Characterizing ( d , h ) (d,h) -elliptic stable irreducible curves Juliana Coelho, Renata Costa Abstract We use admissible covers to characterize irreducible stable curves that are ( d , h ) (d,h) -elliptic, that is, that are limits of smooth curves admiting finite maps of degree- d d to smooth curves of genus h ≥ 1 h\geq 1 . 1 Introduction A smooth curve C is said to be d -gonal if it admits a degree- d map to \mathbb{P}^{1} . The gonality, that is, the minimum d such that C is d -gonal, is an important numerical invariant in the study of algebraic curves. More generally, we say that C is (d,h) -elliptic if it admits a degree- d morphism to a smooth curve of genus h . It follows from de Franchi’s theorem that, if d\geq 2 , then a general curve C of genus g\geq 2 is not (d,h) -elliptic for any h\geq 1 (see [ 3autores , Corollary XXI.8.32] ). It is thus an interesting problem to try to understand how such a smooth curve can degenerate, that is, what are the possible stable limits of (d,h) -elliptic smooth curves for h\geq 1 . In this paper, we deal with this issue using the theory of admissible covers, introduced in [ HarrMum ] , and greatly generalized by many authors such as in [ abracortivistoli ] and [ mochizuki ] . This approach has already been considered for instance in [ faberpaganini ] , where the authors determine the class of the so-called bielliptic locus for genus-3 stable curves, that is, the locus of (2,1) -elliptic curves in \overline{M}_{3} . The case h=1 is related to the classical study of elliptic integrals and was also studied in [ ddelliptic2018 ] , albeit without the use of admissible covers, where the authors considered stable curves of genus 2 admiting two morphisms to a smooth curve of genus 1 . 1.1 Main Results In this paper we work over \mathbb{C} . As we mentioned, we consider a stable curve C to be (d,h) -elliptic if it is a limit of smooth ones or, equivalently, if there exists a d -sheeted admissible cover from a curve stably equivalent to C to a curve of genus h . Rougly speaking, an admissible cover is a finite map of nodal curves satisfying some conditions, one of which is that the target is a pointed stable curve, when considered with the smooth branch points of the map. On Section 3 we define pseudo-admissible covers by switching this condition to the condition that every node of the target curve lies in the intersection of two components. On Proposition 5 we present a gluing for pseudo-admissible covers in the style of those introduced in [ 1 ] , and on Proposition 6 we show that it is possible to characterize (d,h) -ellipticity using pseudo-admissible covers. On Section 4 we focus on irreducible curves. In Theorem 10 we characterize irreducible stable curves that are (d,h) -elliptic for h\geq 1 , in terms of its normalization. This result is similar to those of [ 2 ] , where the case h=0 was addressed for stable curves with two components. As a consequence, we examine in Corollaries 11 and 12 the case of irreducible stable curves with only one node. 2 Technical background In this paper we always work over the field of complex numbers \mathbb{C} . A curve C is a connected, projective and reduced scheme of dimension 1 over \mathbb{C} . We denote by C^{\text{sing}} the singular locus of C , and by C^{\text{sm}} the smooth locus of C . The genus of C is the arithmetic genus g(C):=h^{1}(C,\mathcal{O}_{C}) . A subcurve Y of C is a reduced subscheme of pure dimension 1 , or equivalently, a reduced union of irreducible components of C . If Y\subseteq C is a subcurve, we say Y^{c}:=\overline{C\smallsetminus Y} is the complement of Y in C . A nodal curve C is a curve with at most ordinary double points, called nodes . A node is said to be separating if there is a subcurve Y of C such that Y\cap Y^{c}=\{n\} . Let C be a nodal curve and \nu:{C}_{\nu}\rightarrow C be its normalization. Let n be a node of C . The points n_{1},n_{2}\in{C}_{\nu} such that \nu(n_{1})=\nu(n_{2})=n are said to be branches of n . A rational chain is a nodal curve of genus 0. We remark that a curve is a rational chain if and only if its nodes are all separating and its components are all rational. A cycle is a nodal curve E whose dual graph is a cycle, that is, up to reordering, the components E_{1},\ldots,E_{s} of E are such that E_{i} meets E_{i+1} at a single point, for i=1,\ldots,s-1 , and E_{s} meets E_{1} at a single point. We denote the cycle E as the ordered s -uple (E_{1},\ldots,E_{s}) . A rational cycle is a cycle where all components are rational. Note that a rational cycle is a curve of genus 1. Let g and n be non negative integers such that 2g-2+n>0 . A n -pointed stable curve of genus g is a curve C of genus g together with n distinct marked points p_{1},\ldots,p_{n}\in C such that for every smooth rational component E of C , the number of points in the intersection E\cap E^{c} plus the number of indices i such that p_{i} lies on E is at least three. A stable curve is a 0-pointed stable curve. Let \pi:C\rightarrow B be a finite map between curves and let p\in C and q\in B be smooth points such that \pi(p)=q . Let x be a local parameter of C around p and t be a local parameter of B around q . Then, locally around p , the map \pi is given by t=x^{e} , where e is the ramification index of \pi at p , denoted by e_{\pi}(p) . Note that 1\leq e_{\pi}(p)\leq d , where d is the degree of \pi . We say that \pi is ramified at p , or that p is a ramification point of \pi , if e_{\pi}(p)\geq 2. Moreover, we say \pi is totally ramified at p if e_{\pi}(p)=d . Finally, we say that q is a branch point of \pi if e_{\pi}(p^{\prime})\geq 2 for some p^{\prime}\in\pi^{-1}(q) . 3 Admissible Covers A smooth curve C is (d,h) -elliptic if it admits a degree- d morphism to a smooth curve of genus h . In this work we define a stable curve C to be (d,h) -elliptic if it is a limit of smooth (d,h) -elliptic curves in the moduli space \overline{\mathcal{M}}_{g} of stable curves. More precisely, this occurs if there is a smoothing f:\mathcal{C}\rightarrow S whose generic fiber is a smooth (d,h) -elliptic curve and the special fiber is isomorphic to C . Alternativelly, (d,h) -ellipticity can be characterized in terms of admissible covers. A d -sheeted admissible cover consists of a finite morphism \pi:C\rightarrow B of degree d , such that B and C are nodal curves and: 1. \pi^{-1}(B^{\text{sing}})=C^{\text{sing}}; 2. \pi is simply branched away from C^{\text{sing}} , that is, over each smooth point of B there exists at most one point of C where \pi is ramified and this point has ramification index 2; 3. B is a stable pointed curve, when considered with the smooth branch points of \pi ; 4. for every node q of B and every node n of C lying over it, the two branches of C over n map to the branches of B over q with the same ramification index. Let C and C^{\prime} be nodal curves. We say that C^{\prime} is stably equivalent to C if C can be obtained from C^{\prime} by contracting to a point some of the smooth rational components of C^{\prime} meeting the other components of C^{\prime} in only one or two points. We remark that stable equivalence is not an equivalence relation, but rather a partial order and thus we write C^{\prime}\succ_{stab}C to mean that C^{\prime} is stably equivalent to C . Note moreover that, in this case, g(C)=g(C^{\prime}) and there exists a contraction map \tau:C^{\prime}\rightarrow C . We say that a point p^{\prime}\in C^{\prime} lies over a point p\in C if \tau(p^{\prime})=p . Theorem 1 . A stable curve C is (d,h) -elliptic if and only if there exists a d -sheeted admissible cover {C}^{\prime}\rightarrow B , where {C}^{\prime}\succ_{stab}C and g(B)=h . Proof. Follows from [ mochizuki ] . ∎ Now we introduce a slightly relaxed but, as we’ll see in Proposition 6 , equivalent notion of admissibility. Let C and B be nodal curves. A finite morphism \pi:C\rightarrow B of degree d is called a d -sheeted pseudo-admissible cover if it satisfies conditions (1), (2) and (4) of an admissible cover and also satisfies: {3}^{\prime} . the curve B has no internal nodes. The next result is a pseudo-admissible version of [ 1 , Lemma 3.2] , and we include it here for the sake of completeness. Lemma 2 . Let C be a nodal curve and \pi:{C}^{\prime}\rightarrow B be a d -sheeted pseudo-admissible cover, where C^{\prime}\succ_{stab}C . Then for each irreducible component C_{r} of C there is a unique irreducible component {C}^{\prime}_{r} of {C}^{\prime} such that \tau({C}^{\prime}_{r})=C_{r} and \tau|_{{C}^{\prime}_{r}} is a normalization map, where \tau\colon C^{\prime}\rightarrow C is the contraction map. Proof. As the components of B are smooth and \pi^{-1}(B^{sing})=C^{sing} , then the components of {C}^{\prime} are also smooth. Fixing a component C_{r} of C , by definition, there is an unique component {C}^{\prime}_{r} of \tau^{-1}(C_{r}) dominating C_{r} . Since {C}^{\prime}_{r} is smooth, \tau|_{{C}^{\prime}_{r}} is a normalization map. ∎ The following two results have appeared in [ 1 ] , albeit in the context of admissible covers to curves of genus 0. Lemma 3 . Let \pi:C\rightarrow B be a degree- d finite morphism of nodal curves. If \pi satisfies conditions (1),({3}^{\prime}) and (4) of d -sheeted pseudo-admissible covers, then there is a d -sheeted pseudo-admissible cover {\pi}^{\prime}:{C}^{\prime}\rightarrow{B}^{\prime} such that {C}^{\prime} (respectively {B}^{\prime} ) is stably equivalent to C (respectively B ), contains C (respectively B ) as a subcurve and {\pi}^{\prime}|_{C}=\pi . Proof. The proof is mutatis mutandis that of [ 1 , Theorem 3.3 (a)] . ∎ Proposition 4 . Let C be a nodal curve and n be a node of C such that the normalization C_{n} of C at n is connected. Let \pi\colon C_{n}^{\prime}\rightarrow B be a finite map of degree d between nodal curves, satisfying conditions (1) , (3^{\prime}) and (4) of pseudo-admissible curves, where C_{n}^{\prime}\succ_{stab}C_{n} . Let n^{(1)},n^{(2)}\in C_{n} be the branches of n and let {n}^{\prime(1)} , {n}^{\prime(2)} be smooth points of {C}^{\prime}_{n} lying over n^{(1)} and n^{(2)} , respectively. If \pi(n^{\prime(1)})=\pi(n^{\prime(2)}) then there is a d -sheeted quasi-admissible cover \pi^{\prime}\colon C^{\prime}\rightarrow B^{\prime} where C^{\prime}\succ_{stab}C , C_{n}^{\prime}\subset C^{\prime} and \pi^{\prime}|_{C_{n}}=\pi . Proof. The proof is mutatis mutandis that of [ 1 , Theorem 3.4 (b)] . ∎ The next result describes a gluing of a pseudo-admissible cover, similar to those on [ 1 ] , but more suited to study (d,h) -ellipticity for h\geq 1 . Proposition 5 . Let C be a stable curve and let C_{n} be the normalization of C at nodes n_{1},\ldots,n_{\delta} . Assume that C_{n} is connected and let \pi:{C}^{\prime}_{n}\rightarrow B be a finite map of degree d satisfying conditions (1),({3}^{\prime}) and (4) of pseudo-admissible cover, where B is a curve of genus h and {C}^{\prime}_{n}\succ_{stab}C_{n} . Let n_{i}^{(1)},n_{i}^{(2)}\in C_{n} be the branches of n_{i} and {n}_{i}^{\prime(1)} , {n}_{i}^{\prime(2)} be smooth points of {C}^{\prime}_{n} lying over n_{i}^{(1)} and n_{i}^{(2)} , respectively, for i=1,\ldots,\delta . Assume that (a) e_{\pi}({n}_{i}^{\prime(1)})=e_{\pi}({n}_{i}^{\prime(2)}) , for i=1,\ldots,\delta ; (b) there exist q_{1},q_{2}\in B such that q_{1}\neq q_{2} and \pi^{-1}(q_{j})=\{n^{\prime(j)}_{1},\ldots,n^{\prime(j)}_{\delta}\} , for j=1,2 . Then there exists a d -sheeted pseudo-admissible cover {\pi}^{\prime}:{C}^{\prime}\rightarrow{B}^{\prime} where C^{\prime}\succ_{stab}C , C_{n}^{\prime}\subset C^{\prime} and \pi^{\prime}|_{C_{n}^{\prime}}=\pi . Moreover, g({B}^{\prime})=h+1 and C is (d,h+1) -elliptic. Proof. Set e_{i}:=e_{\pi}({n}_{i}^{\prime(1)})=e_{\pi}({n}_{i}^{\prime(2)}) , for i=1,\ldots,\delta . To obtain the cover {\pi}^{\prime} we proceed as follows. We glue (see Figure 1 ): Figure 1: (d,h+1) -ellipticity for d=3 \bullet a copy of \mathbb{P}^{1} in B passing through q_{1} and q_{2} . Call this copy L ; \bullet a copy of \mathbb{P}^{1} in {C}_{n}^{\prime} passing through n^{\prime(1)}_{i} and n^{\prime(2)}_{i} mapping to L via a map of degree e_{i} , totally ramified at n^{\prime(1)}_{i} and n^{\prime(2)}_{i} , for i=1,\ldots,\delta . Thus, we obtain from {C}^{\prime}_{n} a nodal curve \widetilde{C} stably equivalent to {C} , from B a nodal curve \widetilde{B} such that g(\widetilde{B})=h+1 , and the map \widetilde{\pi}:\widetilde{C}\rightarrow\widetilde{B} given by the map \pi when restricted to {C}^{\prime}_{n} and by the maps described above when restricted to the added rational components of \widetilde{C} . By construction, {C}^{\prime}_{n} is a subcurve of \widetilde{C} and the map \widetilde{\pi} satisfies conditions (1),({3}^{\prime}) and (4) of the definition of pseudo-admissible cover. It follows from Lemma 3 that there exists a d -sheeted pseudo-admissible cover {\pi}^{\prime}:{C}^{\prime}\rightarrow{B}^{\prime} , where {C}^{\prime} is stably equivalent to \widetilde{C} and {B}^{\prime} is stably equivalent to \widetilde{B} . ∎ We finish this section showing that the (d,h) -ellipticity of a stable curve can be characterized in terms of pseudo-admissible covers. Proposition 6 . A stable curve C is (d,h) -elliptic if and only if there exists a d -sheeted pseudo-admissible cover \pi:{C}^{\prime}\rightarrow B where C^{\prime}\succ_{stab}C and g(B)=h . Proof. By Theorem 1 , it is enough to show that for every pseudo-admissible cover \pi:C\rightarrow B there exists an admissible cover \widetilde{\pi}:\widetilde{C}\rightarrow\widetilde{B} and, conversely, for every admissible cover \widetilde{\pi}:\widetilde{C}\rightarrow\widetilde{B} there exists a pseudo-admissible cover \pi:C\rightarrow B , such that B\succ_{stab}\widetilde{B} and C\succ_{stab}\widetilde{C} the following diagram commutes \textstyle{C\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces} \scriptstyle{\tau_{C}} \scriptstyle{\pi} \textstyle{\widetilde{C}\ignorespaces\ignorespaces\ignorespaces\ignorespaces} \scriptstyle{\widetilde{\pi}} \textstyle{B\ignorespaces\ignorespaces\ignorespaces\ignorespaces} \scriptstyle{\tau_{B}} \textstyle{\widetilde{B}} (1) where \tau_{C} and \tau_{B} are the contraction maps. First we let \pi:C\rightarrow B be a d -sheeted pseudo-admissible cover that is not admissible. Then there exist rational components W of B containing at most two points that are nodes of B or are smooth branch points of \pi . Note that, if L is a component of C such that \pi(L)=W , then L meets its complement in at most two points and has genus 0, by Riemann-Hurwitz. Since the map \pi satisfies conditions (1),(2) and (4) of the definition of admissible cover, to satisfy condition (3) it suffices to contract those components W and the components of C mapping to W . Thus, we obtain curves \widetilde{B} and \widetilde{C} and an induced map \widetilde{\pi}:\widetilde{C}\rightarrow\widetilde{B} as in ( 1 ). Conversely, let \widetilde{\pi}:\widetilde{C}\rightarrow\widetilde{B} be a d -sheeted admissible cover that is not pseudo-admissible, that is, such that \widetilde{B} has internal nodes. Now note that for n\in\widetilde{C} and q\in\widetilde{B} such that \widetilde{\pi}(n)=q , we have that q is an internal node of B if and only if n is an internal node of C . So we let C^{\prime} and B^{\prime} be the normalization of C and B at its internal nodes, respectivelly. Then \widetilde{\pi} induces a d -sheeted admissible cover \pi^{\prime}:C^{\prime}\rightarrow B^{\prime} that is also quasi-admissible. Moreover, if q is an internal node of B and \widetilde{\pi}^{-1}(q)=\{n_{1},\ldots,n_{\delta}\} , then \pi^{\prime} satisfies the conditions of Proposition 5 with respect to the branches of n_{1},\ldots,n_{\delta} . Hence, applying Proposition 5 sucessively for each internal node of B , we obtain a pseudo-admissible cover \pi\colon C\rightarrow B as in ( 1 ). ∎ 4 (d,h) -ellipticity for irreducible curves In this section, our goal is to characterize irreducible nodal (d,h) -elliptic curves in terms of their normalizations. First, we need a lemma. Lemma 7 . Let C be a nodal curve and let {C}^{\prime} be a nodal curve stably equivalent to C with contraction map \tau:{C}^{\prime}\rightarrow C . (a) If Z_{1} and Z_{2} are two connected subcurves of {C}^{\prime} such that \tau(Z_{1})=\tau(Z_{2}) is a subcurve of C , then \tau(Z_{1})=\tau(Z_{2})=\tau(Z_{1}\cap Z_{2}) . (b) If C is irreducible then no two subcurves of {C}^{\prime} of positive genus have finite intersection. (c) Let \pi:{C}^{\prime}\rightarrow B be a d -sheeted pseudo-admissible cover. Assume that C is irreducible and let C_{n} be the normalization of C and let B_{0}=\pi(C_{n}) . Then every subcurve Z of B of positive genus must contain B_{0} . In particular, if there exists a component W of B such that g(W)\geq 1 , then W=B_{0} . Proof. First we prove (a). Set Z:=\tau(Z_{1})=\tau(Z_{2}) and assume Z is a subcurve of C . Recall from Lemma 2 that, if W is a component of C , then there exists an unique component W^{\prime} of C^{\prime} such that \tau(W^{\prime})=W . Hence if W\subset Z , then we must have W^{\prime}\subset Z_{1} and also W^{\prime}\subset Z_{2} . Thus W^{\prime}\subset Z_{1}\cap Z_{2} showing that Z=\tau(Z_{1}\cap Z_{2}) . To show (b), let Z_{1} and Z_{2} be two subcurves of {C}^{\prime} of positive genus so that \tau(Z_{1}) and \tau(Z_{2}) are subcurves of C . Since C is irreducible, we must have \tau(Z_{1})=\tau(Z_{2}) which, by (a), implies that Z_{1}\cap Z_{2} cannot be finite. Lastly we show (c). Assume there exists a subcurve Z of B of positive genus not containing B_{0} . Then C_{n} is not contained in \pi^{-1}(Z) . Let W be a connected component of \pi^{-1}(Z) . Since \pi(W)=Z , we must have g(W)\geq g(Z)\geq 1 . In particular, this implies that W is not contracted by \tau . But since C is irreducible, this means \tau(W)=C=\tau(C_{n}) and, by (a), W\cap C_{n} cannot be finite. As C_{n} is irreducible, this implies that C_{n}\subset W , a contradiction. ∎ Proposition 8 . Let C be an irreducible nodal curve and C_{n} be its normalization. Let \pi:{C}^{\prime}\rightarrow B be a d -sheeted pseudo-admissible cover, where {C}^{\prime}\succ_{stab}C . If g(B)\geq 1 , then \pi|_{C_{n}} has degree d . Proof. Set B_{0}:=\pi(C_{n}) and let e be the degree of the restriction \pi|_{C_{n}} . Assume 1\leq e\leq d-1 . Since \pi is a map of degree d , then there exists a connected subcurve Z of {C}^{\prime} not containing C_{n} such that C_{n}\cup Z\subset\pi^{-1}(B_{0}) . If e^{\prime} is the degree of \pi|_{Z} , then e^{\prime}\geq 1 and e+e^{\prime}\leq d . We have two cases to consider. If g(B_{0})\geq 1 then C_{n} and Z must have positive genus as well, since they map to B_{0} . By Lemma 7 , we then have that C_{n}\cap Z is not finite and, since C_{n} is irreducible, we must have C_{n}\subset Z contradicting the choice of Z . Now assume g(B_{0})=0 . Then, by Lemma 7 , every irreducible component of B has genus 0 and, since g(B)\geq 1 , there exists a rational cycle E=(E_{1},\ldots,E_{s}) in B which must contain B_{0} , say E_{1}=B_{0} . Let q_{1},p_{1},\ldots,p_{s-1},q_{2}\in B be such that E_{1}\cap B_{0}={q_{1}},\quad E_{s}\cap B_{0}={q_{2}},\quad\text{ and }\quad E_{i}\cap E_{i+1}={p_{i}},\quad\text{for }i=1,\ldots,s-1. Let q^{1}_{1}\in C_{n} and q_{1}^{2}\in Z such that \pi(q_{1}^{1})=\pi(q_{1}^{2})=q_{1} Since q_{1} is a node and \pi is pseudo-admissible, then q_{1}^{1} and q_{1}^{2} are also nodes. Therefore, there exist rational components E_{1}^{1} and E_{1}^{2} of C^{\prime} such that q_{1}^{1}=C_{n}\cap E_{1}^{1}\quad\text{ and }\quad q_{1}^{2}=Z\cap E_{1}^{2}. Note that we must have \pi(E_{1}^{1})=\pi(E_{1}^{2})=E_{1} . Furthermore, since E_{1} has another node, namely p_{1} , then E_{1}^{1} and E_{1}^{2} also have nodes, say p_{1}^{1} and p_{1}^{2} , respectively, such that \pi(p_{1}^{1})=\pi(p_{1}^{2})=p_{1}. Consequently, there exist rational components E_{2}^{1} and E_{2}^{2} of C^{\prime} such that p_{1}^{1}=E_{1}^{1}\cap E_{2}^{1}\quad\text{ and }\quad p_{1}^{2}=E_{1}^{2}\cap E_{2}^{2} and again we have, \pi(E_{2}^{1})=\pi(E_{2}^{2})=E_{2} . Prceeding in this manner, for j=1,2 we obtain rational components E_{1}^{j},E_{2}^{j},\ldots,E_{s}^{j} of C^{\prime} such that \pi(E_{i}^{j})=E_{i} and E^{j}_{i} meet E^{j}_{i+1} at p^{j}_{i+1} for i=1,\ldots,s-1 . Lastly, since E_{s} meets B_{0} at the node q_{2} , then E_{s}^{1} meets C_{n} and E_{s}^{2} meets Z . Hence C_{n}\cup E_{1}^{1}\cup\ldots\cup E_{s}^{1}\quad\text{ and }\quad Z\cup E_{1}^{2}\cup\ldots\cup E_{s}^{2} are two subcurves of {C}^{\prime} of positive genus and having no common component, contradicting Lemma 7 , thus showing that e=d . ∎ We remark that the previous result does not hold for admissible or pseudo-admissible covers to a curve of genus 0, as attested by [ 1 , Theorem 3.3 and Corollary 3.5] . Lemma 9 . Let C be an irreducible nodal curve and C_{n} be its normalization. Let \pi:{C}^{\prime}\rightarrow B be a d -sheeted pseudo-admissible cover, where {C}^{\prime}\succ_{stab}C . Assume that g(B)\geq 1 . Let n\in C be a node and denote by n^{1},n^{2}\in C_{n} its branches. If \pi(n^{1})\neq\pi(n^{2}) , then there exists a cycle Z in B such that \pi(n^{1}) and \pi(n^{2}) are nodes of Z . Moreover, we have e_{\pi}(n^{1})=e_{\pi}(n^{2}) . Proof. Assume there is no such cycle of B . Then every connected subcurve of B containing q_{1}:=\pi(n^{1}) and q_{2}:=\pi(n^{2}) must contain B_{0}:=\pi(C_{n}) . Now, since C^{\prime} is stably equivalent to C , there is a rational chain Y^{\prime}\subset{C}^{\prime} not containing C_{n} and connecting n^{1} to n^{2} . Hence, since Y^{\prime} is connected then so is \pi(Y^{\prime}) and, since \pi(Y^{\prime}) contains q_{1} and q_{2} , then \pi(Y^{\prime}) must contain B_{0} . Thus there is a component L of Y^{\prime} such that \pi(L)=B_{0} . Since B_{0}=\pi(C_{n}) , then the degree of the restriction \pi|_{C_{n}} must be less than d , contradicting Proposition 8 . Now we show the last assertion. Let Z=(B_{0},E_{1},\ldots,E_{s})\subset B be the stated cycle, such that q_{1}\in E_{1} and q_{2}\in E_{s} . Set Y=E_{1}\cup\cdots\cup E_{s} , that is, Y:=Z\setminus B_{0} . Moreover, set \widetilde{Y}\subset Y^{\prime} such that C_{n}\cup\widetilde{Y} is a cycle in C^{\prime} through n^{1} and n^{2} , say C_{n}\cup\widetilde{Y}=(C_{n},\widetilde{E}_{1},\ldots,\widetilde{E}_{r}) , with n^{1}\in\widetilde{E}_{1} and n^{2}\in\widetilde{E}_{r} . Note that \pi(\widetilde{Y})\subset Y . Assume, without lack of generality, that e_{\pi}(n^{1})