂ H 0 ( N ~ C ) \eta|_{\widetilde{C}}\in H^{0}(\widetilde{N}_{C}\otimes L_{C})\subset H^{0}(\widetilde{N}_{C}) η ∣ C ∈ H 0 ( N C ⊗ L C ) ⊂ H 0 ( N C ) .
(ii) Restriction of homomorphisms from I C I_{C} I C to I C ~ I_{\widetilde{C}} I C induces an injection H 0 ( N C ⊗ L C ) / k s C ↪ H 0 ( N ~ C ⊗ L C ) H^{0}(N_{C}\otimes L_{C})/k\,s_{C}\hookrightarrow H^{0}(\widetilde{N}_{C}\otimes L_{C}) H 0 ( N C ⊗ L C ) / k s C ↪ H 0 ( N C ⊗ L C ) , and η ∣ C ~ \eta|_{\widetilde{C}} η ∣ C lies in its image.
(iii) Under this injection, η ∣ C ~ = − d Ψ ( x ) \eta|_{\widetilde{C}}=-d\Psi(x) η ∣ C = − d Ψ ( x ) .
Proof of the Claim. (i) The image of η ∣ C ~ \eta|_{\widetilde{C}} η ∣ C in H 0 ( N ~ C ⊗ O C ) = H o m ( I C ~ , O C ) H^{0}(\widetilde{N}_{C}\otimes\mathcal{O}_{C})=\mathcal{H}om(I_{\widetilde{C}},\mathcal{O}_{C}) H 0 ( N C ⊗ O C ) = H o m ( I C , O C ) is η ∣ C : f ↦ η ( f ) ∣ C \eta|_{C}:f\mapsto\eta(f)|_{C} η ∣ C : f ↦ η ( f ) ∣ C . By Lemma 3.10(1), η ( f ) ≡ x ( f ) ( m o d I Y ) \eta(f)\equiv x(f)\pmod{I_{Y}} η ( f ) ≡ x ( f ) ( mod I Y ) ; so η ∣ C \eta|_{C} η ∣ C is the restriction to I C ~ I_{\widetilde{C}} I C of the class f ↦ x ( f ) ∣ C f\mapsto x(f)|_{C} f ↦ x ( f ) ∣ C of Y ε ∩ H D Y_{\varepsilon}\cap H_{D} Y ε ∩ H D as a deformation of C C C in H H H , which is 0 0 0 because Y ε ∩ H D = C D Y_{\varepsilon}\cap H_{D}=C_{D} Y ε ∩ H D = C D . Hence η ∣ C = 0 \eta|_{C}=0 η ∣ C = 0 , and by the first sequence of (39), η ∣ C ~ ∈ H 0 ( N ~ C ⊗ L C ) = H o m ( I C ~ , L C ) \eta|_{\widetilde{C}}\in H^{0}(\widetilde{N}_{C}\otimes L_{C})=\mathcal{H}om(I_{\widetilde{C}},L_{C}) η ∣ C ∈ H 0 ( N C ⊗ L C ) = H o m ( I C , L C ) .
(ii) Tensoring the third sequence of (39) with L C L_{C} L C gives 0 → O C → s C N C ⊗ L C → H C ⊗ L C → 0 0\to\mathcal{O}_{C}\xrightarrow{s_{C}}N_{C}\otimes L_{C}\to\mathcal{H}_{C}\otimes L_{C}\to0 0 → O C s C N C ⊗ L C → H C ⊗ L C → 0 , whence H 0 ( N C ⊗ L C ) / k s C ≅ H 0 ( H C ⊗ L C ) H^{0}(N_{C}\otimes L_{C})/k\,s_{C}\cong H^{0}(\mathcal{H}_{C}\otimes L_{C}) H 0 ( N C ⊗ L C ) / k s C ≅ H 0 ( H C ⊗ L C ) as H 1 ( O C ) = 0 H^{1}(\mathcal{O}_{C})=0 H 1 ( O C ) = 0 ; tensoring the second with L C L_{C} L C gives 0 → H C ⊗ L C → N ~ C ⊗ L C → L C − 1 → 0 0\to\mathcal{H}_{C}\otimes L_{C}\to\widetilde{N}_{C}\otimes L_{C}\to L_{C}^{-1}\to0 0 → H C ⊗ L C → N C ⊗ L C → L C − 1 → 0 , whence H 0 ( H C ⊗ L C ) ↪ H 0 ( N ~ C ⊗ L C ) H^{0}(\mathcal{H}_{C}\otimes L_{C})\hookrightarrow H^{0}(\widetilde{N}_{C}\otimes L_{C}) H 0 ( H C ⊗ L C ) ↪ H 0 ( N C ⊗ L C ) . The composite is t ↦ t ∣ I C ~ t\mapsto t|_{I_{\widetilde{C}}} t ↦ t ∣ I C . Now C ~ ε \widetilde{C}_{\varepsilon} C ε is a double structure on C D C_{D} C D , i.e. I C D 2 ⊂ I C ~ ε I_{C_{D}}^{2}\subset I_{\widetilde{C}_{\varepsilon}} I C D 2 ⊂ I C ε (Lemma 3.11(ii)); in terms of the class η ∣ C ~ \eta|_{\widetilde{C}} η ∣ C this says exactly that η ∣ C ~ ∈ H o m ( I C ~ , L C ) \eta|_{\widetilde{C}}\in\mathcal{H}om(I_{\widetilde{C}},L_{C}) η ∣ C ∈ H o m ( I C , L C ) vanishes on I C 2 I_{C}^{2} I C 2 , i.e. η ∣ C ~ ∈ H o m ( I C ~ / I C 2 , L C ) = H 0 ( H C ⊗ L C ) \eta|_{\widetilde{C}}\in\mathcal{H}om(I_{\widetilde{C}}/I_{C}^{2},L_{C})=H^{0}(\mathcal{H}_{C}\otimes L_{C}) η ∣ C ∈ H o m ( I C / I C 2 , L C ) = H 0 ( H C ⊗ L C ) .
(iii) Write d Ψ ( x ) = [ t ] d\Psi(x)=[t] d Ψ ( x ) = [ t ] , i.e. C ~ ε = ker ( s C + ε t ) \widetilde{C}_{\varepsilon}=\ker(s_{C}+\varepsilon t) C ε = ker ( s C + εt ) (Lemma 3.11). For f ∈ I C ~ f\in I_{\widetilde{C}} f ∈ I C the element f + ε g f+\varepsilon g f + ε g lies in I C ~ ε I_{\widetilde{C}_{\varepsilon}} I C ε if and only if s C ( g ) = − t ( f ) s_{C}(g)=-t(f) s C ( g ) = − t ( f ) ; so the class of C ~ ε \widetilde{C}_{\varepsilon} C ε is f ↦ − t ( f ) ∈ L C = I C / I C ~ f\mapsto-t(f)\in L_{C}=I_{C}/I_{\widetilde{C}} f ↦ − t ( f ) ∈ L C = I C / I C , that is, η ∣ C ~ = − t ∣ I C ~ \eta|_{\widetilde{C}}=-t|_{I_{\widetilde{C}}} η ∣ C = − t ∣ I C . This proves the Claim.
Now let δ ~ : H 0 ( N ~ C ⊗ L C ) → H 1 ( N ~ Y ~ ( − H + K Y ) ) \widetilde{\delta}:H^{0}(\widetilde{N}_{C}\otimes L_{C})\to H^{1}(\widetilde{N}_{\widetilde{Y}}(-H+K_{Y})) δ : H 0 ( N C ⊗ L C ) → H 1 ( N Y ( − H + K Y )) be the connecting map of the restriction sequence
0 → N ~ Y ~ ( − H + K Y ) → N ~ Y ~ ( K Y ) → N ~ Y ~ ( K Y ) ∣ C = N ~ C ⊗ L C → 0 0\to\widetilde{N}_{\widetilde{Y}}(-H+K_{Y})\to\widetilde{N}_{\widetilde{Y}}(K_{Y})\to\widetilde{N}_{\widetilde{Y}}(K_{Y})|_{C}=\widetilde{N}_{C}\otimes L_{C}\to0 0 → N Y ( − H + K Y ) → N Y ( K Y ) → N Y ( K Y ) ∣ C = N C ⊗ L C → 0
of N ~ Y ~ ( K Y ) = H o m ( I Y ~ , K Y ) \widetilde{N}_{\widetilde{Y}}(K_{Y})=\mathcal{H}om(I_{\widetilde{Y}},K_{Y}) N Y ( K Y ) = H o m ( I Y , K Y ) , K Y = I Y / I Y ~ K_{Y}=I_{Y}/I_{\widetilde{Y}} K Y = I Y / I Y . Restriction of homomorphisms from I Y I_{Y} I Y to I Y ~ I_{\widetilde{Y}} I Y maps the restriction sequence of N Y ⊗ K Y N_{Y}\otimes K_{Y} N Y ⊗ K Y to this one, inducing j j j on H 1 H^{1} H 1 of the first terms and the injection of (ii) on H 0 H^{0} H 0 of the third terms; so δ ~ \widetilde{\delta} δ restricts to δ \delta δ on H 0 ( N C ⊗ L C ) / k s C H^{0}(N_{C}\otimes L_{C})/k\,s_{C} H 0 ( N C ⊗ L C ) / k s C , and by (iii)
δ ~ ( η ∣ C ~ ) = − δ ‾ ( d Ψ ( x ) ) . \widetilde{\delta}(\eta|_{\widetilde{C}})=-\overline{\delta}(d\Psi(x)). δ ( η ∣ C ) = − δ ( d Ψ ( x )) .
Therefore what we need to show is the following.
Lemma. δ ~ ( η ∣ C ~ ) = − F ( ξ ) \widetilde{\delta}(\eta|_{\widetilde{C}})=-F(\xi) δ ( η ∣ C ) = − F ( ξ ) .
Proof of the Lemma. Multiplication by h h h followed by restriction to C ~ \widetilde{C} C gives a commutative diagram of sheaves on Y ~ \widetilde{Y} Y with exact rows and columns
0 0 0 ↓ ↓ ↓ 0 ⟶ N Y ~ ( − H + K Y ) ⟶ N Y ~ ( − H ~ ) ⟶ N Y ~ ( − H ) ⟶ 0 ↓ h ↓ h ↓ h 0 ⟶ N Y ~ ( K Y ) ⟶ N Y ~ ⟶ N Y ~ ⊗ O Y ⟶ 0 ↓ ↓ ↓ 0 ⟶ N C ⊗ L C ⟶ N C ⟶ N C ⊗ O C ⟶ 0 ↓ ↓ ↓ 0 0 0 (40) \begin{array}{ccccccccc}
& & 0 & & 0 & & 0 & & \\
& & \downarrow& & \downarrow& & \downarrow& & \\
0 & \longrightarrow& N_{\widetilde{Y}}(-H+K_Y) & \longrightarrow& N_{\widetilde{Y}}(-\widetilde{H}) & \longrightarrow& N_{\widetilde{Y}}(-H) & \longrightarrow& 0 \\
& & \downarrow h & & \downarrow h & & \downarrow h & & \\
0 & \longrightarrow& N_{\widetilde{Y}}(K_Y) & \longrightarrow& N_{\widetilde{Y}} & \longrightarrow& N_{\widetilde{Y}}\otimes\mathcal{O}_Y & \longrightarrow& 0 \\
& & \downarrow& & \downarrow& & \downarrow& & \\
0 & \longrightarrow& N_C\otimes L_C & \longrightarrow& N_C & \longrightarrow& N_C\otimes\mathcal{O}_C & \longrightarrow& 0 \\
& & \downarrow& & \downarrow& & \downarrow& & \\
& & 0 & & 0 & & 0 & & \tag*{(40)}
\end{array} 0 0 0 ⟶ ⟶ ⟶ 0 ↓ N Y ( − H + K Y ) ↓ h N Y ( K Y ) ↓ N C ⊗ L C ↓ 0 ⟶ ⟶ ⟶ 0 ↓ N Y ( − H ) ↓ h N Y ↓ N C ↓ 0 ⟶ ⟶ ⟶ 0 ↓ N Y ( − H ) ↓ h N Y ⊗ O Y ↓ N C ⊗ O C ↓ 0 ⟶ ⟶ ⟶ 0 0 0 (40)
whose rows are (2) twisted by − H ~ -\widetilde{H} − H , (2) itself, and the first sequence of (39); the columns are exact because h h h is a nonzerodivisor on O Y ~ \mathcal{O}_{\widetilde{Y}} O Y and on O Y \mathcal{O}_Y O Y . F F F is the connecting map of the top row and δ ~ \widetilde{\delta} δ that of the left column. Lift ξ \xi ξ locally to sections ξ i \xi_i ξ i of N Y ~ ( − H ~ ) N_{\widetilde{Y}}(-\widetilde{H}) N Y ( − H ) over an affine cover { U i } \{U_i\} { U i } ; then F ( ξ ) = [ ξ i − ξ j ] F(\xi)=[\xi_i-\xi_j] F ( ξ ) = [ ξ i − ξ j ] . We make two observations:
(a) View the ξ i \xi_i ξ i as local sections of N Y ~ N_{\widetilde{Y}} N Y (middle column). Then η ∣ U i − ξ i \eta|_{U_i}-\xi_i η ∣ U i − ξ i restricts to zero on Y Y Y , since η \eta η and ξ i \xi_i ξ i both reduce to ξ = x ∣ I Y \xi=x|_{I_Y} ξ = x ∣ I Y (Lemma 3.10(1)). Hence μ i : = η ∣ U i − ξ i \mu_i:=\eta|_{U_i}-\xi_i μ i := η ∣ U i − ξ i are local sections of N Y ~ ( K Y ) N_{\widetilde{Y}}(K_Y) N Y ( K Y ) .
(b) The ξ i \xi_i ξ i vanish on C ~ \widetilde{C} C , being multiples of h h h ; so μ i ∣ C ~ = η ∣ C ~ ∩ U i \mu_i|_{\widetilde{C}}=\eta|_{\widetilde{C}\cap U_i} μ i ∣ C = η ∣ C ∩ U i , i.e. the μ i \mu_i μ i are local lifts of η ∣ C ~ \eta|_{\widetilde{C}} η ∣ C to N Y ~ ( K Y ) N_{\widetilde{Y}}(K_Y) N Y ( K Y ) .
Consequently δ ~ ( η ∣ C ~ ) = [ μ i − μ j ] = [ ξ j − ξ i ] = − F ( ξ ) \widetilde{\delta}(\eta|_{\widetilde{C}})=[\mu_i-\mu_j]=[\xi_j-\xi_i]=-F(\xi) δ ( η ∣ C ) = [ μ i − μ j ] = [ ξ j − ξ i ] = − F ( ξ ) , which proves the Lemma.
By the Lemma, δ ‾ ( d Ψ ( x ) ) = − δ ~ ( η ∣ C ~ ) = F ( ξ ) = a ( x ) \overline{\delta}(d\Psi(x))=-\widetilde{\delta}(\eta|_{\widetilde{C}})=F(\xi)=a(x) δ ( d Ψ ( x )) = − δ ( η ∣ C ) = F ( ξ ) = a ( x ) . □
Construction of local sections of Ψ \Psi Ψ and surjectivity of the differential d Ψ d\Psi d Ψ
In this subsubsection, we construct the local sections of Ψ \Psi Ψ and show that the differential d Ψ d\Psi d Ψ is surjective. Put Q : = K C ⊗ L C − 1 = O C ( C ) = O P 1 ( g − 1 ) Q:=K_C\otimes L_C^{-1}=\mathcal{O}_C(C)=\mathcal{O}_{\mathbb{P}^1}(g-1) Q := K C ⊗ L C − 1 = O C ( C ) = O P 1 ( g − 1 ) . For Y ′ ∈ S C Y'\in\mathcal{S}_C Y ′ ∈ S C let π ′ : Y ′ → C \pi':Y'\to C π ′ : Y ′ → C be the ruling for which C C C is a section and E Y ′ : = π ∗ ′ O Y ′ ( C ) E_{Y'}:=\pi'_* \mathcal{O}_{Y'}(C) E Y ′ := π ∗ ′ O Y ′ ( C ) . Pushing forward we have 0 → O Y ′ → O Y ′ ( C ) → O C ( C ) → 0 0\to\mathcal{O}_{Y'}\to\mathcal{O}_{Y'}(C)\to\mathcal{O}_C(C)\to0 0 → O Y ′ → O Y ′ ( C ) → O C ( C ) → 0 since R 1 π ∗ ′ O Y ′ = 0 R^1\pi'_*\mathcal{O}_{Y'}=0 R 1 π ∗ ′ O Y ′ = 0 . This gives the extension
0 ⟶ O C ⟶ E Y ′ ⟶ Q ⟶ 0. (41) 0\longrightarrow\mathcal{O}_C\longrightarrow E_{Y'}\longrightarrow Q\longrightarrow0. \tag*{(41)} 0 ⟶ O C ⟶ E Y ′ ⟶ Q ⟶ 0. (41)
and Y ′ = P ( E Y ′ ) Y'=\mathbb{P}(E_{Y'}) Y ′ = P ( E Y ′ ) with O P ( E Y ′ ) ( 1 ) = O Y ′ ( C ) \mathcal{O}_{\mathbb{P}(E_{Y'})}(1)=\mathcal{O}_{Y'}(C) O P ( E Y ′ ) ( 1 ) = O Y ′ ( C ) and C C C the section defined by E Y ′ → Q E_{Y'}\to Q E Y ′ → Q . Since Y ′ Y' Y ′ is smooth, E Y ′ ≅ O ( a 1 ′ ) ⊕ O ( a 2 ′ ) E_{Y'}\cong\mathcal{O}(a'_1)\oplus\mathcal{O}(a'_2) E Y ′ ≅ O ( a 1 ′ ) ⊕ O ( a 2 ′ ) with a i ′ ≥ 1 a'_i\ge1 a i ′ ≥ 1 , so (41) is not split. Put V Y ′ : = E Y ′ ⊗ L C V_{Y'}:=E_{Y'}\otimes L_C V Y ′ := E Y ′ ⊗ L C , an extension 0 → L C → V Y ′ → K C → 0 0\to L_C\to V_{Y'}\to K_C\to0 0 → L C → V Y ′ → K C → 0 with class [ V Y ′ ] ∈ Ext 1 ( K C , L C ) = Ext 1 ( Ω C , L C ) [V_{Y'}]\in\operatorname{Ext}^1(K_C,L_C)=\operatorname{Ext}^1(\Omega_C,L_C) [ V Y ′ ] ∈ Ext 1 ( K C , L C ) = Ext 1 ( Ω C , L C ) , [ V Y ′ ] ≠ 0 [V_{Y'}]\ne0 [ V Y ′ ] = 0 . We will also denote the extension class of (41) by [ E Y ′ ] [E_{Y'}] [ E Y ′ ] .
The two spaces R C \mathcal{R}_C R C and P ( Ext 1 ( Ω C , L C ) ) \mathbb{P}(\operatorname{Ext}^1(\Omega_C,L_C)) P ( Ext 1 ( Ω C , L C )) are identified by the classical dictionary of Bayer–Eisenbud:
Lemma 3.14. Let ϵ C ∈ Ext 1 ( Ω C , N C ∨ ) \epsilon_C\in\operatorname{Ext}^1(\Omega_C,N_C^\vee) ϵ C ∈ Ext 1 ( Ω C , N C ∨ ) be the class of the conormal sequence 0 → N C ∨ → Ω H ∣ C → Ω C → 0 0\to N_C^\vee\to\Omega_H|_C\to\Omega_C\to0 0 → N C ∨ → Ω H ∣ C → Ω C → 0 , and let
β : H 0 ( N C ⊗ L C ) = Hom ( N C ∨ , L C ) ⟶ Ext 1 ( Ω C , L C ) , t ⟼ t ∗ ( ϵ C ) \beta:H^0(N_C\otimes L_C)=\operatorname{Hom}(N_C^\vee,L_C)\longrightarrow\operatorname{Ext}^1(\Omega_C,L_C),\qquad t\longmapsto t_*(\epsilon_C) β : H 0 ( N C ⊗ L C ) = Hom ( N C ∨ , L C ) ⟶ Ext 1 ( Ω C , L C ) , t ⟼ t ∗ ( ϵ C ) (push-out along t t t ). Then β ( t ) \beta(t) β ( t ) is the class of the restricted cotangent sequence 0 → L C → Ω C ~ ∣ C → Ω C → 0 0\to L_C\to\Omega_{\widetilde{C}}|_C\to\Omega_C\to0 0 → L C → Ω C ∣ C → Ω C → 0 of the double structure C ~ \widetilde{C} C , and β \beta β is an isomorphism of ( g − 2 ) (g-2) ( g − 2 ) -dimensional vector spaces.
Proof. Tensoring the conormal sequence of C ~ ⊂ H \widetilde{C}\subset H C ⊂ H with O C \mathcal{O}_C O C gives I C ~ / I C I C ~ → Ω H ∣ C → Ω C ~ ∣ C → 0 I_{\widetilde{C}}/I_CI_{\widetilde{C}}\to\Omega_H|_C\to\Omega_{\widetilde{C}}|_C\to0 I C / I C I C → Ω H ∣ C → Ω C ∣ C → 0 , and the image of the first map is the image of ker t = I C ~ / I C 2 ⊂ N C ∨ \ker t=I_{\widetilde{C}}/I_C^2\subset N_C^\vee ker t = I C / I C 2 ⊂ N C ∨ in Ω H ∣ C \Omega_H|_C Ω H ∣ C . Hence Ω C ~ ∣ C = Ω H ∣ C / ker t \Omega_{\widetilde{C}}|_C=\Omega_H|_C/\ker t Ω C ∣ C = Ω H ∣ C / ker t , and 0 → N C ∨ / ker t = L C → Ω H ∣ C / ker t → Ω C → 0 0\to N_C^\vee/\ker t=L_C\to\Omega_H|_C/\ker t\to\Omega_C\to0 0 → N C ∨ / ker t = L C → Ω H ∣ C / ker t → Ω C → 0 is the push-out of the conormal sequence along t t t ; this is the first claim (cf. [3 Theorem 1.2] ). For the second, apply Hom ( − , L C ) \operatorname{Hom}(-,L_C) Hom ( − , L C ) to the conormal sequence: ker β \ker\beta ker β is the image of Hom ( Ω H ∣ C , L C ) = H 0 ( T H ∣ C ⊗ L C ) \operatorname{Hom}(\Omega_{H|C}, L_C)=H^0(T_{H|C}\otimes L_C) Hom ( Ω H ∣ C , L C ) = H 0 ( T H ∣ C ⊗ L C ) . From the Euler sequence of H ≅ P g − 1 H\cong\mathbb{P}^{g-1} H ≅ P g − 1 restricted to C C C and tensored with L C L_C L C ,
0 → L C → L C ( 1 ) ⊕ g → T H ∣ C ⊗ L C → 0 , L C ( 1 ) = O P 1 ( − 2 ) , 0\to L_C\to L_C(1)^{\oplus g}\to T_{H|C}\otimes L_C\to0,\qquad L_C(1)=\mathcal{O}_{\mathbb{P}^1}(-2), 0 → L C → L C ( 1 ) ⊕ g → T H ∣ C ⊗ L C → 0 , L C ( 1 ) = O P 1 ( − 2 ) ,
we get H 0 ( T H ∣ C ⊗ L C ) = ker ( H 1 ( L C ) → H 1 ( L C ( 1 ) ) ⊗ H 0 ( O H ( 1 ) ) ∨ ) H^0(T_{H|C}\otimes L_C)=\ker\left(H^1(L_C)\to H^1(L_C(1))\otimes H^0(\mathcal{O}_H(1))^\vee\right) H 0 ( T H ∣ C ⊗ L C ) = ker ( H 1 ( L C ) → H 1 ( L C ( 1 )) ⊗ H 0 ( O H ( 1 ) ) ∨ ) . This map is adjoint to the multiplication pairing H 1 ( O P 1 ( − g − 1 ) ) ⊗ H 0 ( O P 1 ( g − 1 ) ) → H 1 ( O P 1 ( − 2 ) ) = k H^1(\mathcal{O}_{\mathbb{P}^1}(-g-1))\otimes H^0(\mathcal{O}_{\mathbb{P}^1}(g-1))\to H^1(\mathcal{O}_{\mathbb{P}^1}(-2))=k H 1 ( O P 1 ( − g − 1 )) ⊗ H 0 ( O P 1 ( g − 1 )) → H 1 ( O P 1 ( − 2 )) = k , which is perfect by Serre duality; hence it is injective and H 0 ( T H ∣ C ⊗ L C ) = 0 H^0(T_{H|C}\otimes L_C)=0 H 0 ( T H ∣ C ⊗ L C ) = 0 . So β \beta β is injective, and both spaces have dimension g − 2 g-2 g − 2 . □
In the following theorem, given a point in R C \mathcal{R}_C R C we construct a map Γ U \Gamma_U Γ U defined around a neighbourhood U U U of the point. We will show in Proposition 3.17 , that Γ U \Gamma_U Γ U is a section of Ψ \Psi Ψ over U U U .
Theorem 3.15. Let pr : C × R C → R C \operatorname{pr}: C\times\mathcal{R}_C\to\mathcal{R}_C pr : C × R C → R C be the projection.
(1) The restricted cotangent sequence of the universal ribbon C ~ R C ⊂ H × R C \widetilde{C}_{\mathcal{R}_C}\subset H\times\mathcal{R}_C C R C ⊂ H × R C ,
0 → L C ⊠ O R C ( 1 ) → Ω C ~ R C / R C ∣ C × R C → K C ⊠ O R C → 0 , (42) 0\to L_C\boxtimes\mathcal{O}_{\mathcal{R}_C}(1)\to\Omega_{\widetilde{C}_{\mathcal{R}_C}/\mathcal{R}_C}|_{C\times\mathcal{R}_C}\to K_C\boxtimes\mathcal{O}_{\mathcal{R}_C}\to0, \tag*{(42)} 0 → L C ⊠ O R C ( 1 ) → Ω C R C / R C ∣ C × R C → K C ⊠ O R C → 0 , (42) is exact, and V : = Ω C ~ R C / R C ∣ C × R C ⊗ ( L C − 1 ⊠ O R C ( − 1 ) ) \mathcal{V}:=\Omega_{\widetilde{C}_{\mathcal{R}_C}/\mathcal{R}_C}|_{C\times\mathcal{R}_C}\otimes(L_C^{-1}\boxtimes\mathcal{O}_{\mathcal{R}_C}(-1)) V := Ω C R C / R C ∣ C × R C ⊗ ( L C − 1 ⊠ O R C ( − 1 )) sits in an extension
0 → O C × R C → V → Q ⊠ O R C ( − 1 ) → 0 (43) 0\to\mathcal{O}_{C\times\mathcal{R}_C}\to\mathcal{V}\to Q\boxtimes\mathcal{O}_{\mathcal{R}_C}(-1)\to0 \tag*{(43)} 0 → O C × R C → V → Q ⊠ O R C ( − 1 ) → 0 (43) whose fibre V u \mathcal{V}_u V u over u = [ t ] ∈ R C u=[t]\in\mathcal{R}_C u = [ t ] ∈ R C is isomorphic to Ω C ~ t ∣ C ⊗ L C − 1 \Omega_{\widetilde{C}_t|C}\otimes L_C^{-1} Ω C t ∣ C ⊗ L C − 1 : a non-split extension of Q Q Q by O C \mathcal{O}_C O C with [ V u ⊗ L C ] ∈ k × β ( t ) [\mathcal{V}_u\otimes L_C]\in k^\times\beta(t) [ V u ⊗ L C ] ∈ k × β ( t ) .
(2) Every point of R C \mathcal{R}_C R C has an open neighbourhood U U U with a morphism Γ U : U → S C \Gamma_U:U\to\mathcal{S}_C Γ U : U → S C such that, for every u ∈ U u\in U u ∈ U , Γ U ( u ) \Gamma_U(u) Γ U ( u ) is P ( V u ) \mathbb{P}(\mathcal{V}_u) P ( V u ) embedded by ∣ O P ( V u ) ( 1 ) ∣ |\mathcal{O}_{\mathbb{P}(\mathcal{V}_u)}(1)| ∣ O P ( V u ) ( 1 ) ∣ , with C C C the section defined by V u → Q \mathcal{V}_u\to Q V u → Q . In particular [ E Γ U ( u ) ] = [ V u ] [E_{\Gamma_U(u)}]=[\mathcal{V}_u] [ E Γ U ( u ) ] = [ V u ] as extensions of Q Q Q by O C \mathcal{O}_C O C , and [ V Γ U ( u ) ] ∈ k × β ( t ) [\mathcal{V}_{\Gamma_U(u)}]\in k^\times\beta(t) [ V Γ U ( u ) ] ∈ k × β ( t ) for u = [ t ] u=[t] u = [ t ] .
Proof. (1) The sequence (42) is the conormal sequence of C × R C ⊂ C ~ R C C\times\mathcal{R}_C\subset\widetilde{C}_{\mathcal{R}_C} C × R C ⊂ C R C relative to R C \mathcal{R}_C R C : its first term is I C × R C / I C ~ R C = L C ⊠ O R C ( 1 ) I_{C\times\mathcal{R}_C}/I_{\widetilde{C}_{\mathcal{R}_C}}=L_C\boxtimes\mathcal{O}_{\mathcal{R}_C}(1) I C × R C / I C R C = L C ⊠ O R C ( 1 ) and Ω C × R C / R C = K C ⊠ O R C \Omega_{C\times\mathcal{R}_C/\mathcal{R}_C}=K_C\boxtimes\mathcal{O}_{\mathcal{R}_C} Ω C × R C / R C = K C ⊠ O R C . It is exact on the left because C ~ R C \widetilde{C}_{\mathcal{R}_C} C R C is locally isomorphic to ( C × R C ) × D (C\times\mathcal{R}_C)\times D ( C × R C ) × D (ribbons are locally split, [3 §1] ; the same local computation applies to C ~ R C \widetilde{C}_{\mathcal{R}_C} C R C , which is defined by the surjection τ \tau τ ). Its restriction to C × { [ t ] } C\times\{[t]\} C × {[ t ]} is the restricted cotangent sequence 0 → L C → Ω C ~ t ∣ C → K C → 0 0\to L_C\to\Omega_{\widetilde{C}_t|C}\to K_C\to0 0 → L C → Ω C t ∣ C → K C → 0 of C ~ t \widetilde{C}_t C t , of class β ( t ) ≠ 0 \beta(t)\ne0 β ( t ) = 0 by Lemma 3.14. Tensoring with L C − 1 ⊠ O R C ( − 1 ) L_C^{-1}\boxtimes\mathcal{O}_{\mathcal{R}_C}(-1) L C − 1 ⊠ O R C ( − 1 ) gives (43) and the description of its fibres.
(2) The surjection V → Q ⊠ O R C ( − 1 ) \mathcal{V}\to Q\boxtimes\mathcal{O}_{\mathcal{R}_C}(-1) V → Q ⊠ O R C ( − 1 ) gives a section i : C × R C → P ( V ) i:C\times\mathcal{R}_C\to\mathbb{P}(\mathcal{V}) i : C × R C → P ( V ) of p : P ( V ) → C × R C p:\mathbb{P}(\mathcal{V})\to C\times\mathcal{R}_C p : P ( V ) → C × R C with i ∗ O P ( V ) ( 1 ) = Q ⊠ O R C ( − 1 ) i^*\mathcal{O}_{\mathbb{P}(\mathcal{V})}(1)=Q\boxtimes\mathcal{O}_{\mathcal{R}_C}(-1) i ∗ O P ( V ) ( 1 ) = Q ⊠ O R C ( − 1 ) ; since the kernel of this surjection is O C × R C \mathcal{O}_{C\times\mathcal{R}_C} O C × R C , i ( C × R C ) i(C\times\mathcal{R}_C) i ( C × R C ) is the divisor of a section of O P ( V ) ( 1 ) \mathcal{O}_{\mathbb{P}(\mathcal{V})}(1) O P ( V ) ( 1 ) . Let W : = pr ∗ p ∗ O P ( V ) ( 1 ) = pr ∗ V \mathcal{W}:=\operatorname{pr}_*p_*\mathcal{O}_{\mathbb{P}(\mathcal{V})}(1)=\operatorname{pr}_*\mathcal{V} W := pr ∗ p ∗ O P ( V ) ( 1 ) = pr ∗ V . For u ∈ R C u\in\mathcal{R}_C u ∈ R C , W u = H 0 ( V u ) \mathcal{W}_u=H^0(\mathcal{V}_u) W u = H 0 ( V u ) with V u ≅ O ( a 1 ) ⊕ O ( a 2 ) \mathcal{V}_u\cong\mathcal{O}(a_1)\oplus\mathcal{O}(a_2) V u ≅ O ( a 1 ) ⊕ O ( a 2 ) , a 1 + a 2 = g − 1 a_1+a_2=g-1 a 1 + a 2 = g − 1 , a i ≥ 1 a_i\ge1 a i ≥ 1 (an extension of O ( g − 1 ) \mathcal{O}(g-1) O ( g − 1 ) by O \mathcal{O} O with a 1 = 0 a_1=0 a 1 = 0 is split); so h 0 ( V u ) = g + 1 = M + 1 h^0(\mathcal{V}_u)=g+1=M+1 h 0 ( V u ) = g + 1 = M + 1 , h 1 ( V u ) = 0 h^1(\mathcal{V}_u)=0 h 1 ( V u ) = 0 , and W \mathcal{W} W is a vector bundle of rank M + 1 M+1 M + 1 . Embed P ( V ) \mathbb{P}(\mathcal{V}) P ( V ) in P ( W ) \mathbb{P}(\mathcal{W}) P ( W ) over R C \mathcal{R}_C R C by ∣ O P ( V ) ( 1 ) ∣ |\mathcal{O}_{\mathbb{P}(\mathcal{V})}(1)| ∣ O P ( V ) ( 1 ) ∣ . Pushing forward (43) ,
0 → O R C → W → H 0 ( Q ) ⊗ O R C ( − 1 ) → 0 , 0\to\mathcal{O}_{\mathcal{R}_C}\to\mathcal{W}\to H^0(Q)\otimes\mathcal{O}_{\mathcal{R}_C}(-1)\to0, 0 → O R C → W → H 0 ( Q ) ⊗ O R C ( − 1 ) → 0 ,
and i ( C × R C ) i(C\times\mathcal{R}_C) i ( C × R C ) lies in P ( H 0 ( Q ) ⊗ O R C ( − 1 ) ) = P ( H 0 ( Q ) ) × R C = H × R C \mathbb{P}(H^0(Q)\otimes\mathcal{O}_{\mathcal{R}_C}(-1))=\mathbb{P}(H^0(Q))\times\mathcal{R}_C=H\times\mathcal{R}_C P ( H 0 ( Q ) ⊗ O R C ( − 1 )) = P ( H 0 ( Q )) × R C = H × R C , where it is C × R C C\times\mathcal{R}_C C × R C , because i ( C × { u } ) i(C\times\{u\}) i ( C × { u }) is embedded by ∣ i ∗ O ( 1 ) ∣ = ∣ O C ( 1 ) ∣ |i^*\mathcal{O}(1)|=|\mathcal{O}_C(1)| ∣ i ∗ O ( 1 ) ∣ = ∣ O C ( 1 ) ∣ . Given u ∈ R C u\in\mathcal{R}_C u ∈ R C , let U ∋ u U\ni u U ∋ u be an open set on which O R C ( − 1 ) \mathcal{O}_{\mathcal{R}_C}(-1) O R C ( − 1 ) is trivial, e.g. the complement of a hyperplane, and choose a splitting of the sequence (it exists, as Ext 1 ( H 0 ( Q ) ′ ⊗ O R C ( − 1 ) , O R C ) = H 0 ( Q ) ∨ ⊗ H 1 ( O R C ( 1 ) ) = 0 \operatorname{Ext}^1(H^0(Q)'\otimes\mathcal{O}_{\mathcal{R}_C}(-1),\mathcal{O}_{\mathcal{R}_C})=H^0(Q)^\vee\otimes H^1(\mathcal{O}_{\mathcal{R}_C}(1))=0 Ext 1 ( H 0 ( Q ) ′ ⊗ O R C ( − 1 ) , O R C ) = H 0 ( Q ) ∨ ⊗ H 1 ( O R C ( 1 )) = 0 ). This gives W ∣ U ≅ ( k ⊕ H 0 ( Q ) ) ⊗ O U ≅ H 0 ( O P M ( 1 ) ) ⊗ O U \mathcal{W}|_U\cong(k\oplus H^0(Q))\otimes\mathcal{O}_U\cong H^0(\mathcal{O}_{\mathbb{P}^M}(1))\otimes\mathcal{O}_U W ∣ U ≅ ( k ⊕ H 0 ( Q )) ⊗ O U ≅ H 0 ( O P M ( 1 )) ⊗ O U compatible with the embeddings onto H 0 ( Q ) = H 0 ( O H ( 1 ) ) H^0(Q)=H^0(\mathcal{O}_H(1)) H 0 ( Q ) = H 0 ( O H ( 1 )) , hence P ( W ) ∣ U ≅ P M × U \mathbb{P}(\mathcal{W})|_U\cong\mathbb{P}^M\times U P ( W ) ∣ U ≅ P M × U carrying H × U H\times U H × U to H × U H\times U H × U and C × U C\times U C × U to C × U C\times U C × U . The composite P ( V ) ∣ U ↪ P ( W ) ∣ U ≅ P M × U \mathbb{P}(\mathcal{V})|_U\hookrightarrow\mathbb{P}(\mathcal{W})|_U\cong\mathbb{P}^M\times U P ( V ) ∣ U ↪ P ( W ) ∣ U ≅ P M × U is a flat family of smooth scrolls containing C × U C\times U C × U , i.e. a morphism Γ U : U → S C \Gamma_U:U\to\mathcal{S}_C Γ U : U → S C . By construction Γ U ( u ) = P ( V u ) \Gamma_U(u)=\mathbb{P}(\mathcal{V}_u) Γ U ( u ) = P ( V u ) embedded by ∣ O P ( V u ) ( 1 ) ∣ |\mathcal{O}_{\mathbb{P}(\mathcal{V}_u)}(1)| ∣ O P ( V u ) ( 1 ) ∣ with C = i ( C × { u } ) C=i(C\times\{u\}) C = i ( C × { u }) ; as O P ( V u ) ( 1 ) = O ( C ) \mathcal{O}_{\mathbb{P}(\mathcal{V}_u)}(1)=\mathcal{O}(C) O P ( V u ) ( 1 ) = O ( C ) , E Γ U ( u ) = π ∗ O Γ U ( u ) ( C ) = V u E_{\Gamma_U(u)}=\pi_*\mathcal{O}_{\Gamma_U(u)}(C)=\mathcal{V}_u E Γ U ( u ) = π ∗ O Γ U ( u ) ( C ) = V u , and (41) for Γ U ( u ) \Gamma_U(u) Γ U ( u ) is the fibre of (43) . The last assertion follows from (1). □
In the following lemma we show the G H G_H G H invariance of Ψ \Psi Ψ , where G H G_H G H is the set of automorphisms which induces identity on H H H .
Lemma 3.16. (1) G H G_H G H acts on S C S_C S C , and Ψ ∘ γ = Ψ \Psi\circ\gamma= \Psi Ψ ∘ γ = Ψ for every γ ∈ G H \gamma\in G_H γ ∈ G H .
(2) If Y ′ , Y ′ ′ ∈ S C Y',Y'' \in S_C Y ′ , Y ′′ ∈ S C and the classes [ V Y ′ ] , [ V Y ′ ′ ] ∈ Ext 1 ( K C , L C ) [V_{Y'}],[V_{Y''}] \in\operatorname{Ext}^{1}(K_C,L_C) [ V Y ′ ] , [ V Y ′′ ] ∈ Ext 1 ( K C , L C ) are proportional, then Y ′ ′ = γ Y ′ Y'' = \gamma Y' Y ′′ = γ Y ′ for some γ ∈ G H \gamma\in G_H γ ∈ G H .
Proof. (1) A projectivity γ \gamma γ with γ ∣ H = id \gamma|_H = \operatorname{id} γ ∣ H = id maps Y ′ ∈ S C Y' \in S_C Y ′ ∈ S C to a smooth scroll γ Y ′ ⊃ γ C = C \gamma Y' \supset\gamma C = C γ Y ′ ⊃ γ C = C , so G H G_H G H acts on S C S_C S C . Next, γ Y ~ ′ \gamma\widetilde{Y}' γ Y ′ is a K3 carpet on γ Y ′ \gamma Y' γ Y ′ , hence is the K3 carpet on γ Y ′ \gamma Y' γ Y ′ [20 Prop. 1.7] , and γ Y ~ ′ ∩ H = γ ( Y ~ ′ ∩ H ) = Y ~ ′ ∩ H \gamma\widetilde{Y}' \cap H = \gamma(\widetilde{Y}' \cap H) = \widetilde{Y}' \cap H γ Y ′ ∩ H = γ ( Y ′ ∩ H ) = Y ′ ∩ H as subschemes of H H H because γ ∣ H = id \gamma|_H = \operatorname{id} γ ∣ H = id . Thus Ψ ( γ Y ′ ) = Ψ ( Y ′ ) \Psi(\gamma Y') = \Psi(Y') Ψ ( γ Y ′ ) = Ψ ( Y ′ ) on closed points, and Ψ ∘ γ = Ψ \Psi\circ\gamma= \Psi Ψ ∘ γ = Ψ since S C S_C S C is reduced (proof of Theorem 3.9 ).
(2) Proportional classes give an isomorphism E Y ′ ≅ E Y ′ ′ E_{Y'} \cong E_{Y''} E Y ′ ≅ E Y ′′ inducing the identity on Q Q Q (and a nonzero scalar on O C O_C O C ), hence an isomorphism ( P ( E Y ′ ) , C ) ≅ ( P ( E Y ′ ′ ) , C ) (\mathbb{P}(E_{Y'}),C) \cong(\mathbb{P}(E_{Y''}),C) ( P ( E Y ′ ) , C ) ≅ ( P ( E Y ′′ ) , C ) over C C C . So Y ′ Y' Y ′ and Y ′ ′ Y'' Y ′′ are two embeddings of the same pair by the complete linear series ∣ O ( C ) ∣ |O(C)| ∣ O ( C ) ∣ which agree on C C C ; they differ by a γ ∈ P G L M + 1 ( k ) \gamma\in PGL_{M+1}(k) γ ∈ P G L M + 1 ( k ) with γ ∣ C = id \gamma|_C = \operatorname{id} γ ∣ C = id , hence γ ∣ H = id \gamma|_H = \operatorname{id} γ ∣ H = id since C C C spans H H H , i.e. γ ∈ G H \gamma\in G_H γ ∈ G H . □
In the following Proposition, we show that Γ U \Gamma_U Γ U is a local section of Ψ \Psi Ψ .
Proposition 3.17. Let Γ U : U → S C \Gamma_U: U \to S_C Γ U : U → S C be as in Theorem 3.15 (2).
(1) Ψ ∘ Γ U = id U \Psi\circ\Gamma_U = \operatorname{id}_U Ψ ∘ Γ U = id U .
(2) For every Y ′ ∈ S C Y' \in S_C Y ′ ∈ S C , Ψ ( Y ′ ) = [ t ] \Psi(Y') = [t] Ψ ( Y ′ ) = [ t ] for the t ∈ H 0 ( N C ⊗ L C ) t \in H^0(N_C \otimes L_C) t ∈ H 0 ( N C ⊗ L C ) with β ( t ) ∈ k × [ V Y ′ ] \beta(t) \in k^\times[V_{Y'}] β ( t ) ∈ k × [ V Y ′ ] , i.e., the hyperplane section Y ~ ′ ∩ H \widetilde{Y}' \cap H Y ′ ∩ H of the carpet is the ribbon whose restricted cotangent extension is that of [ V Y ′ ] = [ E Y ′ ⊗ L C ] [V_{Y'}] = [E_{Y'} \otimes L_C] [ V Y ′ ] = [ E Y ′ ⊗ L C ] .
(3) Γ U ( Ψ ( Y ′ ) ) ∈ G H ⋅ Y ′ \Gamma_U(\Psi(Y')) \in G_H \cdot Y' Γ U ( Ψ ( Y ′ )) ∈ G H ⋅ Y ′ for every Y ′ ∈ Ψ − 1 ( U ) Y' \in\Psi^{-1}(U) Y ′ ∈ Ψ − 1 ( U ) .
Proof. (1) Let u = [ t ] ∈ U u = [t] \in U u = [ t ] ∈ U and C ~ : = C ~ t ⊂ H \widetilde{C} := \widetilde{C}_t \subset H C := C t ⊂ H . It is a non-split ribbon on C C C with conormal bundle L C L_C L C (Lemma 3.8 , Lemma 3.14), and it is canonically embedded in H = P g − 1 H = \mathbb{P}^{g-1} H = P g − 1 : it is nondegenerate of degree 2 g − 2 2g-2 2 g − 2 and arithmetic genus g g g in P g − 1 \mathbb{P}^{g-1} P g − 1 , so by Riemann–Roch ω C ~ ⊗ O C ~ ( − 1 ) \omega_{\widetilde{C}} \otimes O_{\widetilde{C}}(-1) ω C ⊗ O C ( − 1 ) is a line bundle of degree 0 0 0 with a nonzero section, hence trivial (its restriction to C C C is trivial, and H 0 ( L C ) = 0 H^0(L_C) = 0 H 0 ( L C ) = 0 ). We apply Theorem A.3 (2) to C ~ ⊂ H \widetilde{C} \subset H C ⊂ H (so g C = 0 g_C = 0 g C = 0 in Set-up A.1). By Theorem 3.15 (1) the restricted cotangent sequence of C ~ \widetilde{C} C is the fibre of (42) at u u u , so V = Ω C ~ ∣ C = V u ⊗ L C V = \Omega_{\widetilde{C}}|_C = V_u \otimes L_C V = Ω C ∣ C = V u ⊗ L C ; hence the ruled surface of Set-up A.1 is P ( V ) = P ( V u ) \mathbb{P}(V) = \mathbb{P}(V_u) P ( V ) = P ( V u ) , with O P ( V ) ( C ) = O P ( V ) ( 1 ) ⊗ p C ∗ L C − 1 = O P ( V u ) ( 1 ) O_{\mathbb{P}(V)}(C) = O_{\mathbb{P}(V)}(1) \otimes p_C^*L_C^{-1} = O_{\mathbb{P}(V_u)}(1) O P ( V ) ( C ) = O P ( V ) ( 1 ) ⊗ p C ∗ L C − 1 = O P ( V u ) ( 1 ) (Lemma A.2 ) and C C C the section defined by V u → Q V_u \to Q V u → Q . By Theorem 3.15 (2), Γ U ( u ) ⊂ P M \Gamma_U(u) \subset\mathbb{P}^M Γ U ( u ) ⊂ P M is P ( V u ) \mathbb{P}(V_u) P ( V u ) embedded by the complete linear series of O P ( V u ) ( 1 ) O_{\mathbb{P}(V_u)}(1) O P ( V u ) ( 1 ) , with H H H the hyperplane such that Γ U ( u ) ∩ H = C \Gamma_U(u) \cap H = C Γ U ( u ) ∩ H = C is the given curve (Lemma 3.7 ); and Y ~ ′ \widetilde{Y}' Y ′ is the K3 carpet on Γ U ( u ) \Gamma_U(u) Γ U ( u ) . Theorem A.3 (2) gives Y ~ ′ ∩ H = C ~ \widetilde{Y}' \cap H = \widetilde{C} Y ′ ∩ H = C , hence Ψ ( Γ U ( u ) ) = [ Y ~ ′ ∩ H ] = [ C ~ ] = u \Psi(\Gamma_U(u)) = [\widetilde{Y}' \cap H] = [\widetilde{C}] = u Ψ ( Γ U ( u )) = [ Y ′ ∩ H ] = [ C ] = u .
(2), (3) Let Y ′ ∈ S C Y' \in S_C Y ′ ∈ S C . Since [ E Y ′ ] [E_{Y'}] [ E Y ′ ] is non-split, [ V Y ′ ] ≠ 0 [V_{Y'}] \ne0 [ V Y ′ ] = 0 , and since β \beta β is an isomorphism (Lemma 3.14) there is a t ≠ 0 t \ne0 t = 0 with β ( t ) = [ V Y ′ ] \beta(t) = [V_{Y'}] β ( t ) = [ V Y ′ ] . Put u : = [ t ] u := [t] u := [ t ] and choose Γ U ′ \Gamma_{U'} Γ U ′ with u ∈ U ′ u \in U' u ∈ U ′ (Theorem 3.15 (2)). Then [ V Γ U ′ ( u ) ] ∈ k × β ( t ) = k × [ V Y ′ ] [V_{\Gamma_{U'}(u)}] \in k^\times\beta(t) = k^\times[V_{Y'}] [ V Γ U ′ ( u ) ] ∈ k × β ( t ) = k × [ V Y ′ ] , so Y ′ ∈ G H ⋅ Γ U ′ ( u ) Y' \in G_H \cdot\Gamma_{U'}(u) Y ′ ∈ G H ⋅ Γ U ′ ( u ) by Lemma 3.16 (2), and by Lemma 3.16 (1) and part (1),
Ψ ( Y ′ ) = Ψ ( Γ U ′ ( u ) ) = u = [ t ] . \Psi(Y') = \Psi(\Gamma_{U'}(u)) = u = [t]. Ψ ( Y ′ ) = Ψ ( Γ U ′ ( u )) = u = [ t ] .
This is (2). If moreover Ψ ( Y ′ ) ∈ U \Psi(Y') \in U Ψ ( Y ′ ) ∈ U , then u = Ψ ( Y ′ ) ∈ U u = \Psi(Y') \in U u = Ψ ( Y ′ ) ∈ U , and [ V Γ U ( u ) ] [V_{\Gamma_U(u)}] [ V Γ U ( u ) ] also lies in k × β ( t ) k^\times\beta(t) k × β ( t ) , so Γ U ( u ) ∈ G H ⋅ Γ U ′ ( u ) = G H ⋅ Y ′ \Gamma_U(u) \in G_H \cdot\Gamma_{U'}(u) = G_H \cdot Y' Γ U ( u ) ∈ G H ⋅ Γ U ′ ( u ) = G H ⋅ Y ′ by Lemma 3.16 (2). This is (3). □
Finally in the following Proposition, we show that d Ψ d\Psi d Ψ is surjective.
Proposition 3.18. Let d Ψ : H 0 ( N Y ( − H ) ) = T [ Y ] S C → T [ C ~ ] R C = H 0 ( N C ⊗ L C ) / k S C d\Psi: H^0(N_{Y}(-H)) = T_{[Y]}S_C \to T_{[\widetilde{C}]}R_C = H^0(N_C \otimes L_C)/kS_C d Ψ : H 0 ( N Y ( − H )) = T [ Y ] S C → T [ C ] R C = H 0 ( N C ⊗ L C ) / k S C be the differential of Ψ \Psi Ψ at [ Y ] [Y] [ Y ] (see Lemma 3.11 ).
(1) d Ψ d\Psi d Ψ is surjective.
(2) ker d Ψ = Im c = T [ Y ] ( G H ⋅ Y ) \ker d\Psi= \operatorname{Im}\mathfrak{c} = T_{[Y]}(G_H \cdot Y) ker d Ψ = Im c = T [ Y ] ( G H ⋅ Y ) .
Proof. (1) Choose an open U ∋ Ψ ( Y ) U \ni\Psi(Y) U ∋ Ψ ( Y ) and Γ U \Gamma_U Γ U as in Theorem 3.15 (2). By Proposition 3.17 (1), Ψ ∘ Γ U = id U \Psi\circ\Gamma_U = \operatorname{id}_U Ψ ∘ Γ U = id U ; differentiating at Ψ ( Y ) \Psi(Y) Ψ ( Y ) ,
( d Ψ ) Γ U ( Ψ ( Y ) ) ∘ ( d Γ U ) Ψ ( Y ) = id : T Ψ ( Y ) R C → T Ψ ( Y ) R C , (d\Psi)_{\Gamma_U(\Psi(Y))} \circ(d\Gamma_U)_{\Psi(Y)} = \operatorname{id}: T_{\Psi(Y)}\mathcal{R}_C \to T_{\Psi(Y)}\mathcal{R}_C, ( d Ψ ) Γ U ( Ψ ( Y )) ∘ ( d Γ U ) Ψ ( Y ) = id : T Ψ ( Y ) R C → T Ψ ( Y ) R C ,
so d Ψ d\Psi d Ψ is surjective at Γ U ( Ψ ( Y ) ) = γ Y \Gamma_U(\Psi(Y)) = \gamma_Y Γ U ( Ψ ( Y )) = γ Y , γ ∈ G H \gamma\in G_H γ ∈ G H (Proposition 3.17 (3)). By Lemma 3.16 (1), d Ψ Y = d Ψ γ Y ∘ d γ Y d\Psi_Y = d\Psi_{\gamma_Y} \circ d\gamma_Y d Ψ Y = d Ψ γ Y ∘ d γ Y , and d γ Y : T Y S C → T γ Y S C d\gamma_Y: T_Y S_C \to T_{\gamma_Y}S_C d γ Y : T Y S C → T γ Y S C is an isomorphism since γ \gamma γ is an automorphism of S C S_C S C ; hence d Ψ Y d\Psi_Y d Ψ Y is surjective.
(2) By Lemma 3.16 (1), the two morphisms G H × S C → R C G_H \times S_C \to\mathcal{R}_C G H × S C → R C , ( γ , Y ′ ) ↦ Ψ ( γ Y ′ ) (\gamma,Y') \mapsto\Psi(\gamma Y') ( γ , Y ′ ) ↦ Ψ ( γ Y ′ ) and ( γ , Y ′ ) ↦ Ψ ( Y ′ ) (\gamma,Y') \mapsto\Psi(Y') ( γ , Y ′ ) ↦ Ψ ( Y ′ ) , agree on closed points, and hence agree on G H × S C G_H \times S_C G H × S C since it is reduced. Differentiating at ( 1 , Y ) (1,Y) ( 1 , Y ) along G H G_H G H gives d Ψ ∘ c = 0 d\Psi\circ c = 0 d Ψ ∘ c = 0 , i.e. Im c ⊂ ker d Ψ \operatorname{Im} c \subset\ker d\Psi Im c ⊂ ker d Ψ . By (1) and Lemma 3.7 ,
dim ker d Ψ = h 0 ( N Y ( − H ) ) − dim R C = ( M + 1 ) + ( 2 b − e − 2 ) − ( 2 b − e − 2 ) = M + 1 = dim Im c , \dim\ker d\Psi= h^0(N_Y(-H)) - \dim\mathcal{R}_C = (M+1) + (2b-e-2) - (2b-e-2) = M+1 = \dim\operatorname{Im} c, dim ker d Ψ = h 0 ( N Y ( − H )) − dim R C = ( M + 1 ) + ( 2 b − e − 2 ) − ( 2 b − e − 2 ) = M + 1 = dim Im c ,
so ker d Ψ = Im c \ker d\Psi= \operatorname{Im} c ker d Ψ = Im c . □
Finally, we conclude that the map a a a is surjective.
Proposition 3.19. The map δ ˉ − 1 ∘ a = d Ψ : H 0 ( N Y ( − H ) ) → H 0 ( N C ⊗ K Y ∣ C ) / k s C \bar{\delta}^{-1} \circ a = d\Psi: H^0(N_Y(-H)) \to H^0(N_C \otimes K_Y|_C)/k s_C δ ˉ − 1 ∘ a = d Ψ : H 0 ( N Y ( − H )) → H 0 ( N C ⊗ K Y ∣ C ) / k s C is surjective. Consequently a a a and F F F are surjective, and ker a = Im c \ker a = \operatorname{Im} c ker a = Im c .
Proof. By Proposition 3.18 , d Ψ d\Psi d Ψ is surjective with kernel Im c \operatorname{Im} c Im c . By Proposition 3.13, a = δ ˉ ∘ d Ψ a = \bar{\delta} \circ d\Psi a = δ ˉ ∘ d Ψ is surjective with kernel Im c \operatorname{Im} c Im c . Surjectivity of F F F follows from diagram (35) and the sequence (28), as explained in the proof of Theorem 3.2. □
Prime K3 surfaces of genus g ≥ 6 g \ge6 g ≥ 6 and Fano threefolds of Picard rank one
In this section we give sharp bounds to α ( X ) \alpha(X) α ( X ) , where X X X is a prime K3 surface of genus g ≥ 6 g \ge6 g ≥ 6 .
Set-up
For a smooth K3 surface X ⊂ P M X \subset\mathbb{P}^M X ⊂ P M we write N X : = N X / P M N_X := N_{X/\mathbb{P}^M} N X := N X / P M and, as in §3.1, α ( X ) : = h 0 ( N X ( − 1 ) ) − ( M + 1 ) \alpha(X) := h^0(N_X(-1)) - (M+1) α ( X ) := h 0 ( N X ( − 1 )) − ( M + 1 ) . Recall [Zak91, Lvo92] that if α ( X ) ≤ 0 \alpha(X) \le0 α ( X ) ≤ 0 then X X X is not extendable: every V ⊂ P M + 1 V \subset\mathbb{P}^{M+1} V ⊂ P M + 1 having X X X as a hyperplane section is a cone over X X X .
Fix g ≥ 3 g \ge3 g ≥ 3 . Let H g \mathcal{H}_g H g be the locus, in the Hilbert scheme of P g \mathbb{P}^g P g , of smooth K3 surfaces X ⊂ P g X \subset\mathbb{P}^g X ⊂ P g embedded by the complete linear series of a very ample line bundle H H H with H 2 = 2 g − 2 H^2 = 2g - 2 H 2 = 2 g − 2 and H H H primitive in Pic X \operatorname{Pic} X Pic X . We call them prime K3 surfaces of genus g g g . Since the moduli space of primitively polarized K3 surfaces of genus g g g is irreducible [21 Ch. 6] , H g \mathcal{H}_g H g is irreducible. It is smooth of dimension 18 + ( g + 1 ) 2 18 + (g+1)^2 18 + ( g + 1 ) 2 , because for X ∈ H g X \in\mathcal{H}_g X ∈ H g the Euler sequence gives h 0 ( T P g ∣ X ) = ( g + 1 ) 2 − 1 h^0(T_{\mathbb{P}^g}|_X) = (g+1)^2 - 1 h 0 ( T P g ∣ X ) = ( g + 1 ) 2 − 1 and H 1 ( T P g ∣ X ) = H 2 ( O X ) = k H^1(T_{\mathbb{P}^g}|_X) = H^2(\mathcal{O}_X) = k H 1 ( T P g ∣ X ) = H 2 ( O X ) = k , the map H 1 ( T X ) → H 2 ( O X ) H^1(T_X) \to H^2(\mathcal{O}_X) H 1 ( T X ) → H 2 ( O X ) , cup product with c 1 ( H ) c_1(H) c 1 ( H ) , is surjective, and H 2 ( T X ) = 0 H^2(T_X) = 0 H 2 ( T X ) = 0 , so that h 1 ( N X ) = 0 h^1(N_X) = 0 h 1 ( N X ) = 0 and h 0 ( N X ) = ( g + 1 ) 2 − 1 + 19 h^0(N_X) = (g+1)^2 - 1 + 19 h 0 ( N X ) = ( g + 1 ) 2 − 1 + 19 . A point of H g \mathcal{H}_g H g is called general if it lies in a suitable dense open subset.
Let e ∈ { 0 , 1 } e \in\{0,1\} e ∈ { 0 , 1 } and b ≥ e + 2 b \ge e+2 b ≥ e + 2 , and let Y ~ ⊂ P g \widetilde{Y} \subset\mathbb{P}^g Y ⊂ P g be the K3 carpet on Y = F e Y = F_e Y = F e embedded by ∣ H ∣ |H| ∣ H ∣ , H = C 0 + b f H = C_0 + bf H = C 0 + b f , so that g = 2 b − e + 1 g = 2b-e+1 g = 2 b − e + 1 (this is the set-up of §3.1, except that we allow b = e + 2 b=e+2 b = e + 2 ). By Theorem 2.4 (see also [20 Theorems 3.5 and 4.4] ), [ Y ~ ] [\widetilde{Y}] [ Y ] lies in the closure of H g \mathcal{H}_g H g . Since Y ~ \widetilde{Y} Y and the smooth K3 surfaces are local complete intersections, the normal sheaf of a flat family in P g \mathbb{P}^g P g degenerating a general X ∈ H g X \in\mathcal{H}_g X ∈ H g to Y ~ \widetilde{Y} Y is locally free and commutes with base change; by upper semicontinuity, for X X X general in H g \mathcal{H}_g H g and every k ≥ 0 k \ge0 k ≥ 0 ,
h 0 ( N X ( − k ) ) ≤ h 0 ( N Y ~ ( − k H ~ ) ) , in particular α ( X ) ≤ α ( Y ~ ) . (44) h^0(N_X(-k)) \le h^0(N_{\widetilde{Y}}(-k\widetilde{H})), \qquad\text{in particular} \qquad\alpha(X) \le\alpha(\widetilde{Y}). \tag*{(44)} h 0 ( N X ( − k )) ≤ h 0 ( N Y ( − k H )) , in particular α ( X ) ≤ α ( Y ) . (44)
The invariant α \alpha α for $g \ge 6
Theorem 4.1. Let X ⊂ P g X \subset\mathbb{P}^g X ⊂ P g be a general prime K3 surface of genus g g g .
(1) If g = 12 g = 12 g = 12 , then α ( X ) ≤ 1 \alpha(X) \le1 α ( X ) ≤ 1 .
(2) If g = 10 g = 10 g = 10 , then α ( X ) ≤ 3 \alpha(X) \le3 α ( X ) ≤ 3 .
(3) If g = 9 g = 9 g = 9 , then α ( X ) ≤ 4 \alpha(X) \le4 α ( X ) ≤ 4 .
(4) If g = 8 g = 8 g = 8 , then α ( X ) ≤ 6 \alpha(X) \le6 α ( X ) ≤ 6 .
(5) If g = 7 g = 7 g = 7 , then α ( X ) ≤ 8 \alpha(X) \le8 α ( X ) ≤ 8 .
(6) If g = 6 g = 6 g = 6 , then α ( X ) ≤ 10 \alpha(X) \le10 α ( X ) ≤ 10 .
(7) If g = 11 g = 11 g = 11 or g ≥ 13 g \ge13 g ≥ 13 , then α ( X ) ≤ 0 \alpha(X) \le0 α ( X ) ≤ 0 ; consequently X X X is not extendable.
Proof. We can degenerate such a K3 surface to a K3 carpet Y ~ ⊂ P g \widetilde{Y} \subset\mathbb{P}^{g} Y ⊂ P g , extending Y ⊂ P g Y \subset\mathbb{P}^{g} Y ⊂ P g , where Y = F e Y = F_e Y = F e is embedded by the complete linear series of C 0 + b f C_0 + bf C 0 + b f with e = 0 e = 0 e = 0 or e = 1 e = 1 e = 1 depending on whether g g g is odd or even respectively, and g = 2 b − e + 1 g = 2b - e + 1 g = 2 b − e + 1 . By (44) it is enough to find an upper bound for α ( Y ~ ) \alpha(\widetilde{Y}) α ( Y ) . In cases (1)–(5) the pair ( e , b ) (e,b) ( e , b ) is ( 1 , 6 ) (1,6) ( 1 , 6 ) , ( 1 , 5 ) (1,5) ( 1 , 5 ) , ( 0 , 4 ) (0,4) ( 0 , 4 ) , ( 1 , 4 ) (1,4) ( 1 , 4 ) , ( 0 , 3 ) (0,3) ( 0 , 3 ) respectively, and in case (7) it is ( 0 , b ) (0,b) ( 0 , b ) with b ≥ 5 b \ge5 b ≥ 5 or ( 1 , b ) (1,b) ( 1 , b ) with b ≥ 7 b \ge7 b ≥ 7 . In all these cases b ≥ e + 3 b \ge e + 3 b ≥ e + 3 , so by Theorem 3.2
α ( Y ~ ) = h 0 ( O Y ( 3 C 0 + ( 2 e + 4 − b ) f ) ) = ∑ i = 0 3 h 0 ( O P 1 ( 2 e + 4 − b − i e ) ) , \alpha(\widetilde{Y}) = h^{0}\bigl(\mathcal{O}_{Y}(3C_{0} + (2e + 4 - b)f)\bigr) = \sum_{i=0}^{3} h^{0}\bigl(\mathcal{O}_{\mathbb{P}^{1}}(2e + 4 - b - ie)\bigr), α ( Y ) = h 0 ( O Y ( 3 C 0 + ( 2 e + 4 − b ) f ) ) = i = 0 ∑ 3 h 0 ( O P 1 ( 2 e + 4 − b − i e ) ) ,
which equals 1 , 3 , 4 , 6 , 8 1, 3, 4, 6, 8 1 , 3 , 4 , 6 , 8 in cases (1)–(5) and 0 0 0 in case (7). In case (6), ( e , b ) = ( 1 , 3 ) (e,b) = (1,3) ( e , b ) = ( 1 , 3 ) and α ( Y ~ ) = 10 \alpha(\widetilde{Y}) = 10 α ( Y ) = 10 by Proposition 4.2 below. The last assertion in (7) is the theorem of Zak and L’vovsky. □
Proposition 4.2 (The case g = 6 g = 6 g = 6 ). Let Y = F 1 ⊂ P 6 Y = F_{1} \subset\mathbb{P}^{6} Y = F 1 ⊂ P 6 be embedded by ∣ C 0 + 3 f ∣ |C_{0} + 3f| ∣ C 0 + 3 f ∣ and let Y ~ \widetilde{Y} Y be the K3 carpet on Y Y Y . Then α ( Y ~ ) = h 0 ( O Y ( − H − 2 K Y ) ) = h 0 ( O Y ( 3 C 0 + 3 f ) ) = 10 \alpha(\widetilde{Y}) = h^{0}\bigl(\mathcal{O}_{Y}(-H - 2K_{Y})\bigr) = h^{0}\bigl(\mathcal{O}_{Y}(3C_{0} + 3f)\bigr) = 10 α ( Y ) = h 0 ( O Y ( − H − 2 K Y ) ) = h 0 ( O Y ( 3 C 0 + 3 f ) ) = 10 .
Proof. Here b = e + 2 b = e + 2 b = e + 2 , M = 6 M = 6 M = 6 , and O Y ( − H − K Y ) = O Y ( C 0 ) \mathcal{O}_{Y}(-H - K_{Y}) = \mathcal{O}_{Y}(C_{0}) O Y ( − H − K Y ) = O Y ( C 0 ) , with h 0 ( O Y ( C 0 ) ) = 1 h^{0}\bigl(\mathcal{O}_{Y}(C_{0})\bigr) = 1 h 0 ( O Y ( C 0 ) ) = 1 and h 1 ( O Y ( C 0 ) ) = h 2 ( O Y ( C 0 ) ) = 0 h^{1}\bigl(\mathcal{O}_{Y}(C_{0})\bigr) = h^{2}\bigl(\mathcal{O}_{Y}(C_{0})\bigr) = 0 h 1 ( O Y ( C 0 ) ) = h 2 ( O Y ( C 0 ) ) = 0 . We go through the proof of Theorem 3.2, where the assumption b ≥ e + 3 b \ge e + 3 b ≥ e + 3 was used only to have H 0 ( − H − K Y ) = 0 H^{0}(-H - K_{Y}) = 0 H 0 ( − H − K Y ) = 0 in (8) and (19) ; everything else, in particular (14) , (15) , (23) –(26) and (28) , holds for b ≥ e + 1 b \ge e + 1 b ≥ e + 1 . From (8) and (15) we get an exact sequence
0 → H 0 ( N Y ~ ( − H + K Y ) ) → H 0 ( O Y ( C 0 ) ) = k → δ 1 H 1 ( N Y ( − H + K Y ) ) = k 3 0 \to H^{0}\bigl(N_{\widetilde{Y}}(-H + K_{Y})\bigr) \to H^{0}\bigl(\mathcal{O}_{Y}(C_{0})\bigr) = k \xrightarrow{\delta_{1}} H^{1}\bigl(N_{Y}(-H + K_{Y})\bigr) = k^{3} 0 → H 0 ( N Y ( − H + K Y ) ) → H 0 ( O Y ( C 0 ) ) = k δ 1 H 1 ( N Y ( − H + K Y ) ) = k 3
→ j H 1 ( N Y ~ ( − H + K Y ) ) → 0 \xrightarrow{j} H^{1}\bigl(N_{\widetilde{Y}}(-H + K_{Y})\bigr) \to0 j H 1 ( N Y ( − H + K Y ) ) → 0
and H 2 ( N Y ~ ( − H + K Y ) ) = 0 H^{2}\bigl(N_{\widetilde{Y}}(-H + K_{Y})\bigr) = 0 H 2 ( N Y ( − H + K Y ) ) = 0 ; thus h 0 ( N Y ~ ( − H + K Y ) ) = 1 − rank δ 1 h^{0}\bigl(N_{\widetilde{Y}}(-H + K_{Y})\bigr) = 1 - \operatorname{rank}\delta_{1} h 0 ( N Y ( − H + K Y ) ) = 1 − rank δ 1 , h 1 ( N Y ~ ( − H + K Y ) ) = 3 − rank δ 1 h^{1}\bigl(N_{\widetilde{Y}}(-H + K_{Y})\bigr) = 3 - \operatorname{rank}\delta_{1} h 1 ( N Y ( − H + K Y ) ) = 3 − rank δ 1 , and j j j is surjective. From (19) and (25) , h 0 ( H ( − H ) ) = h 0 ( N Y ( − H ) ) − 1 = ( M + 1 ) + h 1 ( T Y ( − H ) ) − 1 = 7 + 3 − 1 = 9 h^{0}\bigl(\mathcal{H}(-H)\bigr) = h^{0}\bigl(N_{Y}(-H)\bigr) - 1 = (M + 1) + h^{1}\bigl(T_{Y}(-H)\bigr) - 1 = 7 + 3 - 1 = 9 h 0 ( H ( − H ) ) = h 0 ( N Y ( − H ) ) − 1 = ( M + 1 ) + h 1 ( T Y ( − H ) ) − 1 = 7 + 3 − 1 = 9 , and from (28) , h 0 ( N Y ~ ( − H ) ) = 9 + h 0 ( O Y ( 3 C 0 + 3 f ) ) = 19 h^{0}\bigl(N_{\widetilde{Y}}(-H)\bigr) = 9 + h^{0}\bigl(\mathcal{O}_{Y}(3C_{0} + 3f)\bigr) = 19 h 0 ( N Y ( − H ) ) = 9 + h 0 ( O Y ( 3 C 0 + 3 f ) ) = 19 and H 1 ( N Y ~ ( − H ) ) = H 1 ( O Y ( 3 C 0 + 3 f ) ) = 0 H^{1}\bigl(N_{\widetilde{Y}}(-H)\bigr) = H^{1}\bigl(\mathcal{O}_{Y}(3C_{0} + 3f)\bigr) = 0 H 1 ( N Y ( − H ) ) = H 1 ( O Y ( 3 C 0 + 3 f ) ) = 0 . Sequence (7) therefore gives
h 0 ( N Y ~ ( − H ~ ) ) = h 0 ( N Y ~ ( − H + K Y ) ) + h 0 ( N Y ~ ( − H ) ) − rank F = 20 − rank δ 1 − rank F , h^{0}\bigl(N_{\widetilde{Y}}(-\widetilde{H})\bigr) = h^{0}\bigl(N_{\widetilde{Y}}(-H + K_{Y})\bigr) + h^{0}\bigl(N_{\widetilde{Y}}(-H)\bigr) - \operatorname{rank} F = 20 - \operatorname{rank}\delta_{1} - \operatorname{rank} F, h 0 ( N Y ( − H ) ) = h 0 ( N Y ( − H + K Y ) ) + h 0 ( N Y ( − H ) ) − rank F = 20 − rank δ 1 − rank F ,
where F : H 0 ( N Y ~ ( − H ) ) → H 1 ( N Y ~ ( − H + K Y ) ) F \colon H^{0}\bigl(N_{\widetilde{Y}}(-H)\bigr) \to H^{1}\bigl(N_{\widetilde{Y}}(-H + K_{Y})\bigr) F : H 0 ( N Y ( − H ) ) → H 1 ( N Y ( − H + K Y ) ) is the connecting map, and it remains to show that F F F is surjective. Since then, rank F = 3 − rank δ 1 \operatorname{rank} F = 3 - \operatorname{rank}\delta_{1} rank F = 3 − rank δ 1 and α ( Y ~ ) = 17 − 7 = 10 \alpha(\widetilde{Y}) = 17 - 7 = 10 α ( Y ) = 17 − 7 = 10 .
The proof of Proposition 3.13 shows, without using b ≥ e + 3 b \ge e + 3 b ≥ e + 3 , that F ( ι ( x ) ) = j ( δ ( t ) ) F(\iota(x)) = j(\delta(t)) F ( ι ( x )) = j ( δ ( t )) for every x ∈ H 0 ( N Y ( − H ) ) x \in H^{0}\bigl(N_{Y}(-H)\bigr) x ∈ H 0 ( N Y ( − H ) ) , where ι : H 0 ( N Y ( − H ) ) → H 0 ( N Y ~ ( − H ) ) \iota\colon H^{0}\bigl(N_{Y}(-H)\bigr) \to H^{0}\bigl(N_{\widetilde{Y}}(-H)\bigr) ι : H 0 ( N Y ( − H ) ) → H 0 ( N Y ( − H ) ) is induced by restriction of homomorphisms from I Y I_{Y} I Y to I Y ~ I_{\widetilde{Y}} I Y , d Ψ ( x ) = [ t ] d\Psi(x) = [t] d Ψ ( x ) = [ t ] , and δ \delta δ is the map of Lemma 3.6. Now δ \delta δ is surjective (Lemma 3.6), d Ψ d\Psi d Ψ is surjective (Proposition 3.18 ; its proof, and those of the results of §§3.3.2–3.3.3 on which it relies, only use b ≥ e + 1 b \ge e + 1 b ≥ e + 1 ), and j j j is surjective; hence the image of F F F is all of H 1 ( N Y ~ ( − H + K Y ) ) H^{1}\bigl(N_{\widetilde{Y}}(-H + K_{Y})\bigr) H 1 ( N Y ( − H + K Y ) ) . □
Sections of N(-2)
In this subsection we give upper bounds to the dimension of the space H 0 ( N Y ~ ( − 2 ) ) H^{0}\bigl(N_{\widetilde{Y}}(-2)\bigr) H 0 ( N Y ( − 2 ) ) where Y ~ ⊂ P g \widetilde{Y} \subset\mathbb{P}^{g} Y ⊂ P g is a K3 carpet on a rational normal scroll Y ⊂ P g Y \subset\mathbb{P}^{g} Y ⊂ P g . This consequently gives upper bounds to the dimension of the space H 0 ( N X ( − 2 ) ) H^{0}\bigl(N_{X}(-2)\bigr) H 0 ( N X ( − 2 ) ) , where X ⊂ P g X \subset\mathbb{P}^{g} X ⊂ P g is a prime K3 surface of genus g g g .
Proposition 4.3. Let Y = F e ⊂ P M Y = F_{e} \subset\mathbb{P}^{M} Y = F e ⊂ P M be embedded by ∣ H ∣ |H| ∣ H ∣ , H = C 0 + b f H = C_{0} + bf H = C 0 + b f with b ≥ e + 2 b \ge e + 2 b ≥ e + 2 , and let Y ~ \widetilde{Y} Y be the K3 carpet on Y Y Y . Then
h 0 ( N Y ~ ( − 2 H ~ ) ) ≤ h 0 ( O Y ( − 2 H − 2 K Y ) ) = h 0 ( O Y ( 2 C 0 + ( 2 e + 4 − 2 b ) f ) ) . h^{0}\bigl(N_{\widetilde{Y}}(-2\widetilde{H})\bigr) \le h^{0}\bigl(\mathcal{O}_{Y}(-2H - 2K_{Y})\bigr) = h^{0}\bigl(\mathcal{O}_{Y}(2C_{0} + (2e + 4 - 2b)f)\bigr). h 0 ( N Y ( − 2 H ) ) ≤ h 0 ( O Y ( − 2 H − 2 K Y ) ) = h 0 ( O Y ( 2 C 0 + ( 2 e + 4 − 2 b ) f ) ) . In particular h 0 ( N Y ~ ( − 2 H ~ ) ) = 0 h^{0}\bigl(N_{\widetilde{Y}}(-2\widetilde{H})\bigr) = 0 h 0 ( N Y ( − 2 H ) ) = 0 if b ≥ e + 3 b \ge e + 3 b ≥ e + 3 , and h 0 ( N Y ~ ( − 2 H ~ ) ) ≤ 1 h^{0}\bigl(N_{\widetilde{Y}}(-2\widetilde{H})\bigr) \le1 h 0 ( N Y ( − 2 H ) ) ≤ 1 if ( e , b ) = ( 1 , 3 ) (e,b) = (1,3) ( e , b ) = ( 1 , 3 ) .
Proof. Tensoring (2) by O Y ~ ( − 2 H ~ ) \mathcal{O}_{\widetilde{Y}}(-2\widetilde{H}) O Y ( − 2 H ) gives h 0 ( N Y ~ ( − 2 H ~ ) ) ≤ h 0 ( N Y ~ ( − 2 H + K Y ) ) + h 0 ( N Y ~ ( − 2 H ) ) h^0(\mathcal{N}_{\widetilde{Y}}(-2\widetilde{H})) \leq h^0(\mathcal{N}_{\widetilde{Y}}(-2H+K_Y))+h^0(\mathcal{N}_{\widetilde{Y}}(-2H)) h 0 ( N Y ( − 2 H )) ≤ h 0 ( N Y ( − 2 H + K Y )) + h 0 ( N Y ( − 2 H )) . We use throughout that H 0 ( O Y ( − H + K Y ) ) = 0 H^0(\mathcal{O}_Y(-H+K_Y))=0 H 0 ( O Y ( − H + K Y )) = 0 , H 0 ( O Y ( − 2 H − K Y ) ) = 0 H^0(\mathcal{O}_Y(-2H-K_Y))=0 H 0 ( O Y ( − 2 H − K Y )) = 0 (as − 2 H − K Y = ( e + 2 − 2 b ) f -2H-K_Y=(e+2-2b)f − 2 H − K Y = ( e + 2 − 2 b ) f and 2 b > e + 2 2b>e+2 2 b > e + 2 ), and, by Kodaira vanishing, H 1 ( O Y ( − 2 H ) ) = H 1 ( O Y ( − 2 H + K Y ) ) = 0 H^1(\mathcal{O}_Y(-2H))=H^1(\mathcal{O}_Y(-2H+K_Y))=0 H 1 ( O Y ( − 2 H )) = H 1 ( O Y ( − 2 H + K Y )) = 0 and H i ( O Y ( − H ) ) = 0 H^i(\mathcal{O}_Y(-H))=0 H i ( O Y ( − H )) = 0 for all i i i (see the derivation of (21) ).
Step 1: H 0 ( N Y ~ ( − 2 H + K Y ) ) = 0 H^0(\mathcal{N}_{\widetilde{Y}}(-2H+K_Y))=0 H 0 ( N Y ( − 2 H + K Y )) = 0 . By (3) tensored with − 2 H + K Y -2H+K_Y − 2 H + K Y , H 0 ( N Y ~ ( − 2 H + K Y ) ) = H 0 ( H ( − 2 H + K Y ) ) H^0(\mathcal{N}_{\widetilde{Y}}(-2H+K_Y))=H^0(\mathcal{H}(-2H+K_Y)) H 0 ( N Y ( − 2 H + K Y )) = H 0 ( H ( − 2 H + K Y )) , and by (4) tensored with − 2 H + K Y -2H+K_Y − 2 H + K Y the latter is a quotient of H 0 ( N Y ( − 2 H + K Y ) ) H^0(\mathcal{N}_Y(-2H+K_Y)) H 0 ( N Y ( − 2 H + K Y )) . By (6) tensored with − 2 H + K Y -2H+K_Y − 2 H + K Y , H 0 ( T P M ∣ Y ( − 2 H + K Y ) ) = 0 H^0(T_{\mathbb{P}^M|Y}(-2H+K_Y))=0 H 0 ( T P M ∣ Y ( − 2 H + K Y )) = 0 , so by (5) H 0 ( N Y ( − 2 H + K Y ) ) H^0(\mathcal{N}_Y(-2H+K_Y)) H 0 ( N Y ( − 2 H + K Y )) injects into H 1 ( T Y ( − 2 H + K Y ) ) H^1(T_Y(-2H+K_Y)) H 1 ( T Y ( − 2 H + K Y )) . Finally H 1 ( T Y ( − 2 H + K Y ) ) = 0 H^1(T_Y(-2H+K_Y))=0 H 1 ( T Y ( − 2 H + K Y )) = 0 : in
0 ⟶ T Y / P 1 ( − 2 H + K Y ) ⟶ T Y ( − 2 H + K Y ) ⟶ p ∗ T P 1 ( − 2 H + K Y ) ⟶ 0 0 \longrightarrow T_{Y/\mathbb{P}^1}(-2H+K_Y) \longrightarrow T_Y(-2H+K_Y) \longrightarrow p^*T_{\mathbb{P}^1}(-2H+K_Y) \longrightarrow0 0 ⟶ T Y / P 1 ( − 2 H + K Y ) ⟶ T Y ( − 2 H + K Y ) ⟶ p ∗ T P 1 ( − 2 H + K Y ) ⟶ 0
we have T Y / P 1 ( − 2 H + K Y ) = O Y ( − 2 C 0 − ( 2 b + 2 ) f ) T_{Y/\mathbb{P}^1}(-2H+K_Y)=\mathcal{O}_Y(-2C_0-(2b+2)f) T Y / P 1 ( − 2 H + K Y ) = O Y ( − 2 C 0 − ( 2 b + 2 ) f ) and p ∗ T P 1 ( − 2 H + K Y ) = O Y ( − 4 C 0 − ( 2 b + e ) f ) p^*T_{\mathbb{P}^1}(-2H+K_Y)=\mathcal{O}_Y(-4C_0-(2b+e)f) p ∗ T P 1 ( − 2 H + K Y ) = O Y ( − 4 C 0 − ( 2 b + e ) f ) ; both have vanishing p ∗ p_* p ∗ , while by relative duality R 1 p ∗ O Y ( − 2 C 0 ) = O P 1 ( e ) R^1p_*\mathcal{O}_Y(-2C_0)=\mathcal{O}_{\mathbb{P}^1}(e) R 1 p ∗ O Y ( − 2 C 0 ) = O P 1 ( e ) and R 1 p ∗ O Y ( − 4 C 0 ) = O P 1 ( e ) ⊕ O P 1 ( 2 e ) ⊕ O P 1 ( 3 e ) R^1p_*\mathcal{O}_Y(-4C_0)=\mathcal{O}_{\mathbb{P}^1}(e)\oplus\mathcal{O}_{\mathbb{P}^1}(2e)\oplus\mathcal{O}_{\mathbb{P}^1}(3e) R 1 p ∗ O Y ( − 4 C 0 ) = O P 1 ( e ) ⊕ O P 1 ( 2 e ) ⊕ O P 1 ( 3 e ) . Hence
H 1 ( T Y / P 1 ( − 2 H + K Y ) ) = H 0 ( O P 1 ( e − 2 b − 2 ) ) = 0 , H^1(T_{Y/\mathbb{P}^1}(-2H+K_Y))=H^0(\mathcal{O}_{\mathbb{P}^1}(e-2b-2))=0, H 1 ( T Y / P 1 ( − 2 H + K Y )) = H 0 ( O P 1 ( e − 2 b − 2 )) = 0 ,
H 1 ( p ∗ T P 1 ( − 2 H + K Y ) ) = H 0 ( O P 1 ( − 2 b ) ⊕ O P 1 ( e − 2 b ) ⊕ O P 1 ( 2 e − 2 b ) ) = 0. H^1(p^*T_{\mathbb{P}^1}(-2H+K_Y))=H^0(\mathcal{O}_{\mathbb{P}^1}(-2b)\oplus\mathcal{O}_{\mathbb{P}^1}(e-2b)\oplus\mathcal{O}_{\mathbb{P}^1}(2e-2b))=0. H 1 ( p ∗ T P 1 ( − 2 H + K Y )) = H 0 ( O P 1 ( − 2 b ) ⊕ O P 1 ( e − 2 b ) ⊕ O P 1 ( 2 e − 2 b )) = 0.
Step 2: H 0 ( H ( − 2 H ) ) = 0 H^0(\mathcal{H}(-2H))=0 H 0 ( H ( − 2 H )) = 0 . Granting this, (3) tensored with − 2 H -2H − 2 H gives h 0 ( N Y ~ ( − 2 H ) ) ≤ h 0 ( O Y ( − 2 H − 2 K Y ) ) h^0(\mathcal{N}_{\widetilde{Y}}(-2H))\leq h^0(\mathcal{O}_Y(-2H-2K_Y)) h 0 ( N Y ( − 2 H )) ≤ h 0 ( O Y ( − 2 H − 2 K Y )) , which together with Step 1 proves the inequality. By (4) tensored with − 2 H -2H − 2 H we have an exact sequence
0 ⟶ H 0 ( N Y ( − 2 H ) ) ⟶ H 0 ( H ( − 2 H ) ) ⟶ H 1 ( O Y ( − 2 H − K Y ) ) → s ∨ H 1 ( N Y ( − 2 H ) ) , 0\longrightarrow H^0(\mathcal{N}_Y(-2H))\longrightarrow H^0(\mathcal{H}(-2H))\longrightarrow H^1(\mathcal{O}_Y(-2H-K_Y))\xrightarrow{s^\vee} H^1(\mathcal{N}_Y(-2H)), 0 ⟶ H 0 ( N Y ( − 2 H )) ⟶ H 0 ( H ( − 2 H )) ⟶ H 1 ( O Y ( − 2 H − K Y )) s ∨ H 1 ( N Y ( − 2 H )) ,
so it suffices to show that (a) H 0 ( N Y ( − 2 H ) ) = 0 H^0(\mathcal{N}_Y(-2H))=0 H 0 ( N Y ( − 2 H )) = 0 and that (b) s ∨ s^\vee s ∨ is injective.
(a) Twisting the sequence (5) by − 2 H -2H − 2 H and taking cohomology, we have
0 ⟶ H 0 ( T Y ( − 2 H ) ) ⟶ H 0 ( T P M ∣ Y ( − 2 H ) ) ⟶ H 0 ( N Y ( − 2 H ) ) ⟶ H 1 ( T Y ( − 2 H ) ) → s H 1 ( T P M ∣ Y ( − 2 H ) ) . 0\longrightarrow H^0(T_Y(-2H))\longrightarrow H^0(T_{\mathbb{P}^M|Y}(-2H))\longrightarrow H^0(\mathcal{N}_Y(-2H))\longrightarrow H^1(T_Y(-2H))\xrightarrow{s} H^1(T_{\mathbb{P}^M|Y}(-2H)). 0 ⟶ H 0 ( T Y ( − 2 H )) ⟶ H 0 ( T P M ∣ Y ( − 2 H )) ⟶ H 0 ( N Y ( − 2 H )) ⟶ H 1 ( T Y ( − 2 H )) s H 1 ( T P M ∣ Y ( − 2 H )) .
Turning to the Euler sequence (6) , twisting by − 2 H -2H − 2 H and taking cohomology, we get
0 ⟶ H 0 ( O Y ( − H ) ⊕ M + 1 ) = ( 0 ) ⟶ H 0 ( T P M ∣ Y ( − 2 H ) ) ⟶ H 1 ( O Y ( − 2 H ) ) = ( 0 ) . 0\longrightarrow H^0(\mathcal{O}_Y(-H)^{\oplus M+1})=(0)\longrightarrow H^0(T_{\mathbb{P}^M|Y}(-2H))\longrightarrow H^1(\mathcal{O}_Y(-2H))=(0). 0 ⟶ H 0 ( O Y ( − H ) ⊕ M + 1 ) = ( 0 ) ⟶ H 0 ( T P M ∣ Y ( − 2 H )) ⟶ H 1 ( O Y ( − 2 H )) = ( 0 ) .
Hence H 0 ( T P M ∣ Y ( − 2 H ) ) = 0 H^0(T_{\mathbb{P}^M|Y}(-2H))=0 H 0 ( T P M ∣ Y ( − 2 H )) = 0 . So it is enough to show that the map H 1 ( T Y ( − 2 H ) ) → s H 1 ( T P M ∣ Y ( − 2 H ) ) H^1(T_Y(-2H))\xrightarrow{s}H^1(T_{\mathbb{P}^M|Y}(-2H)) H 1 ( T Y ( − 2 H )) s H 1 ( T P M ∣ Y ( − 2 H )) is injective.
First consider the two exact sequences, the first one is the relative Euler sequence on Y = P ( O P 1 ⊕ O P 1 ( − e ) ) Y=\mathbb{P}(\mathcal{O}_{\mathbb{P}^1}\oplus\mathcal{O}_{\mathbb{P}^1}(-e)) Y = P ( O P 1 ⊕ O P 1 ( − e )) and the second one is the Euler sequence of P M \mathbb{P}^M P M pulled back to Y Y Y .
0 → O Y → p ∗ ( O P 1 ⊕ O P 1 ( − e ) ) ∗ ⊗ O Y ( C 0 ) → x T Y / P 1 → 0 y ↓ 0 → O Y → O Y ( H ) ⊕ M + 1 → T P M ∣ Y → 0 \begin{CD}
0 @>>> \mathcal{O}_Y @>>> p^*(\mathcal{O}_{\mathbb{P}^1}\oplus\mathcal{O}_{\mathbb{P}^1}(-e))^*\otimes\mathcal{O}_Y(C_0) @>{x}>> T_{Y/\mathbb{P}^1} @>>> 0 \\
@. @. @. @V{y}VV @. \\
0 @>>> \mathcal{O}_Y @>>> \mathcal{O}_Y(H)^{\oplus M+1} @>>> T_{\mathbb{P}^M|Y} @>>> 0
\end{CD} 0 0 O Y O Y p ∗ ( O P 1 ⊕ O P 1 ( − e ) ) ∗ ⊗ O Y ( C 0 ) O Y ( H ) ⊕ M + 1 x T Y / P 1 y ↓ ⏐ T P M ∣ Y 0 0
The map y y y is the injective map induced by the composition of two injective maps given by 0 → T Y / P 1 → T Y 0\to T_{Y/\mathbb{P}^1}\to T_Y 0 → T Y / P 1 → T Y and 0 → T Y → T P M ∣ Y 0\to T_Y\to T_{\mathbb{P}^M|Y} 0 → T Y → T P M ∣ Y . We show that the map y y y has a unique lift to a map
z : p ∗ ( O P 1 ⊕ O P 1 ( − e ) ) ∗ ⊗ O Y ( C 0 ) ⟶ O Y ( H ) ⊕ M + 1 . z:p^*(\mathcal{O}_{\mathbb{P}^1}\oplus\mathcal{O}_{\mathbb{P}^1}(-e))^*\otimes\mathcal{O}_Y(C_0)\longrightarrow\mathcal{O}_Y(H)^{\oplus M+1}. z : p ∗ ( O P 1 ⊕ O P 1 ( − e ) ) ∗ ⊗ O Y ( C 0 ) ⟶ O Y ( H ) ⊕ M + 1 .
Apply the functor Hom ( p ∗ ( O P 1 ⊕ O P 1 ( − e ) ) ∗ ⊗ O Y ( C 0 ) , − ) \operatorname{Hom}(p^*(\mathcal{O}_{\mathbb{P}^1}\oplus\mathcal{O}_{\mathbb{P}^1}(-e))^*\otimes\mathcal{O}_Y(C_0),-) Hom ( p ∗ ( O P 1 ⊕ O P 1 ( − e ) ) ∗ ⊗ O Y ( C 0 ) , − ) to the bottom exact sequence we have
0 ⟶ Hom ( p ∗ ( O P 1 ⊕ O P 1 ( − e ) ) ∗ ⊗ O Y ( C 0 ) , O Y ) ⟶ Hom ( p ∗ ( O P 1 ⊕ O P 1 ( − e ) ) ∗ ⊗ O Y ( C 0 ) , O Y ( H ) ⊕ M + 1 ) ⟶ Hom ( p ∗ ( O P 1 ⊕ O P 1 ( − e ) ) ∗ ⊗ O Y ( C 0 ) , T P M ∣ Y ) ⟶ Ext 1 ( p ∗ ( O P 1 ⊕ O P 1 ( − e ) ) ∗ ⊗ O Y ( C 0 ) , O Y ) \begin{aligned}
0\longrightarrow{}&\operatorname{Hom}(p^*(\mathcal{O}_{\mathbb{P}^1}\oplus\mathcal{O}_{\mathbb{P}^1}(-e))^*\otimes\mathcal{O}_Y(C_0),\mathcal{O}_Y) \\
\longrightarrow{}&\operatorname{Hom}(p^*(\mathcal{O}_{\mathbb{P}^1}\oplus\mathcal{O}_{\mathbb{P}^1}(-e))^*\otimes\mathcal{O}_Y(C_0),\mathcal{O}_Y(H)^{\oplus M+1}) \\
\longrightarrow{}&\operatorname{Hom}(p^*(\mathcal{O}_{\mathbb{P}^1}\oplus\mathcal{O}_{\mathbb{P}^1}(-e))^*\otimes\mathcal{O}_Y(C_0),T_{\mathbb{P}^M|Y}) \\
\longrightarrow{}&\operatorname{Ext}^1(p^*(\mathcal{O}_{\mathbb{P}^1}\oplus\mathcal{O}_{\mathbb{P}^1}(-e))^*\otimes\mathcal{O}_Y(C_0),\mathcal{O}_Y)
\end{aligned} 0 ⟶ ⟶ ⟶ ⟶ Hom ( p ∗ ( O P 1 ⊕ O P 1 ( − e ) ) ∗ ⊗ O Y ( C 0 ) , O Y ) Hom ( p ∗ ( O P 1 ⊕ O P 1 ( − e ) ) ∗ ⊗ O Y ( C 0 ) , O Y ( H ) ⊕ M + 1 ) Hom ( p ∗ ( O P 1 ⊕ O P 1 ( − e ) ) ∗ ⊗ O Y ( C 0 ) , T P M ∣ Y ) Ext 1 ( p ∗ ( O P 1 ⊕ O P 1 ( − e ) ) ∗ ⊗ O Y ( C 0 ) , O Y )
Now, Ext 1 ( p ∗ ( O P 1 ⊕ O P 1 ( − e ) ) ∗ ⊗ O Y ( C 0 ) , O Y ) = Ext 1 ( O Y ( C 0 ) ⊕ O Y ( C 0 + e f ) , O Y ) = H 1 ( O Y ( − C 0 ) ) ⊕ O Y ( − C 0 − e f ) ) = 0 \operatorname{Ext}^1(p^*(\mathcal{O}_{\mathbb{P}^1}\oplus\mathcal{O}_{\mathbb{P}^1}(-e))^*\otimes\mathcal{O}_Y(C_0),\mathcal{O}_Y)=\operatorname{Ext}^1(\mathcal{O}_Y(C_0)\oplus\mathcal{O}_Y(C_0+ef),\mathcal{O}_Y)=H^1(\mathcal{O}_Y(-C_0))\oplus\mathcal{O}_Y(-C_0-ef))=0 Ext 1 ( p ∗ ( O P 1 ⊕ O P 1 ( − e ) ) ∗ ⊗ O Y ( C 0 ) , O Y ) = Ext 1 ( O Y ( C 0 ) ⊕ O Y ( C 0 + e f ) , O Y ) = H 1 ( O Y ( − C 0 )) ⊕ O Y ( − C 0 − e f )) = 0 . Further, $\operatorname{Hom}(p^(\mathcal{O}_{\mathbb{P}^1}\oplus\mathcal{O}_{\mathbb{P}^1}(-e))^ \otimes\mathcal{O}_Y(C_0),\mathcal{O}_Y)=\operatorname{Hom}(\mathcal{O}_Y(C_0)\oplus\mathcal{O}_Y(C_0+ e f ) ef) e f ) , O Y ) = H 0 ( O Y ( − C 0 ) ⊕ O Y ( − C 0 − e f ) ) = 0 \mathcal{O}_Y) = H^0(\mathcal{O}_Y(-C_0) \oplus\mathcal{O}_Y(-C_0-ef)) = 0 O Y ) = H 0 ( O Y ( − C 0 ) ⊕ O Y ( − C 0 − e f )) = 0 . Hence the map y y y has a unique lift to the map z z z , which induces the following commutative diagram involving the two previous exact sequences
0 ⟶ O Y ⟶ p ∗ ( O P 1 ⊕ O P 1 ( − e ) ) ∗ ⊗ O Y ( C 0 ) → x T Y / P 1 ⟶ 0 0 ↓ ≅ w ↓ y 0 ⟶ O Y ⟶ O Y ( H ) ⊕ M + 1 ⟶ T P M ∣ Y ⟶ 0 (45) \begin{aligned}
0 &\longrightarrow\mathcal{O}_Y \longrightarrow p^*(\mathcal{O}_{\mathbb{P}^1}\oplus\mathcal{O}_{\mathbb{P}^1}(-e))^*\otimes\mathcal{O}_Y(C_0) \xrightarrow{x} T_{Y/\mathbb{P}^1} \longrightarrow0\\
&\phantom{0}\quad\ \ \downarrow^{\cong w}\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\downarrow^{y}\\
0 &\longrightarrow\mathcal{O}_Y \longrightarrow\mathcal{O}_Y(H)^{\oplus M+1} \longrightarrow T_{\mathbb{P}^M|Y} \longrightarrow0 \tag*{(45)}
\end{aligned} 0 0 ⟶ O Y ⟶ p ∗ ( O P 1 ⊕ O P 1 ( − e ) ) ∗ ⊗ O Y ( C 0 ) x T Y / P 1 ⟶ 0 0 ↓ ≅ w ↓ y ⟶ O Y ⟶ O Y ( H ) ⊕ M + 1 ⟶ T P M ∣ Y ⟶ 0 (45)
To see that the map w w w is an isomorphism, we only need to show that w w w is non-zero. To show this restrict the above diagram to a fiber of the projection map p : Y → P 1 p:Y\to\mathbb{P}^1 p : Y → P 1 . Since O Y ( H ) = O Y ( C 0 + b f ) \mathcal{O}_Y(H)=\mathcal{O}_Y(C_0+bf) O Y ( H ) = O Y ( C 0 + b f ) , we have that under the embedding, the fiber P 1 \mathbb{P}^1 P 1 is embedded by the complete linear series of O P 1 ( 1 ) \mathcal{O}_{\mathbb{P}^1}(1) O P 1 ( 1 ) , i.e. by a linear embedding i : P 1 ↪ P M i:\mathbb{P}^1\hookrightarrow\mathbb{P}^M i : P 1 ↪ P M . Hence after restricting the above diagram to the fiber, we get
0 ⟶ O P 1 → σ P 1 O P 1 ( 1 ) ⊕ 2 → ρ P 1 T P 1 ⟶ 0 0 ↓ ≅ w ~ ↓ α ~ ↓ β = d i 0 ⟶ O P 1 → i ∗ σ P M O P 1 ( 1 ) ⊕ M + 1 → i ∗ ρ P M T P 1 ⊕ O P 1 ( 1 ) ⊕ M − 1 = T P M ∣ P 1 ⟶ 0 \begin{aligned}
0 &\longrightarrow\mathcal{O}_{\mathbb{P}^1}\xrightarrow{\sigma_{\mathbb{P}^1}}\mathcal{O}_{\mathbb{P}^1}(1)^{\oplus2}\xrightarrow{\rho_{\mathbb{P}^1}}T_{\mathbb{P}^1}\longrightarrow0\\
&\phantom{0}\quad\ \ \downarrow^{\cong\widetilde w}\qquad\qquad\downarrow^{\widetilde\alpha}\qquad\qquad\downarrow^{\beta=di}\\
0 &\longrightarrow\mathcal{O}_{\mathbb{P}^1}\xrightarrow{i^*\sigma_{\mathbb{P}^M}}\mathcal{O}_{\mathbb{P}^1}(1)^{\oplus M+1}\xrightarrow{i^*\rho_{\mathbb{P}^M}}T_{\mathbb{P}^1}\oplus\mathcal{O}_{\mathbb{P}^1}(1)^{\oplus M-1}=T_{\mathbb{P}^M|{\mathbb{P}^1}}\longrightarrow0
\end{aligned} 0 0 ⟶ O P 1 σ P 1 O P 1 ( 1 ) ⊕ 2 ρ P 1 T P 1 ⟶ 0 0 ↓ ≅ w ↓ α ↓ β = d i ⟶ O P 1 i ∗ σ P M O P 1 ( 1 ) ⊕ M + 1 i ∗ ρ P M T P 1 ⊕ O P 1 ( 1 ) ⊕ M − 1 = T P M ∣ P 1 ⟶ 0
where σ P 1 \sigma_{\mathbb{P}^1} σ P 1 and ρ P 1 \rho_{\mathbb{P}^1} ρ P 1 are the maps of the Euler exact sequence on P 1 \mathbb{P}^1 P 1 , while i ∗ σ P M i^*\sigma_{\mathbb{P}^M} i ∗ σ P M and i ∗ ρ P M i^*\rho_{\mathbb{P}^M} i ∗ ρ P M are the maps of the Euler exact sequence on P M \mathbb{P}^M P M , pulled back to the line P 1 \mathbb{P}^1 P 1 , and β = d i \beta=di β = d i is the differential map of the embedding i i i . As before, since Hom ( O P 1 ( 1 ) ⊕ 2 , O P 1 ) = Ext 1 ( O P 1 ( 1 ) ⊕ 2 , O P 1 ) = 0 \operatorname{Hom}(\mathcal{O}_{\mathbb{P}^1}(1)^{\oplus2},\mathcal{O}_{\mathbb{P}^1})=\operatorname{Ext}^1(\mathcal{O}_{\mathbb{P}^1}(1)^{\oplus2},\mathcal{O}_{\mathbb{P}^1})=0 Hom ( O P 1 ( 1 ) ⊕ 2 , O P 1 ) = Ext 1 ( O P 1 ( 1 ) ⊕ 2 , O P 1 ) = 0 , the map α ~ \widetilde\alpha α is the unique map lifting d i ∘ ρ P 1 di\circ\rho_{\mathbb{P}^1} d i ∘ ρ P 1 . Choose coordinates [ u : v ] [u:v] [ u : v ] on P 1 \mathbb{P}^1 P 1 and [ X 0 : ⋯ : X M ] [X_0:\cdots:X_M] [ X 0 : ⋯ : X M ] on P M \mathbb{P}^M P M such that i ( [ u : v ] ) = [ u : v : 0 : ⋯ : 0 ] i([u:v])=[u:v:0:\cdots:0] i ([ u : v ]) = [ u : v : 0 : ⋯ : 0 ] . Then d i di d i is the inclusion of the first summand T P 1 → T P 1 ⊕ O P 1 ( 1 ) ⊕ M − 1 T_{\mathbb{P}^1}\to T_{\mathbb{P}^1}\oplus\mathcal{O}_{\mathbb{P}^1}(1)^{\oplus M-1} T P 1 → T P 1 ⊕ O P 1 ( 1 ) ⊕ M − 1 , and the map
α ~ : O P 1 ( 1 ) ⊕ 2 ⟶ O P 1 ( 1 ) ⊕ M + 1 , α ~ ( a , b ) = ( a , b , 0 , … , 0 ) , \widetilde\alpha:\mathcal{O}_{\mathbb{P}^1}(1)^{\oplus2}\longrightarrow\mathcal{O}_{\mathbb{P}^1}(1)^{\oplus M+1},\qquad\widetilde\alpha(a,b)=(a,b,0,\ldots,0), α : O P 1 ( 1 ) ⊕ 2 ⟶ O P 1 ( 1 ) ⊕ M + 1 , α ( a , b ) = ( a , b , 0 , … , 0 ) ,
makes the diagram commutative, since both ρ P 1 ( a , b ) \rho_{\mathbb{P}^1}(a,b) ρ P 1 ( a , b ) and i ∗ ρ P M ( a , b , 0 , … , 0 ) i^*\rho_{\mathbb{P}^M}(a,b,0,\ldots,0) i ∗ ρ P M ( a , b , 0 , … , 0 ) are the vector field a ∂ / ∂ X 0 + b ∂ / ∂ X 1 a\,\partial/\partial X_0+b\,\partial/\partial X_1 a ∂ / ∂ X 0 + b ∂ / ∂ X 1 along the line. Hence this is α ~ \widetilde\alpha α , and α ~ ∘ σ P 1 ( 1 ) = α ~ ( u , v ) = ( u , v , 0 , … , 0 ) = i ∗ σ P M ( 1 ) \widetilde\alpha\circ\sigma_{\mathbb{P}^1}(1)=\widetilde\alpha(u,v)=(u,v,0,\ldots,0)=i^*\sigma_{\mathbb{P}^M}(1) α ∘ σ P 1 ( 1 ) = α ( u , v ) = ( u , v , 0 , … , 0 ) = i ∗ σ P M ( 1 ) . Therefore w ~ ( 1 ) = 1 \widetilde w(1)=1 w ( 1 ) = 1 . Hence w w w is an isomorphism.
Now twisting the commutative diagram (45) by O Y ( − 2 H ) \mathcal{O}_Y(-2H) O Y ( − 2 H ) , we have the following
0 ⟶ O Y ( − 2 H ) ⟶ p ∗ ( O P 1 ⊕ O P 1 ( − e ) ) ∗ ⊗ O Y ( C 0 − 2 H ) → x T Y / P 1 ( − 2 H ) ⟶ 0 0 ↓ ≅ w ( − 2 H ) ↓ y 0 ⟶ O Y ( − 2 H ) ⟶ O Y ( − H ) ⊕ M + 1 ⟶ T P M ∣ Y ( − 2 H ) ⟶ 0 (46) \begin{aligned}
0 &\longrightarrow\mathcal{O}_Y(-2H)\longrightarrow p^*(\mathcal{O}_{\mathbb{P}^1}\oplus\mathcal{O}_{\mathbb{P}^1}(-e))^*\otimes\mathcal{O}_Y(C_0-2H)\xrightarrow{x}T_{Y/\mathbb{P}^1}(-2H)\longrightarrow0\\
&\phantom{0}\quad\ \ \downarrow^{\cong w(-2H)}\qquad\qquad\qquad\qquad\qquad\qquad\downarrow^{y}\\
0 &\longrightarrow\mathcal{O}_Y(-2H)\longrightarrow\mathcal{O}_Y(-H)^{\oplus M+1}\longrightarrow T_{\mathbb{P}^M|Y}(-2H)\longrightarrow0 \tag*{(46)}
\end{aligned} 0 0 ⟶ O Y ( − 2 H ) ⟶ p ∗ ( O P 1 ⊕ O P 1 ( − e ) ) ∗ ⊗ O Y ( C 0 − 2 H ) x T Y / P 1 ( − 2 H ) ⟶ 0 0 ↓ ≅ w ( − 2 H ) ↓ y ⟶ O Y ( − 2 H ) ⟶ O Y ( − H ) ⊕ M + 1 ⟶ T P M ∣ Y ( − 2 H ) ⟶ 0 (46)
Taking cohomology of the bottom row of diagram (46) , since H 1 ( O Y ( − H ) ) = H 2 ( O Y ( − H ) ) = 0 H^1(\mathcal{O}_Y(-H))=H^2(\mathcal{O}_Y(-H))=0 H 1 ( O Y ( − H )) = H 2 ( O Y ( − H )) = 0 , we get
0 ⟶ H 1 ( T P M ∣ Y ( − 2 H ) ) ⟶ H 2 ( O Y ( − 2 H ) ) ⟶ 0. 0\longrightarrow H^1(T_{\mathbb{P}^M|Y}(-2H))\longrightarrow H^2(\mathcal{O}_Y(-2H))\longrightarrow0. 0 ⟶ H 1 ( T P M ∣ Y ( − 2 H )) ⟶ H 2 ( O Y ( − 2 H )) ⟶ 0.
Similarly tensoring the exact sequence of horizontal and vertical tangent bundles of Y Y Y by − 2 H -2H − 2 H , we have
0 ⟶ T Y / P 1 ( − 2 H ) ⟶ T Y ( − 2 H ) ⟶ p ∗ ( T P 1 ) ( − 2 H ) ⟶ 0 0\longrightarrow T_{Y/\mathbb{P}^1}(-2H)\longrightarrow T_Y(-2H)\longrightarrow p^*(T_{\mathbb{P}^1})(-2H)\longrightarrow0 0 ⟶ T Y / P 1 ( − 2 H ) ⟶ T Y ( − 2 H ) ⟶ p ∗ ( T P 1 ) ( − 2 H ) ⟶ 0
and taking the cohomology, we have that
H 0 ( p ∗ ( T P 1 ) ( − 2 H ) ) ⟶ H 1 ( T Y / P 1 ( − 2 H ) ) ⟶ H 1 ( T Y ( − 2 H ) ) ⟶ H 1 ( p ∗ ( T P 1 ) ( − 2 H ) ) ⟶ 0. H^0(p^*(T_{\mathbb{P}^1})(-2H))\longrightarrow H^1(T_{Y/\mathbb{P}^1}(-2H))\longrightarrow H^1(T_Y(-2H))\longrightarrow H^1(p^*(T_{\mathbb{P}^1})(-2H))\longrightarrow0. H 0 ( p ∗ ( T P 1 ) ( − 2 H )) ⟶ H 1 ( T Y / P 1 ( − 2 H )) ⟶ H 1 ( T Y ( − 2 H )) ⟶ H 1 ( p ∗ ( T P 1 ) ( − 2 H )) ⟶ 0.
Now note that p ∗ ( T P 1 ) ( − 2 H ) = − 2 C 0 + ( 2 − 2 b ) f p^*(T_{\mathbb{P}^1})(-2H)=-2C_0+(2-2b)f p ∗ ( T P 1 ) ( − 2 H ) = − 2 C 0 + ( 2 − 2 b ) f . We have that H 0 ( − 2 C 0 + ( 2 − 2 b ) f ) = 0 H^0(-2C_0+(2-2b)f)=0 H 0 ( − 2 C 0 + ( 2 − 2 b ) f ) = 0 and H 1 ( − 2 C 0 + ( 2 − 2 b ) f ) = H 1 ( ( 2 b − e − 4 ) f ) ∗ = 0 H^1(-2C_0+(2-2b)f)=H^1((2b-e-4)f)^*=0 H 1 ( − 2 C 0 + ( 2 − 2 b ) f ) = H 1 (( 2 b − e − 4 ) f ) ∗ = 0 if b ≥ e + 2 b\geq e+2 b ≥ e + 2 . Hence we have
0 ⟶ H 1 ( T Y / P 1 ( − 2 H ) ) ⟶ H 1 ( T Y ( − 2 H ) ) ⟶ 0. 0\longrightarrow H^1(T_{Y/\mathbb{P}^1}(-2H))\longrightarrow H^1(T_Y(-2H))\longrightarrow0. 0 ⟶ H 1 ( T Y / P 1 ( − 2 H )) ⟶ H 1 ( T Y ( − 2 H )) ⟶ 0.
On the other hand, p ∗ ( O P 1 ⊕ O P 1 ( − e ) ) ∗ ⊗ O Y ( C 0 − 2 H ) = O Y ( − C 0 − 2 b f ) ⊕ O Y ( − C 0 + ( e − 2 b ) f ) p^{*}(\mathcal{O}_{\mathbb{P}^{1}} \oplus\mathcal{O}_{\mathbb{P}^{1}}(-e))^{*} \otimes\mathcal{O}_{Y}(C_{0}-2H)=\mathcal{O}_{Y}(-C_{0}-2bf)\oplus\mathcal{O}_{Y}(-C_{0}+(e-2b)f) p ∗ ( O P 1 ⊕ O P 1 ( − e ) ) ∗ ⊗ O Y ( C 0 − 2 H ) = O Y ( − C 0 − 2 b f ) ⊕ O Y ( − C 0 + ( e − 2 b ) f ) has no cohomology, since p ∗ O Y ( − C 0 ) = R 1 p ∗ O Y ( − C 0 ) = 0 p_{*}\mathcal{O}_{Y}(-C_{0})=R^{1}p_{*}\mathcal{O}_{Y}(-C_{0})=0 p ∗ O Y ( − C 0 ) = R 1 p ∗ O Y ( − C 0 ) = 0 ; so taking the cohomology of the top row of the diagram (46) , we get
0 → H 1 ( T Y / P 1 ( − 2 H ) ) → H 2 ( O Y ( − 2 H ) ) → 0. 0 \to H^{1}(T_{Y/\mathbb{P}^{1}}(-2H)) \to H^{2}(\mathcal{O}_{Y}(-2H)) \to0. 0 → H 1 ( T Y / P 1 ( − 2 H )) → H 2 ( O Y ( − 2 H )) → 0.
Hence taking the cohomology of the diagram (46) , we have a commutative diagram
H 1 ( T Y / P 1 ( − 2 H ) ) → ≃ H 2 ( O Y ( − 2 H ) ) ↓ ≃ ↓ ≃ H 1 ( T Y ( − 2 H ) ) ↓ s H 1 ( T P M ∣ Y ( − 2 H ) ) → ≃ H 2 ( O Y ( − 2 H ) ) \begin{aligned}
H^{1}(T_{Y/\mathbb{P}^{1}}(-2H)) &\xrightarrow{\simeq} H^{2}(\mathcal{O}_{Y}(-2H)) \\
\downarrow{\simeq} & \qquad\qquad\qquad\qquad\downarrow{\simeq} \\
H^{1}(T_{Y}(-2H)) & \\
\downarrow{s} & \\
H^{1}(T_{\mathbb{P}^{M}|Y}(-2H)) &\xrightarrow{\simeq} H^{2}(\mathcal{O}_{Y}(-2H))
\end{aligned} H 1 ( T Y / P 1 ( − 2 H )) ↓ ≃ H 1 ( T Y ( − 2 H )) ↓ s H 1 ( T P M ∣ Y ( − 2 H )) ≃ H 2 ( O Y ( − 2 H )) ↓ ≃ ≃ H 2 ( O Y ( − 2 H ))
This shows that the map s s s is injective and is in fact an isomorphism. Hence H 0 ( N Y ( − 2 H ) ) = 0 H^{0}(N_{Y}(-2H))=0 H 0 ( N Y ( − 2 H )) = 0 .
(b) Consider the exact sequence given by twisting (4) by − 2 H -2H − 2 H .
0 → O Y ( − 2 H − K Y ) → s ∨ N Y ( − 2 H ) → H ( − 2 H ) → 0. (47) 0 \to\mathcal{O}_{Y}(-2H-K_{Y}) \xrightarrow{s^{\vee}} N_{Y}(-2H) \to\mathcal{H}(-2H) \to0. \tag*{(47)} 0 → O Y ( − 2 H − K Y ) s ∨ N Y ( − 2 H ) → H ( − 2 H ) → 0. (47)
Let p : Y → P 1 p:Y\to\mathbb{P}^{1} p : Y → P 1 denote the smooth fibration. Now note that − 2 H − K Y = ( e + 2 − 2 b ) f -2H-K_{Y}=(e+2-2b)f − 2 H − K Y = ( e + 2 − 2 b ) f and hence R 1 p ∗ ( − 2 H − K Y ) = 0 R^{1}p_{*}(-2H-K_{Y})=0 R 1 p ∗ ( − 2 H − K Y ) = 0 . Hence we have the following exact sequence by pushing forward the exact sequence (47) .
0 → p ∗ ( − 2 H − K Y ) → β p ∗ ( N Y ( − 2 H ) ) → p ∗ ( H ( − 2 H ) ) → 0. (48) 0 \to p_{*}(-2H-K_{Y}) \xrightarrow{\beta} p_{*}(N_{Y}(-2H)) \to p_{*}(\mathcal{H}(-2H)) \to0. \tag*{(48)} 0 → p ∗ ( − 2 H − K Y ) β p ∗ ( N Y ( − 2 H )) → p ∗ ( H ( − 2 H )) → 0. (48)
It is enough to show that the exact sequence (48) is split: then H 1 ( p ∗ ( − 2 H − K Y ) ) → H 1 ( p ∗ ( N Y ( − 2 H ) ) ) H^{1}(p_{*}(-2H-K_{Y}))\to H^{1}(p_{*}(N_{Y}(-2H))) H 1 ( p ∗ ( − 2 H − K Y )) → H 1 ( p ∗ ( N Y ( − 2 H ))) is injective, and since H 1 ( O Y ( − 2 H − K Y ) ) = H 1 ( p ∗ ( − 2 H − K Y ) ) H^{1}(\mathcal{O}_{Y}(-2H-K_{Y}))=H^{1}(p_{*}(-2H-K_{Y})) H 1 ( O Y ( − 2 H − K Y )) = H 1 ( p ∗ ( − 2 H − K Y )) and H 1 ( p ∗ ( N Y ( − 2 H ) ) ) → H 1 ( N Y ( − 2 H ) ) H^{1}(p_{*}(N_{Y}(-2H)))\to H^{1}(N_{Y}(-2H)) H 1 ( p ∗ ( N Y ( − 2 H ))) → H 1 ( N Y ( − 2 H )) is injective by the Leray spectral sequence, the map s ∨ : H 1 ( O Y ( − 2 H − K Y ) ) → H 1 ( N Y ( − 2 H ) ) s^{\vee}:H^{1}(\mathcal{O}_{Y}(-2H-K_{Y}))\to H^{1}(N_{Y}(-2H)) s ∨ : H 1 ( O Y ( − 2 H − K Y )) → H 1 ( N Y ( − 2 H )) is injective. To see this we first show that there is a surjection from p ∗ ( N Y ( − 2 H ) ) p_{*}(N_{Y}(-2H)) p ∗ ( N Y ( − 2 H )) to p ∗ ( − 2 H − K Y ) p_{*}(-2H-K_{Y}) p ∗ ( − 2 H − K Y ) . To see this consider the exact sequence (5) twisted by − 2 H -2H − 2 H to get
0 → T Y ( − 2 H ) → T P M ∣ Y ( − 2 H ) → N Y ( − 2 H ) → 0. (49) 0 \to T_{Y}(-2H) \to T_{\mathbb{P}^{M}|Y}(-2H) \to N_{Y}(-2H) \to0. \tag*{(49)} 0 → T Y ( − 2 H ) → T P M ∣ Y ( − 2 H ) → N Y ( − 2 H ) → 0. (49)
We show that
R 1 p ∗ ( T P M ∣ Y ( − 2 H ) ) = 0 , p ∗ ( T Y ( − 2 H ) ) = O P 1 ( e − 2 b ) and R 1 p ∗ ( T Y ( − 2 H ) ) = O P 1 ( e + 2 − 2 b ) . R^{1}p_{*}(T_{\mathbb{P}^{M}|Y}(-2H))=0,\qquad p_{*}(T_{Y}(-2H))=\mathcal{O}_{\mathbb{P}^{1}}(e-2b)\quad\text{and}\quad R^{1}p_{*}(T_{Y}(-2H))=\mathcal{O}_{\mathbb{P}^{1}}(e+2-2b). R 1 p ∗ ( T P M ∣ Y ( − 2 H )) = 0 , p ∗ ( T Y ( − 2 H )) = O P 1 ( e − 2 b ) and R 1 p ∗ ( T Y ( − 2 H )) = O P 1 ( e + 2 − 2 b ) .
To see the first vanishing once again note that from the Euler exact sequence (6) twisted by − 2 H -2H − 2 H , we have
0 → O Y ( − 2 H ) → O Y ( − H ) ⊕ M + 1 → T P M ∣ Y ( − 2 H ) → 0. 0 \to\mathcal{O}_{Y}(-2H) \to\mathcal{O}_{Y}(-H)^{\oplus M+1} \to T_{\mathbb{P}^{M}|Y}(-2H) \to0. 0 → O Y ( − 2 H ) → O Y ( − H ) ⊕ M + 1 → T P M ∣ Y ( − 2 H ) → 0.
Since R 1 p ∗ ( O Y ( − H ) ) = R 2 p ∗ ( O Y ( − 2 H ) ) = 0 R^{1}p_{*}(\mathcal{O}_{Y}(-H))=R^{2}p_{*}(\mathcal{O}_{Y}(-2H))=0 R 1 p ∗ ( O Y ( − H )) = R 2 p ∗ ( O Y ( − 2 H )) = 0 , we have that R 1 p ∗ ( T P M ∣ Y ( − 2 H ) ) = 0 R^{1}p_{*}(T_{\mathbb{P}^{M}|Y}(-2H))=0 R 1 p ∗ ( T P M ∣ Y ( − 2 H )) = 0 . Now we turn to R i p ∗ ( T Y ( − 2 H ) ) R^{i}p_{*}(T_{Y}(-2H)) R i p ∗ ( T Y ( − 2 H )) . Consider the exact sequence of horizontal and vertical tangent bundles on Y Y Y twisted by − 2 H -2H − 2 H .
0 → T Y / P 1 ( − 2 H ) → T Y ( − 2 H ) → p ∗ T P 1 ( − 2 H ) → 0. (50) 0 \to T_{Y/\mathbb{P}^{1}}(-2H) \to T_{Y}(-2H) \to p^{*}T_{\mathbb{P}^{1}}(-2H) \to0. \tag*{(50)} 0 → T Y / P 1 ( − 2 H ) → T Y ( − 2 H ) → p ∗ T P 1 ( − 2 H ) → 0. (50)
Using the relative Euler exact sequence we have that T Y / P 1 = 2 C 0 + e f T_{Y/\mathbb{P}^{1}}=2C_{0}+ef T Y / P 1 = 2 C 0 + e f , so that T Y / P 1 ( − 2 H ) = ( e − 2 b ) f T_{Y/\mathbb{P}^{1}}(-2H)=(e-2b)f T Y / P 1 ( − 2 H ) = ( e − 2 b ) f . Therefore p ∗ ( T Y / P 1 ( − 2 H ) ) = O P 1 ( e − 2 b ) p_{*}(T_{Y/\mathbb{P}^{1}}(-2H))=\mathcal{O}_{\mathbb{P}^{1}}(e-2b) p ∗ ( T Y / P 1 ( − 2 H )) = O P 1 ( e − 2 b ) and R i p ∗ ( T Y / P 1 ( − 2 H ) ) = R i p ∗ ( ( e − 2 b ) f ) = 0 R^{i}p_{*}(T_{Y/\mathbb{P}^{1}}(-2H))=R^{i}p_{*}((e-2b)f)=0 R i p ∗ ( T Y / P 1 ( − 2 H )) = R i p ∗ (( e − 2 b ) f ) = 0 for i = 1 , 2 i=1,2 i = 1 , 2 . Now p ∗ T P 1 ( − 2 H ) = − 2 C 0 + ( 2 − 2 b ) f p^{*}T_{\mathbb{P}^{1}}(-2H)=-2C_{0}+(2-2b)f p ∗ T P 1 ( − 2 H ) = − 2 C 0 + ( 2 − 2 b ) f . Therefore, p ∗ p ∗ T P 1 ( − 2 H ) = p ∗ ( − 2 C 0 + ( 2 − 2 b ) f ) = p_{*}p^{*}T_{\mathbb{P}^{1}}(-2H)=p_{*}(-2C_{0}+(2-2b)f)= p ∗ p ∗ T P 1 ( − 2 H ) = p ∗ ( − 2 C 0 + ( 2 − 2 b ) f ) = O P 1 ( 2 − 2 b ) ⊗ p ∗ ( − 2 C 0 ) = 0 \mathcal{O}_{\mathbb{P}^1}(2-2b)\otimes p_*(-2C_0)=0 O P 1 ( 2 − 2 b ) ⊗ p ∗ ( − 2 C 0 ) = 0 and
R 1 p ∗ ( p ∗ T P 1 ( − 2 H ) ) = R 1 p ∗ ( − 2 C 0 + ( 2 − 2 b ) f ) = O P 1 ( 2 − 2 b ) ⊗ R 1 p ∗ ( − 2 C 0 ) = O P 1 ( 2 − 2 b ) ⊗ ( p ∗ ( O Y ) ) ∗ ⊗ ( ⋀ 2 ( O P 1 ⊕ O P 1 ( − e ) ) ) ∗ = O P 1 ( e + 2 − 2 b ) . \begin{aligned}
R^1p_*(p^*T_{\mathbb{P}^1}(-2H))&=R^1p_*(-2C_0+(2-2b)f)\\
&=\mathcal{O}_{\mathbb{P}^1}(2-2b)\otimes R^1p_*(-2C_0)\\
&=\mathcal{O}_{\mathbb{P}^1}(2-2b)\otimes(p_*(\mathcal{O}_Y))^*\otimes\left(\bigwedge^2(\mathcal{O}_{\mathbb{P}^1}\oplus\mathcal{O}_{\mathbb{P}^1}(-e))\right)^*\\
&=\mathcal{O}_{\mathbb{P}^1}(e+2-2b).
\end{aligned} R 1 p ∗ ( p ∗ T P 1 ( − 2 H )) = R 1 p ∗ ( − 2 C 0 + ( 2 − 2 b ) f ) = O P 1 ( 2 − 2 b ) ⊗ R 1 p ∗ ( − 2 C 0 ) = O P 1 ( 2 − 2 b ) ⊗ ( p ∗ ( O Y ) ) ∗ ⊗ ( ⋀ 2 ( O P 1 ⊕ O P 1 ( − e )) ) ∗ = O P 1 ( e + 2 − 2 b ) .
Pushing forward the sequence (50) , we have that
0 → p ∗ ( T Y / P 1 ( − 2 H ) ) = O P 1 ( e − 2 b ) → p ∗ T Y ( − 2 H ) → 0 0\to p_*(T_{Y/\mathbb{P}^1}(-2H))=\mathcal{O}_{\mathbb{P}^1}(e-2b)\to p_*T_Y(-2H)\to0 0 → p ∗ ( T Y / P 1 ( − 2 H )) = O P 1 ( e − 2 b ) → p ∗ T Y ( − 2 H ) → 0
and
0 → R 1 p ∗ T Y ( − 2 H ) → O P 1 ( e + 2 − 2 b ) → 0. 0\to R^1p_*T_Y(-2H)\to\mathcal{O}_{\mathbb{P}^1}(e+2-2b)\to0. 0 → R 1 p ∗ T Y ( − 2 H ) → O P 1 ( e + 2 − 2 b ) → 0.
Now pushing forward the exact sequence (49) , we have that
0 → O P 1 ( e − 2 b ) = p ∗ T Y ( − 2 H ) = p ∗ ( T Y / P 1 ( − 2 H ) ) → p ∗ ( T P M ∣ Y ( − 2 H ) ) → p ∗ ( N Y ( − 2 H ) ) → λ O P 1 ( e + 2 − 2 b ) = p ∗ ( − 2 H − K Y ) → 0. (51) 0\to\mathcal{O}_{\mathbb{P}^1}(e-2b)=p_*T_Y(-2H)=p_*(T_{Y/\mathbb{P}^1}(-2H))\to p_*(T_{\mathbb{P}M|Y}(-2H))\to p_*(N_Y(-2H))
\xrightarrow{\lambda}\mathcal{O}_{\mathbb{P}^1}(e+2-2b)=p_*(-2H-K_Y)\to0. \tag*{(51)} 0 → O P 1 ( e − 2 b ) = p ∗ T Y ( − 2 H ) = p ∗ ( T Y / P 1 ( − 2 H )) → p ∗ ( T P M ∣ Y ( − 2 H )) → p ∗ ( N Y ( − 2 H )) λ O P 1 ( e + 2 − 2 b ) = p ∗ ( − 2 H − K Y ) → 0. (51)
We now end the proof by showing that for the K3 carpet Y ~ \widetilde{Y} Y , which is defined by the injection s ∨ : O Y ( − K Y ) → N Y s^\vee:\mathcal{O}_Y(-K_Y)\to N_Y s ∨ : O Y ( − K Y ) → N Y of (1) , the map λ \lambda λ is, up to a nonzero scalar, a section to the injective map β \beta β induced by s ∨ s^\vee s ∨ after twisting by − 2 H -2H − 2 H and pushing forward, which sits in the following diagram (52)
0 p ∗ ( − 2 H − K Y ) 0 → O P 1 ( e − 2 b ) → p ∗ ( T P M ∣ Y ( − 2 H ) ) → δ p ∗ ( N Y ( − 2 H ) ) → λ p ∗ ( − 2 H − K Y ) → 0 \begin{CD}
&&0\\
&&\downarrow\\
&&p_*(-2H-K_Y)\\
&&\downarrow\beta\\
0 @>>> \mathcal{O}_{\mathbb{P}^1}(e-2b) @>>> p_*(T_{\mathbb{P}M|Y}(-2H)) @>{\delta}>> p_*(N_Y(-2H)) @>{\lambda}>> p_*(-2H-K_Y) @>>> 0
\end{CD} 0 p ∗ ( − 2 H − K Y ) 0 O P 1 ( e − 2 b ) p ∗ ( T P M ∣ Y ( − 2 H )) δ p ∗ ( N Y ( − 2 H )) λ p ∗ ( − 2 H − K Y ) 0
Since p ∗ ( − 2 H − K Y ) = O P 1 ( e + 2 − 2 b ) p_*(-2H-K_Y)=\mathcal{O}_{\mathbb{P}^1}(e+2-2b) p ∗ ( − 2 H − K Y ) = O P 1 ( e + 2 − 2 b ) is a line bundle, it is enough to show that λ ∘ β \lambda\circ\beta λ ∘ β is non-zero. This happens if and only if Im ( β ) \operatorname{Im}(\beta) Im ( β ) is not contained in Ker ( λ ) \operatorname{Ker}(\lambda) Ker ( λ ) . So we need to show that β \beta β does not factor as
p ∗ ( − 2 H − K Y ) ↙ 0 → Ker ( λ ) → p ∗ ( N Y ( − 2 H ) ) \begin{CD}
&&p_*(-2H-K_Y)\\
&\swarrow&\downarrow\beta\\
0 @>>> \operatorname{Ker}(\lambda) @>>> p_*(N_Y(-2H))
\end{CD} p ∗ ( − 2 H − K Y ) ↙ 0 Ker ( λ ) p ∗ ( N Y ( − 2 H ))
i.e. that β \beta β is not in the image of
Hom ( p ∗ ( − 2 H − K Y ) , Ker ( λ ) ) → Hom ( p ∗ ( − 2 H − K Y ) , p ∗ ( N Y ( − 2 H ) ) ) . (52) \operatorname{Hom}(p_*(-2H-K_Y),\operatorname{Ker}(\lambda))\to\operatorname{Hom}(p_*(-2H-K_Y),p_*(N_Y(-2H))). \tag*{(52)} Hom ( p ∗ ( − 2 H − K Y ) , Ker ( λ )) → Hom ( p ∗ ( − 2 H − K Y ) , p ∗ ( N Y ( − 2 H ))) . (52)
To see this note that (by sequence (51) and diagram (52)), we have an exact sequence
0 → O P 1 ( e − 2 b ) = p ∗ T Y / P 1 ( − 2 H ) → p ∗ ( T P M ∣ Y ( − 2 H ) ) → Ker ( λ ) → 0. 0\to\mathcal{O}_{\mathbb{P}^1}(e-2b)=p_*T_{Y/\mathbb{P}^1}(-2H)\to p_*(T_{\mathbb{P}M|Y}(-2H))\to\operatorname{Ker}(\lambda)\to0. 0 → O P 1 ( e − 2 b ) = p ∗ T Y / P 1 ( − 2 H ) → p ∗ ( T P M ∣ Y ( − 2 H )) → Ker ( λ ) → 0.
Applying the functor Hom ( p ∗ ( − 2 H − K Y ) , − ) \operatorname{Hom}(p_*(-2H-K_Y),-) Hom ( p ∗ ( − 2 H − K Y ) , − ) we get
0 → Hom ( p ∗ ( − 2 H − K Y ) , p ∗ T Y / P 1 ( − 2 H ) ) → Hom ( p ∗ ( − 2 H − K Y ) , p ∗ ( T P M ∣ Y ( − 2 H ) ) ) → Hom ( p ∗ ( − 2 H − K Y ) , Ker ( λ ) ) → Ext 1 ( p ∗ ( − 2 H − K Y ) , p ∗ T Y / P 1 ( − 2 H ) ) → η Ext 1 ( p ∗ ( − 2 H − K Y ) , p ∗ ( T P M ∣ Y ( − 2 H ) ) ) . (53) \begin{aligned}
0\to{}&\operatorname{Hom}(p_*(-2H-K_Y),p_*T_{Y/\mathbb{P}^1}(-2H))\\
\to{}&\operatorname{Hom}(p_*(-2H-K_Y),p_*(T_{\mathbb{P}M|Y}(-2H)))\\
\to{}&\operatorname{Hom}(p_*(-2H-K_Y),\operatorname{Ker}(\lambda))\\
\to{}&\operatorname{Ext}^1(p_*(-2H-K_Y),p_*T_{Y/\mathbb{P}^1}(-2H))\\
\xrightarrow{\eta}{}&\operatorname{Ext}^1(p_*(-2H-K_Y),p_*(T_{\mathbb{P}M|Y}(-2H))). \tag*{(53)}
\end{aligned} 0 → → → → η Hom ( p ∗ ( − 2 H − K Y ) , p ∗ T Y / P 1 ( − 2 H )) Hom ( p ∗ ( − 2 H − K Y ) , p ∗ ( T P M ∣ Y ( − 2 H ))) Hom ( p ∗ ( − 2 H − K Y ) , Ker ( λ )) Ext 1 ( p ∗ ( − 2 H − K Y ) , p ∗ T Y / P 1 ( − 2 H )) Ext 1 ( p ∗ ( − 2 H − K Y ) , p ∗ ( T P M ∣ Y ( − 2 H ))) . (53)
We first show that the map η \eta η is an injective map and is in fact an isomorphism. Pushing forward the bottom row of diagram (46) , since p ∗ ( O Y ( − H ) ) = R 1 p ∗ ( O Y ( − H ) ) = 0 p_*(\mathcal{O}_Y(-H))=R^1p_*(\mathcal{O}_Y(-H))=0 p ∗ ( O Y ( − H )) = R 1 p ∗ ( O Y ( − H )) = 0 , we get
0 → p ∗ ( T P M ∣ Y ( − 2 H ) ) → R 1 p ∗ ( − 2 H ) → 0 , (54) 0\to p_*(T_{\mathbb{P}M|Y}(-2H))\to R^1p_*(-2H)\to0, \tag*{(54)} 0 → p ∗ ( T P M ∣ Y ( − 2 H )) → R 1 p ∗ ( − 2 H ) → 0 , (54)
and pushing forward the top row of diagram (46) , since p ∗ ( O P 1 ⊕ O P 1 ( − e ) ) ∗ ⊗ O Y ( C 0 − 2 H ) p^*(\mathcal{O}_{\mathbb{P}^1}\oplus\mathcal{O}_{\mathbb{P}^1}(-e))^*\otimes\mathcal{O}_Y(C_0-2H) p ∗ ( O P 1 ⊕ O P 1 ( − e ) ) ∗ ⊗ O Y ( C 0 − 2 H ) has no direct images, we get
0 ⟶ p ∗ T Y / P 1 ( − 2 H ) ⟶ R 1 p ∗ ( − 2 H ) ⟶ 0. (55) 0\longrightarrow p_*T_Y/\mathbb{P}^1(-2H)\longrightarrow R^1p_*(-2H)\longrightarrow0. \tag*{(55)} 0 ⟶ p ∗ T Y / P 1 ( − 2 H ) ⟶ R 1 p ∗ ( − 2 H ) ⟶ 0. (55)
Because of the commutative diagram (46) , (54) and (55) sit in a commutative diagram
0 ⟶ Ext 1 ( p ∗ ( − 2 H − K Y ) , p ∗ T Y / P 1 ( − 2 H ) ) ⟶ Ext 1 ( p ∗ ( − 2 H − K Y ) , R 1 p ∗ ( − 2 H ) ) ⟶ 0 ⟶ ↓ η ↓ ≅ 0 ⟶ Ext 1 ( p ∗ ( − 2 H − K Y ) , p ∗ ( T P M ∣ Y ( − 2 H ) ) ) ⟶ Ext 1 ( p ∗ ( − 2 H − K Y ) , R 1 p ∗ ( − 2 H ) ) ⟶ 0. \begin{aligned}
0&\longrightarrow\operatorname{Ext}^1(p_*(-2H-K_Y),p_*T_Y/\mathbb{P}^1(-2H))\longrightarrow\operatorname{Ext}^1(p_*(-2H-K_Y),R^1p_*(-2H))\longrightarrow0\\
&\phantom{\longrightarrow}\downarrow\eta\hspace{7.5cm}\downarrow\cong\\
0&\longrightarrow\operatorname{Ext}^1(p_*(-2H-K_Y),p_*(T_{\mathbb{P}^M}|_Y(-2H)))\longrightarrow\operatorname{Ext}^1(p_*(-2H-K_Y),R^1p_*(-2H))\longrightarrow0.
\end{aligned} 0 0 ⟶ Ext 1 ( p ∗ ( − 2 H − K Y ) , p ∗ T Y / P 1 ( − 2 H )) ⟶ Ext 1 ( p ∗ ( − 2 H − K Y ) , R 1 p ∗ ( − 2 H )) ⟶ 0 ⟶ ↓ η ↓≅ ⟶ Ext 1 ( p ∗ ( − 2 H − K Y ) , p ∗ ( T P M ∣ Y ( − 2 H ))) ⟶ Ext 1 ( p ∗ ( − 2 H − K Y ) , R 1 p ∗ ( − 2 H )) ⟶ 0.
The right hand vertical map above is an isomorphism because it is induced by w ( − 2 H ) w(-2H) w ( − 2 H ) in diagram (46) which is an isomorphism. Hence the map η \eta η is an isomorphism. Going back to sequence (53) , we have
0 ⟶ Hom ( p ∗ ( − 2 H − K Y ) , p ∗ T Y / P 1 ( − 2 H ) ) ⟶ Hom ( p ∗ ( − 2 H − K Y ) , p ∗ ( T P M ∣ Y ( − 2 H ) ) ) ⟶ Hom ( p ∗ ( − 2 H − K Y ) , Ker ( λ ) ) ⟶ 0. (56) \begin{aligned}
0\longrightarrow&\operatorname{Hom}(p_*(-2H-K_Y),p_*T_Y/\mathbb{P}^1(-2H))\\
\longrightarrow&\operatorname{Hom}(p_*(-2H-K_Y),p_*(T_{\mathbb{P}^M}|_Y(-2H)))\\
\longrightarrow&\operatorname{Hom}(p_*(-2H-K_Y),\operatorname{Ker}(\lambda))\longrightarrow0. \tag*{(56)}
\end{aligned} 0 ⟶ ⟶ ⟶ Hom ( p ∗ ( − 2 H − K Y ) , p ∗ T Y / P 1 ( − 2 H )) Hom ( p ∗ ( − 2 H − K Y ) , p ∗ ( T P M ∣ Y ( − 2 H ))) Hom ( p ∗ ( − 2 H − K Y ) , Ker ( λ )) ⟶ 0. (56)
Therefore, the image of the map (52) is the image of
Hom ( p ∗ ( − 2 H − K Y ) , p ∗ ( T P M ∣ Y ( − 2 H ) ) ) ⟶ Hom ( p ∗ ( − 2 H − K Y ) , p ∗ ( N Y ( − 2 H ) ) ) . (57) \operatorname{Hom}(p_*(-2H-K_Y),p_*(T_{\mathbb{P}^M}|_Y(-2H)))\longrightarrow\operatorname{Hom}(p_*(-2H-K_Y),p_*(N_Y(-2H))). \tag*{(57)} Hom ( p ∗ ( − 2 H − K Y ) , p ∗ ( T P M ∣ Y ( − 2 H ))) ⟶ Hom ( p ∗ ( − 2 H − K Y ) , p ∗ ( N Y ( − 2 H ))) . (57)
Since p ∗ p ∗ ( − 2 H − K Y ) = − 2 H − K Y p^*p_*(-2H-K_Y)=-2H-K_Y p ∗ p ∗ ( − 2 H − K Y ) = − 2 H − K Y , by the projection formula, this is the map
H 0 ( T P M ∣ Y ⊗ K Y ) ⟶ H 0 ( N Y ⊗ K Y ) . (58) H^0(T_{\mathbb{P}^M}|_Y\otimes K_Y)\longrightarrow H^0(N_Y\otimes K_Y). \tag*{(58)} H 0 ( T P M ∣ Y ⊗ K Y ) ⟶ H 0 ( N Y ⊗ K Y ) . (58)
and β \beta β corresponds to s ∈ H 0 ( N Y ⊗ K Y ) s\in H^0(N_Y\otimes K_Y) s ∈ H 0 ( N Y ⊗ K Y ) . Now by taking the cohomology of the exact sequence
0 ⟶ T Y ⊗ K Y ⟶ T P M ∣ Y ⊗ K Y ⟶ N Y ⊗ K Y ⟶ 0 0\longrightarrow T_Y\otimes K_Y\longrightarrow T_{\mathbb{P}^M}|_Y\otimes K_Y\longrightarrow N_Y\otimes K_Y\longrightarrow0 0 ⟶ T Y ⊗ K Y ⟶ T P M ∣ Y ⊗ K Y ⟶ N Y ⊗ K Y ⟶ 0
we have
0 ⟶ H 0 ( T Y ⊗ K Y ) ⟶ H 0 ( T P M ∣ Y ⊗ K Y ) ⟶ H 0 ( N Y ⊗ K Y ) → ∂ H 1 ( T Y ⊗ K Y ) . 0\longrightarrow H^0(T_Y\otimes K_Y)\longrightarrow H^0(T_{\mathbb{P}^M}|_Y\otimes K_Y)\longrightarrow H^0(N_Y\otimes K_Y)\xrightarrow{\partial} H^1(T_Y\otimes K_Y). 0 ⟶ H 0 ( T Y ⊗ K Y ) ⟶ H 0 ( T P M ∣ Y ⊗ K Y ) ⟶ H 0 ( N Y ⊗ K Y ) ∂ H 1 ( T Y ⊗ K Y ) .
Now s s s is not in the image of (58) , because ∂ ( s ) ≠ 0 \partial(s)\ne0 ∂ ( s ) = 0 by Lemma 3.5. Hence λ ∘ β ≠ 0 \lambda\circ\beta\ne0 λ ∘ β = 0 , the sequence (48) is split, and s ∨ : H 1 ( O Y ( − 2 H − K Y ) ) ⟶ H 1 ( N Y ( − 2 H ) ) s^\vee:H^1(\mathcal{O}_Y(-2H-K_Y))\longrightarrow H^1(N_Y(-2H)) s ∨ : H 1 ( O Y ( − 2 H − K Y )) ⟶ H 1 ( N Y ( − 2 H )) is injective. Therefore H 0 ( H ( − 2 H ) ) = 0 H^0(\mathcal{H}(-2H))=0 H 0 ( H ( − 2 H )) = 0 . □
Classification of Fano threefolds of Picard rank one
Definition 4.4. For g ≥ 3 g\ge3 g ≥ 3 let V g V_g V g denote the locus, in the Hilbert scheme of P g + 1 \mathbb{P}^{g+1} P g + 1 , of smooth Fano threefolds V ⊂ P g + 1 V\subset\mathbb{P}^{g+1} V ⊂ P g + 1 of Picard rank one, index one and genus g g g , i.e. Pic V = Z ( − K V ) \operatorname{Pic}V=\mathbb{Z}(-K_V) Pic V = Z ( − K V ) and ( − K V ) 3 = 2 g − 2 (-K_V)^3=2g-2 ( − K V ) 3 = 2 g − 2 , embedded by the complete linear series ∣ − K V ∣ \lvert-K_V\rvert ∣ − K V ∣ ; it is open in the Hilbert scheme, and V ‾ g \overline{V}_g V g denotes its closure. For X ⊂ P g X\subset\mathbb{P}^g X ⊂ P g we denote by C ( X ) ⊂ P g + 1 C(X)\subset\mathbb{P}^{g+1} C ( X ) ⊂ P g + 1 the cone over X X X .
Corollary 4.5. Let X ⊂ P g X\subset\mathbb{P}^g X ⊂ P g be a general prime K3 surface of genus g ≥ 6 g\ge6 g ≥ 6 . Then the tangent space to the Hilbert scheme of P g + 1 \mathbb{P}^{g+1} P g + 1 at [ C ( X ) ] [C(X)] [ C ( X )] has dimension
dim T [ C ( X ) ] = h 0 ( N X ) + h 0 ( N X ( − 1 ) ) + h 0 ( N X ( − 2 ) ) = 18 + ( g + 1 ) 2 + ( g + 1 ) + α ( X ) + h 0 ( N X ( − 2 ) ) , \dim T_{[C(X)]}=h^0(N_X)+h^0(N_X(-1))+h^0(N_X(-2))=18+(g+1)^2+(g+1)+\alpha(X)+h^0(N_X(-2)), dim T [ C ( X )] = h 0 ( N X ) + h 0 ( N X ( − 1 )) + h 0 ( N X ( − 2 )) = 18 + ( g + 1 ) 2 + ( g + 1 ) + α ( X ) + h 0 ( N X ( − 2 )) ,
where h 0 ( N X ( − 2 ) ) = 0 h^0(N_X(-2))=0 h 0 ( N X ( − 2 )) = 0 if g ≥ 7 g\ge7 g ≥ 7 and h 0 ( N X ( − 2 ) ) ≤ 1 h^0(N_X(-2))\le1 h 0 ( N X ( − 2 )) ≤ 1 if g = 6 g=6 g = 6 .
Proof. Since X X X is projectively normal and H 1 ( O X ( k ) ) = 0 H^1(\mathcal{O}_X(k))=0 H 1 ( O X ( k )) = 0 for all k k k , one has H 0 ( N C ( X ) / P g + 1 ) = ⨁ k ≥ 0 H 0 ( N X ( − k ) ) H^0(N_{C(X)/\mathbb{P}^{g+1}})=\bigoplus_{k\ge0}H^0(N_X(-k)) H 0 ( N C ( X ) / P g + 1 ) = ⨁ k ≥ 0 H 0 ( N X ( − k )) (see [13 §2] ). We have h 0 ( N X ) = 18 + ( g + 1 ) 2 h^0(N_X)=18+(g+1)^2 h 0 ( N X ) = 18 + ( g + 1 ) 2 (§ 5.1 ) and h 0 ( N X ( − 1 ) ) = ( g + 1 ) + α ( X ) h^0(N_X(-1))=(g+1)+\alpha(X) h 0 ( N X ( − 1 )) = ( g + 1 ) + α ( X ) by definition. Since Pic X = Z H \operatorname{Pic}X=\mathbb{Z}H Pic X = Z H and g ≥ 5 g\ge5 g ≥ 5 , the homogeneous ideal of X X X is generated by quadrics [32 ] ; hence N X = H o m ( I X / I X 2 , O X ) N_X=\mathcal{H}om(I_X/I_X^2,\mathcal{O}_X) N X = H o m ( I X / I X 2 , O X ) is a subsheaf of a direct sum of copies of O X ( 2 ) \mathcal{O}_X(2) O X ( 2 ) , and H 0 ( N X ( − k ) ) = 0 H^0(N_X(-k))=0 H 0 ( N X ( − k )) = 0 for k ≥ 3 k\ge3 k ≥ 3 . Finally h 0 ( N X ( − 2 ) ) ≤ h 0 ( N Y ~ ( − 2 H ~ ) ) h^0(N_X(-2))\le h^0(N_{\widetilde{Y}}(-2\widetilde{H})) h 0 ( N X ( − 2 )) ≤ h 0 ( N Y ( − 2 H )) by (44) , and the right-hand side is 0 for g ≥ 7 g\ge7 g ≥ 7 (then b ≥ e + 3 b\ge e+3 b ≥ e + 3 ) and at most 1 for g = 6 g=6 g = 6 (( e , b ) = ( 1 , 3 ) (e,b)=(1,3) ( e , b ) = ( 1 , 3 ) ) by Proposition 4.3. □
Lemma 4.6. Let W W W be an irreducible component of V ‾ g \overline{\mathcal V}_g V g and V V V a general element of W W W . Then a general hyperplane section of V V V corresponds to a general K3 surface in H g \mathcal H_g H g .
Proof. The proof follows from standard deformation theory arguments (see for example [14 Lemma 3.7] ). □
Theorem 4.7. Let g ≥ 6 g \ge6 g ≥ 6 .
V g = ∅ \mathcal V_g = \varnothing V g = ∅ if g = 11 g = 11 g = 11 or g ≥ 13 g \ge13 g ≥ 13 .
For 6 ≤ g ≤ 10 6 \le g \le10 6 ≤ g ≤ 10 and g = 12 g = 12 g = 12 , V g \mathcal V_g V g is irreducible, and the families of Fano and Iskovskikh [22 , 23 ] (see [31 ] and [14 Table 3.1] ) form an open dense subset of V g \mathcal V_g V g .
Proof. Proof of (1). If V g ≠ ∅ \mathcal V_g \ne\varnothing V g = ∅ , let V V V be a general element of an irreducible component of V ‾ g \overline{\mathcal V}_g V g . By Lemma 4.6 a general hyperplane section of V V V is a general prime K3 surface X X X of genus g g g , and V V V is not a cone; this contradicts Theorem 4.1(7).
Proof of (2). First one notes that H g \mathcal H_g H g is irreducible. Now the list of examples of Fano–Iskovskikh–Mori–Mukai (see for example [14 Table 3.1] ) shows that in each of the cases mentioned in part (2), there exist K3 surfaces in H g \mathcal H_g H g which are extendable to smooth Fano threefolds. Hence by Lemma 4.6 , in each of the cases, H g \mathcal H_g H g is dominated by V g \mathcal V_g V g . We end the proof by showing that V g \mathcal V_g V g has a unique irreducible component and that the examples of Fano–Iskovskikh–Mori–Mukai form a dense subset of V g \mathcal V_g V g . To see this, we observe first, as in [13 ] and [14 ] , that a projectively Cohen–Macaulay scheme degenerates along a flat family to the cone over its hyperplane section. Hence, once again using Lemma 4.6 , it is enough to show that for a general K3 surface X ⊂ P g X \subset\mathbb{P}^g X ⊂ P g in H g \mathcal H_g H g , the cone C ( X ) ⊂ P g + 1 C(X) \subset\mathbb{P}^{g+1} C ( X ) ⊂ P g + 1 , which lies on every irreducible component of V ‾ g \overline{\mathcal V}_g V g , is a smooth point of the Hilbert scheme. For the stated values of g g g , Corollary 4.5 along with Theorem 4.1 gives an upper bound to dim T [ C ( X ) ] \dim T_{[C(X)]} dim T [ C ( X )] , which is listed as follows:
Table 1.
Now for a smooth Fano threefold V ∈ V g V \in\mathcal V_g V ∈ V g , V V V represents a smooth point of the Hilbert scheme, since embedded deformations of an anticanonically embedded Fano threefold are unobstructed by the vanishing of h 1 ( N V / P g + 1 ) h^1(N_{V/\mathbb{P}^{g+1}}) h 1 ( N V / P g + 1 ) , which follows from the Kodaira–Nakano vanishing theorem. Therefore the dimension of V g \mathcal V_g V g at V V V is given by dim T [ V ] = h 0 ( N V / P g + 1 ) \dim T_{[V]} = h^0(N_{V/\mathbb{P}^{g+1}}) dim T [ V ] = h 0 ( N V / P g + 1 ) . Now using the descriptions of the examples of Fano–Iskovskikh–Mori–Mukai, one can calculate the dimension h 0 ( N V / P g + 1 ) h^0(N_{V/\mathbb{P}^{g+1}}) h 0 ( N V / P g + 1 ) (see for example [13 Table 2, p. 666] and [14 Table 3.1] ) to check that in each case dim T [ V ] \dim T_{[V]} dim T [ V ] is exactly the upper bound just obtained. Since C ( X ) C(X) C ( X ) lies on the irreducible component of V ‾ g \overline{\mathcal V}_g V g containing V V V , whose dimension is h 0 ( N V / P g + 1 ) h^0(N_{V/\mathbb{P}^{g+1}}) h 0 ( N V / P g + 1 ) , C ( X ) C(X) C ( X ) is also a smooth point of the Hilbert scheme. Hence V ‾ g \overline{\mathcal V}_g V g has a unique irreducible component. Finally, the examples of Fano–Iskovskikh–Mori–Mukai form an irreducible family (in each case they are the linear sections of a fixed homogeneous variety, see [31 ] ) of dimension h 0 ( N V / P g + 1 = dim V g h^0(N_{V/\mathbb{P}^{g+1}} = \dim\mathcal V_g h 0 ( N V / P g + 1 = dim V g , hence a dense subset of V g \mathcal V_g V g . □
Corollary 4.8. Let X ⊂ P g X \subset\mathbb{P}^g X ⊂ P g be a general prime K3 surface of genus g g g . Then the upper bounds of Theorem 4.1 and Corollary 4.5 are equalities, namely
If g = 12 g = 12 g = 12 , then α ( X ) = 1 \alpha(X) = 1 α ( X ) = 1 .
If g = 10 g = 10 g = 10 , then α ( X ) = 3 \alpha(X) = 3 α ( X ) = 3 .
If g = 9 g = 9 g = 9 , then α ( X ) = 4 \alpha(X) = 4 α ( X ) = 4 .
If g = 8 g = 8 g = 8 , then α ( X ) = 6 \alpha(X) = 6 α ( X ) = 6 .
If g = 7 g = 7 g = 7 , then α ( X ) = 8 \alpha(X) = 8 α ( X ) = 8 .
If g = 6 g = 6 g = 6 , then α ( X ) = 10 \alpha(X) = 10 α ( X ) = 10 and h 0 ( N X ( − 2 ) ) = 1 h^0(N_{X(-2)}) = 1 h 0 ( N X ( − 2 ) ) = 1 .
Proof. By Corollary 4.5, dim T [ C ( X ) ] = 18 + ( g + 1 ) 2 + ( g + 1 ) + α ( X ) + h 0 ( N X ( − 2 ) ) \dim T_{[C(X)]} = 18+(g+1)^2+(g+1)+\alpha(X)+h^0(N_{X(-2)}) dim T [ C ( X )] = 18 + ( g + 1 ) 2 + ( g + 1 ) + α ( X ) + h 0 ( N X ( − 2 ) ) , and α ( X ) ≤ α ( Y ~ ) \alpha(X) \le\alpha(\widetilde{Y}) α ( X ) ≤ α ( Y ) and h 0 ( N X ( − 2 ) ) ≤ h 0 ( N Y ~ ( − 2 H ~ ) ) h^0(N_{X(-2)}) \le h^0(N_{\widetilde{Y}(-2\widetilde{H})}) h 0 ( N X ( − 2 ) ) ≤ h 0 ( N Y ( − 2 H ) ) by (44) . Hence dim T [ C ( X ) ] ≤ 18 + ( g + 1 ) 2 + ( g + 1 ) + α ( Y ~ ) + \dim T_{[C(X)]} \le18+(g+1)^2+(g+1)+\alpha(\widetilde{Y})+ dim T [ C ( X )] ≤ 18 + ( g + 1 ) 2 + ( g + 1 ) + α ( Y ) + h 0 ( N ~ Y ~ ( − 2 H ~ ) ) = h 0 ( N V / P g + 1 ) h^{0}(\widetilde{N}_{\widetilde{Y}}(-2\widetilde{H})) = h^{0}(N_{V/\mathbb{P}^{g+1}}) h 0 ( N Y ( − 2 H )) = h 0 ( N V / P g + 1 ) by the proof of part (2) of Theorem 4.7 . But a smooth V V V is a smooth point of the Hilbert scheme and hence α ( X ) = α ( Y ~ ) \alpha(X) = \alpha(\widetilde{Y}) α ( X ) = α ( Y ) and h 0 ( N X ( − 2 ) ) = h 0 ( N ~ Y ~ ( − 2 H ~ ) ) h^{0}(N_{X}(-2)) = h^{0}(\widetilde{N}_{\widetilde{Y}}(-2\widetilde{H})) h 0 ( N X ( − 2 )) = h 0 ( N Y ( − 2 H )) . Now the conclusion follows from Theorem 3.2, Proposition 4.2 and Proposition 4.3 . □
Appendix A. Extensions of ribbons to K K K -trivial carpets
by Anand Deopurkar and Jayan Mukherjee
In this section we prove Theorem A.3 , that is used to show that Γ U \Gamma_{U} Γ U is a section to Ψ \Psi Ψ over U U U in Proposition 3.17 .
Set-up A.1. Let C ~ \widetilde{C} C be a ribbon of arithmetic genus g g g and conormal bundle L C L_{C} L C over a smooth curve C C C of genus g C g_{C} g C . Then C ~ \widetilde{C} C is defined by an extension
0 → L C → V → K C → 0 0 \to L_{C} \to V \to K_{C} \to0 0 → L C → V → K C → 0 where V V V is a rank two vector bundle on C C C . Further V = Ω C ~ ∣ C V = \Omega_{\widetilde{C}}|_{C} V = Ω C ∣ C . Now the data of the extension canonically gives
a ruled surface Y = P ( V ) → C Y = \mathbb{P}(V) \to C Y = P ( V ) → C and
an embedding C ↪ Y = P ( V ) C \hookrightarrow Y = \mathbb{P}(V) C ↪ Y = P ( V ) as a section of the ruled surface Y Y Y defined by the surjection V → K C V \to K_{C} V → K C .
Note that deg L C = 2 g C − 1 − g \deg L_{C} = 2g_{C} - 1 - g deg L C = 2 g C − 1 − g , since χ ( O C ~ ) = χ ( O C ) + χ ( L C ) \chi(\mathcal{O}_{\widetilde{C}}) = \chi(\mathcal{O}_{C}) + \chi(L_{C}) χ ( O C ) = χ ( O C ) + χ ( L C ) .
Lemma A.2. In Set-up A.1, the line bundle O Y ( C ) = O P ( V ) ( 1 ) ⊗ p ∗ L C − 1 \mathcal{O}_{Y}(C) = \mathcal{O}_{\mathbb{P}(V)}(1) \otimes p^{*}L_{C}^{-1} O Y ( C ) = O P ( V ) ( 1 ) ⊗ p ∗ L C − 1 .
Proof. Note that O P ( V ) ( 1 ) ∣ C = K C \mathcal{O}_{\mathbb{P}(V)}(1)|_{C} = K_{C} O P ( V ) ( 1 ) ∣ C = K C . Further since C C C is a section, we have that O Y ( C ) = O P ( V ) ( 1 ) ⊗ p ∗ M \mathcal{O}_{Y}(C) = \mathcal{O}_{\mathbb{P}(V)}(1) \otimes p^{*}M O Y ( C ) = O P ( V ) ( 1 ) ⊗ p ∗ M for some line bundle M M M on C C C . From the exact sequence
0 → O P ( V ) ( 1 ) ⊗ O P ( V ) ( − C ) → O P ( V ) ( 1 ) → O P ( V ) ( 1 ) ∣ C = K C → 0 0 \to\mathcal{O}_{\mathbb{P}(V)}(1) \otimes\mathcal{O}_{\mathbb{P}(V)}(-C) \to\mathcal{O}_{\mathbb{P}(V)}(1) \to\mathcal{O}_{\mathbb{P}(V)}(1)|_{C} = K_{C} \to0 0 → O P ( V ) ( 1 ) ⊗ O P ( V ) ( − C ) → O P ( V ) ( 1 ) → O P ( V ) ( 1 ) ∣ C = K C → 0
we get
0 → p ∗ M − 1 → O P ( V ) ( 1 ) → O P ( V ) ( 1 ) ∣ C = K C → 0 0 \to p^{*}M^{-1} \to\mathcal{O}_{\mathbb{P}(V)}(1) \to\mathcal{O}_{\mathbb{P}(V)}(1)|_{C} = K_{C} \to0 0 → p ∗ M − 1 → O P ( V ) ( 1 ) → O P ( V ) ( 1 ) ∣ C = K C → 0
Pushing forward, we have
0 → M − 1 → V → K C → 0 0 \to M^{-1} \to V \to K_{C} \to0 0 → M − 1 → V → K C → 0
This gives, M − 1 = L C M^{-1} = L_{C} M − 1 = L C . Hence O Y ( C ) = O P ( V ) ( 1 ) ⊗ p ∗ L C − 1 \mathcal{O}_{Y}(C) = \mathcal{O}_{\mathbb{P}(V)}(1) \otimes p^{*}L_{C}^{-1} O Y ( C ) = O P ( V ) ( 1 ) ⊗ p ∗ L C − 1 . □
Theorem A.3. In Set-up A.1, assume that C ~ \widetilde{C} C is non-split. For y ~ ∈ Ext 1 ( Ω Y , K Y ) \widetilde{y} \in\operatorname{Ext}^{1}(\Omega_{Y}, K_{Y}) y ∈ Ext 1 ( Ω Y , K Y ) let Y ~ y ~ \widetilde{Y}_{\widetilde{y}} Y y be the K K K -trivial carpet on Y Y Y of class y ~ \widetilde{y} y .
(1) There is a unique y ~ 0 ∈ Ext 1 ( Ω Y , K Y ) \widetilde{y}_{0} \in\operatorname{Ext}^{1}(\Omega_{Y}, K_{Y}) y 0 ∈ Ext 1 ( Ω Y , K Y ) , y ~ 0 ≠ 0 \widetilde{y}_{0} \ne0 y 0 = 0 , such that the embedding i : C ↪ Y i: C \hookrightarrow Y i : C ↪ Y extends to a closed embedding C ~ ↪ Y ~ y ~ 0 \widetilde{C} \hookrightarrow\widetilde{Y}_{\widetilde{y}_{0}} C ↪ Y y 0 inducing the identity K Y ∣ C = L C K_{Y}|_{C} = L_{C} K Y ∣ C = L C on conormal bundles.
(2) Let g C = 0 g_{C} = 0 g C = 0 and let C ~ ⊂ P M − 1 \widetilde{C} \subset\mathbb{P}^{M-1} C ⊂ P M − 1 be a canonically embedded ribbon, M = g M = g M = g , so that C ⊂ P M − 1 C \subset\mathbb{P}^{M-1} C ⊂ P M − 1 is a rational normal curve. Embed Y Y Y in P M \mathbb{P}^{M} P M by the complete linear series of O Y ( C ) \mathcal{O}_{Y}(C) O Y ( C ) and identify P M − 1 \mathbb{P}^{M-1} P M − 1 with a hyperplane H ⊂ P M H \subset\mathbb{P}^{M} H ⊂ P M so that Y ∩ H = C Y \cap H = C Y ∩ H = C . If Y ~ ⊂ P M \widetilde{Y} \subset\mathbb{P}^{M} Y ⊂ P M is the K 3 K3 K 3 carpet on Y Y Y , then
Y ~ ∩ H = C ~ \widetilde{Y} \cap H = \widetilde{C} Y ∩ H = C in P M − 1 = H \mathbb{P}^{M-1} = H P M − 1 = H .
Proof. Proof of (1). Let C C C be a smooth curve and L C L_{C} L C a line bundle on C C C . Fix a ribbon C ~ \widetilde{C} C on C C C with conormal bundle L C L_{C} L C . Then C ~ \widetilde{C} C defines an element [ C ~ ] ∈ Ext 1 ( Ω C , L C ) [\widetilde{C}] \in\operatorname{Ext}^{1}(\Omega_{C}, L_{C}) [ C ] ∈ Ext 1 ( Ω C , L C ) . Set V = Ω C ~ ∣ C V = \Omega_{\widetilde{C}}|_{C} V = Ω C ∣ C . Then we have the exact sequence
0 → L C → V → Ω C → 0 0 \to L_{C} \to V \to\Omega_{C} \to0 0 → L C → V → Ω C → 0
Let Y = P V Y = \mathbb{P}V Y = P V . The exact sequence above gives an embedding
i : C ↪ Y i: C \hookrightarrow Y i : C ↪ Y
which is a section of the projection p : Y → C p: Y \to C p : Y → C . We have
K Y = O P ( V ) ( − 2 ) ⊗ p ∗ ( K C ⊗ det V ) K_Y = \mathcal{O}_{\mathbb{P}(V)}(-2) \otimes p^*(K_C \otimes\det V) K Y = O P ( V ) ( − 2 ) ⊗ p ∗ ( K C ⊗ det V )
and hence
K Y ∣ C = Ω C − 2 ⊗ ( Ω C ⊗ L C ) ⊗ K C = L C . K_Y|_C = \Omega_C^{-2} \otimes(\Omega_C \otimes L_C) \otimes K_C = L_C. K Y ∣ C = Ω C − 2 ⊗ ( Ω C ⊗ L C ) ⊗ K C = L C .
A K K K -trivial carpet Y ~ \widetilde{Y} Y on Y Y Y is given by an element y ~ ∈ Ext 1 ( Ω Y , K Y ) \widetilde{y} \in\operatorname{Ext}^1(\Omega_Y,K_Y) y ∈ Ext 1 ( Ω Y , K Y ) . To show the existence of a K K K -trivial carpet as claimed in part (1), we want to choose Y ~ = Y ~ y ~ \widetilde{Y}=\widetilde{Y}_{\widetilde{y}} Y = Y y so that the embedding
C → i Y \begin{array}{c}
C \xrightarrow{i} Y
\end{array} C i Y
extends to
C ~ → i ~ Y ~ \begin{array}{c}
\widetilde{C} \xrightarrow{\widetilde{i}} \widetilde{Y}
\end{array} C i Y
inducing the identity on conormal bundles. This is equivalent to the equality i ~ [ C ~ ] = r ( y ~ ) \widetilde{i}[\widetilde{C}]=r(\widetilde{y}) i [ C ] = r ( y ) in the following diagram:
Ext 1 ( Ω Y , K Y ) ↓ r Ext 1 ( Ω C , L C ) → i ‾ Ext 1 ( Ω Y ∣ C , L C ) (59) \begin{array}{c}
\operatorname{Ext}^1(\Omega_Y,K_Y)\\
\downarrow r\\
\operatorname{Ext}^1(\Omega_C,L_C) \xrightarrow{\overline{i}} \operatorname{Ext}^1(\Omega_Y|_C,L_C) \tag*{(59)}
\end{array} Ext 1 ( Ω Y , K Y ) ↓ r Ext 1 ( Ω C , L C ) i Ext 1 ( Ω Y ∣ C , L C ) (59)
where r r r is the map induced by the restriction of the short exact sequence 0 → K Y → Ω Y ~ ∣ Y → Ω Y → 0 0 \to K_Y \to\Omega_{\widetilde{Y}}|_Y \to\Omega_Y \to0 0 → K Y → Ω Y ∣ Y → Ω Y → 0 to C C C and i ‾ : Ext 1 ( Ω C , L C ) → Ext 1 ( Ω Y ∣ C , L C ) \overline{i}: \operatorname{Ext}^1(\Omega_C,L_C) \to\operatorname{Ext}^1(\Omega_Y|_C,L_C) i : Ext 1 ( Ω C , L C ) → Ext 1 ( Ω Y ∣ C , L C ) is the map induced by the embedding i : C ↪ Y i:C\hookrightarrow Y i : C ↪ Y .
Now the projection p : Y → C p:Y\to C p : Y → C induces a map p ∗ Ω C → Ω Y p^*\Omega_C\to\Omega_Y p ∗ Ω C → Ω Y which when restricted to the section i : C ↪ Y i:C\hookrightarrow Y i : C ↪ Y induces a map Ω C → Ω Y ∣ C \Omega_C\to\Omega_Y|_C Ω C → Ω Y ∣ C . This further induces the map p ‾ : Ext 1 ( Ω Y ∣ C , L C ) → Ext 1 ( Ω C , L C ) \overline{p}: \operatorname{Ext}^1(\Omega_Y|_C,L_C)\to\operatorname{Ext}^1(\Omega_C,L_C) p : Ext 1 ( Ω Y ∣ C , L C ) → Ext 1 ( Ω C , L C ) which is a left inverse to the map i ‾ \overline{i} i : p ‾ ∘ i ‾ = id \overline{p}\circ\overline{i}=\operatorname{id} p ∘ i = id . Thus the conormal sequence 0 → N C / Y ∨ → Ω Y ∣ C → Ω C → 0 0\to N^\vee_{C/Y}\to\Omega_Y|_C\to\Omega_C\to0 0 → N C / Y ∨ → Ω Y ∣ C → Ω C → 0 is split by Ω C → Ω Y ∣ C \Omega_C\to\Omega_Y|_C Ω C → Ω Y ∣ C , and
Ext 1 ( Ω Y ∣ C , L C ) = Ext 1 ( N C / Y ∨ , L C ) ⊕ Ext 1 ( Ω C , L C ) . (60) \operatorname{Ext}^1(\Omega_Y|_C,L_C)=\operatorname{Ext}^1(N^\vee_{C/Y},L_C)\oplus\operatorname{Ext}^1(\Omega_C,L_C). \tag*{(60)} Ext 1 ( Ω Y ∣ C , L C ) = Ext 1 ( N C / Y ∨ , L C ) ⊕ Ext 1 ( Ω C , L C ) . (60)
where i ‾ \overline{i} i is the inclusion of the second summand, p ‾ \overline{p} p the projection onto it, and the projection onto the first summand is the map
r N : Ext 1 ( Ω Y ∣ C , L C ) → Ext 1 ( N C / Y ∨ , L C ) r_N:\operatorname{Ext}^1(\Omega_Y|_C,L_C)\to\operatorname{Ext}^1(N^\vee_{C/Y},L_C) r N : Ext 1 ( Ω Y ∣ C , L C ) → Ext 1 ( N C / Y ∨ , L C )
induced by N C / Y ∨ ↪ Ω Y ∣ C N^\vee_{C/Y}\hookrightarrow\Omega_Y|_C N C / Y ∨ ↪ Ω Y ∣ C . So it suffices to prove that there is a unique y ~ 0 ∈ Ext 1 ( Ω Y , K Y ) \widetilde{y}_0\in\operatorname{Ext}^1(\Omega_Y,K_Y) y 0 ∈ Ext 1 ( Ω Y , K Y ) satisfying the two conditions
[ C ~ ] = p ‾ r ( y ~ 0 ) , (a) r N r ( y ~ 0 ) = 0. (b) (61) \begin{aligned}
[\widetilde{C}]&=\overline{p}r(\widetilde{y}_0), &&\text{(a)}\\
r_Nr(\widetilde{y}_0)&=0. &&\text{(b)} \tag*{(61)}
\end{aligned} [ C ] r N r ( y 0 ) = p r ( y 0 ) , = 0. (a) (b) (61)
Condition (a). We have to describe the y ~ \widetilde{y} y for which [ C ~ ] [\widetilde{C}] [ C ] is the image of y ~ \widetilde{y} y under the map
Ext 1 ( Ω Y , K Y ) → r Ext 1 ( Ω Y ∣ C , L C ) → p ‾ Ext 1 ( Ω C , L C ) \operatorname{Ext}^1(\Omega_Y,K_Y)\xrightarrow{r}\operatorname{Ext}^1(\Omega_Y|_C,L_C)\xrightarrow{\overline{p}}\operatorname{Ext}^1(\Omega_C,L_C) Ext 1 ( Ω Y , K Y ) r Ext 1 ( Ω Y ∣ C , L C ) p Ext 1 ( Ω C , L C )
We have a diagram as follows:
Ext 1 ( Ω Y / C , K Y ) ⟶ Ext 1 ( Ω Y , K Y ) ⟶ Ext 1 ( p ∗ Ω C , K Y ) = k ↓ ↓ r ↓ Ext 1 ( Ω Y / C ∣ C , L C ) ⟵ Ext 1 ( Ω Y ∣ C , L C ) → p ‾ Ext 1 ( Ω C , L C ) (62) \begin{array}{ccccc}
\operatorname{Ext}^1(\Omega_{Y/C},K_Y)&\longrightarrow&\operatorname{Ext}^1(\Omega_Y,K_Y)&\longrightarrow&\operatorname{Ext}^1(p^*\Omega_C,K_Y)=k\\
\downarrow&&\downarrow r&&\downarrow\\
\operatorname{Ext}^1(\Omega_{Y/C}|_C,L_C)&\longleftarrow&\operatorname{Ext}^1(\Omega_Y|_C,L_C)&\xrightarrow{\overline{p}}&\operatorname{Ext}^1(\Omega_C,L_C) \tag*{(62)}
\end{array} Ext 1 ( Ω Y / C , K Y ) ↓ Ext 1 ( Ω Y / C ∣ C , L C ) ⟶ ⟵ Ext 1 ( Ω Y , K Y ) ↓ r Ext 1 ( Ω Y ∣ C , L C ) ⟶ p Ext 1 ( p ∗ Ω C , K Y ) = k ↓ Ext 1 ( Ω C , L C ) (62)
whose vertical maps are restriction to C C C . Hence the map the map
Ext 1 ( Ω Y , K Y ) → r Ext 1 ( Ω Y ∣ C , L C ) → p ‾ Ext 1 ( Ω C , L C ) \operatorname{Ext}^1(\Omega_Y,K_Y)\xrightarrow{r}\operatorname{Ext}^1(\Omega_Y|_C,L_C)\xrightarrow{\overline{p}}\operatorname{Ext}^1(\Omega_C,L_C) Ext 1 ( Ω Y , K Y ) r Ext 1 ( Ω Y ∣ C , L C ) p Ext 1 ( Ω C , L C )
also factors as
Ext 1 ( Ω Y , K Y ) → Ext 1 ( p ∗ Ω C , K Y ) → Ext 1 ( Ω C , L C ) . \operatorname{Ext}^{1}(\Omega_Y,K_Y)\to\operatorname{Ext}^{1}(p^{*}\Omega_C,K_Y)\to\operatorname{Ext}^{1}(\Omega_C,L_C). Ext 1 ( Ω Y , K Y ) → Ext 1 ( p ∗ Ω C , K Y ) → Ext 1 ( Ω C , L C ) .
We first show that the map Ext 1 ( Ω Y , K Y ) → Ext 1 ( p ∗ Ω C , K Y ) \operatorname{Ext}^{1}(\Omega_Y,K_Y)\to\operatorname{Ext}^{1}(p^{*}\Omega_C,K_Y) Ext 1 ( Ω Y , K Y ) → Ext 1 ( p ∗ Ω C , K Y ) is surjective. Considering the exact sequence
0 → p ∗ Ω C → Ω Y → Ω Y / C → 0 , 0\to p^{*}\Omega_C\to\Omega_Y\to\Omega_{Y/C}\to0, 0 → p ∗ Ω C → Ω Y → Ω Y / C → 0 ,
it is enough to show that Ext 2 ( Ω Y / C , K Y ) = 0 \operatorname{Ext}^{2}(\Omega_{Y/C},K_Y)=0 Ext 2 ( Ω Y / C , K Y ) = 0 . By Serre duality,
Ext 2 ( Ω Y / C , K Y ) = Hom ( O Y , Ω Y / C ) ∨ = 0 \operatorname{Ext}^{2}(\Omega_{Y/C},K_Y)=\operatorname{Hom}(\mathcal{O}_Y,\Omega_{Y/C})^{\vee}=0 Ext 2 ( Ω Y / C , K Y ) = Hom ( O Y , Ω Y / C ) ∨ = 0
since Ω Y / C \Omega_{Y/C} Ω Y / C is O P ( V ) ( − 2 ) ⊗ p ∗ ( det V ) \mathcal{O}_{\mathbb{P}(V)}(-2)\otimes p^{*}(\det V) O P ( V ) ( − 2 ) ⊗ p ∗ ( det V ) . First note that since p ∗ K Y = 0 p_*K_Y=0 p ∗ K Y = 0 , we have
Ext 1 ( p ∗ Ω C , K Y ) = H 1 ( K Y ⊗ p ∗ T C ) = H 0 ( R 1 p ∗ K Y ⊗ T C ) = H 0 ( O C ) , \operatorname{Ext}^{1}(p^{*}\Omega_C,K_Y)=H^{1}(K_Y\otimes p^{*}T_C)=H^{0}(R^{1}p_*K_Y\otimes T_C)=H^{0}(\mathcal{O}_C), Ext 1 ( p ∗ Ω C , K Y ) = H 1 ( K Y ⊗ p ∗ T C ) = H 0 ( R 1 p ∗ K Y ⊗ T C ) = H 0 ( O C ) ,
because
R 1 p ∗ K Y = R 1 p ∗ ( K Y / C ⊗ p ∗ K C ) = K C . R^{1}p_*K_Y=R^{1}p_*(K_{Y/C}\otimes p^{*}K_C)=K_C. R 1 p ∗ K Y = R 1 p ∗ ( K Y / C ⊗ p ∗ K C ) = K C .
So Ext 1 ( p ∗ Ω C , K Y ) = k \operatorname{Ext}^{1}(p^{*}\Omega_C,K_Y)=k Ext 1 ( p ∗ Ω C , K Y ) = k . We claim that the image of 1 ∈ Ext 1 ( p ∗ Ω C , K Y ) = H 1 ( Ω Y / C ) 1\in\operatorname{Ext}^{1}(p^{*}\Omega_C,K_Y)=H^{1}(\Omega_{Y/C}) 1 ∈ Ext 1 ( p ∗ Ω C , K Y ) = H 1 ( Ω Y / C ) (note that K Y ⊗ p ∗ T C = Ω Y / C K_Y\otimes p^{*}T_C=\Omega_{Y/C} K Y ⊗ p ∗ T C = Ω Y / C ) in Ext 1 ( Ω C , L C ) = H 1 ( T C ⊗ L C ) \operatorname{Ext}^{1}(\Omega_C,L_C)=H^{1}(T_C\otimes L_C) Ext 1 ( Ω C , L C ) = H 1 ( T C ⊗ L C ) is [ C ~ ] [\widetilde{C}] [ C ] . This follows from the following general claim.
Claim A.4. Let 0 → L → V → A → 0 0\to L\to V\to A\to0 0 → L → V → A → 0 be an extension where L L L and A A A are line bundles on C C C . Consider the section i : C ↪ P ( V ) i:C\hookrightarrow\mathbb{P}(V) i : C ↪ P ( V ) defined by the extension. This induces a map Ω P ( V ) / C → Ω P ( V ) / C ∣ C \Omega_{\mathbb{P}(V)/C}\to\Omega_{\mathbb{P}(V)/C}|_C Ω P ( V ) / C → Ω P ( V ) / C ∣ C . Then the image of 1 ∈ H 1 ( Ω P ( V ) / C ) = k 1\in H^{1}(\Omega_{\mathbb{P}(V)/C})=k 1 ∈ H 1 ( Ω P ( V ) / C ) = k inside H 1 ( Ω P ( V ) / C ∣ C ) = H 1 ( L ⊗ A − 1 ) H^{1}(\Omega_{\mathbb{P}(V)/C}|_C)=H^{1}(L\otimes A^{-1}) H 1 ( Ω P ( V ) / C ∣ C ) = H 1 ( L ⊗ A − 1 ) is the class of the extension.
Proof of the claim. 1 ∈ H 1 ( Ω P V / C ) 1\in H^{1}(\Omega_{\mathbb{P}V/C}) 1 ∈ H 1 ( Ω P V / C ) is represented by the Euler sequence
0 → Ω P V / C → p ∗ V ⊗ O P V ( − 1 ) → O P V → 0. 0\to\Omega_{\mathbb{P}V/C}\to p^{*}V\otimes\mathcal{O}_{\mathbb{P}V}(-1)\to\mathcal{O}_{\mathbb{P}V}\to0. 0 → Ω P V / C → p ∗ V ⊗ O P V ( − 1 ) → O P V → 0.
Restricted to C C C via i i i , it is
0 → L ⊗ A − 1 → V ⊗ A − 1 → O C → 0 , 0\to L\otimes A^{-1}\to V\otimes A^{-1}\to\mathcal{O}_C\to0, 0 → L ⊗ A − 1 → V ⊗ A − 1 → O C → 0 ,
which is the class of the extension twisted by A − 1 A^{-1} A − 1 . This proves the claim.
For y ~ ∈ Ext 1 ( Ω Y , K Y ) \widetilde{y}\in\operatorname{Ext}^{1}(\Omega_Y,K_Y) y ∈ Ext 1 ( Ω Y , K Y ) let y ~ h ∈ Ext 1 ( p ∗ Ω C , K Y ) = k \widetilde{y}_h\in\operatorname{Ext}^{1}(p^{*}\Omega_C,K_Y)=k y h ∈ Ext 1 ( p ∗ Ω C , K Y ) = k be its image. By the claim,
p r ‾ ( y ~ ) = y ~ h [ C ~ ] , (63) \overline{p r}(\widetilde{y})=\widetilde{y}_h[\widetilde{C}], \tag*{(63)} p r ( y ) = y h [ C ] , (63)
and since C ~ \widetilde{C} C is non-split, [ C ~ ] ≠ 0 [\widetilde{C}]\ne0 [ C ] = 0 ; so condition (a) holds for y ~ \widetilde{y} y if and only if y ~ h = 1 \widetilde{y}_h=1 y h = 1 . The kernel of the map Ext 1 ( Ω Y , K Y ) → Ext 1 ( p ∗ Ω C , K Y ) \operatorname{Ext}^{1}(\Omega_Y,K_Y)\to\operatorname{Ext}^{1}(p^{*}\Omega_C,K_Y) Ext 1 ( Ω Y , K Y ) → Ext 1 ( p ∗ Ω C , K Y ) is
Ext 1 ( Ω Y / C , K Y ) = H 1 ( T Y / C ⊗ K Y ) = H 1 ( p ∗ K C ) = H 1 ( K C ) = k . \operatorname{Ext}^{1}(\Omega_{Y/C},K_Y)=H^{1}(T_{Y/C}\otimes K_Y)=H^{1}(p^{*}K_C)=H^{1}(K_C)=k. Ext 1 ( Ω Y / C , K Y ) = H 1 ( T Y / C ⊗ K Y ) = H 1 ( p ∗ K C ) = H 1 ( K C ) = k .
Hence the set of y ~ \widetilde{y} y satisfying condition (a) is the affine line
B : = { y ~ ∈ Ext 1 ( Ω Y , K Y ) : y ~ h = 1 } , B:=\{\widetilde{y}\in\operatorname{Ext}^{1}(\Omega_Y,K_Y):\widetilde{y}_h=1\}, B := { y ∈ Ext 1 ( Ω Y , K Y ) : y h = 1 } ,
the inverse image of 1 ∈ Ext 1 ( p ∗ Ω C , K Y ) 1\in\operatorname{Ext}^{1}(p^{*}\Omega_C,K_Y) 1 ∈ Ext 1 ( p ∗ Ω C , K Y ) .
Condition (b). Consider again the diagram:
Ext 1 ( Ω Y / C , K Y ) ⟶ Ext 1 ( Ω Y , K Y ) ⟶ Ext 1 ( p ∗ Ω C , K Y ) = k ↓ ↓ r ↓ Ext 1 ( Ω Y / C ∣ C , L C ) ⟵ Ext 1 ( Ω Y ∣ C , L C ) → p ‾ Ext 1 ( Ω C , L C ) (64) \begin{array}{ccccc}
\operatorname{Ext}^{1}(\Omega_{Y/C},K_Y)&\longrightarrow&\operatorname{Ext}^{1}(\Omega_Y,K_Y)&\longrightarrow&\operatorname{Ext}^{1}(p^{*}\Omega_C,K_Y)=k\\
\downarrow&&\downarrow r&&\downarrow\\
\operatorname{Ext}^{1}(\Omega_{Y/C}|_C,L_C)&\longleftarrow&\operatorname{Ext}^{1}(\Omega_Y|_C,L_C)&\xrightarrow{\overline{p}}&\operatorname{Ext}^{1}(\Omega_C,L_C) \tag*{(64)}
\end{array} Ext 1 ( Ω Y / C , K Y ) ↓ Ext 1 ( Ω Y / C ∣ C , L C ) ⟶ ⟵ Ext 1 ( Ω Y , K Y ) ↓ r Ext 1 ( Ω Y ∣ C , L C ) ⟶ p Ext 1 ( p ∗ Ω C , K Y ) = k ↓ Ext 1 ( Ω C , L C ) (64)
whose vertical maps are restriction to C C C . Let B ⊂ Ext 1 ( Ω Y , K Y ) B\subset\operatorname{Ext}^{1}(\Omega_Y,K_Y) B ⊂ Ext 1 ( Ω Y , K Y ) be the affine line above, and let K K K be the kernel of r N : Ext 1 ( Ω Y ∣ C , L C ) → Ext 1 ( N C / Y ∨ , L C ) = k r_N:\operatorname{Ext}^{1}(\Omega_Y|_C,L_C)\to\operatorname{Ext}^{1}(N^\vee_{C/Y},L_C)=k r N : Ext 1 ( Ω Y ∣ C , L C ) → Ext 1 ( N C / Y ∨ , L C ) = k . We show that r ( B ) ∩ K r(B)\cap K r ( B ) ∩ K consists of a single point, i.e. that r N ∘ r r_N\circ r r N ∘ r vanishes at exactly one point of $B.
Two distinct elements of B B B differ by a nonzero element of Ext 1 ( Ω Y / C , K Y ) \operatorname{Ext}^{1}(\Omega_{Y/C},K_Y) Ext 1 ( Ω Y / C , K Y ) . This is because
Hom ( p ∗ Ω C , K Y ) = H 0 ( K Y ⊗ p ∗ Ω C ∗ ) = 0. \operatorname{Hom}(p^{*}\Omega_C,K_Y)=H^{0}(K_Y\otimes p^{*}\Omega_C^{*})=0. Hom ( p ∗ Ω C , K Y ) = H 0 ( K Y ⊗ p ∗ Ω C ∗ ) = 0.
Now note that the conormal sequence
0 → N C / Y ∨ → Ω Y ∣ C → Ω C → 0 0\to N_{C/Y}^{\vee}\to\Omega_Y|_C\to\Omega_C\to0 0 → N C / Y ∨ → Ω Y ∣ C → Ω C → 0
has a splitting given by
0 → p ∗ Ω C ∣ C = Ω C → Ω Y ∣ C → Ω Y / C ∣ C → 0 , 0\to p^{*}\Omega_C|_C=\Omega_C\to\Omega_Y|_C\to\Omega_{Y/C}|_C\to0, 0 → p ∗ Ω C ∣ C = Ω C → Ω Y ∣ C → Ω Y / C ∣ C → 0 ,
the restriction to C C C of 0 → p ∗ Ω C → Ω Y → Ω Y / C → 0 0\to p^{*}\Omega_C\to\Omega_Y\to\Omega_{Y/C}\to0 0 → p ∗ Ω C → Ω Y → Ω Y / C → 0 . Therefore Ω Y / C ∣ C ≅ N C / Y ∨ \Omega_{Y/C}|_C\cong N_{C/Y}^{\vee} Ω Y / C ∣ C ≅ N C / Y ∨ , and N C / Y ∨ ↪ Ω Y ∣ C → Ω Y / C ∣ C N_{C/Y}^{\vee}\hookrightarrow\Omega_Y|_C\to\Omega_{Y/C}|_C N C / Y ∨ ↪ Ω Y ∣ C → Ω Y / C ∣ C is the identity N C / Y ∨ → N C / Y ∨ N_{C/Y}^{\vee}\to N_{C/Y}^{\vee} N C / Y ∨ → N C / Y ∨ . Hence, for w ∈ Ext 1 ( Ω Y / C , K Y ) w\in\operatorname{Ext}^{1}(\Omega_{Y/C},K_Y) w ∈ Ext 1 ( Ω Y / C , K Y ) , r N r ( w ) r_Nr(w) r N r ( w ) is the image of w w w under the left vertical map of (62) ,
Ext 1 ( Ω Y / C , K Y ) → Ext 1 ( Ω Y / C ∣ C , L C ) = Ext 1 ( N C / Y ∨ , L C ) , \operatorname{Ext}^{1}(\Omega_{Y/C},K_Y)\to\operatorname{Ext}^{1}(\Omega_{Y/C}|_C,L_C)=\operatorname{Ext}^{1}(N_{C/Y}^{\vee},L_C), Ext 1 ( Ω Y / C , K Y ) → Ext 1 ( Ω Y / C ∣ C , L C ) = Ext 1 ( N C / Y ∨ , L C ) ,
and it is enough to show that this map is injective. Tensoring
0 → O Y ( − C ) → O Y → O C → 0 0\to\mathcal{O}_Y(-C)\to\mathcal{O}_Y\to\mathcal{O}_C\to0 0 → O Y ( − C ) → O Y → O C → 0
with Ω Y / C ∗ ⊗ K Y \Omega_{Y/C}^{*}\otimes K_Y Ω Y / C ∗ ⊗ K Y we get
0 → Ω Y / C ∗ ⊗ K Y ⊗ O Y ( − C ) → Ω Y / C ∗ ⊗ K Y → Ω Y / C ∗ ∣ C ⊗ L C → 0. 0\to\Omega_{Y/C}^{*}\otimes K_Y\otimes\mathcal{O}_Y(-C)\to\Omega_{Y/C}^{*}\otimes K_Y\to\Omega_{Y/C}^{*}|_C\otimes L_C\to0. 0 → Ω Y / C ∗ ⊗ K Y ⊗ O Y ( − C ) → Ω Y / C ∗ ⊗ K Y → Ω Y / C ∗ ∣ C ⊗ L C → 0.
Since Ω Y / C = K Y ⊗ p ∗ K C ∗ \Omega_{Y/C}=K_Y\otimes p^{*}K_C^{*} Ω Y / C = K Y ⊗ p ∗ K C ∗ , we have Ω Y / C ∗ = K Y ∗ ⊗ p ∗ K C \Omega_{Y/C}^{*}=K_Y^{*}\otimes p^{*}K_C Ω Y / C ∗ = K Y ∗ ⊗ p ∗ K C , and hence, by Lemma A.2,
Ω Y / C ∗ ⊗ K Y ⊗ O Y ( − C ) = p ∗ K C ⊗ O Y ( − C ) = O P V ( − 1 ) ⊗ p ∗ ( K C ⊗ L C ) , \Omega_{Y/C}^{*}\otimes K_Y\otimes\mathcal{O}_Y(-C)=p^{*}K_C\otimes\mathcal{O}_Y(-C)=\mathcal{O}_{\mathbb{P}V}(-1)\otimes p^{*}(K_C\otimes L_C), Ω Y / C ∗ ⊗ K Y ⊗ O Y ( − C ) = p ∗ K C ⊗ O Y ( − C ) = O P V ( − 1 ) ⊗ p ∗ ( K C ⊗ L C ) ,
whose H 1 H^{1} H 1 vanishes because R j p ∗ O P V ( − 1 ) = 0 R^{j}p_{*}\mathcal{O}_{\mathbb{P}V}(-1)=0 R j p ∗ O P V ( − 1 ) = 0 for all j j j . Taking cohomology of the last exact sequence gives the injectivity of H 1 ( Ω Y / C ∗ ⊗ K Y ) → H 1 ( Ω Y / C ∗ ∣ C ⊗ L C ) H^{1}(\Omega_{Y/C}^{*}\otimes K_Y)\to H^{1}(\Omega_{Y/C}^{*}|_C\otimes L_C) H 1 ( Ω Y / C ∗ ⊗ K Y ) → H 1 ( Ω Y / C ∗ ∣ C ⊗ L C ) , i.e. of the left vertical map of (62) .
Therefore r N ∘ r r_N\circ r r N ∘ r is injective on B B B . Being an injective affine map from the line B B B to the line Ext 1 ( N C / Y ∨ , L C ) = k \operatorname{Ext}^{1}(N_{C/Y}^{\vee},L_C)=k Ext 1 ( N C / Y ∨ , L C ) = k , it is bijective, so exactly one point y ~ 0 \widetilde{y}_0 y 0 of B B B satisfies condition (b). This proves (1).
Proof of (2). Since C ~ \widetilde{C} C is non-split, V ⊗ L C − 1 ≅ O P 1 ( a ) ⊕ O P 1 ( a ′ ) V\otimes L_C^{-1}\cong\mathcal{O}_{\mathbb{P}^{1}}(a)\oplus\mathcal{O}_{\mathbb{P}^{1}}(a') V ⊗ L C − 1 ≅ O P 1 ( a ) ⊕ O P 1 ( a ′ ) with a , a ′ ≥ 1 a,a'\ge1 a , a ′ ≥ 1 (an extension 0 → O C → V ⊗ L C − 1 → K C ⊗ L C − 1 → 0 0\to\mathcal{O}_C\to V\otimes L_C^{-1}\to K_C\otimes L_C^{-1}\to0 0 → O C → V ⊗ L C − 1 → K C ⊗ L C − 1 → 0 with a = 0 a=0 a = 0 is split), so O Y ( C ) = O P ( V ⊗ L C − 1 ) ( 1 ) \mathcal{O}_Y(C)=\mathcal{O}_{\mathbb{P}(V\otimes L_C^{-1})}(1) O Y ( C ) = O P ( V ⊗ L C − 1 ) ( 1 ) is very ample; and h 0 ( O Y ( C ) ) = M + 1 h^{0}(\mathcal{O}_Y(C))=M+1 h 0 ( O Y ( C )) = M + 1 , since H 1 ( O Y ) = 0 H^{1}(\mathcal{O}_Y)=0 H 1 ( O Y ) = 0 and N C / Y = K C ⊗ L C − 1 = O P 1 ( g − 1 ) N_{C/Y}=K_C\otimes L_C^{-1}=\mathcal{O}_{\mathbb{P}^{1}}(g-1) N C / Y = K C ⊗ L C − 1 = O P 1 ( g − 1 ) .
Let Y ~ y ~ 0 \widetilde{Y}_{\widetilde{y}_0} Y y 0 and C ~ ⊂ Y ~ y ~ 0 \widetilde{C}\subset\widetilde{Y}_{\widetilde{y}_0} C ⊂ Y y 0 be as in (1). Then C ~ \widetilde{C} C is a Cartier divisor on Y ~ y ~ 0 \widetilde{Y}_{\widetilde{y}_0} Y y 0 with O Y ~ y ~ 0 ( C ~ ) ∣ Y = O Y ( C ) \mathcal{O}_{\widetilde{Y}_{\widetilde{y}_0}}(\widetilde{C})|_Y=\mathcal{O}_Y(C) O Y y 0 ( C ) ∣ Y = O Y ( C ) , and from
0 → K Y ( C ) → O Y ~ y ~ 0 ( C ~ ) → O Y ( C ) → 0 0\to K_Y(C)\to\mathcal{O}_{\widetilde{Y}_{\widetilde{y}_0}}(\widetilde{C})\to\mathcal{O}_Y(C)\to0 0 → K Y ( C ) → O Y y 0 ( C ) → O Y ( C ) → 0
and H i ( K Y ( C ) ) = H 2 − i ( O Y ( − C ) ) ∨ = 0 H^{i}(K_Y(C))=H^{2-i}(\mathcal{O}_Y(-C))^{\vee}=0 H i ( K Y ( C )) = H 2 − i ( O Y ( − C ) ) ∨ = 0 for all i i i (Lemma A.2: O Y ( − C ) = O P V ( − 1 ) ⊗ p ∗ L C \mathcal{O}_Y(-C)=\mathcal{O}_{\mathbb{P}V}(-1)\otimes p^{*}L_C O Y ( − C ) = O P V ( − 1 ) ⊗ p ∗ L C ) we get H 0 ( O Y ~ y ~ 0 ( C ~ ) ) ≅ H 0 ( O Y ( C ) ) H^{0}(\mathcal{O}_{\widetilde{Y}_{\widetilde{y}_0}}(\widetilde{C}))\cong H^{0}(\mathcal{O}_Y(C)) H 0 ( O Y y 0 ( C )) ≅ H 0 ( O Y ( C )) . Let
φ : Y ~ y ~ 0 → P 1 M : = P ( H 0 ( O Y ~ y ~ 0 ( C ~ ) ) ∨ ) \varphi:\widetilde{Y}_{\widetilde{y}_0}\to\mathbb{P}^{M}_{1}:=\mathbb{P}\left(H^{0}\left(\mathcal{O}_{\widetilde{Y}_{\widetilde{y}_0}}(\widetilde{C})\right)^{\vee}\right) φ : Y y 0 → P 1 M := P ( H 0 ( O Y y 0 ( C ) ) ∨ )
be the morphism defined by the complete linear series of O Y ~ y ~ 0 ( C ~ ) \mathcal{O}_{\widetilde{Y}_{\widetilde{y}_0}}(\widetilde{C}) O Y y 0 ( C ) . Its restriction to Y Y Y is the embedding by ∣ O Y ( C ) ∣ |\mathcal{O}_Y(C)| ∣ O Y ( C ) ∣ , and φ − 1 ( H 1 ) = C ~ \varphi^{-1}(H_1)=\widetilde{C} φ − 1 ( H 1 ) = C for the hyperplane H 1 H_1 H 1 defined by the section of O Y ~ y ~ 0 ( C ~ ) \mathcal{O}_{\widetilde{Y}_{\widetilde{y}_0}}(\widetilde{C}) O Y y 0 ( C ) vanishing on C ~ \widetilde{C} C .
We claim that φ \varphi φ is a closed embedding. It is finite (proper and injective on points), so it is enough to show that O P 1 M → φ ∗ O Y ~ y ~ 0 \mathcal{O}_{\mathbb{P}^{M}_{1}}\to\varphi_{*}\mathcal{O}_{\widetilde{Y}_{\widetilde{y}_0}} O P 1 M → φ ∗ O Y y 0 is surjective. Its image is a subsheaf of algebras mapping onto O Y = φ ∗ O Y ~ y ~ 0 / K Y \mathcal{O}_Y=\varphi_{*}\mathcal{O}_{\widetilde{Y}_{\widetilde{y}_0}}/K_Y O Y = φ ∗ O Y y 0 / K Y , so it is everything if and only if it contains K Y K_Y K Y , i.e. if and only if the map I Y → K Y I_Y\to K_Y I Y → K Y induced by O P 1 M → φ ∗ O Y ~ y ~ 0 \mathcal{O}_{\mathbb{P}^{M}_{1}}\to\varphi_{*}\mathcal{O}_{\widetilde{Y}_{\widetilde{y}_0}} O P 1 M → φ ∗ O Y y 0 is surjective. Since K Y 2 = 0 K_Y^{2}=0 K Y 2 = 0 this map factors through N Y = I Y / I Y 2 N_Y^{} = I_Y/I_Y^2 N Y = I Y / I Y 2 , so it is a section of N Y ⊗ K Y N_Y \otimes K_Y N Y ⊗ K Y . But H 0 ( N Y ⊗ K Y ) = k s H^0(N_Y \otimes K_Y)=k\,s H 0 ( N Y ⊗ K Y ) = k s with s s s nowhere vanishing by [20 Proposition 1.7] . Thus the map is λ s \lambda s λ s for some λ ∈ k \lambda\in k λ ∈ k , and it is surjective if λ ≠ 0 \lambda\ne0 λ = 0 . If λ = 0 \lambda=0 λ = 0 , then I Y I_Y I Y maps to 0 0 0 , so O P M → φ ∗ O Y ~ y ~ 0 \mathcal{O}_{\mathbb{P}^M} \to\varphi_*\mathcal{O}_{\widetilde{Y}_{\widetilde{y}_0}} O P M → φ ∗ O Y y 0 factors through O Y \mathcal{O}_Y O Y , i.e. φ \varphi φ factors through a retraction of Y ⊂ Y ~ y ~ 0 Y \subset\widetilde{Y}_{\widetilde{y}_0} Y ⊂ Y y 0 . But a retraction splits 0 → K Y → O Y ~ y ~ 0 → O Y → 0 0 \to K_Y \to\mathcal{O}_{\widetilde{Y}_{\widetilde{y}_0}} \to\mathcal{O}_Y \to0 0 → K Y → O Y y 0 → O Y → 0 as O Y \mathcal{O}_Y O Y -algebras, so Y ~ y ~ 0 \widetilde{Y}_{\widetilde{y}_0} Y y 0 would be the split carpet, contradicting y ~ 0 ≠ 0 \widetilde{y}_0 \ne0 y 0 = 0 . Hence φ \varphi φ is a closed embedding; φ ( Y ) ⊂ P M \varphi(Y) \subset\mathbb{P}^M φ ( Y ) ⊂ P M is embedded by ∣ O Y ( C ) ∣ |\mathcal{O}_Y(C)| ∣ O Y ( C ) ∣ , φ ( Y ~ y ~ 0 ) \varphi(\widetilde{Y}_{\widetilde{y}_0}) φ ( Y y 0 ) is a K3 carpet on it, and φ ( Y ~ y ~ 0 ) ∩ H 1 = φ ( C ~ ) \varphi(\widetilde{Y}_{\widetilde{y}_0}) \cap H_1 = \varphi(\widetilde{C}) φ ( Y y 0 ) ∩ H 1 = φ ( C ) .
It is clear that φ ( C ~ ) ⊂ H 1 = P g − 1 \varphi(\widetilde{C}) \subset H_1 = \mathbb{P}^{g-1} φ ( C ) ⊂ H 1 = P g − 1 is canonically embedded.
Now the given C ~ ⊂ P M − 1 \widetilde{C} \subset\mathbb{P}^{M-1} C ⊂ P M − 1 and φ ( C ~ ) ⊂ H 1 \varphi(\widetilde{C}) \subset H_1 φ ( C ) ⊂ H 1 are two embeddings of C ~ \widetilde{C} C by the complete linear series of ω C ~ \omega_{\widetilde{C}} ω C , so there is a projectivity T : H 1 → P M − 1 T:H_1 \to\mathbb{P}^{M-1} T : H 1 → P M − 1 with T ∘ φ ∣ C ~ T \circ\varphi|_{\widetilde{C}} T ∘ φ ∣ C the given inclusion; extend T T T to a projectivity T ^ : P M − 1 → P M \widehat{T}:\mathbb{P}^{M-1} \to\mathbb{P}^M T : P M − 1 → P M with T ^ ( H 1 ) = H \widehat{T}(H_1)=H T ( H 1 ) = H . Then T ^ ∘ φ ∣ Y \widehat{T} \circ\varphi|_Y T ∘ φ ∣ Y and the given embedding Y ⊂ P M Y \subset\mathbb{P}^M Y ⊂ P M are two embeddings of Y Y Y by the complete linear series ∣ O Y ( C ) ∣ |\mathcal{O}_Y(C)| ∣ O Y ( C ) ∣ which agree on C C C , so they differ by a projectivity γ \gamma γ of P M \mathbb{P}^M P M with γ ∣ C = i d \gamma|_C=\mathrm{id} γ ∣ C = id , hence γ ∣ H = i d \gamma|_H=\mathrm{id} γ ∣ H = id as C C C spans H H H . Finally γ T ^ φ ( Y ~ y ~ 0 ) \gamma\widehat{T}\varphi(\widetilde{Y}_{\widetilde{y}_0}) γ T φ ( Y y 0 ) is a K3 carpet on Y ⊂ P M Y \subset\mathbb{P}^M Y ⊂ P M , hence is Y ~ \widetilde{Y} Y [20 Proposition 1.7] , and
Y ~ ∩ H = γ T ^ ( φ ( Y ~ y ~ 0 ) ∩ H 1 ) = γ T ^ φ ( C ~ ) = γ ( C ~ ) = C ~ . \widetilde{Y} \cap H = \gamma\widehat{T}\bigl(\varphi(\widetilde{Y}_{\widetilde{y}_0}) \cap H_1\bigr) = \gamma\widehat{T}\varphi(\widetilde{C}) = \gamma(\widetilde{C}) = \widetilde{C}. Y ∩ H = γ T ( φ ( Y y 0 ) ∩ H 1 ) = γ T φ ( C ) = γ ( C ) = C .
Department of Mathematics, Oklahoma State University
Email address: jayan.mukherjee@okstate.edu
Mathematical Sciences Institute, Australian National University
Email address: anand.deopurkar@anu.edu.au