CAUCHY-SCHWARZ-TYPE INEQUALITIES ON KA¨HLER MANIFOLDS-II PINGLI Abstract. We establish in this note some Cauchy-Schwarz-type inequalities on compact 5 Ka¨hler manifolds, which generalize the classical Khovanskii-Teissier inequalities to higher- 1 dimensional cases. Our proof is to make full use of the mixed Hodge-Riemann bilinear 0 relations due to Dinh and Nguyˆen. A proportionality problem related to our main result is 2 also proposed. n a J 0 3 1. Introduction and main results ] G D SupposeX isann-dimensionalalgebraicmanifoldandD ,D ,...,D aren(notnecessarily 1 2 n . distinct) ample divisors on X. Then we have the following opposite Cauchy-Schwarz-type h t inequality, a m (1.1) ([D D D D ])2 [D D D D ] [D D D D ], [ 1 2 3··· n ≥ 1 1 3··· n · 2 2 3··· n 1 where [] denotes the intersection number of the divisors inside it and the equality holds if v · and only if the two divisors D and D are numerically proportional. 4 1 2 5 (1.1) was discovered independently by Khovanskii and Teissier around in 1979 ([9],[11]) 6 7 and now is called Khovanskii-Teissier inequality. This equality is indeed a generalization of 0 the classical Aleksandrov-Fenchel inequalities and thus present a nice relationship between . 1 the theory of mixed volumes and algebraic geometry ([6, p. 114]). The proof of (1.1) is 0 to apply the usual Hodge-Riemann bilinear relations ([7, p. 122-123]) to the Ka¨hler classes 5 1 determined by these divisors and an induction argument. The approach also suggests that : the usual Hodge-Riemann bilinear relations may be extended to the mixed case. After some v i partial results towards this direction ([8],[13]), this aim was achieved in its full generality by X Dinh and Nguyˆen in [2]. r a We would like to point out a fact, which was not mentioned explicitly in [2], that (1.1) can now be extended by the mixed Hodge-Riemann bilinear relations as follows. Suppose ω,ω ,ω ,...,ω are n 1 Ka¨hler classes and α H1,1(M,R) an arbitrary real-valued 1 2 n−2 − ∈ (1,1)-form on an n-dimensional compact connected Ka¨hler manifold M. Then we have (1.2) α ω ω ω 2 α2 ω ω ω2 ω ω , Z ∧ ∧ 1∧···∧ n−2 ≥ Z ∧ 1∧···∧ n−2 · Z ∧ 1∧···∧ n−2 (cid:0) M (cid:1) (cid:0) M (cid:1) (cid:0) M (cid:1) 2010 Mathematics Subject Classification. 32Q15, 58A14, 14F2XX.. Key words and phrases. Cauchy-Schwarz-typeinequality, compact K¨ahler manifold, Hodge-Riemann bilin- ear relation. The author was partially supported by the National Natural Science Foundation of China (Grant No. 11471247) and and the FundamentalResearch Fundsfor theCentral Universities. 1 2 PINGLI where the equality holds if and only if α Rω. Indeed, [2, Theorem A] tells us that the index ∈ of the following bilinear form (1.3) Q(u,v) := u v ω ω , u,v H1,1(M,R), Z ∧ ∧ 1∧···∧ n−2 ∈ M is of the form (+, , , ), i.e., the positive and negative indices are 1 and h1,1 1 respec- − ··· − − tively,whereh1,1 isthecorrespondingHodgenumberofM thedimensionofH1,1(M,R) . De- fineareal-valued functionf(t):= Q(ω+tα,ω+tα)(t R)(cid:0). Thenf(0) > 0as ω,ω , (cid:1),ω 1 n−2 ∈ ··· are all Ka¨hler classes and thus their product is strictly positive. ω+tα spans a 2-dimensional subspaceinH1,1(M,R)ifαandω arelinearly independentandthusf(t ) < 0forsomet R 0 0 ∈ in view of the index of Q(, ). Then the discriminant of f(t) gives (1.2) with strict sign “> ”. · · When these Ka¨hler classes are all equal: ω = ω = = ω , (1.2) degenerates to the 1 n−2 ··· following special case: (1.4) α ωn−1 2 α2 ωn−2 ωn , α H1,1(M,R), Z ∧ ≥ Z ∧ · Z ∀ ∈ (cid:0) M (cid:1) (cid:0) M (cid:1) (cid:0) M (cid:1) which is quite well-known and, to the author’s best knowledge, should be due to Apte in [1]. Inspired by (1.4), the author asked in [10] whether or not there exists a similar inequality to (1.4)forthoseα Hp,p(M,R)(1 p [n])andobtainedarelated result([10,Theorem1.3]), ∈ ≤ ≤ 2 whose proof is also based on the usual Hodge-Riemann bilinear relations. As an application we presented some Chern number inequalities when the Hodge numbers of the manifolds satisfysomeconstraints([10,Corollary1.5]). NowkeepingthemixedHodge-Riemannbilinear relations established in [2] in mind, we may also ask if the main idea of the proof in [10] can be carried over to the mixed case to extend the α in (1.2) to Hp,p(M,R) for 1 p [n]. The ≤ ≤ 2 answer is affirmative and this is the main goal of our current article. So this article can be viewed as a sequel to [10], which explains its title either. Our main result (Theorem 1.3) will be stated in the rest of this section. In Section 2 we briefly review the mixed Hodge-Riemann bilinear relations and then present the proof of Theorem 1.3. In Section 3 we discuss a proportionality problem related to (1.1) posed by Teissier and propose a similar problem related to our main result. In order to state our result as general as possible, we would like to investigate the elements in Hp,p(M,C), i.e., complex-valued (p,p)-forms on M. The following definition is inspired by (1.2) and is a mixed analogue to [10, Definition 1.1]. Definition 1.1. Suppose M is an n-dimensional compact connected Ka¨hler manifold. For 1 p [n], α Hp,p(M,C) and n 2p+1 Ka¨hler calsses ω,ω ,...,ω , we put Ω := ≤ ≤ 2 ∈ − 1 n−2p p ω ω and define 1 n−2p ∧···∧ g(α,ω;Ω ):= α α¯ Ω ω2p Ω α ωp Ω α¯ ωp Ω . p p p p p Z ∧ ∧ · Z ∧ − Z ∧ ∧ · Z ∧ ∧ (cid:0) M (cid:1) (cid:0) M (cid:1) (cid:0) M (cid:1) (cid:0) M (cid:1) α is said to satisfy Cauchy-Schwarz (resp. opposite Cauchy-Schwarz) inequality with respect to the Ka¨hler classes ω and (ω ,...,ω ) if g(α,ω;Ω ) 0 (resp. g(α,ω;Ω ) 0). 1 n−2p p p ≥ ≤ Remark 1.2. Note that α Hp,p(M,R) if and only if α = α¯. Also note that g(α,ω;Ω ) in p ∈ theabovedefinitionisarealnumberandsowecandiscussitsnon-negativity ornon-positivity. The main result of this note, which extends [10, Theorem 1.3] to the mixed case, is the following Theorem 1.3. Suppose M is an n-dimensional compact connected Ka¨hler manifold. CAUCHY-SCHWARZ-TYPE INEQUALITIES ON KA¨HLER MANIFOLDS-II 3 (1) Given 1 p [n], all elements in Hp,p(M,C) satisfy Cauchy-Schwarz inequality with ≤ ≤ 2 respect to any K¨ahler classes ω and (ω ,...,ω ) (in the sense of Definition 1.1) if 1 n−2p and only if the Hodge numbers of M satisfy p+1 (1.5) h2i,2i = h2i+1,2i+1, 0 i [ ] 1. ≤ ≤ 2 − (2) All elements in H1,1(M,C) satisfy opposite Cauchy-Schwarz inequality with respect to any K¨ahler classes ω and (ω ,...,ω ). 1 n−2p (3) Given 2 p [n], all elements in Hp,p(M,C) satisfy opposite Cauchy-Schwarz in- ≤ ≤ 2 equality with respect to any K¨ahler classes ω and (ω ,...,ω ) if and only if the 1 n−2p Hodge numbers of M satisfy p (1.6) h2i−1,2i−1 = h2i,2i, 1 i [ ]. ≤ ≤ 2 Moreover, in all the cases mentioned above, the equalities hold if and only if these α are proportional to ωp. The first part of the following corollary extends the Khovanskii-Teissier inequalities (1.1) and (1.2). Corollary 1.4. (1) α ω ω ω α¯ ω ω ω Z ∧ ∧ 1∧···∧ n−2 · Z ∧ ∧ 1∧···∧ n−2 (cid:0) M (cid:1) (cid:0) M (cid:1) α α¯ ω ω ω2 ω ω ≥ Z ∧ ∧ 1∧···∧ n−2 · Z ∧ 1∧···∧ n−2 (cid:0) M (cid:1) (cid:0) M (cid:1) for any Ka¨hler classes ω,ω ,...,ω and any α H1,1(M,C), where the equality 1 n−2 ∈ holds if and only if α Cω. ∈ (2) If h1,1 = 1, then all elements in H2,2(M,C) (n 4) satisfy Cauchy-Schwarz inequality ≥ withrespecttoanyKa¨hlerclassesinthesenseofDefinition1.1. Moreover, theequality case holds if and only if the element is proportional to ω2. (3) If h1,1 = h2,2, then all elements in H2,2(M,C) (n 4) and H3,3(M,C) (n 6) ≥ ≥ satisfy opposite Cauchy-Schwarz inequality with respect to any Ka¨hler classes in the sense of Definition 1.1. Moreover, the equality case holds if and only if the element is proportional to ω2 or ω3 respectively. Remark 1.5. In [10, Example 1.7], the author describedin detail many examples of compact connected Ka¨hler manifolds whose Hodge numbers satisfy h1,1 = 1 and h1,1 = h2,2 respec- tively. These include the complete intersections in complex projective spaces, the complex flag manifolds G/P (G is asemisimple complex Liegroup and P is a maximal parabolic max max subgroup of G), the one point blow-up of complex projective spaces and so on. 2. Proof of the main result 2.1. The mixed Hodge-Riemann bilinear relations. In this subsection we briefly recall the mixed Hodge-Riemann biliner relations established in [2] by Dinh and Nguyˆen. As before denote by M an n-dimensional compact connected Ka¨hler manifold. We arbi- trarily fix two non-negative integers p,q such that p,q [n] and n p q+1 Ka¨hler classes ≤ 2 − − 4 PINGLI ω,ω ,...,ω on M. Put Ω := ω ω . Define the mixed primitive subspace of 1 n−p−q 1 n−p−q ∧···∧ Hp,q(M,C) with respect to ω and Ω by (2.1) Pp,q(M;ω,Ω) := α Hp,q(M,C) α ω Ω = 0 . { ∈ | ∧ ∧ } Define the mixed Hodge-Riemann bilinear form Q (, ) with respect to Ω on Hp,q(M,C) by Ω · · (2.2) QΩ(α,β) := (√−1)q−p(−1)(p+q)(p2+q+1) Z α∧β¯∧Ω, α,β ∈ Hp,q(M,C). M Remark 2.1. (1) Note that this definition of Q (, ) differs from that in (1.3) by asign whenp = q = 1. Ω · · (2) Clearly when ω = ω = ω , Pp,q(M;ω,Ω) and Q (, ) are nothing but the 1 n−p−q Ω ··· · · usual primitive cohomology group and Hodge-Riemann bilinear form with respect to the Ka¨hler class ω. (3) The symbols Pp,q(M;ω,Ω) and Q (, ) we use here are simply denoted by Pp,q(M) Ω · · and Q(, ) respectively in [2]. We use the current symbols to avoid confusion as they · · stress the dependence on the choices of ω and Ω, whose advantage will be clear in the process of our proof in Theorem 1.3 in the next section. With the above notation understood, we have the following remarkable result due to Dinh and Nguyˆen in [2, Theorems A,B,C], which extends the classical Hodge-Riemann biliner re- lations. Theorem 2.2 (mixed Hodge-Riemann bilinear relations). (1) (mixed Hard Lefschetz theorem) The linear map τ : Hp,q(M,C) Hn−q,n−p(M,C) → given by (2.3) τ(α) := α Ω, α Hp,q(M,C) ∧ ∈ is an isomorphism. (2) (mixed Lefschetz decomposition) We have the following canonical decomposition: (2.4) Hp,q(M,C) = Pp,q(M;ω,Ω) ω Hp−1,q−1(M,C) . ⊕ ∧ (cid:0) (cid:1) Here Hp−1,q−1(M,C) := 0 if either p = 0 or q = 0. (3) (Positive-definiteness) The mixed Hodge-Riemann bilinear form Q (, ) is positive- Ω · · definite on the mixed primitive subspace Pp,q(M;ω,Ω). Remark 2.3. Note that Pp,q(M;ω,Ω) depends on ω and Ω while Hp,q(M,C) is clearly independent of them. This means, if we fix ω but change ω ,...,ω , then Ω is also 1 n−p−q changedrespectively andsoisPp,q(M;ω,Ω). ButthemixedLefschetzdecompositiontheorem tells us that (2.4) remains true. So the reference Ka¨hler classes ω,ω ,...,ω in context 1 n−p−q should be clear when we apply (2.4). For unambiguity we shall use the sentence ”We apply (2.4) to α Hp,q(M,C) with respect to the reference Ka¨hler classes ω and (ω ,...,ω )” 1 n−2p ∈ to emphasize it. This notation will play a key role in the proof of (2.6) Lemma 2.4. 2.2. Proof of Theorem 1.3. We now apply the mixed Hodge-Riemann bilinear relations to prove our Theorem 1.3. The following lemma uses the full power of (2.4). Lemma 2.4. CAUCHY-SCHWARZ-TYPE INEQUALITIES ON KA¨HLER MANIFOLDS-II 5 (1) n (2.5) dimCPp,q(M;ω,Ω) = hp,q hp−1,q−1, 0 p,q [ ], − ≤ ≤ 2 where hp−1,q−1 := 0 if either p = 0 or q = 0. This means that the dimension of Pp,q(M;ω,Ω) is independent of ω and Ω and only depends on the complex structure of M. (2) Let α Hp,p(M,C) with 1 p [n] and ω,ω ,...,ω be n 2p+1Ka¨hler classes. ∈ ≤ ≤ 2 1 n−2p − Put Ω := ω ω . Then this α can be written as follows. p 1 n−2p ∧···∧ p (2.6) α = λωp+ α ωp−i, i ∧ X i=1 where λ C and ∈ (2.7) α Pi,i(M;ω,ω2(p−i) Ω ) α ω2(p−i)+1 Ω = 0 . i p i p ∈ ∧ ⇔ ∧ ∧ (cid:0) (cid:1) (3) g(α,ω;Ω ) given in Definition 1.1 has the following expression in terms of α : p i p (2.8) g(α,ω;Ω ) = ω2p Ω α α¯ ω2(p−i) Ω . p p i i p Z ∧ · Z ∧ ∧ ∧ (cid:0) M (cid:1) (cid:0)Xi=1 M (cid:1) Proof. (1) The usual Hard Lefschetz theorem ([7, p. 122]) tells us that the map ωn−p−q () : Hp,q(M,C) Hn−q,n−p(M,C) ∧ · → is an isomorphism. This means that, for p+q n 1, the map ≤ − ω () : Hp,q(M,C) Hp+1,q+1(M,C) ∧ · → is injective and consequently dimC ω Hp−1,q−1(M,C) = hp−1,q−1 ∧ (cid:0) (cid:1) for 1 p,q [n], which, together with (2.4), leads to (2.5). ≤ ≤ 2 (2) The strategy for proving (2.6) is to apply (2.4) repeatedly to yield the desired α . i We first apply (2.4) to this α with respect to the reference Ka¨hler classes ω and (ω ,...,ω ) to yield α : 1 n−2p p α = α +ω α˜ with α Pp,p(M;ω,Ω ). p p−1 p p ∧ ∈ We continue to apply (2.4) to α˜ Hp−1,p−1(M,C) with respect to the reference p−1 ∈ Ka¨hler classes ω and (ω,ω,ω ,...,ω ) to yield α : 1 n−2p p−1 α˜ = α +ω α˜ with α Pp−1,p−1(M;ω,ω2 Ω ). p−1 p−1 p−2 p−1 p ∧ ∈ ∧ Obviously the next step is to apply (2.4) to α˜ Hp−2,p−2(M,C) with respect to p−2 ∈ the reference Ka¨hler classes ω and (ω,ω,ω,ω,ω ,...,ω ) to obtain 1 n−2p α˜ = α +ω α˜ with α Pp−2,p−2(M;ω,ω4 Ω ). p−2 p−2 p−3 p−2 p ∧ ∈ ∧ Now it is easy to see that repeated use of (2.4) to α˜ determined by α˜ with p−i p−i+1 respect to the Ka¨hler classes ω and (ω,...,ω,ω ,...,ω ) yields the desired α : 1 n−2p p−i 2icopies | {z } α˜ = α +ω α˜ with α Pp−i,p−i(M;ω,ω4 Ω ). p−i p−i p−i−1 p−i p ∧ ∈ ∧ This completes the proof of (2.6). 6 PINGLI (3) We now know from (2.6) that p (2.9) α ωp Ω = λω2p Ω + α ω2p−i Ω p p i p ∧ ∧ ∧ ∧ ∧ X i=1 and α α¯ Ω p ∧ ∧ p p =(λωp+ α ωp−i) (λ¯ωp+ α¯ ωp−j) Ω i j p ∧ ∧ ∧ ∧ (2.10) Xi=1 Xj=1 p p p = λ 2ω2p+λ α¯ ω2p−j +λ¯ α ω2p−i+ α α¯ ω2p−(i+j) Ω . j i i j p | | ∧ ∧ ∧ ∧ ∧ (cid:0) Xj=1 Xi=1 iX,j=1 (cid:1) Note that α ω2(p−i)+1 Ω = 0 by (2.7), 1 i p, i p ∧ ∧ ≤ ≤ (2.11)  2p i 2(p i)+1, 1 i p,  − ≥ − ≤ ≤ 2p (i+j) 2(p max i,j )+1, 1 i= j p. − ≥ − { } ≤ 6 ≤  This means that (2.9) and (2.10) can be simplified via (2.11) as follows. (2.12) α ωp Ω = λω2p Ω p p ∧ ∧ ∧ and p (2.13) α α¯ Ω = λ 2ω2p Ω + α α¯ ω2(p−i) Ω . p p i i p ∧ ∧ | | ∧ ∧ ∧ ∧ X i=1 Integrating (2.12) and (2.13) over M deduces that α ωp Ω α¯ ωp Ω (2.14) λ = M ∧ ∧ p, λ¯ = M ∧ ∧ p, R ω2p Ω R ω2p Ω M ∧ p M ∧ p R R and p (2.15) α α¯ Ω = λ λ¯ ω2p Ω + α α¯ ω2(p−i) Ω . p p i i p Z ∧ ∧ · ·Z ∧ Z ∧ ∧ ∧ M M X M i=1 (2.8) now follows from substituting the two expressions in (2.14) for the λ and λ¯ in (2.15). (cid:3) Now we are ready to prove Theorem 1.3, our main result in this article. Proof. It suffices to prove the first part in Theorem 1.3 as the resulting two cases are similar. Since α Pi,i(M;ω,ω2(p−i) Ω ), the positive-definiteness of the mixed Hodge-Riemann i p ∈ ∧ bilinear forms guarantees that Q(ω2(p−i)∧Ωp)(αi,αi) ≥ 0 with equality if and only if αi = 0. This, together with the definition of Q (, ) in (2.2), implies that (·) · · (2.16) ( 1)i α α¯ ω2(p−i) Ω 0, 1 i p, i i p − Z ∧ ∧ ∧ ≥ ≤ ≤ M with the equality holds if and only if α = 0. i We first show the “if” part of (1) in Theorem 1.3. CAUCHY-SCHWARZ-TYPE INEQUALITIES ON KA¨HLER MANIFOLDS-II 7 The dimension formula (2.5) in Lemma 2.4 and the assumption (1.5) imply that p+1 α = 0, 0 i [ ] 1, 2i+1 ≤ ≤ 2 − which, together with (2.8) and (2.16), give us g(α,ω;Ω ) = ω2p Ω α α¯ ω2(p−i) Ω 0, p p i i p Z ∧ · Z ∧ ∧ ∧ ≥ (cid:0) M (cid:1) (cid:0)1≤Xi≤p M (cid:1) i even with equality if and only if all α = 0 and thus α Cωp by the decomposition formula (2.6). i ∈ The proof of the “only if” part. Suppose on the contrary that there exists some 1 i [p+1] 1 such that h2i0+1,2i0+1 > ≤ 0 ≤ 2 − h2i0,2i0. Then we can choose a 0 = α(i ) P2i0+1,2i0+1(M;ω,ω2 p−(2i0+1) Ω ) 0 p 6 ∈ (cid:0) (cid:1)∧ by (2.5) and set θ := ωp+α(i ) ωp−(2i0+1). 0 ∧ But in this case g(θ,ω;Ω )= ω2p Ω α(i ) α(¯i ) ω2 p−(2i0+1) Ω < 0 p Z ∧ p · Z 0 ∧ 0 ∧ (cid:0) (cid:1) ∧ p (cid:0) M (cid:1) (cid:0) M (cid:1) and thus this θ does not satisfy Cauchy-Schwarz inequality with respect to the Ka¨hler classes ω and (ω ,...,ω ) in the sense of Definition 1.1, which contradicts to the assumption. 1 n−2p This gives the desired proof. (cid:3) The corollary below follows from the process of the above proof. Corollary 2.5. Suppose α Hp,p(M,C) with 1 p [n]. Then this α satisfies Cauchy- ∈ ≤ ≤ 2 Schwarz(resp. oppositeCauchy-Schwarz)inequalitywithrespecttoKa¨hlerclassesω,ω ,...,ω 1 n−2p if those α with i odd (resp. even) determined by the decomposition formula (2.8) all vanish. i 3. A proportionality problem Let H1,1(M,R) bethe Ka¨hler cone of M, which consists of all the Ka¨hler classes of M. K ∈ Recall that c H1,1(M,R) is called a nef class if c ¯, the closure of the Ka¨hler cone. So ∈ ∈ K nef classes can beapproximated by Ka¨hler classes. This means Theorem 1.3 has the following corollary. Corollary 3.1. Suppose M is an n-dimensional compact connected Ka¨hler manifold. (1) Given 1 p [n], if the Hodge numbers of M satisfy ≤ ≤ 2 p+1 h2i,2i = h2i+1,2i+1, 0 i [ ] 1, ≤ ≤ 2 − then for any α Hp,p(M,C) and any nef classes c,c ,...,c we have 1 n−2p ∈ (3.1) α α¯ C c2p C α cp C α¯ cp C , p p p p Z ∧ ∧ · Z ∧ ≥ Z ∧ ∧ · Z ∧ ∧ (cid:0) M (cid:1) (cid:0) M (cid:1) (cid:0) M (cid:1) (cid:0) M (cid:1) where C =c c . p 1 n−2p ∧···∧ 8 PINGLI (2) For any α H1,1(M,C) and any nef classes c,c ,...,c we have 1 n−2 ∈ (3.2) α α¯ C) c2p C α cp C α¯ cp C , Z ∧ ∧ · Z ∧ ≤ Z ∧ ∧ · Z ∧ ∧ (cid:0) M (cid:0) M (cid:1) (cid:0) M (cid:1) (cid:0) M (cid:1) where C = c c . 1 n−2 ∧···∧ (3) Given 2 p [n], if the Hodge numbers of M satisfy ≤ ≤ 2 p h2i−1,2i−1 = h2i,2i, 1 i [ ], ≤ ≤ 2 then for any α Hp,p(M,C) and any nef classes c,c ,...,c we have 1 n−2p ∈ (3.3) α α¯ C c2p C α cp C α¯ cp C , p p p p Z ∧ ∧ · Z ∧ ≤ Z ∧ ∧ · Z ∧ ∧ (cid:0) M (cid:1) (cid:0) M (cid:1) (cid:0) M (cid:1) (cid:0) M (cid:1) where C =c c . p 1 n−2p ∧···∧ However, unlike Theorem 1.3 for Ka¨hler classes, we can not conclude directly in this case that the equalities in (3.1), (3.2) and (3.3) hold if and only if α and the nef class c are proportional. So a natural question is to characterize the equalities in (3.1)-(3.3), a very special case of which has been proposed by Teissier in [12] as a further question related to his inequality (1.1) and we shall briefly reivew it in what follows. (1.1) or (3.2) gives us that, for two nef divisors D and D on an algebraic manifold and 1 2 for 1 k n 1, we have ≤ ≤ − (3.4) [DkDn−k] 2 [Dk−1Dn−k+1] [Dk+1Dn−k−1]. 1 2 ≥ 1 2 · 1 2 (cid:0) (cid:1) Teissier considered in [12] that how to characterize the equality case in (3.4) for nef and big divisors D and D (recall that a nef divisor D is called big if moreover [Dn] > 0). This 1 2 problem was solved in [3, Theorem D] by Boucksom, Favre and Jonsson, whose result asserts that for two nef and big divisors D and D the equality in (3.4) holds if and only if D and 1 2 1 D are numerically proportional. Very recently Fu and Xiao obtained the same type result in 2 the context of Ka¨hler manifolds ([5, Theorem 2.1]), some of whose ideas are based on their previous work in [4]. Remark 3.2. The expression used in [3, Theorem D, (2)] is slightly different from our (3.4) but they are indeed equivalent (see, for instance, the equivalent statements in ([5, Theorem 2.1]). WiththeCauchy-Schwarz-type inequalitiesinitsfullgenerality inCorollary3.1inhand,we can now end our article by posing the following problem, whose solution is obviously beyond the content of this note. Question3.3. Howtocharacterizethethreeinequalitycasesin(3.1),(3.2)and(3.3)? Clearly α being proportional to cp is a sufficient condition. Is this also a necessary condition? Or weakly how to establish such a necessary condition by imposing more constraints on the element α, the nef classes c,c ,...,c and/or the underlying Ka¨hler manifold M? 1 n−2p Acknowledgements The author thanks Ji-Xiang Fu for drawing his attention to the reference [2] in a talk on [5] in a workshop held in Center of Mathematical Science, Zhejiang University, inspired by which the author realized that his idea in [10] could be carried over to extend the classical Khovanskii-Teissier inequalities to higher-dimensional cases. CAUCHY-SCHWARZ-TYPE INEQUALITIES ON KA¨HLER MANIFOLDS-II 9 References 1. M. Apte: Sur certaines classes characte´ristiques des varie´te´s K¨ahle´riennes compactes, C. R. Acad. Sci. Paris 240 (1955), 149-151. 2. T.-C. Dinh, V.-A. Nguyeˆn: The mixed Hodge-Riemann bilinear relations for compact K¨ahler manifolds, Geom. Funct.Anal. 16 (2006), 838-849. 3. S. Boucksom, C. Favre, M. Jonsson: Differentiaility of volumes of divisors and a problem of Teissier, J. Algebraic Geom. 18 (2009), 279-308. 4. J.-X. Fu, J. Xiao: Relations between the K¨ahler cone and the balanced cone of a K¨ahler manifold, Adv. Math. 263 (2014), 230-252. 5. J.-X. Fu, J. Xiao: Teissier’s problem on proportionality of nef and big classes over a compat K¨ahler manifold,arXiv:1410.4878v1. 6. W.Fulton: Introduction to toric varieties, Princeton UniversityPress, Princeton, NJ, 1993. 7. P.Griffiths,J.Harris: Principles of algebraic geometry, PureandAppliedMathematics,Wiley,NewYork, 1978. 8. M. Gromov: Convex sets and K¨ahler manifolds, Advances in Differential Geometry and Topology World Sci. Publishing, Teaneck, NJ (1990), 1-38. 9. A.G. Khovanskii: Newton polyhedra, and the genus of the complete interstions,, Funktsional. Anal. i Prilozhen. 12 (1978), 51-61. 10. P.Li: Cauchy-Schwarz-type inequalities on K¨ahler manifolds,J. Geom. Phys. 68 (2013), 76-80. 11. B. Teissier: Du the´ore`me de l’index de Hodge aux ine´ga lite´s isope´rime´triques, C.R. Acad. Sci. Paris Se´r. A-B,288 (1979), A287-A289. 12. B. Teissier: Bonnesen-type inequalities in algebraic geometry. I. Introduction to the problem. In Seminar on Differential Geometry, pp.85-105. Ann.Math. Stud.,102. Princeton University Press, 1982. 13. V.A.Timorin: Mixed Hodge-Riemann libinear relations in a linear context, Funct.Anal.Appl. 32(1998), 268-272. Department of Mathematics, Tongji University, Siping Road 1239, Shanghai 200092, China E-mail address: [email protected]