1 Introduction

Let \(S={\mathbb {P}}^1\times {\mathbb {P}}^1\), let C be a smooth curve in S of degree (5, 1), and let \(\epsilon :C\rightarrow {\mathbb {P}}^1\) be the morphism induced by the projection \(S\rightarrow {\mathbb {P}}^1\) to the first factor. Then \(\epsilon \) is a finite morphism of degree five, and we may assume that the points ([1 : 0], [0 : 1]) and ([0 : 1], [1 : 0]) are among its ramifications points. This assumption implies that the curve C is given by

$$\begin{aligned} u\big (x^5+a_1x^{4}y+a_2x^{3}y^2+a_3x^{2}y^3\big )=v\big (y^5+b_1xy^4+b_2x^{2}y^3+b_3x^{3}y^2\big ) \end{aligned}$$

for some \(a_1\), \(a_2\), \(a_3\), \(b_1\), \(b_2\), \(b_3\), where ([u : v], [x : y]) are coordinates on S. Note that the ramification index of the point ([1 : 0], [0 : 1]) can be computed as follows:

$$\begin{aligned} \left\{ \begin{aligned}&2\ \text {if }a_3\ne 0,\\&3\ \text {if }a_3=0\text { and }a_2\ne 0,\\&4\ \text {if }a_3=a_2=0\text { and }a_1\ne 0,\\&5\ \text {if }a_3=a_2=a_1=0. \end{aligned} \right. \end{aligned}$$

Likewise, we can compute the ramification index of the point ([0 : 1], [1 : 0]). We may assume that

  • ([1 : 0], [0 : 1]) has the largest ramification index among all ramifications points of \(\epsilon \)

  • the ramification index of the point ([0 : 1], [1 : 0]) is the second largest index.

If both these indices are 5, then \(a_1=a_2=a_3=b_1=b_2=b_3=0\), the morphism \(\epsilon \) does not have other ramification points, and the equation of the curve C simplifies as

$$\begin{aligned} ux^5=vy^5. \end{aligned}$$

In this case, we have \(\textrm{Aut}(S,C)\cong {\mathbb {C}}^*\rtimes {\mathbb {Z}}/2{\mathbb {Z}}\). In all other cases, this group is finite [5, Corollary 2.7].

Now, we consider embedding \(S\hookrightarrow {\mathbb {P}}^1\times {\mathbb {P}}^2\) given by

$$\begin{aligned} \big ([u:v],[x:y]\big )\mapsto \big ([u:v],[x^2:xy:y^2]\big ), \end{aligned}$$

and identify S and C with their images in \({\mathbb {P}}^1\times {\mathbb {P}}^2\). Let \(\pi :X\rightarrow {\mathbb {P}}^1\times {\mathbb {P}}^2\) be the blow up of the curve C. Then X is a smooth Fano threefold in the deformation family № 3.5 in the Mori–Mukai list and every smooth member of this family can be obtained in this way. We know from [2, Section 5.14], that

  • X is K-stable if the numbers \(a_1\), \(a_2\), \(a_3\), \(b_1\), \(b_2\), \(b_3\) are general enough,

  • X is K-polystable if \(a_1=a_2=a_3=b_1=b_2=b_3=0\).

However, for some \(a_1\), \(a_2\), \(a_3\), \(b_1\), \(b_2\), \(b_3\), the threefold X is not K-polystable.

Example 1

If \((a_1,a_2,a_3)=(0,0,0)\ne (b_1,b_2,b_3)\), then X is not K-polystable [2, Lemma 7.6].

Note also that it follows from the proof of [5, Lemma 8.7] that \(\textrm{Aut}(X)\cong \textrm{Aut}(S,C)\). In particular, we conclude the group \(\textrm{Aut}(X)\) is finite if and only if \((a_1,a_2,a_3,b_1,b_2,b_3)\ne (0,0,0,0,0,0)\). In this case, the threefold X is K-polystable if and only if it is K-stable. Moreover, we have

Conjecture 1

([2]) The Fano threefold X is K-stable if and only if \((a_1,a_2,a_3)\ne (0,0,0)\).

Geometrically, this conjecture says that the following two conditions are equivalent:

  1. (1)

    the threefold X is K-stable,

  2. (2)

    the morphism \(\epsilon :C\rightarrow {\mathbb {P}}^1\) does not have ramification points of ramification index five.

The goal of this paper is to prove the following (slightly weaker) result:

MainTheorem

If all ramification points of \(\epsilon \) have ramification index two, then X is K-stable.

Let \(\textrm{pr}_1: {\mathbb {P}}^1\times {\mathbb {P}}^2 \rightarrow {\mathbb {P}}^1\) be the projection to the first factor and \(\phi _1=\textrm{pr}_1\circ \pi \). Then \(\phi _1\) is a fibration into del Pezzo surfaces of degree four, and every singular fiber of this fibration has Du Val singular points of types \({\mathbb {A}}_1\), \({\mathbb {A}}_2\), \({\mathbb {A}}_3\) or \({\mathbb {A}}_4\), and we have the following possibilities for the singularities of a given singular fiber

  1. (1)

    one singular point of type \({\mathbb {A}}_1\),

  2. (2)

    two singular points of type \({\mathbb {A}}_1\),

  3. (3)

    one singular point of type \({\mathbb {A}}_2\),

  4. (4)

    one singular point of type \({\mathbb {A}}_1\) and one singular point of type \({\mathbb {A}}_2\),

  5. (5)

    one singular point of type \({\mathbb {A}}_3\),

  6. (6)

    one singular point of type \({\mathbb {A}}_4\).

Note that \(\phi _1\) has at most two singular fibers that have singular points of type \({\mathbb {A}}_4\). Moreover, if \(\phi _1\) has two singular fibers with singular points of type \({\mathbb {A}}_4\) then all numbers \(a_i\) and \(b_j\) vanish, so that X is K-polystable. Vice versa, if \(\phi _1\) has exactly one singular fiber with a point type \({\mathbb {A}}_1\), then the authors of [2] proved that X is not K-polystable. Moreover, they conjectured that X is K-stable in all remaining cases. Now Main Theorem and Conjecture 1 can be restated as follows:

MainTheorem

If every singular fiber of \(\phi _1\) has only singular points of type \({\mathbb {A}}_1\), then X is K-stable.

Conjecture 2

The Fano threefold X is K-stable if and only if every singular fiber of \(\phi _1\) has only singular points of type \({\mathbb {A}}_1\), \({\mathbb {A}}_2\) or \({\mathbb {A}}_3\).

2 The Proof

To prove Main Theorem, we suppose that each singular fiber of the fibration \(\phi _1\) has one or two singular points of type \({\mathbb {A}}_1\). Note that this fiber is a del Pezzo surface of degree 4 with Du Val singularities. We know ( [7, 9]) that the Fano threefold X is K-stable if and only if for every prime divisor \({\textbf{F}}\) over X we have

$$\begin{aligned} \beta ({\textbf{F}})=A_X({\textbf{F}})-S_X({\textbf{F}})>0 \end{aligned}$$

where \(A_X({\textbf{F}})\) is the log discrepancy of the divisor \({\textbf{F}}\), and

$$\begin{aligned} S_X\big ({\textbf{F}}\big )=\frac{1}{(-K_X)^3}\int \limits _0^{\infty }\textrm{vol}\big (-K_X-u{\textbf{F}}\big )du. \end{aligned}$$

To show this, we fix a prime divisor \({\textbf{F}}\) over X. Then we set \(Z=C_{X}({\textbf{F}})\). If Z is an irreducible surface, then it follows from [8] that \(\beta ({\textbf{F}})>0\), see also [2, Theorem 3.17]. Therefore, we may assume that

  • either Z is an irreducible curve in X,

  • or Z is a point in X.

In both cases, we fix a point \(O\in Z\). Let \({\overline{T}}\) be the fiber of \(\phi _1\) which contains O. Then \({\overline{T}}\) is a del Pezzo surface with at most Du Val singularities. Set

$$\begin{aligned} \tau ({\overline{T}}) = \textrm{sup}\Big \{u\in {\mathbb {R}}_{>0}\big |\text { the divisor }-K_X - u{\overline{T}} \text { is pseudo-effective}\Big \} \end{aligned}$$

For \(u \in [0, \tau ({\overline{T}}) ]\) let P(u) be the positive part of the Zariski decomposition of the divisor \(-K_X-u{\overline{T}}\), and let N(u) be its negative part. We denote \({\widetilde{S}}\) to be the proper transform on X of the surface S. Then we have

$$\begin{aligned}&P(u)=\left\{ \begin{aligned}&-K_X-u{\overline{T}} \ \text { if } u\in [0,1], \\&-K_X-u{\overline{T}}-(u-1){\widetilde{S}}\ \text { if } u\in [1,2], \end{aligned} \right. \text { and }\\&N(u)= \left\{ \begin{aligned}&0\ \text { if } u\in [0,1], \\&(u-1){\widetilde{S}}\ \text { if } u\in [1,2], \end{aligned} \right. \end{aligned}$$

which gives

$$\begin{aligned} S_{X}({\overline{T}})=\frac{1}{20}\int \limits _{0}^{2}P(u)^3du=\frac{69}{80}<1 \end{aligned}$$

Now, for every prime divisor F over the surface \({\overline{T}}\), we set

$$\begin{aligned} S\big (W^{{\overline{T}}}_{\bullet ,\bullet };F\big )= & {} \frac{3}{(-K_X)^3}\int \limits _0^{\tau }\textrm{ord}_F \big (N(u)\vert _{{\overline{T}}}\big )\big (P(u)\vert _{{\overline{T}}}\big )^2du\\{} & {} +\frac{3}{(-K_X)^3}\int \limits _0^\tau \int \limits _0^{\infty }\textrm{vol}\big (P(u)\big \vert _{{\overline{T}}}-vF\big )dvdu. \end{aligned}$$

Then, following [1, 2], we let

$$\begin{aligned} \delta _O\big ({\overline{T}},W^{{\overline{T}}}_{\bullet ,\bullet }\big )=\inf _{\begin{array}{c} F/{\overline{T}}\\ O\in C_{{\overline{T}}}(F) \end{array}}\frac{A_T(F)}{S\big (W^{{\overline{T}}}_{\bullet ,\bullet };F\big )}, \end{aligned}$$

where the infimum is taken by all prime divisors over the surface \({\overline{T}}\) whose center on \({\overline{T}}\) contains O. Then it follows from [1, 2] that

$$\begin{aligned} \frac{A_X({\textbf{F}})}{S_X({\textbf{F}})}\geqslant \min \Bigg \{\frac{1}{S_X({\overline{T}})},\delta _O\big ({\overline{T}},W^{{\overline{T}}}_{\bullet ,\bullet }\big )\Bigg \}. \end{aligned}$$

Therefore, if \(\beta ({\textbf{F}})\leqslant 0\), then \(\delta _O({\overline{T}},W^{{\overline{T}}}_{\bullet ,\bullet })\leqslant 1\).

Let’s prove that \(\delta _O({\overline{T}},W^{{\overline{T}}}_{\bullet ,\bullet }) > 1\). To estimate \(\delta _O(T,W^{{\overline{T}}}_{\bullet ,\bullet })\), we set \({\overline{D}}=P(u)\vert _{{\overline{T}}}\). We have

$$\begin{aligned} {\overline{D}}=\left\{ \begin{aligned}&-K_{{\overline{T}}} \ \text { if } u\in [0,1], \\&-K_{{\overline{T}}}-(u-1){\overline{C}}_2\ \text { if } u\in [1,2], \end{aligned} \right. \end{aligned}$$

where \({\overline{C}}_2:={\widetilde{S}}|_{{\overline{T}}}\). Then \({\overline{D}}\) is ample for \(u\in [0,2)\), and

$$\begin{aligned} {\overline{D}}^2=\left\{ \begin{aligned}&4\ \text { if } u\in [0,1], \\&5-u^2 \ \text { if } u\in [1,2]. \end{aligned} \right. \end{aligned}$$

By [2, Lemma 5.68] and [2, Lemma 5.69] we have

Lemma 1

If \(O\in {\widetilde{S}}\) then \(\delta _O(X) > 1\).

Lemma 2

If \({\overline{T}}\) is smooth then \(\delta _O(X)>1\).

Thus, to prove Main Theorem, we may assume that \(O\not \in {\widetilde{S}}\) and \({\overline{T}}\) is singular. Recall that

$$\begin{aligned} \delta _O\big ({\overline{T}},{\overline{D}}\big )=\inf _{\begin{array}{c} F/{\overline{T}}\\ O\in C_{{\overline{T}}}(F) \end{array}}\frac{A_{{\overline{T}}}(F)}{S_{{\overline{D}}}\big (F\big )}\end{aligned}$$

where the infimum is taken by all prime divisors over \({\overline{T}}\) whose center on \({\overline{T}}\) contain O, and \(S_{{\overline{D}}}(F)~=~\frac{1}{{\overline{D}}^2}\int \limits _0^{\infty }\textrm{vol}\big ({\overline{D}}-vF\big )dv\). Usually \(\delta _O({\overline{T}},-K_{{\overline{T}}})\) is denoted by \(\delta _O({\overline{T}})\).

Note that since \(O\not \in {\widetilde{S}}\) then for any divisor F over \({\overline{T}}\) then we get

$$\begin{aligned} S\big (W^{{\overline{T}}}_{\bullet ,\bullet };F\big )&=\frac{3}{(-K_X)^3}\Bigg (\int _0^\tau \big (P(u)^{2}\cdot {\overline{T}}\big )\cdot \textrm{ord}_{O}\Big (N(u)\big \vert _{{\overline{T}}}\Big )du\\&\quad +\int _0^\tau \int _0^\infty \textrm{vol}\big (P(u)\big \vert _{{\overline{T}}}-vF\big )dvdu\Bigg )\\&= \frac{3}{20}\int _0^\tau \int _0^\infty \textrm{vol}\big (P(u)\big \vert _{{\overline{T}}}-vF\big )dvdu\\&= \frac{3}{20}\Bigg (\int _0^1\int _0^\infty \textrm{vol}\big (-K_{{\overline{T}}}-vF\big )dvdu\\&\quad +\int _1^2\int _0^\infty \textrm{vol}\big (-K_{{\overline{T}}}-(u-1){\overline{C}}_2-vF\big )dvdu\Bigg )\\&= \frac{3}{20}\Bigg (\int _0^\infty \textrm{vol}\big (-K_{{\overline{T}}}-vF\big )dv\\&\quad +\int _0^\infty \textrm{vol}\big (-K_T-(u-1){\overline{C}}_2-vF\big )dv\Bigg )\\&= \frac{3}{20}\Bigg (\int _0^\infty \textrm{vol}\big (-K_{{\overline{T}}}-vF\big )dv+\int _0^\infty \textrm{vol}\big (-K_{{\overline{T}}}-vF\big )dv\Bigg )\\&= \frac{3}{10}\Bigg (\int _0^\infty \textrm{vol}\big (-K_{{\overline{T}}}-vF\big )dv\Bigg )\\&=\frac{6}{5}\Bigg (\frac{1}{4}\int _0^\infty \textrm{vol}\big (-K_{{\overline{T}}}-vF\big )dv\Bigg )\\&=\frac{6}{5}S_{{\overline{T}}}(F)\le \frac{6}{5}\cdot \frac{A_{{\overline{T}}}(F)}{\delta _O({\overline{T}})} \end{aligned}$$

Thus, if \(\delta _O({\overline{T}})>6/5\), then \(\delta _O({\overline{T}},W^{{\overline{T}}}_{\bullet ,\bullet })>1\). To estimate \(\delta _O({\overline{T}},W^{{\overline{T}}}_{\bullet ,\bullet })\) in the case when \(\delta _O({\overline{T}})\le 6/5\), we define the following positive continuous function on [1, 2]:

$$\begin{aligned}f(u):= \left\{ \begin{aligned}&\frac{15 - 3 u^2}{16 + 3 u - 9 u^2 + 2 u^3}\text { if }u\in [1,a],\\&\frac{15 - 3 u^2}{11 - u^3}\text { if }u\in [a,2] \end{aligned} \right. \end{aligned}$$

where a is a root of \(3u^3 - 9u^2 + 3u + 5\) on [1, 2]. More precisely, \(a\in [1.355,1.356]\). In the appendix we prove that for each O such that \(\delta _O({\overline{T}})\le \frac{6}{5}\) we have \(\delta _{O}({\overline{T}},{\overline{D}})\ge f(u)\) for every \(u\in [1,2]\). So we obtain

$$\begin{aligned} S\big (W^{{\overline{T}}}_{\bullet ,\bullet };F\big )&=\frac{3}{(-K_X)^3}\int \limits _1^{2} \int \limits _0^{\infty }\textrm{vol}\big (P(u)\big \vert _{{\overline{T}}}-vF\big )dvdu\\&\quad +\frac{3}{(-K_X)^3}\int \limits _0^{1}\int \limits _0^{\infty }\textrm{vol}\big (P(u)\big \vert _{{\overline{T}}}-vF\big )dvdu\\&\le \frac{3}{20} \Bigg (\int \limits _1^{2}\frac{(5-u^2)}{\delta _O({\overline{T}},{\overline{D}})}du\Bigg )A_{{\overline{T}}}(F) +\frac{3}{20}\cdot \frac{4A_{{\overline{T}}}(F)}{\delta _O({\overline{T}})}\\&\le \frac{3}{20} \Bigg (\int \limits _1^{2}\frac{(5-u^2)}{f(u)}du\Bigg )A_{{\overline{T}}}(F)+\frac{3}{5}A_{{\overline{T}}}(F)\\&\le \frac{3}{20} \Bigg (\int \limits _1^{1.356}(5-u^2)\frac{16 + 3 u - 9 u^2 + 2 u^3}{15 - 3 u^2} du\\&\quad +\int \limits _{1.355}^{2}(5-u^2)\frac{11 - u^3}{15 - 3 u^2} du\Bigg )A_{{\overline{T}}}(F)+\frac{3}{5}A_{{\overline{T}}}(F)\le \frac{99}{100}A_{{\overline{T}}}(F) \end{aligned}$$

Thus \(\frac{A_{{\overline{T}}}(F)}{S\big (W^{{\overline{T}}}_{\bullet ,\bullet };F\big )}\ge \frac{100}{99}\) for every prime divisor F over \({\overline{T}}\) whose support on F contains O, so that \(\delta _O(W^{{\overline{T}}},F)\ge \frac{100}{99}\), which implies \(\beta ({\textbf{F}})>0\) and X is K-stable.

Remark 1

If O were a singular point of type \({\mathbb {A}}_2\) in \({\overline{T}}\), this approach would not work, because as is shown in Appendix A.3 we have \(\delta _O({\overline{T}}, {\overline{D}}) = \frac{15-3u^2}{u^3 - 6u^2 + 19}\) and there is prime divisor F over \({\overline{T}}\) such that \(A_{{\overline{T}}}(F) = 1\) and \(S(W^{{\overline{T}}}_{\bullet ,\bullet }; F) = \frac{83}{80}\), which implies that \(\delta _O({\overline{T}},W^{{\overline{T}}}_{\bullet ,\bullet }) \le \frac{80}{83}\).