Soliton-type metrics associated with weighted CSCK metrics on Fano manifolds Satoshi Nakamura Abstract . We study weighted constant scalar curvature Kähler metrics, introduced by Lahdili as ( v , w ) (v,w) -CSCK metrics, on Fano manifolds and their relationship with soliton-type metrics. In this paper, we introduce a weight function g ⁡ ( v , w ) g(v,w) associated with a pair of weight functions ( v , w ) (v,w) . Assuming that v v and g ⁡ ( v , w ) g(v,w) are positive and log-concave on the moment polytope, we prove that the existence of a ( v , w ) (v,w) -CSCK metric in the first Chern class is equivalent to the existence of a g ⁡ ( v , w ) g(v,w) -soliton. We also explain that a g ⁡ ( v , w ) g(v,w) -soliton arises naturally from Sasaki geometry. More precisely, let ( v , w ) (v,w) be the weight functions defining a weighted CSCK metric in 2 ​ π ​ c 1 ​ ( X ) 2\pi c_{1}(X) which gives rise to a ξ ^ \hat{\xi} -transverse extremal metric on an S 1 S^{1} -bundle N N in the canonical bundle of a Fano manifold X X , where ξ ^ \hat{\xi} is a possibly irregular Reeb field on N N . We prove that the associated g ⁡ ( v , w ) g(v,w) -soliton on X X gives rise to a ξ ^ \hat{\xi} -transverse Mabuchi soliton on N N . 2010 Mathematics Subject Classification Primary 53C25; Secondary 53C55, 58E11. 1. Introduction In his paper [ 28 ] , Lahdili introduced the notion of a weighted constant scalar curvature Kähler metric (a weighted CSCK metric for short) formulated as follows. Let X be an n -dimensional compact Kähler manifold with a Kähler class \Omega , and let T be a maximal compact torus in the reduced automorphism group \mathrm{Aut}_{r}(X) . Fix two smooth functions v>0 and w defined on the moment map image P_{X} . A T -invariant Kähler metric \omega\in\Omega is called a (v,w) -constant scalar curvature Kähler metric (a (v,w) -CSCK metric for short) if S_{v}(\omega)=w(m_{\omega}) , where S_{v}(\omega) is the v -weighted scalar curvature of \omega and m_{\omega} is the moment map for the T -action with respect to \omega . When v\equiv 1 , the weighted scalar curvature equals the ordinary scalar curvature. The (v,w) -CSCK metrics provide a unifying framework that encompasses several well-known classes of canonical metrics, including constant scalar curvature Kähler metrics, extremal Kähler metrics, extremal Sasaki metrics, and so on, as reviewed in Section 2 . In recent years, (v,w) -CSCK metrics itself have attracted considerable attention. The existence of (v,w) -CSCK metrics with log-concave weight v was shown by Han-Liu [ 21 ] and Di Nezza-Jubert-Lahdili [ 15 , 16 ] to be equivalent to the coercivity of the associated Mabuchi-type energy functional. Moreover a Yau-Tian-Donaldson-type correspondence was established very recently by Boucksom-Jonsson [ 10 ] , building on earlier work of Li [ 30 ] (see also [ 22 , 23 ] ), characterizing the existence of (v,w) -CSCK metrics with log-concave weight v in terms of an appropriate notion of K-stability. In the case where X is a Fano manifold and \Omega=2\pi c_{1}(X) , soliton-type metrics called g -solitons, where g is a smooth positive function defined on P_{X} , play a central role within the class of weighted CSCK metrics. A T -invariant metric \omega\in 2\pi c_{1}(X) is called a g -soliton if \mathrm{Ric}_{g}(\omega)=\omega , where \mathrm{Ric}_{g}(\omega) is the g -weighted Ricci curvature form of \omega . When g\equiv 1 , the wighted Ricci form equals the ordinary Ricci form. A g -soliton equals a (g,w) -CSCK metric with w(x)=n+\langle(\log g)^{\prime}(x),x\rangle for x\in P_{X} [ 4 ] . The notion of g -solitons provides a unified framework encompassing Kähler-Einstein metrics, Kähler-Ricci solitons, Mabuchi solitons, Sasaki-Einstein metrics, and so on, as reviewed in Section 2 . Originating in Mabuchi’s work [ 33 ] and further developed by Berman-Witt Nysröm [ 9 ] and Han-Li [ 20 ] , the existence of g -solitons with the log-concave weight g is by now well understood to be characterized by the coercivity of the associated Ding-type functional as well as by an appropriate Ding-type stability condition. Recently, there has been growing progress in the study of relationships between the existence of two, in general different, canonical Kähler metrics on a Fano manifold. For instance, it was proved by Apostolov-Lahdili-Nitta [ 6 ] and by Hisamoto and the author [ 25 ] (see also [ 34 , Section 9.6] ) that the existence of a T -invariant Calabi’s extremal metric [ 14 ] in 2\pi c_{1}(X) whose scalar curvature is strictly less than n+1 is equivalent to that of a T -invariant Mabuchi’s soliton [ 32 ] in 2\pi c_{1}(X) . Here an extremal metric means a (1,n+l_{\mathrm{ext}}) -CSCK metric for an affine function l_{\mathrm{ext}} on P_{X} corresponds to the extremal vector field [ 18 ] on X , and a Mabuchi soliton means a (1-l_{\mathrm{ext}}) -soliton. More generally we already know the following. Theorem 1.1 . ( [ 6 , 25 ] , see also [ 38 ] ) Let X be an n -dimensional Fano manifold. Assume the positive weight function g is log-concave on P_{X} . Then the following conditions are equivalent. (1) There exists a T -invariant (1,1+n-g) -CSCK metric (i.e. g -extremal metric in the sense of [ 25 , 38 ] ) in 2\pi c_{1}(X) . (2) There exists a T -invariant g -soliton in 2\pi c_{1}(X) . The main result of this article is an extension of Theorem 1.1 , formulated within the framework of (v,w) -CSCK metrics on Fano manifolds, which we state as follows. Theorem 1.2 . Let X be an n -dimensional Fano manifold. Assume the positive weight function v is log-concave on P_{X} . Assume also the associated weight function (1.1) g(v,w)(x):=v(x)(1+n+\langle(\log v)^{\prime}(x),x\rangle-w(x)) is positive and log-concave on P_{X} . Then the following conditions are equivalent. (1) There exists a T -invariant (v,w) -CSCK metric in 2\pi c_{1}(X) . (2) There exists a T -invariant g(v,w) -soliton in 2\pi c_{1}(X) . In particular, Theorem 1.2 shows that the existence of a (v,w) -CSCK metric on a Fano manifold is governed by the existence of a soliton-type metric, namely the g(v,w) -soliton. In view of Theorem 1.1 , one obtains the following. Corollary 1.3 . Under the same assumptions as in Theorem 1.2 , the following conditions are equivalent. (1) There exists a T -invariant (v,w) -CSCK metric in 2\pi c_{1}(X) . (2) There exists a T -invariant (1,1+n-g(v,w)) -CSCK metric in 2\pi c_{1}(X) . Remark 1.4 . The author has been informed by V. Apostolov and Y. Nitta that the same results as Theorem 1.2 and Corollary 1.3 were obtained independently in a joint work with Apostolov-Lahdili-Nitta [ 7 ] . One piece of evidence for the equivalence between the existence of (v,w) -CSCK metrics and that of g(v,w) -solitons is that the Futaki-type invariant M_{v,w}^{\prime} of X , which provides an obstruction to the existence of (v,w) -CSCK metrics, coincides with another Futaki-type invariant D_{g(v,w)}^{\prime} of X , which obstructs the existence of g(v,w) -solitons. See Proposition 3.2 . With a suitable formulation of the weight function g(v,w) , the arguments developed by Hisamoto and the author [ 25 ] can be extended. A crucial step in our proof of Theorem 1.2 is to establish Theorem 4.2 which shows the uniform boundedness of the Mabuchi-type functional M_{v,w} associated with (v,w) -CSCK metrics along a continuity path used to construct a g(v,w) -soliton. In Theorem 1.2 , we assume the positivity and log-concavity of the weight functions v and g(v,w) . The positivity assumption is essential in order for the weighted Monge-Ampère measure to be positive. On the other hand, the log-concavity assumption is used in the regularity argument for weak solutions [ 15 , 16 , 20 , 21 ] . At present, it is not clear to the author whether this log-concavity assumption can be relaxed. One of the main significances of Theorem 1.2 is that the existence of (v,w) -CSCK metrics on Fano manifolds can be captured by the existence of g(v,w) -solitons, that is, by the solvability of a complex Monge-Ampère equation, and hence can be characterized by a Ding-type stability condition, which are deeply connected with the theory of minimal model program in birational geometry [ 20 ] . The author expects that Theorem 1.2 may be applied to the moduli theory of Fano manifolds admitting (v,w) -CSCK metrics. An important and interesting application of (v,w) -CSCK metrics and g -solitons is that they allow one to describe canonical Sasaki metrics on a Sasaki manifold with a possibly irregular Reeb field via (quasi-) regular quotients [ 2 , 4 ] . Let X be a Fano manifold and \omega_{0}\in 2\pi c_{1}(X) a Kähler metric invariant under the action of a maximal compact torus T . Let N be the Sasaki manifold with the regular Reeb filed \hat{\chi} associated with (X,\omega_{0}) . Namely \pi:N_{\omega_{0}}\to X is the S^{1} -bundle contained in the canonical bundle K_{X} with respect to the Hermitian metric on K_{X} whose curvature is -\omega_{0} , and \hat{\chi} is the generator of the S^{1} -action. Then one obtains a strongly pseudo-convex CR manifold (N_{\omega_{0}},D_{0},J_{0}) , where D_{0} is the horizontal distribution and J_{0} is induced complex structure on D_{0} . Fix the new torus \hat{T}:=T\times S^{1} acting on N_{\omega_{0}} , where T acts on K_{X} as a lift and S^{1} is generated by \hat{\chi} . Let \xi\in\mathrm{Lie}(T)\subset\mathrm{Lie}(\hat{T}) and consider \hat{\xi}:=\xi+\hat{\chi}\in\mathrm{Lie}(\hat{T}) . Now we assume that \hat{\xi} defines a Reeb field for (N_{\omega_{0}},D_{0},J_{0}) . One has the corresponding contact 1 -form \eta_{0}^{\hat{\xi}} on N_{\omega_{0}} characterized by the conditions \eta_{0}^{\hat{\xi}}(\hat{\xi})=1 and \operatorname{Ker}\eta_{0}^{\hat{\xi}}=D_{0} . In these setting, as observed in Section 5 , the defining equation of a \hat{\xi} -transverse extremal metric [ 13 ] in the basic cohomology class [d\eta_{0}^{\hat{\xi}}]_{B,\hat{\xi}} , which is in fact equal to the basic first Chern class 2\pi c_{1}^{B,\hat{\xi}}(N_{\omega_{0}}) , can be rewritten as the defining equation of a weighted CSCK metric on X in 2\pi c_{1}(X) for an appropriate weights (v_{\xi},w_{\xi}) , and conversely. Similarly, the defining equation of a \hat{\xi} -transverse Mabuchi soliton in [d\eta_{0}^{\hat{\xi}}]_{B,\hat{\xi}} can also be rewritten as that of an appropriate soliton-type metric on X , and conversely. As shown in the following result, an example of a g(v,w) -soliton appears naturally in Sasaki geometry of N_{\omega_{0}} for the possibly irregular Reeb field \hat{\xi} . Moreover as observed in Section 5 this g(v,w) -soliton is associated with weights satisfying v(x)\not\equiv 1 and w(x)\neq n+\langle(\log v)^{\prime}(x),x\rangle in general. Theorem 1.5 . Let (v_{\xi},w_{\xi}) be the weights of a weighted CSCK metric in 2\pi c_{1}(X) giving rise to a \hat{\xi} -transverse extremal metric in [d\eta_{0}^{\hat{\xi}}]_{B,\hat{\xi}} . Then the g(v_{\xi},w_{\xi}) -soliton in 2\pi c_{1}(X) gives rise to a \hat{\xi} -transverse Mabuchi soliton in [d\eta_{0}^{\hat{\xi}}]_{B,\hat{\xi}} . We shall give a proof from a general viewpoint of a \hat{\xi} -transversal g -extremal metric and a \hat{\xi} -transversal g -soliton. See Theorem 5.5 . As an extension of the Kähler case, it is a natural problem to ask whether the existence of a transverse extremal metric whose transverse scalar curvature is strictly less than n+1 is equivalent to that of a transverse Mabuchi soliton. Unfortunately Theorem 1.2 and 1.5 do not currently provide an answer to this question, since the weight functions v_{\xi} and g(v_{\xi},w_{\xi}) are not log-concave in general. However recall that the log-concavity assumption on the weights in Theorem 1.2 is a technical condition used to obtain regularity of weak solutions [ 15 , 16 , 20 , 21 ] . Therefore Theorem 1.2 and 1.5 provide strong evidence supporting the validity of the above, as yet unproven, equivalence. Organization . In Section 2, we recall basic notions on weighted CSCK metrics and soliton-type metrics, including weighted scalar curvature and weighted Ricci curvature. Typical situations in which these metrics appear are also reviewed. Section 3 introduces g(v,w) -solitons and explains their relationship with (v,w) -CSCK metrics from the viewpoint of Futaki-type invariants and weighted energy functionals. In Section 4, we prove Theorem 1.2 by combining coercivity properties of weighted energy functionals with a continuity method for g(v,w) -solitons. Finally, in Section 5, we prove Theorem 1.5 (and Theorem 5.5 for more general viewpoint). For this, we discuss how (v,w) -CSCK metrics and g(v,w) -solitons arise from Sasaki-extremal metrics and Sasaki-Mabuchi solitons via regular quotients. Acknowledgment . The author would like to thank Vestislav Apostolov and Yasufumi Nitta for sharing their research article and for helpful discussions and comments. He would like to thank Tomoyuki Hisamoto for many stimulating discussions. He also would like to thank Eiji Inoue and Abdellah Lahdili for helpful comments. He is supported by JSPS Grant-in-Aid for Early-Career Scientists No. 24K16917. 2. Preliminaries on (v,w) -CSCK metrics and g -solitons In this section we fix some notation to give precise definitions of weighted CSCK metrics and soliton-type metrics. Let X be an n -dimensional compact Kähler manifold with a Kähler class \Omega . Let T be a maximal compact torus in the reduced automorphism group \mathrm{Aut}_{r}(X) , and \mathfrak{t} be the Lie algebra of T . Fix a T -invariant metric \omega_{0}\in\Omega to define the space of T -invariant Kähler potentials \mathcal{H}^{T} with respect to \omega_{0} . For any \varphi\in\mathcal{H}^{T} , the T -invariant metric \omega_{\varphi}:=\omega_{0}+\sqrt{-1}\partial\overline{\partial}\varphi gives rise to the moment map (2.1) m_{\varphi}:X\to\mathfrak{t}^{*} defined by the equation i_{\xi}\omega_{\varphi}=-d\langle m_{\varphi},\xi\rangle and the normalization \int_{X}\langle m_{\varphi},\xi\rangle\omega_{\varphi}^{n}=0 for any \xi\in\mathfrak{t} . It is well known that the image P_{X}:=m_{\varphi}(X) is a convex polytope independent of the choice of \varphi\in\mathcal{H}^{T} . Fix a strictly positive and smooth function v on P_{X} . One can assume that the weighted Monge-Ampère measure (2.2) \mathrm{MA}_{v}(\varphi):=\frac{v(m_{\varphi})\omega_{\varphi}^{n}}{\int_{X}\omega_{0}^{n}} defines a probability measure on X without loss of generality, since the integral \int_{X}v(m_{\varphi})\omega_{\varphi}^{n} is independent of the choice of \varphi\in\mathcal{H}^{T} [ 28 ] . We denote the ordinary Monge-Ampère measure \omega_{\varphi}^{n}/\int_{X}\omega_{0}^{n} by \mathrm{MA}(\varphi) . Following [ 15 ] , we introduce generalized notions of the trace and the Laplacian in the weighted setting. Definition 2.1 . ( [ 15 , Appendix A] ) For any T -invariant (1,1) -form \theta with a moment map m_{\theta} , the v -weighted trace of \theta is defined by (2.3) \mathrm{tr}_{v,\varphi}(\theta):=\mathrm{tr}_{\varphi}(\theta)+\langle(\log v)^{\prime}(m_{\varphi}),m_{\theta}\rangle, where \mathrm{tr}_{\varphi}(\theta) is the ordinary trace defined by the equation \mathrm{tr}_{\varphi}(\theta)\omega_{\varphi}^{n}=n\theta\wedge\omega_{\varphi}^{n-1} . Equivalently, in a fixed basis (\xi_{1},\dots,\xi_{r}) of \mathfrak{t} , (2.4) \mathrm{tr}_{v,\varphi}(\theta)=\mathrm{tr}_{\varphi}(\theta)+\frac{1}{v(m_{\varphi})}\sum_{j}v_{,j}(m_{\varphi})m^{j}_{\theta}, where v_{,j}:=\frac{\partial v}{\partial x_{j}} and m_{\theta}=(m_{\theta}^{1},\dots,m_{\theta}^{r}). Definition 2.2 . ( [ 15 , Appendix A] ) For any smooth function f on X , the v -weighted Laplacian for \varphi\in\mathcal{H}^{T} is defined by (2.5) \displaystyle\Delta_{v,\varphi}f \displaystyle:= \displaystyle-\frac{1}{v(m_{\varphi})}\overline{\partial}^{*}(v(m_{\varphi})\overline{\partial}f)=\Delta_{\varphi}f+\frac{1}{v(m_{\varphi})}\langle\overline{\partial}v(m_{\varphi}),\overline{\partial}f\rangle, where \Delta_{\varphi} is the ordinary (negative) Laplacian for \varphi . In particular \Delta_{v,\varphi}f=\mathrm{tr}_{v,\varphi}(\sqrt{-1}\partial\overline{\partial}f) . It follows from the definition that the operator \Delta_{v,\varphi} is elliptic and self-adjoint with respect to the v -twisted Hermitian inner product \left\langle f_{1},f_{2}\right\rangle_{v,\varphi}:=\int_{X}f_{1}\overline{f_{2}}\,\mathrm{MA}_{v}(\varphi) on C^{\infty}(X;\mathbb{C}) , and the kernel of \Delta_{v,\varphi} consists of the constant functions on X . By considering the weighted Monge-Ampère measure and the weighted trace, one can define the notions of weighted curvatures. Definition 2.3 . The v -weighted Ricci curvature for \varphi\in\mathcal{H}^{T} is defined by the formula (2.6) \mathrm{Ric}_{v}(\omega_{\varphi}):=-\sqrt{-1}\partial\overline{\partial}\log\mathrm{MA}_{v}(\varphi)=\mathrm{Ric}(\omega_{\varphi})-\sqrt{-1}\partial\overline{\partial}\log v(m_{\varphi}), where \mathrm{Ric}(\omega_{\varphi}):=-\sqrt{-1}\partial\overline{\partial}\log\mathrm{MA}(\varphi) is the ordinary Ricci curvature for \varphi . The v -weighted scalar curvature for \varphi\in\mathcal{H}^{T} is defined by the formula (2.7) S_{v}(\varphi):=\mathrm{tr}_{v,\varphi}\mathrm{Ric}_{v}(\omega_{\varphi}). 2.1. Weighted CSCK metrics In order to define a weighted CSCK metric, let w be a smooth function on P_{X} satisfying \int_{X}S_{v}(\varphi)\mathrm{MA}_{v}(\varphi)=\int_{X}w(m_{\varphi})\mathrm{MA}_{v}(\varphi) . Note that the integrals on both sides are independent of the choice of a metric \varphi\in\mathcal{H}^{T} [ 28 ] . Definition 2.4 . ( [ 28 ] ) A metric \varphi\in\mathcal{H}^{T} is called a (v,w) constant scalar curvature Kähler metric (a (v,w) -CSCK metric for short) if (2.8) S_{v}(\varphi)=w(m_{\varphi}). Remark 2.5 . Lahdili [ 28 , Section 2] originally defined the weighted scalar curvature as v(m_{\varphi})\mathrm{tr}_{v,\varphi}\mathrm{Ric}_{v}(\omega_{\varphi}) . We adopted the definition introduced by Boucksom-Jonsson-Trusiani [ 11 , Section 3] . Namely, a (v,w) -CSCK metric as we adopted in this article is equivalent to a (v,vw) -CSCK metric in the sense of Lahdili. Typical situations in which a weighted CSCK metric appears are as follows. • If v and w are constant function, the weighted CSCK metric is the ordinary CSCK metric. • If v is constant and w is an affine function, the weighted CSCK metric is the extremal Kähler metric introduced by Calabi [ 14 ] . • If v and w are appropriate polynomials, the weighted CSCK metric describes Calabi’s extremal metric on the total space of an holomorphic fibration Y with fiber X , called semisimple principle fibration [ 3 , 4 , 27 ] . • If v=e^{l} for an affine function and w=l+a for a constant a\in\mathbb{R} determined by the relation ( 2.14 ), the weighted CSCK metric is the \mu -CSCK metric, in the sense of Inoue [ 26 ] , extending the notion of Kähler-Ricci solitons on Fano manifolds to general Kähler manifolds. • If \Omega=2\pi c_{1}(L) for an ample line bundle over X , v=l^{-n-1} for a positive affine function and w=al^{-1} for a constant a\in\mathbb{R} , then the weighted CSCK metric describes a constant scalar curvature Sasaki metric on the unit circle bundle associated with (L^{-1})^{\times} . In the same setting as above, if one instead takes w=l^{\prime}l^{-1} for an affine function l^{\prime} , the weighted CSCK metric describes a Sasaki extremal metric on the unit circle bundle associated with (L^{-1})^{\times} . See [ 1 , 2 ] . • If v=l^{-2n+1} and w=l^{-2} for a positive affine function l , the weighted CSCK metric describes a Kähler metric which is conformal to an Einstein-Maxwell metric [ 8 ] . 2.2. Soliton-type metrics Let us assume that X is Fano and \Omega=2\pi c_{1}(X) . For any metric \varphi\in\mathcal{H}^{T} , one obtains the Ricci potential function \rho_{\varphi} defined by (2.9) \mathrm{Ric}(\omega_{\varphi})-\omega_{\varphi}=\sqrt{-1}\partial\overline{\partial}\rho_{\varphi}\quad\text{and}\quad\int_{X}(e^{\rho_{\varphi}}-1)\mathrm{MA}(\varphi)=0. In order to introduce soliton-type metrics, fix a strictly positive and smooth function g on P_{X} such that \mathrm{MA}_{g}(\varphi) is a probability measure on X for any \varphi\in\mathcal{H}^{T} . One can also define the g -weighted Ricci potential function \rho_{g,\varphi} by the formulas (2.10) \mathrm{Ric}_{g}(\omega_{\varphi})-\omega_{\varphi}=\sqrt{-1}\partial\overline{\partial}\rho_{g,\varphi}\quad\text{and}\quad\int_{X}(e^{\rho_{g,\varphi}}-1)\mathrm{MA}_{g}(\varphi)=0. It follows from the definitions that \rho_{g,\varphi}=\rho_{\varphi}-\log g(m_{\varphi}) . Definition 2.6 . ( [ 9 , 20 ] ) A metric \varphi\in\mathcal{H}^{T} is called a g -soliton if (2.11) \mathrm{Ric}_{g}(\omega_{\varphi})=\omega_{\varphi},\quad\text{equivalently,}\quad\rho_{g,\varphi}=0. Let \mu_{\varphi} be the canonical measure for \varphi\in\mathcal{H}^{T} defined by (2.12) \mu_{\varphi}=\frac{e^{-\varphi+\rho_{0}}\mathrm{MA}(0)}{\int_{X}e^{-\varphi+\rho_{0}}\mathrm{MA}(0)}. A g -soliton \varphi is characterized by the complex Monge-Ampère equation \mathrm{MA}_{g}(\varphi)=\mu_{\varphi} . We will use the following formula frequently later in this paper. (2.13) \mu_{\varphi}=e^{\rho_{\varphi}}\mathrm{MA}(\varphi)=e^{\rho_{g,\varphi}}\mathrm{MA}_{g}(\varphi). A g -soliton is a weighted CSCK metric in 2\pi c_{1}(X) with respect to appropriate weight functions. Proposition 2.7 . ( [ 4 ] ) A metric \varphi\in\mathcal{H}^{T} is a g -soliton if and only if it is a (v,w) -CSCK metric with the weights defined by, for any x\in P_{X} , (2.14) v(x)=g(x)\quad\text{and}\quad w(x)=n+\langle(\log g)^{\prime}(x),x\rangle. Proof.. In general, by taking the g -weighted trace of the defining equation ( 2.10 ) of the g -weighted Ricci potential \rho_{g,\varphi} , one has the formula (2.15) \Delta_{g,\varphi}\rho_{g,\varphi}=S_{g}(\varphi)-n-\langle(\log g)^{\prime}(m_{\varphi}),m_{\varphi}\rangle. The proposition follows directly from the formula ( 2.15 ). ∎ Typical situations in which a g -soliton appears are as follows. • If g is constant, the g -soliton is a Kähler-Einstein metric. • If g=e^{l} for an affine function l , the g -soliton is a Kähler-Ricci soliton [ 39 ] . • If g is a positive affine function, the g -soliton is a Mabuchi soliton [ 32 ] . • If X is toric Fano and g is an appropriate polynomial, the g -soliton describes a Kähler-Einstein metric on a total space of fibration Y with fiber X , called KSM manifolds [ 36 , 37 ] . • If g=l^{-n-2} for a positive affine function l , then the g -soliton describes a Sasaki-Einstein metric on the unit circle bundle associated with (K_{X})^{\times} [ 4 ] . 3. g(v,w) -solitons In this section we explain some background for introducing g(v,w) -solitons from viewpoints of Futaki-type invariants and energy functionals. 3.1. Futaki-type invariants Let (X,\Omega) be a compact Kähler manifold. For any affine function l on \mathfrak{t}^{*} , we define (3.1) M^{\prime}_{v,w}(l):=-\int_{X}l(m_{0})\Big(S_{v}(0)-w(m_{0})\Big)\mathrm{MA}_{v}(0). Lahdili [ 28 ] showed that the quantity M^{\prime}_{v,w}(l) is independent of the choice of a T -invariant metric \omega_{0}\in\Omega . Thus if there exists a T -invariant (v,w) -CSCK metric in \Omega then M^{\prime}_{v,w} must vanish identically. The invariant M^{\prime}_{v,w} gives an obstruction to the existence of a T -invariant (v,w) -CSCK metric. We call M^{\prime}_{v,w} the (v,w) -Futaki invariant. Let us assume that X is Fano and \Omega=2\pi c_{1}(X) . Let us consider describing the (v,w) -Futaki invariant using the v -weighted Ricci potential \rho_{v} . For this we use the following lemma. Lemma 3.1 . Fix a holomorphic vector field \xi\in\mathfrak{t} with the potential function l(m_{\varphi}) for an affine function l . The following equation holds. (3.2) \Delta_{v,\varphi}l(m_{\varphi})+l(m_{\varphi})+\xi(\rho_{v,\varphi})=\int_{X}l(m_{\varphi})e^{\rho_{v,\varphi}}\mathrm{MA}_{v}(\varphi). Proof.. Note that L_{\xi}\omega_{\varphi}=\sqrt{-1}\partial\overline{\partial}l(m_{\varphi}) by definition of the moment map. By the Lie derivative of the defining equation of the weighted Ricci potential ( 2.10 ), (3.3) \Delta_{v,\varphi}l(m_{\varphi})+l(m_{\varphi})+\xi(\rho_{v,\varphi})=C for a constant C which is determined by the normalization (3.4) 0=\int_{X}L_{\xi}\Big(e^{\rho_{v,\varphi}}\mathrm{MA}_{v}(\varphi)\Big)=\int_{X}\Big(\Delta_{v,\varphi}l(m_{\varphi})+\xi(\rho_{v,\varphi})\Big)e^{\rho_{v,\varphi}}\mathrm{MA}_{v}(\varphi). ∎ Then the weight function (3.5) g(v,w)(x):=v(x)\left((1+n+\langle(\log v)^{\prime}(x),x\rangle-w(x)\right) naturally appears, as shown in the following proposition. Proposition 3.2 . For any affine function l on \mathfrak{t}^{*} , we define (3.6) \displaystyle D^{\prime}_{g(v,w)}(l) \displaystyle:= \displaystyle-\int_{X}l(m_{0})\Big(\frac{g(v,w)(m_{0})}{v(m_{0})}-e^{\rho_{v,0}}\Big)\mathrm{MA}_{v}(0) (3.7) \displaystyle= \displaystyle-\int_{X}l(m_{0})\Big(g(v,w)(m_{0})-e^{\rho_{0}}\Big)\mathrm{MA}(0). Then we have M^{\prime}_{v,w}(l)=D^{\prime}_{g(v,w)}(l) . Proof.. Use the formula ( 2.15 ) and Lemma 3.1 to show \displaystyle M^{\prime}_{v,w}(l) \displaystyle= \displaystyle\int_{X}-\langle\overline{\partial}l(m_{0}),\overline{\partial}\rho_{v,0}\rangle\mathrm{MA}_{v}(0)+\int_{X}l(m_{0})\Big(\frac{g(v,w)(m_{0})}{v(m_{0})}-1\Big)\mathrm{MA}_{v}(0) \displaystyle= \displaystyle\int_{X}-\xi(\rho_{v,0})\mathrm{MA}_{v}(0)+\int_{X}l(m_{0})\Big(\frac{g(v,w)(m_{0})}{v(m_{0})}-1\Big)\mathrm{MA}_{v}(0) \displaystyle= \displaystyle\int_{X}\Big(\Delta_{v,0}l(m_{0})+l(m_{0})-\int_{X}l(m_{0})e^{\rho_{v,0}}\mathrm{MA}_{v}(0)\Big)\mathrm{MA}_{v}(0) \displaystyle+\int_{X}l(m_{0})\Big(\frac{g(v,w)(m_{0})}{v(m_{0})}-1\Big)\mathrm{MA}_{v}(0) \displaystyle= \displaystyle D^{\prime}_{g(v,w)}(l). ∎ In other words, the existence of a (v,w) -CSCK metric and that of a g(v,w) -soliton are obstructed by the same Futaki-type invariant for a Fano manifold X . Proposition 3.2 provides evidence for the validity of the statement that the existence conditions for a (v,w) -CSCK metric and a g(v,w) -soliton are equivalent. 3.2. Energy functionals Let (X,\Omega) be a compact Kähler manifold. Analogously to Mabuchi’s definition of the Mabuchi functional [ 31 ] by integrating the Futaki invariant [ 17 ] , one has a functional on \mathcal{H}^{T} admitting a (v,w) -CSCK metric as its critical point by integrating M_{v,w}^{\prime} . The (v,w) -weighted Mabuchi functional M_{v,w}:\mathcal{H}^{T}\to\mathbb{R} is defined by (3.8) M_{v,w}(\varphi):=-\int_{0}^{1}dt\int_{X}\dot{\varphi}_{t}\Big(S_{v}(\varphi_{t})-w(m_{\varphi_{t}})\Big)\mathrm{MA}_{v}(\varphi_{t}), where \varphi_{t} is a path in \mathcal{H}^{T} connecting 0 and \varphi . The value M_{v,w}(\varphi) is independent of the choice of this path. Indeed, a so-called Chen-Tian-type formula, which gives a path-independent expression, is known [ 11 , 28 ] . Let us assume that X is Fano and \Omega=2\pi c_{1}(X) . Fix a smooth positive function g on P_{X} . Similar to M_{v,w} , one has a functional admitting a g -soliton as its critical point. The g -weighted Ding functional D_{g}:\mathcal{H}^{T}\to\mathbb{R} is defined by (3.9) D_{g}(\varphi):=-\int_{0}^{1}dt\int_{X}\dot{\varphi}_{t}\Big(g(m_{\varphi_{t}})-e^{\rho_{\varphi_{t}}}\Big)\mathrm{MA}(\varphi_{t}), where \varphi_{t} is a path in \mathcal{H}^{T} connecting 0 and \varphi . Put (3.10) E_{g}(\varphi):=\int_{0}^{1}dt\int_{X}\dot{\varphi}_{t}\mathrm{MA}_{g}(\varphi_{t})\quad\text{and}\quad L(\varphi):=-\log\int_{X}e^{\rho_{0}-\varphi}\mathrm{MA}(0). Then D_{g}=-E_{g}+L . These functionals are defined independently of the choice of a path \varphi_{t} [ 9 , 20 ] . The following proposition, which plays important roles in our proof of Theorem 1.2 , may be viewed as an energy functionals version of Proposition 3.2 . Proposition 3.3 . For any \varphi\in\mathcal{H}^{T} , one has (3.11) M_{v,w}(\varphi)-D_{g(v,w)}(\varphi)=-\int_{X}\rho_{v,\varphi}\mathrm{MA}_{v}(\varphi)+\int_{X}\rho_{v,0}\mathrm{MA}_{v}(0). In particular, (3.12) M_{v,w}(\varphi)-D_{g(v,w)}(\varphi)\geqslant\int_{X}\rho_{v,0}\mathrm{MA}_{v}(0). Proof.. Use the variation formula \delta\rho_{v,\varphi}=-\Delta_{v,\varphi}\delta\varphi-\delta\varphi+\int_{X}\delta\varphi d\mu_{\varphi} , where \mu_{\varphi} is the canonical measure defined as ( 2.12 ), to show \displaystyle\delta\Big(-\int_{X}\rho_{v,\varphi}\mathrm{MA}_{v}(\varphi)\Big) \displaystyle= \displaystyle\int_{X}\Big(\Delta_{v,\varphi}\delta\varphi+\delta\varphi-\int_{X}\delta\varphi d\mu_{\varphi}\Big)\mathrm{MA}_{v}(\varphi)-\int_{X}\rho_{v,\varphi}(\Delta_{v,\varphi}\delta\varphi)\mathrm{MA}_{v}(\varphi) \displaystyle= \displaystyle\int_{X}\delta\varphi\Big[\Big(\frac{g(v,w)(m_{\varphi})}{v(m_{\varphi})}+w(m_{\varphi})\Big)\mathrm{MA}_{v}(\varphi)-d\mu_{\varphi}\Big]-\int_{X}\delta\varphi S_{v}(\varphi)\mathrm{MA}_{v}(\varphi) \displaystyle= \displaystyle\delta\left(M_{v,w}(\varphi)-D_{g(v,w)}(\varphi)\right). By integration, one obtains the equality ( 3.11 ). The inequality ( 3.12 ) follows from the elementary inequality e^{x}\geqslant 1+x and the normalization of \rho_{v,\varphi} , that is, (3.13) -\int_{X}\rho_{v,\varphi}\mathrm{MA}_{v}(\varphi)\geqslant\int_{X}(1-e^{\rho_{v,\varphi}})\mathrm{MA}_{v}(\varphi)=0. ∎ 4. Proof of Theorem 1.2 In this section we prove Theorem 1.2 . Our proof uses the coercivity theorems which characterizes the existence of a canonical metric in terms of a behavior of a corresponding functional on the space of Kähler metrics. Let I:\mathcal{H}^{T}\to\mathbb{R} be the Aubin’s I -functional defined by (4.1) I(\varphi):=\int_{X}\varphi\left(\mathrm{MA}(0)-\mathrm{MA}(\varphi)\right), which plays a role of a distance function in \mathcal{H}^{T} . In the following theorem, we say that a functional F:\mathcal{H}^{T}\to\mathbb{R} is T^{\mathbb{C}} -coercive if there exists a positive constant \varepsilon>0 such that the inequality (4.2) F(\varphi)\geqslant\varepsilon\inf_{\sigma\in T^{\mathbb{C}}}I(\sigma[\varphi])-\varepsilon^{-1} holds for any \varphi\in\mathcal{H}^{T} , where \sigma[\varphi]\in\mathcal{H}^{T} is defined by \sigma^{*}\omega_{\varphi}=\omega_{0}+\sqrt{-1}\partial\overline{\partial}\sigma[\varphi] . Theorem 4.1 . Assume that the weight v is log-concave on P_{X} . • ( [ 15 , 16 , 21 ] ) There exists a (v,w) -CSCK metric in \mathcal{H}^{T} if and only if the weighted Mabuchi functional M_{v,w} is T^{\mathbb{C}} -coercive. • ( [ 20 ] ) Let X be a Fano manifold and \Omega=2\pi c_{1}(X) . There exists a v -soliton in \mathcal{H}^{T} if and only if the weighted Ding functional D_{v} is T^{\mathbb{C}} -coercive. Let us focus on the case where X is Fano and \Omega=2\pi c_{1}(X) to prove Theorem 1.2 . One direction is easier to follow. In fact it follows from Theorem 4.1 and the inequality ( 3.12 ) in Proposition 3.3 that the existence of a g(v,w) -soliton implies that of a (v,w) -CSCK metric. As for the converse direction, there is currently no known method to directly deduce the coercivity of D_{g(v,w)} from that of M_{v,w} . Our approach is to construct a g(v,w) -soliton by a continuity method under the assumption on the T^{\mathbb{C}} -coercivity of M_{v,w} . Let us consider the following PDE for \varphi\in\mathcal{H}^{T} : (4.3) \mathrm{MA}_{g(v,w)}(\varphi)=e^{-t\varphi+\rho_{0}}\mathrm{MA}(0), where t\in[0,1] . A solution of the equation ( 4.3 ) at t=1 is nothing but a g(v,w) -soliton. Let \mathcal{T} be the set of t\in[0,1] such that the equation ( 4.3 ) has a solution in \mathcal{H}^{T} . According to [ 25 , Section 3] , the set \mathcal{T} is non-empty and open in general. Moreover, according to [ 25 , Section 5] , if for any \varepsilon\in(0,1) there exists a constant C>0 independent of t such that (4.4) \inf_{\sigma\in T^{\mathbb{C}}}I(\sigma[\varphi_{t}])\leqslant C holds for any solution \varphi_{t}\in\mathcal{H}^{T} of ( 4.3 ) at any t\in\mathcal{T}\cap(\varepsilon,1) , then the set \mathcal{T} is closed, which implies 1\in\mathcal{T} . In view of the assumption on the T^{\mathbb{C}} -coercivity of M_{v,w} , the rest of our proof of Theorem 1.2 is deduced to that of the following. Theorem 4.2 . Fix any \varepsilon\in(0,1) . Let \varphi_{t} be a T -invariant solution of ( 4.3 ) at t\in(\varepsilon,1) . There exists a constant C>0 independent of \varepsilon and t satisfying (4.5) M_{v,w}(\varphi_{t})\leqslant C\varepsilon^{-1}. In the rest of this section we prove Theorem 4.2 . Let \varphi_{t}\in\mathcal{H}^{T} be a solution of the equation ( 4.3 ) at t . In order to obtain the upper bound of M_{v,w}(\varphi_{t}) , it suffices to control D_{g(v,w)}(\varphi_{t})-\int_{X}\rho_{v,\varphi_{t}}\mathrm{MA}_{v}(\varphi_{t}) from above by the formula ( 3.11 ) in Proposition 3.3 . Put \omega_{\varphi_{t}}:=\omega_{t} , \rho_{t}:=\rho_{\varphi_{t}} , \rho_{v,t}:=\rho_{v,\varphi_{t}} , v_{t}:=v(m_{\varphi_{t}}) and g_{t}:=g(v,w)(m_{\varphi_{t}}) for simplicity. By definition of the weighted Ding functional D_{g(v,w)} and [ 25 , Proposition 4.4] , (4.6) D_{g(v,w)}(\varphi_{t})=-E_{g(v,w)}(\varphi_{t})+L(\varphi_{t})\leqslant L(\varphi_{t}). On the other hand, we observe the following to control -\int_{X}\rho_{v,\varphi_{t}}\mathrm{MA}_{v}(\varphi_{t}) . Lemma 4.3 . Let \varphi_{t}\in\mathcal{H}^{T} be a solution of ( 4.3 ) at t . One has the formula (4.7) \rho_{v,t}=-(1-t)\varphi_{t}+\log\frac{g_{t}}{v_{t}}+L(\varphi_{t}), where L is the functional defined in ( 3.10 ). Note that by the compactness of X , there exists a uniform constant C>0 independent of the choice of \varphi\in\mathcal{H}^{T} such that (4.8) C^{-1}\leqslant\frac{g(v,w)(m_{\varphi})}{v(m_{\varphi})}\leqslant C. Proof.. By definition of the weighted Ricci curvature, (4.9) \mathrm{Ric}_{v}\omega_{t}=\sqrt{-1}\partial\overline{\partial}\log\frac{g_{t}}{v_{t}}+\mathrm{Ric}_{g(v,w)}(\omega_{t}). By the formula ( 4.9 ), definition of the v -weighted Ricci potential and the equation ( 4.3 ), (4.10) \sqrt{-1}\partial\overline{\partial}\rho_{v,t}=\sqrt{-1}\partial\overline{\partial}\left(\log\frac{g_{t}}{v_{t}}-(1-t)\varphi_{t}\right). Thus \rho_{v,t}=-(1-t)\varphi_{t}+\log g_{t}/v_{t}+c_{t} for some constant c_{t} depending on t . The constant c_{t} is determined by the normalization of \rho_{v,t} and the equation ( 4.3 ). In fact c_{t}=L(\varphi_{t}) . This completes the proof. ∎ Integrating the both side of the relation in Lemma 4.3 , one thus obtains (4.11) -\int_{X}\rho_{v,\varphi_{t}}\mathrm{MA}_{v}(\varphi_{t})\leqslant(1-t)\int_{X}\varphi_{t}\mathrm{MA}_{v}(\varphi_{t})-\log\left(\inf_{P_{X}}\frac{g(v,w)}{v}\right)-L(\varphi_{t}). Lemma 4.4 . For a fixed \varepsilon\in(0,1) , a solution \varphi_{t} at t>\varepsilon satisfies (4.12) \int_{X}\varphi_{t}\mathrm{MA}_{v}(\varphi_{t})\leqslant e^{\sup\rho_{0}-1}\left(\inf_{P_{X}}\frac{g(v,w)}{v}\right)^{-1}\varepsilon^{-1}. Proof.. Using the equation ( 4.3 ) of the continuity method, one obtains \displaystyle\int_{X}\varphi_{t}\,\mathrm{MA}_{v}(\varphi_{t}) \displaystyle= \displaystyle\int_{X}\varphi_{t}\frac{v_{t}}{g_{t}}\,\mathrm{MA}_{g}(\varphi_{t}) \displaystyle\leqslant \displaystyle\int_{\{\varphi_{t}>0\}}\varphi_{t}\frac{v_{t}}{g_{t}}e^{-t\varphi_{t}+\rho_{0}}\,\mathrm{MA}(0) \displaystyle\leqslant \displaystyle e^{\sup\rho_{0}}\left(\inf_{P_{X}}\frac{g(v,w)}{v}\right)^{-1}\int_{\{\varphi_{t}>0\}}\varphi_{t}e^{-t\varphi_{t}}\mathrm{MA}(0). Note that for any t>\varepsilon , the function \mathbb{R}_{>0}\ni x\mapsto xe^{-tx} is bounded from above by the constant (e\varepsilon)^{-1} . This completes the proof. ∎ Putting together the above arguments, Theorem 4.2 follows from the relation ( 3.11 ), the estimate ( 4.6 ) and Lemma 4.4 . This completes the proof of Theorem 1.2 . 5. (v,w) -CSCK metrics and g(v,w) -solitons in Sasaki geometry The purpose of this section is to prove Theorem 1.5 (and Theorem 5.5 for more general viewpoint) explaining how the weight function g(v,w) arises naturally from Sasaki geometry. For this, we discuss Sasaki-extremal metrics and Sasaki-Mabuchi solitons (for a possibly irregular Reeb field) on a Sasaki manifold admitting a regular Reeb field. We shall see that a Sasaki-extremal metric and a Sasaki-Mabuchi soliton give rise, via the regular quotient, to a (v,w) -CSCK metric and a g(v,w) -soliton respectively on the underlying Fano manifold with the same weights (v,w) . More generally our argument also applies to a transverse g -extremal metric and a transverse g -soliton. We refer to [ 12 , 35 ] for basics of Sasaki manifolds. We also refer to [ 19 , 29 ] for differential geometry in Sasaki manifolds including transverse Kähler geometry. 5.1. Basic setup Let (N,g) be a compact Sasaki manifold of real dimension 2n+1 for n\geqslant 1 i.e. the cone (C(N),\bar{g}):=(N\times\mathbb{R}_{+},r^{2}g+dr^{2}) is a Kähler manifold of complex dimension n+1 . The manifold N is often identified with the submanifold \{r=1\}\subset C(N) . A Sasaki manifold (N,g) admits a contact structure (\hat{\xi},\eta,\Phi) and also admits a one dimensional foliation \mathcal{F}_{\hat{\xi}} . Here the killing vector field \hat{\xi} on (N,g) is called the Reeb vector field, \eta is called the contact 1 -form N , \Phi is a (1,1) tensor on N which defines a complex structure on the contact sub-bundle D:=\operatorname{Ker}\eta . A Sasaki manifold is also denoted by (N,\hat{\xi},\eta,D,\Phi) . The data (\hat{\xi},\eta,D,\Phi) is called a Sasaki structure on N . A Sasaki manifold (N,\hat{\xi},\eta,D,\Phi) is said to be (i) regular if \mathcal{F}_{\hat{\xi}} is obtained by a free S^{1} -action, (ii) quasi-regular if all the leaves of \mathcal{F}_{\hat{\xi}} is compact and (iii) irregular if it is not quasi-regular. The Sasaki structure (\hat{\xi},\eta,D,\Phi) on N yields a transverse holomorphic structure and a transverse Kähler structure on \mathcal{F}_{\hat{\xi}} . One has the splitting of complex valued basic p -forms and the derivative: (5.1) \Lambda_{B}(N)^{p}=\oplus_{i+j=p}\Lambda_{B}^{i,j}(N),\quad d_{B}:=d|_{\Lambda_{B}^{p}(N)}=\partial_{B}+\overline{\partial}_{B}. Put d_{B}^{c}=\frac{\sqrt{-1}}{2}(\overline{\partial}_{B}-\partial_{B}) so that d_{B}d_{B}^{c}=\sqrt{-1}\partial_{B}\overline{\partial_{B}} . If we emphasize the Reeb field \hat{\xi} , we write d_{\hat{\xi}},\partial_{\hat{\xi}},\overline{\partial}_{\hat{\xi}} and d^{c}_{\hat{\xi}} . Let \omega be a basic Kähler form on N . The transverse Ricci form \mathrm{Ric}^{T} for \omega is defined as \mathrm{Ric}^{T}=-\sqrt{-1}\partial_{B}\overline{\partial_{B}}\log\det\omega^{n} . The basic cohomology class represented by \mathrm{Ric}^{T}/2\pi is independent of the choice of \omega . This basic cohomology class [\mathrm{Ric}^{T}/2\pi]_{B,\hat{\xi}} is called the basic first Chern class and is denoted by c^{B,\hat{\xi}}_{1}(N) . 5.2. Transverse g -solitons and g -extremal metrics Let \mathrm{Ric}^{T} is the transverse Ricci form for \omega^{T}:=d\eta . The transverse scalar curvature S^{T} for \omega^{T} is defined by S^{T}(\omega^{T})^{n}\wedge\eta=n\mathrm{Ric}^{T}\wedge(\omega^{T})^{n-1}\wedge\eta . Let V:=\int_{N}(\omega^{T})^{n}\wedge\eta be the volume of N which is independent of the choice of the transverse Kähler metric in [d\eta]_{B,\hat{\xi}} . The average \underline{S}^{T}=\int_{N}S^{T}(\omega^{T})^{n}\wedge\eta/V of S^{T} is also independent of the choice of \eta . Let \hat{T} be a real maximal compact torus acting effectively as CR-automorphisms of (D=\ker\eta,\Phi|_{D}) such that \hat{\xi}\in\hat{\mathfrak{t}}:=\mathrm{Lie}(\hat{T}) . One has the contact moment map \mu^{\hat{\xi}}:C(N)\to\hat{\mathfrak{t}}^{*} defined by (5.2) \langle\mu^{\hat{\xi}},u\rangle=r^{2}\eta(u). Let \mathcal{C}^{*}\subset\hat{\mathfrak{t}^{*}} be the image of \mu^{\hat{\xi}} which is a convex rational polyhedral cone. The image of the restriction \mu^{\hat{\xi}}|_{N} is given by (5.3) P_{\hat{\xi}}:=\Set{x\in\mathcal{C}^{*}}{\langle x,\hat{\xi}\rangle=1} and is called the characteristic hyperplane. Let \mathcal{C}:=\Set{y\in\hat{\mathfrak{t}}}{\langle x,y\rangle>0\text{ for any }x\in\mathcal{C}^{*}} be the dual cone of \mathcal{C}^{*} . An element \hat{\chi}\in\hat{\mathfrak{t}} lies in \mathcal{C} if and only if \eta(\hat{\chi})>0 . Thus \hat{\chi}\in\mathcal{C} defines a Reeb filed on N [ 24 ] . The cone \mathcal{C} is called the Reeb cone or the Sasaki cone. For any Reeb field \hat{\chi}\in\mathcal{C} , there exists a natural projection (5.4) P_{\hat{\xi}}\ni x\mapsto\frac{x}{\langle x,\hat{\chi}\rangle}\in P_{\hat{\chi}}. In order to define the notions of a transverse g -extremal metric and a transverse g -soliton, let g be a smooth positive function on P_{\hat{\xi}} satisfying \int_{N}g(\mu^{\hat{\xi}})(\omega^{T})^{n}\wedge\eta=V . The positivity of the weight function g is not necessarily required to define transverse g -extremal metrics, whereas it is essential for defining transverse g -solitons. Definition 5.1 . A transverse Kähler metric \omega^{T} is called a transverse g -extremal metric if (5.5) S^{T}-\underline{S}^{T}=1-g(\mu^{\hat{\xi}}|_{N}). If g is an affine function, the transverse g -extremal metric is an extremal Sasaki metric in the sense of Boyer-Galicki-Simanca [ 13 ] . Let us now assume transversely Fano condition 2\pi c^{B,\hat{\xi}}_{1}(N)=[d\eta]_{B,\hat{\xi}} . In this case \underline{S}^{T}=n . The condition 2\pi c^{B,\hat{\xi}}_{1}(N)=[d\eta]_{B,\hat{\xi}} follows from the topological assumptions c^{B,\hat{\xi}}_{1}(N)>0 and c_{1}(D)=0 and so-called the D -homothetic transformation for the radial function on the Riemannian cone C(N) . In fact c_{1}(D)=0 if and only if 2\pi c^{B,\hat{\xi}}_{1}(N) is represented by \tau d\eta for some constant \tau [ 19 ] . Definition 5.2 . A transverse Kähler metric \omega^{T} is called a transverse g -soliton if (5.6) \mathrm{Ric}^{T}-\omega^{T}=\sqrt{-1}\partial_{B}\overline{\partial_{B}}\log g(\mu^{\hat{\xi}}|_{N}) If g=e^{l} for an affine function l , the transverse g -soliton is a Sasaki Ricci soliton in the sense of Futaki-Ono-Wang [ 19 ] . If g is an affine function, the transverse g -soliton should be called a Sasaki Mabuchi soliton. 5.3. Regular quotient Let X be a Fano manifold, T a maximal compact torus in the automorphism group \mathrm{Aut}(X) and \omega_{0}\in 2\pi c_{1}(X) a T -invariant Kähler metric. One has the associated moment map m_{0}:X\to\mathfrak{t}^{*} with the moment polytope P_{X} . Let us consider the unitary S^{1} -bundle \pi:N_{\omega_{0}}\to X contained in the canonical bundle K_{X} with respect to the Hermitian metric on K_{X} whose curvature is -\omega_{0} . The new torus \hat{T}:=T\times S^{1} acts on N_{\omega_{0}} , where S^{1} acts on the fibers, preserving the induced CR structure (D_{0},J_{0}) . Here D_{0} is the horizontal distribution and J_{0} is induced complex structure on D_{0} . In fact \hat{T} is a maximal compact torus acting effectively as CR automorphisms. Let \hat{\chi}\in\hat{\mathfrak{t}}:=\mathrm{Lie}(\hat{T}) be the generator of the S^{1} -action. The corresponding 1 -form \eta_{0}^{\hat{\chi}} is uniquely determined as the connection 1 -form on N_{\omega_{0}} with the curvature form d\eta_{0}^{\hat{\chi}}=\pi^{*}\omega_{0} . The kernel of \eta^{\hat{\chi}} is nothing but the horizontal distribution D_{0} . Thus the transverse Kähler structure for the Sasaki structure (\hat{\chi},\eta_{0}^{\hat{\chi}},D_{0},J_{0}) on N_{\omega_{0}} is the pull back of \omega_{0} to (D_{0},J_{0}) . In particular the transverse Fano condition 2\pi c^{B,\hat{\chi}}_{1}(N_{\omega_{0}})=[d\eta_{0}^{\hat{\chi}}]_{B,\hat{\chi}} holds. Let \xi\in\mathfrak{t} to define the affine linear function l_{\xi}(x)=\langle x,\xi\rangle+1 on P_{X} . Let us consider (5.7) \hat{\xi}:=\xi+\hat{\chi}\in\hat{\mathfrak{t}}. An element \xi\in\mathfrak{t}\subset\hat{\mathfrak{t}} gives rise to a vector field on N_{\omega_{0}} given by the horizontal lift of \xi to D_{0} plus \langle\pi^{*}m_{0},\xi\rangle\hat{\chi} . It follows that \eta_{0}^{\hat{\chi}}(\hat{\xi})=l_{\xi}(\pi^{*}m_{0}) . Thus \hat{\xi} defines a Reeb field of (N_{\omega_{0}},D_{0},J_{0}) if and only if l_{\xi} is a positive function on P_{X} . Moreover, if \hat{\xi} defines a Reeb field then the corresponding contact form \eta_{0}^{\hat{\xi}} satisfies the transverse Fano condition 2\pi c^{B,\hat{\xi}}_{1}(N)=[d\eta_{0}^{\hat{\xi}}]_{B,\hat{\xi}} as explained in [ 5 , Section 4.6] for example. Assume that \hat{\xi}=\xi+\hat{\chi} defines a Reeb field of (N_{\omega_{0}},D_{0},J_{0}) . Following [ 2 , Section 2.3] (building on an idea in [ 24 , Proof of Lemma 2.2] ), we introduce the variation space of Sasaki structures as follows. (KRS11) \Xi_{\hat{\xi},\eta^{\hat{\xi}}_{0},J^{\hat{\xi}}}^{\hat{T}}(N_{\omega_{0}}):=\left\{\varphi\in C^{\infty}(N_{\omega_{0}};\mathbb{R})^{\hat{T}}\ \middle|\ \eta^{\hat{\xi}}_{\varphi}:=\eta^{\hat{\xi}}_{0}+d^{c}_{\hat{\xi}}\varphi\ \text{satisfies }d\eta^{\hat{\xi}}_{\varphi}>0\ \text{on }D_{\varphi}:=\ker(\eta^{\hat{\xi}}_{\varphi})\right\}, where J^{\hat{\xi}}\in\mathrm{End}(TN_{\omega_{0}}) extends J_{0}\in\mathrm{End}(D_{0}) by assuming the conditions J^{\hat{\xi}}(\hat{\xi})=0 , d^{c}_{\hat{\xi}}\varphi:=-\,d\varphi\circ J^{\hat{\xi}} and that the positivity of d\eta^{\hat{\xi}}_{\varphi} on D_{\varphi} is defined with respect to J^{\hat{\xi}} . We call an element in \Xi_{\hat{\xi},\eta^{\hat{\xi}}_{0},J^{\hat{\xi}}}^{\hat{T}}(N_{\omega_{0}}) a \hat{\xi} -transverse Kähler potential. As explained in [ 2 , Section 2.3] , there is a bijection \Theta_{\hat{\chi},\hat{\xi}}:\Xi_{\hat{\xi},\eta^{\hat{\xi}}_{0},J^{\hat{\xi}}}^{\hat{T}}(N_{\omega_{0}})\to\Xi_{\hat{\chi},\eta^{\hat{\chi}}_{0},J^{\hat{\chi}}}^{\hat{T}}(N_{\omega_{0}}) , and \varphi and \Theta_{\hat{\chi},\hat{\xi}}(\varphi) induce Sasaki structures with the same underlying CR structure. Note that \Xi_{\hat{\chi},\eta^{\hat{\chi}}_{0},J^{\hat{\chi}}}^{\hat{T}}(N_{\omega_{0}}) is identified with \mathcal{H}^{T} via the pullback by \pi:N_{\omega_{0}}\to X . Now let’s see how a \hat{\xi} -transverse g -soliton and a \hat{\xi} -transverse g -extremal metric are translated into Kähler structures associated with \hat{\chi} , namely, into the Kähler structure of X . For a \hat{\xi} -transverse Kähler potential \hat{\varphi} , one obtains the unique Kähler potential \varphi\in\mathcal{H}^{T} on X satisfying \Theta_{\hat{\chi},\hat{\xi}}(\hat{\varphi})=\pi^{*}\varphi . It follows from the formula (5.8) \eta^{\hat{\xi}}_{\hat{\varphi}}=(\eta^{\hat{\chi}}_{\pi^{*}\varphi}(\hat{\xi}))^{-1}\eta^{\hat{\chi}}_{\pi^{*}\varphi} that the contact moment map \mu^{\hat{\xi}}_{\hat{\varphi}} defined by \eta^{\hat{\xi}}_{\hat{\varphi}} , the contact moment map \mu^{\hat{\chi}}_{\pi^{*\varphi}} defined by \eta^{\hat{\chi}}_{\pi^{*}\varphi} and the moment map m_{\varphi} defined by \varphi\in\mathcal{H}^{T} satisfy the relation (5.9) \mu^{\hat{\xi}}_{\hat{\varphi}}=\frac{\mu^{\hat{\chi}}_{\pi^{*\varphi}}}{\eta^{\hat{\chi}}_{\pi^{*}\varphi}(\hat{\xi})}=\frac{\pi^{*}m_{\varphi}+\hat{\chi}^{*}}{l_{\xi}(\pi^{*}m_{\varphi})}, where \hat{\chi}^{*}\in\hat{\mathfrak{t}}^{*} stands as the dual element of \hat{\chi}\in\hat{\mathfrak{t}} . See [ 29 , Section 2.4] for example. Thus one obtains a smooth positive function g_{0} on P_{X} satisfying (5.10) g_{0}(\pi^{*}m_{\varphi})=g(\mu^{\hat{\xi}}_{\hat{\varphi}}) on N_{\omega_{0}} , where g is the weight function on P_{\hat{\xi}} in the defining equation of a transverse g -extremal metric and a transverse g -soliton. This terminology is used in the following propositions, which show how a \hat{\xi} -transverse g -soliton and a \hat{\xi} - transverse g -extremal metric on N_{\omega_{0}} are translated into a weighted CSCK metric and a soliton-type metric on X respectively. Proposition 5.3 . A \hat{\xi} -transverse Kähler potential \hat{\varphi}\in\Xi_{\hat{\xi},\eta^{\hat{\xi}}_{0},J^{\hat{\xi}}}^{\hat{T}}(N_{\omega_{0}}) defines a \hat{\xi} -transverse g -soliton \omega_{\hat{\xi}}^{T}:=d\eta^{\hat{\xi}}_{\hat{\varphi}} in 2\pi c^{B,\hat{\xi}}_{1}(N_{\omega_{0}}) if and only if the Kähler potential \varphi\in\mathcal{H}^{T} given by \Theta_{\hat{\chi},\hat{\xi}}(\hat{\varphi})=\pi^{*}\varphi defines a g_{\xi} -soliton \omega_{\varphi} with the weight (5.11) g_{\xi}=l_{\xi}^{-n-2}g_{0}. Proof.. We follow an argument in [ 5 ] (see also [ 4 , 29 ] ). It follows from [ 5 , Lemma 4.3] and the formula ( 5.8 ) that \omega_{\hat{\xi}}^{T} is a \hat{\xi} -transverse g -soliton if and only if the \hat{\chi} -transverse Kähler metric \pi^{*}\omega_{\varphi}=d\eta_{\pi^{*}\varphi}^{\hat{\chi}} satisfies (5.12) \mathrm{Ric}^{T}(\pi^{*}\omega_{\varphi})-\pi^{*}\omega_{\varphi}=dd^{c}_{\hat{\chi}}g_{0}(\pi^{*}m_{\varphi})-(n+2)dd^{c}_{\hat{\chi}}\log l_{\xi}(\pi^{*}m_{\varphi}). Thus the Kähler metric \omega_{\varphi} is a g_{\xi} -soliton on X . ∎ Proposition 5.4 . A \hat{\xi} -transverse Kähler potential \hat{\varphi}\in\Xi_{\hat{\xi},\eta^{\hat{\xi}}_{0},J^{\hat{\xi}}}^{\hat{T}}(N_{\omega_{0}}) defines a \hat{\xi} -transverse g -extremal metric \omega_{\hat{\xi}}^{T}:=d\eta^{\hat{\xi}}_{\hat{\varphi}} in 2\pi c^{B,\hat{\xi}}_{1}(N_{\omega_{0}}) if and only if the Kähler potential \varphi\in\mathcal{H}^{T} given by \Theta_{\hat{\chi},\hat{\xi}}(\hat{\varphi})=\pi^{*}\varphi defines a (v_{\xi},w_{\xi}) -CSCK metric with the weights (5.13) v_{\xi}=l_{\xi}^{-n-1}\quad\text{and}\quad w_{\xi}=\left(n+1-g_{0}\right)l_{\xi}^{-1}. Proof.. We apply the result of [ 1 , Lemma 3] proving that (5.14) S^{T}(\hat{\varphi})\hat{\xi}=l_{\xi}(\pi^{*}m_{\varphi})^{2}(\pi^{*}S_{v_{\xi}}(\varphi))\hat{\chi}, where S^{T}(\hat{\varphi}) is the transverse scalar curvature of \omega^{T}_{\hat{\xi}} and S_{v_{\xi}} is the v_{\xi} -weighted scalar curvature of \omega_{\varphi} . Recall Remark 2.5 for the notation. It follows from the formula ( 5.8 ) that S^{T}(\hat{\varphi})=l_{\xi}(\pi^{*}m_{\varphi})\pi^{*}S_{v_{\xi}}(\varphi) . Thus the defining equation of a \hat{\xi} -transverse g -extremal metric is equivalent to (5.15) \pi^{*}S_{v_{\xi}}(\varphi)=w_{\xi}(\pi^{*}m_{\varphi}), which shows that \omega_{\varphi} is a (v_{\xi},w_{\xi}) -CSCK metric in 2\pi c_{1}(X) . ∎ Summarizing the above arguments, we have the following theorem, showing that a g(v,w) -soliton appears naturally in Sasaki geometry of N_{\omega_{0}} . Theorem 5.5 . The weight g(v_{\xi},w_{\xi}) associated with the weight (v_{\xi},w_{\xi}) in Proposition 5.4 equals the wight g_{\xi} in Proposition 5.3 , that is, for any x\in P_{X} , (5.16) g(v_{\xi},w_{\xi})(x)=g_{\xi}(x). Proof.. By definition of g(v,w) -weight ( 3.5 ), one obtains \displaystyle g(v_{\xi},w_{\xi})(x) \displaystyle= \displaystyle l_{\xi}^{-n-1}\left(1+n-(n+1)\langle(\log l_{\xi})^{\prime}(x),x\rangle-\left(n+1-g_{0}\right)l_{\xi}^{-1}\right) \displaystyle= \displaystyle l_{\xi}^{-n-1}\left(1+n-(n+1)\frac{l_{\xi}(x)-1}{l_{\xi}}-\left(n+1-g_{0}\right)l_{\xi}^{-1}\right) \displaystyle= \displaystyle g_{\xi}(x). ∎ Theorem 1.5 stated in the introduction corresponds to Theorem 5.5 , where the weight function g defining a \hat{\xi} -transverse g -extremal metric and a \hat{\xi} -transverse g -soliton is assumed to be affine.