MORI GEOMETRY MEETS CARTAN GEOMETRY: VARIETIES OF MINIMAL RATIONAL TANGENTS 5 1 1 JUN-MUK HWANG 0 2 n Abstract. We give an introduction to the theory of varieties of minimal rational tan- a gents, emphasizing its aspect as a fusion of algebraic geometry and differential geometry, J more specifically, a fusionof Morigeometry ofminimal rationalcurvesand Cartangeom- 0 etry of cone structures. 2 ] G A 1. Introduction: a brief prehistory . h Lines have been champion figures in classical geometry. Together with circles, they t a dominate the entire geometric contents of Euclid. Their dominance is no less strong in m projective geometry. Classical projective geometry is full of fascinating results about intri- [ cate combinations of lines. As geometry entered the modern era, lines evolved into objects 1 of greater flexibility and generality while retaining all the beauty and brilliance of classical v lines. As Euclidean geometry developed into Riemannian geometry, for example, lines were 0 2 replaced by geodesics which then inherited all the glory of Euclidean lines. 7 In the transition from classical projective geometry to complex projective geometry, real 4 0 lines have been replaced by complex lines. Lines over complex numbers have all the power 1. of lines in classical projective geometry and even more: results of greater elegance and 0 harmony are obtained over complex numbers. A large number of results on lines and 5 their interactions with other varieties have been obtained in complex projective geometry, 1 : their dazzling beauty no less impressive than that of classical geometry. But as complex v i projective geometry develops further into complex geometry and abstract algebraic geom- X etry, which emphasize intrinsic properties of complex manifolds and abstract varieties, the r a notion of lines in projective space seems to be too limited for it to keep its leading role. Firstly, to be useful in intrinsic geometry of projective varieties in projective space, lines should lie on the projective varieties. But most projective varieties do not contain lines. Even when a projective manifold contains lines, the locus of lines is, often, small and then such a locus is usually regarded as an exceptional part. Of course, there are many important varieties that are covered by lines, but they belong to a limited class from the general perspective of classification theory of varieties. In short, (*) the class of projective manifolds covered by lines seems to be too special from the perspective of the general theory of complex manifolds or algebraic varieties. 1 supported by National Researcher Program2010-0020413of NRF. 1 2 JUN-MUKHWANG Secondly, many of the methods employed to use lines on varieties in projective space depend on the extrinsic geometry of ambient projective space. They do not truly belong to intrinsic geometry of the varieties. Such geometric arguments are undoubtedly useful in fathoming deeper geometric properties of varieties which are described explicitly, at least to some extent. But can such methods yield results on a priori unknown varieties, defined abstractly by intrinsic conditions? In short, (**) tools employed in line geometry are not intrinsic enough to handle intrinsic problems on abstractly described varieties. These concerns show that lines in projective space have a rather limited role in the modern development of complex algebraic geometry. Is there a more general and more powerful notion in complex algebraic geometry that can replace the role of lines, as geodesics do in Riemannian geometry? No serious candidate had emerged until Mori’s groundbreaking work [37]. In the celebrated paper [37], Mori shows that a large class of projective manifolds, including all Fano manifolds, are covered by certain intrinsically defined rational curves that behave like lines in many respects. Let us call these rational curves ‘minimal rational curves’. If a projective manifold embedded in projective space is covered by lines, these lines are minimal rational curves of the projective manifold, so the notion of minimal rational curves can be viewed as an intrinsic generalization of lines. The class of projective manifolds covered by minimal rational curves are called uniruled projectivemanifolds. GeneralizingMori’sresult, MiyaokaandMorihaveprovedin[32]that a projective manifold is uniruled if its anti-canonical bundle satisfies a certain positivity condition. This implies that uniruled projective manifolds form a large class of algebraic varieties. Furthermore, the minimal model program, a modern structure theory of higher- dimensional algebraic varieties, predicts that uniruled projective manifolds are precisely those projective manifolds that do not admit minimal models. Thus projective manifolds covered by minimal rational curves form a distinguished class of manifolds, worthy of independent study from the view-point of clas- sification theory of general projective varieties, and at the same time, large enough to contain examples of great diversity. This overcomes the limitation (*) of the class of projective manifolds covered by lines. Furthermore, Mori’s work exhibits how to use minimal rational curves in an intrinsic way to obtain geometric information on uniruled projective manifolds. The main tool here is the deformation theory of curves, a machinery of modern complex algebraic geometry- somewhat reminiscent of the use of variational calculus in the local study of geodesics in Riemannian geometry. An example is the property that a minimal rational curve cannot be deformed when two distinct points on the curve are fixed. This result generalizes the fundamental postulate of classical geometry that ”two points determine one line”. The important point is that such a classical property of lines can be recovered by modern deformation theory in an abstract setting. VMRT-STRUCTURES 3 Deformation theory of rational curves is a powerful technique applicable to conceptual problems on varieties defined in abstract intrinsic terms. In[37], Mori, infact, hasresolved one of thetoughest problems ofthis kind, the Hartshorne conjecture, characterizing projective space by the positivity of the tangent bundle. The methods employed in the theory of minimal rational curves are certainly free of the concern (**) on the tools of line geometry. These considerations indicate that minimal rational curves can serve as the natural generalization of lines, overcoming the limitations of lines, while inheriting their powerful and elegant features. Our main interest is the geometry of minimal rational curves in uniruled projective manifolds. As in Mori’s work, we would like to see how minimal rational curves can be used to control the intrinsic geometry of uniruled projective manifolds. One guiding problem is the following question on recognizing a given uniruled projective manifolds by minimal rational curves. Problem 1.1. Let S be a (well-known) uniruled projective manifold. Given another uniruled projective manifold X, what properties of minimal rationalcurves onX guarantee that X is biregular (i.e. isomorphic as abstract algebraic varieties) to S? Here, the setting of the problem is algebraic geometry and by properties of minimal rational curves, we mean algebro-geometric properties. When S is projective space, a version of this problem is precisely what Mori solved in [37]. The initial goal of [37] was to prove the Hartshorne conjecture, which characterizes projective space by certain positivity property of the tangent bundle. After showing that the projective manifold in question is uniruled, Mori used the tangent directions of minimal rational curves to finish the proof. This part of Mori’s proof has been greatly strengthened by the later work of Cho-Miyaoka- Shepherd-Barron [4], which says roughly the following (see Theorem 3.16 for a precise statement). Theorem 1.2. Suppose for a general point x on a uniruled projective manifold X and a general tangent direction α ∈ PT (X), there exists a minimal rational curve through x x tangent to α. Then X is projective space. This is a very satisfactory answer to Problem 1.1 when S is projective space. It includes, as special cases, many previously known characterizations of projective space. One may wonder why the condition on minimal rational curves here is formulated in terms of their tangent directions, not in terms of some other properties of minimal rational curves. The essential reason is because the main technical tool to handle minimal rational curves is the deformation theory of curves, as mentioned before. The tangential information of curves is essential in deformation theory. For this reason, it is natural and also useful to give conditions in terms of tangent directions of minimal rational curves. What about other uniruled projective manifolds? When S is different from projective space, Theorem 1.2 says that minimal rational curves on S exist only in some distinguished directions. Thus in the setting of Problem 1.1, it is natural to consider 4 JUN-MUKHWANG the subvariety C ⊂ PT (S) consisting of the directions of minimal rational s s curves through s ∈ S and the corresponding subvariety C ⊂ PT (X). x x When S is projective space, we have C = PT (S) for any s ∈ S. Theorem 1.2 says that s s a uniruled projective manifold X is projective space if and only if C = PT (X) at some x x point x ∈ X. In other words, if a uniruled projective manifold has the same type of C as x projective space, then it is projective space. Based on this observation, we can refine our guiding Problem 1.1 as follows. Problem 1.3. Let S be a (well-known) uniruled projective manifold. Given another uniruled projective manifold X, what properties of C ⊂ PT (X) for general points x ∈ X x x guarantee that X is biregular to S? Comparing this with Theorem 1.2, one may wonder why we are asking for information on C for general points x ∈ X, instead of a single point x ∈ X as in Theorem 1.2. This is x because the information at one point x ∈ X seems to be too weak to characterize X when C 6= PT (X). The equality C = PT (X) implies that minimal rational curves through one x x x x point x cover the whole of X. This is why in Theorem 1.2 the information at one point is sufficient to control the whole of X. If C 6= PT (X), minimal rational curves through one x x point x cover only small part of X. Besides, the subvariety C ⊂ PT (S) may change as s s the point s ∈ S varies and so the expected condition is not just on C for a single x, but x on the family {C ⊂ PT (X), general x ∈ X}. x x This is why we are asking for the data C for all general x ∈ X. x Now as in Problem 1.1, the properties of C that we are looking for in Problem 1.3 are x algebro-geometric properties. In algebraic geometry, however, to use properties of such a family of varieties to control the whole of X, we usually need to have good information not only on general members of the family, but also on the potential degeneration of the family. Thus it may look more reasonable to require, in Problem 1.3, some additional properties on the behavior of the family C under degeneration. But such additional conditions would x diminish the true interest of Problem 1.3. This is because in the context of intrinsic geometry of uniruled manifolds, the properties of C we are looking for should be checkable x by deformation theory of curves. Deformation theory of rational curves works well at general points of nonsingular varieties, but not so at special points. Thus it is important to find conditions for C only for general x ∈ X in Problem 1.3. But then controlling the x whole of X using the algebraic behavior of {C ⊂ PT (X), general x ∈ X} becomes a x x serious issue. This was exactly the issue puzzling me when I first encountered a version of Problem 1.3 about twenty years ago. At that time, I was working on the deformation rigidity of Hermitian symmetric spaces in the setting of algebraic geometry. I refer the reader to [11] for the details on this rigidity problem. Here it suffices to say that the deformation rigidity of Hermitian symmetric spaces was a question originated from Kodaira-Spencer’s work in 1950’s and the question itself did not involve rational curves. I was trying to attack this VMRT-STRUCTURES 5 question employing Mori’s approach of minimal rational curves, which naturally led to a version of Problem 1.3 when S is an irreducible Hermitian symmetric space. In the setting of this rigidity question, I could derive a certain amount of algebro-geometric information on C for general x ∈ X, but I was unable to figure out how to proceed from there, x essentially because of the above difficulty, that it is hard to control the whole of X by algebro-geometric information on C for general x ∈ X. x There was one hope. A few years earlier, Ngaiming Mok had overcome an obstacle of a similar kind in [33]. In that work, Mok solved what is called the generalized Frankel conjecture, which asks for a characterization of Hermitian symmetric spaces among Ka¨hler manifolds in terms of a curvature condition. The Frankel conjecture itself is the Ka¨hler version of the Hartshorne conjecture and was settled by Siu-Yau [40] around the time Mori solved the Hartshorne conjecture. Since the method used by Siu and Yau was rather restrictive, Mok naturally took the approach of Mori and encountered a situation similar to Problem 1.3. Now in his situation, there is a Riemannian metric on X and Mok could relate C to a suitably deformed Riemannian metric. This enables him to show that X x is Hermitian symmetric space using Berger’s work on Riemannian holonomy. Roughly speaking, in [33] the difficulty in Problem 1.3 was overcome by relating C to a Riemannian x structure. This shows that differential geometry can be a recourse for Problem 1.3 when S is a Hermitian symmetric space. Indeed, compared with tools in algebraic geometry, methods of differential geometry tend to be more effective when the available data are only at general points of a manifold. Motivated by this, I tried to imitate Mok’s argument in the setting of the deformation rigidity problem. However, the nature of the deformation rigidity problem is purely algebro-geometric and it is very hard to relate it to Riemannian structures. As a matter of fact, there had already been some unsuccessful attempts in 1960’s to use Riemannian structures for the deformation rigidity question. This was precisely the problem I was agonizing over when I attended my first ICM: Zu¨rich 1994. Having come to the congress just to have fun listening to the new develop- ments in mathematics, I found that Mok was there as a speaker and managed to have a chat with him. When I told him about the above difficulty in applying the approach of [33] to the deformation rigidity problem, he gave me an enlightening comment: besides the Riemannian metric, there is another differential geometric structure, a certain holomorphic G-structure, which can be used to characterize a Hermitian symmetric space. His sugges- tion was that one might be able to construct these G-structures using the information on C for general x ∈ X and fromthis to recover Hermitian symmetric spaces. This suggestion x looked promising because algebro-geometric data are closer to holomorphic structures than Riemannian structures. Soon after the congress, I started looking into G-structures. I realized that there is a far-reaching generalization of Riemannian structures by Elie Cartan and the G-structures modeled on Hermitian symmetric spaces are special examples of Cartan’s general theory 6 JUN-MUKHWANG of geometric structures. To recover these G-structures, it was necessary to investigate the geometry of C ’s in depth in the setting of the deformation rigidity problem. I had x subsequent communications with Mok, and we started working together on this problem. Our collaboration was successful, leading to a solution of the deformation rigidity problem in [18]. But the most exciting point in our work was not the deformation rigidity itself. As mentioned, the essential part of [18] is to construct on X the G-structures modeled on Hermitian symmetric spaces. It turns out that a crucial point of this construction lies in a study of the behavior of C ’s not just as a family of projective algebraic varieties, but as x data imposed on the tangent bundle of an open subset of X. In other words, we had to treat these C ’s as if x the union of C ’s for general x ∈ X is a differential geometric structure. x And why not? Such a family of subvarieties in PT (X) is a legitimate example of Cartan’s x general geometric structures! So what happened canbe summarized asfollows. Initially we had been trying to relate C ’s to some differential geometric structures. These differential x geometric structures were Riemannian structures in [33] and then G-structures in [18]. But actually, they have been there all along, namely, C ’s themselves! x Now once we accept C ’s as a differential geometric structure, there is no need to restrict x ourselves to Hermitian symmetric spaces. This geometric structure exists for any uniruled projective manifold S and its minimal rational curves! This was an epiphany for me. We realized that the variety C deserves a name of its own and endowed it with the appellation, x somewhat uncharming, ‘variety of minimal rational tangents’. Realizing the varieties of minimal rational tangents as geometric structures opens up an approach to Problem 1.3 via Cartan geometry. In fact, Mok and I were able to show in [20] (see Theorem 4.6 for a precise statement) Theorem 1.4. Assume that S is a fixed uniruled projective manifold with b (S) = 1 and 2 C at a general point s ∈ S is a smooth irreducible variety of positive dimension. If X s is a uniruled projective manifold with b (X) = 1 and the differential geometric structures 2 defined by C ’s and C ’s are locally equivalent in the sense of Cartan, then S and X are s x biregular. This means that for a large and interesting class of uniruled projective manifolds, Prob- lem 1.3 can be solved by studying the Cartan geometry of the structures defined by C ’s. x By Theorem 1.4, the essence of Problem 1.3 has become searching for algebro-geometric properties of varieties of minimal rational tangents which make it possible to control the Cartan geometry of the geo- metric structures defined by them. As we will see in Section 4, this search has been successful in a number of cases and Prob- lem 1.3 has been answered for some uniruled projective manifolds, including irreducible Hermitian symmetric spaces. Since [18], the theory of varieties of minimal rational tangents has seen exciting devel- opments and has found a wide range of applications in algebraic geometry. For surveys VMRT-STRUCTURES 7 on these developments and applications, we refer the reader to [9], [19], [27] and [36]. The purpose of this article is to give an introduction to one special aspect of the theory, the development centered around Problem 1.3. This is a special aspect, because many results on varieties of minimal rational tangents and their applications are not directly related to it. Yet, this is the most fascinating aspect: it offers an area for a fusion of algebraic geometry and differential geometry, more specifically, a fusion of Mori geometry of minimal rational curves and Cartan geometry of cone structures. We will stick to the core of this aspect and will not go into the diverse issues arising from it. Those interested in further directions of explorations may find my MSRI article [13] useful. Conventions We will work over the complex numbers and all our objects are holomor- phic. Open sets refer to the Euclidean topology, unless otherwise stated. All manifolds are connected. A projective manifold is a smooth irreducible projective variety. A variety is a complex analytic set which is not necessarily irreducible, but has finitely many irreducible components. A general point of a manifold or an irreducible variety means a point in a dense open subset. 2. Cartan geometry: Cone structures A priori, this section is about local differential geometry and has nothing to do with rational curves. We will introduce a class of geometric structures, cone structures, and some related notions. In a simpler form, cone structures have already appeared in twistor theory(see[31]), butastheyarenotwidelyknown, Iwilltrytogiveadetailedintroduction. Definition 2.1. For a complex manifold M, let π : PT(M) → M be the projectivized tangent bundle. A smooth cone structure on M is a closed nonsingular subvariety C ⊂ PT(M)such thatallcomponentsofC have thesamedimension andtherestriction̟ := π|C is a submersion. We may restrict our discussion to smooth cone structures. Understanding the geome- try of smooth cone structures is already challenging and lots of examples of smooth cone structures remain uninvestigated. To have a satisfactory general theory, however, we need to allow certain singularity in C. This necessitates the following somewhat technical def- inition. (Readers not familiar with singularities may skip this definition and just stick to e Definition 2.1, regarding ν : C → C as an identity map in the subsequent discussion.) Definition 2.2. A cone structure on a complex manifold M is a closed subvariety C ⊂ e PT(M) the normalization ν : C → C of which satisfies the following conditions. e (1) All components of C are smooth and have the same dimension. e (2) The composition ̟ := π ◦ν : C → M is a submersion. In particular, the relative e tangent bundle T̟ ⊂ T(C) is a vector subbundle. e (3) There is a vector subbundle T ⊂ T(C) with T̟ ⊂ T and rank(T ) = rank(T̟)+1, e such thatforanyα ∈ C andanyv ∈ T \T̟, thenonzero vector d̟ (v) ∈ T (M) α α α ̟(α) 8 JUN-MUKHWANG satisfies [d̟ (v)] = ν(α) as elements of PT (M). α ̟(α) The conditions (1) and (2) say that C is allowed to be singular but it becomes smooth after normalization and the natural projection to M becomes a submersion. Note that on PT(M), we have the tautological line bundle ξ ⊂ π∗T(M). The condition (3) says that the quotient line bundle T /T̟ is naturally isomorphic to ν∗ξ. Another useful interpretation of the condition (3) is in terms of the following Definition 2.3. Given a cone structure C ⊂ PT(M), let Sm(C) ⊂ C be the maximal dense open subset such that π| : Sm(C) → π(Sm(C)) Sm(C) is a submersion. DenotebyTPT(M) ⊂ T(PT(M))theinverseimageofthetautologicalbundleξ ⊂ π∗T(M) under dπ : T(PT(M)) → π∗T(M). Then the condition (3) means that the vector bundle TPT(M)∩T(Sm(C)) on Sm(C), after pulling back to Ceby ν, extends to a vector subbundle e of T(C). From this interpretation of (3), it is easy to see that Proposition 2.4. A cone structure C ⊂ PT(M) is a smooth cone structure if and only if e C is normal, i.e., the normalization ν : C → C is biholomorphic. All three conditions (1)-(3) for cone structures are of local nature on M. This implies Proposition 2.5. Given a cone structure C ⊂ PT(M) and a connected open subsetU ⊂ M, the restriction C| := C ∩PT(U) ⊂ PT(U) U is a cone structure on the complex manifold U. By Proposition 2.5, we can view a cone structure as a geometric structure on M. We are interested in Cartan geometry of cone structures. In particular, isomorphisms in cone structures are given by the following Definition 2.6. A cone structures C ⊂ PT(M) on a complex manifold M is equivalent to a cone structure C′ ⊂ PT(M′) on a complex manifold M′ if there exists a biholomorphic map ϕ : M → M′ such that the projective bundle isomorphism Pdϕ : PT(M) → PT(M′) induced by the differential dϕ : T(M) → T(M′) of ϕ satisfies Pdϕ(C) = C′. It is convenient to have a localized version of this: Definition 2.7. For a cone structure C ⊂ PT(M) (resp. C′ ⊂ PT(M′)) and a point x ∈ M ′ ′ ′ ′ (resp. x ∈ M ), we say that C at x is equivalent to C at x if there exists a neighborhood ′ ′ ′ U ⊂ M of x and a neighborhood U ⊂ M of x such that the restriction C| is equivalent U ′ ′ to C |U′ as cone structures. We say that C is locally equivalent to C if there are points ′ ′ ′ ′ x ∈ M and x ∈ M such that C at x is equivalent to C at x. VMRT-STRUCTURES 9 Let us give one simple example of a cone structure. Let V be a vector space and Z ⊂ PV be a projective variety all components of which have the same dimension such e that the normalization Z is nonsingular. Via the canonical isomorphism T(V) = V ×V, the projectivized tangent bundle PT(V) = V ×PV contains the subvariety C := V ×Z ⊂ e e PT(V). This is a cone structure. Indeed the normalization C is just V ×Z which is smooth e e and ̟ : C → V is just the projection V × Z → V which is a submersion, verifying the conditions (1) and (2) of Definition 2.2. The tautological line bundle of Z ⊂ PV induces e e a line bundle χ in T(V ×Z) via the normalization morphism Z → Z and the subbundle e T = T̟ +χ of T(C) satisfies the condition (3). Definition 2.8. The cone structure V × Z ⊂ PT(V) on V defined above is called the flat cone structure with a fiber Z ⊂ PV. We will denote it by FlatZ ⊂ PT(V). A cone V structure on a complex manifold M is locally flat if it is locally equivalent to FlatZ for V some Z ⊂ PV with dimV = dimM. Definition 2.9. Let Z ⊂ PV be a projective variety. A cone structure C ⊂ PT(M) is Z-isotrivial if for a general x ∈ M, the fiber C = C ∩PT (M) ⊂ PT (M) x x x isisomorphictoZ ⊂ PV asaprojectivevariety, i.e.,asuitablelinearisomorphismT (M) → x V sends C to Z. A cone structure is isotrivial if it is Z-isotrivial for some Z. x A locally flat cone structure is isotrivial. But an isotrivial cone structure needs not be locally flat. Some isotrivial smooth cone structures are very familiar objects in differential geometry. When Z ⊂ PV is a linear subspace of dimension p, a Z-isotrivial cone structure on M is just a Pfaffian system of rank p + 1 on M. It is locally flat if and only if the Pfaffian system is involutive, i.e., it comes from a foliation. When Z ⊂ PV is a nonsingular quadric hypersurface, a Z-isotrivial cone structure is a conformal structure on M. It is locally flat if and only if it is locally conformally flat. A natural generalization of the conformal structure is the cone structure modeled on an irreducible Hermitian symmetric space S = G/P. The isotropy action of P on the tangent space T (S) at the base point o o ∈ S has a unique closed orbit C ⊂ PT (S). A Z-isotrivial cone structure where Z ⊂ PV o o is isomorphic to C ⊂ PT (S) is called an almost S-structure. A conformal structure is o o exactly an almost S-structure where S is a nonsingular quadric hypersurface, equivalently, an irreducible Hermitian symmetric space of type IV. The natural almost S-structure C ⊂ PT(S) given by the translate of C by G-action is locally flat, which can be seen by o Harish-ChandracoordinatesofirreducibleHermitiansymmetric spaces(see Section(1.2)in [35]forapresentationintermsofexplicit coordinatesforGrassmannians). TheG-structure on an irreducible Hermitian symmetric space S referred to in Section 1 is essentially equal to the cone structure C ⊂ PT(S). How do we check the local equivalence of two cone structures? A general method of checking equivalence of geometric structures has been formulated by Elie Cartan [2]. The fundamental apparatus in Cartan’s method is a coframe. 10 JUN-MUKHWANG Definition 2.10. Let V be a vector space and let M be a complex manifold with dimV = dimM. A coframe on M is a trivialization ω : T(M) → M ×V, equivalently, a V-valued 1-formon M such that ω : T (M) → V is an isomorphism for each x ∈ M. We will denote x x by Pω : PT(M) → M ×PV the trivialization of the projectivized tangent bundle induced by ω. Given a coframe, there exists a Hom(∧2V,V)-valued function σω on M, called the structure function of ω, such that dω = σω(ω ∧ω). • A coframe is closed if dω = 0, i.e., the structure function σω is identically zero. A coframe is conformally closed if there exists a holomorphic function f on an open subset U ⊂ M such that fω is closed on U. The following is a simple consequence of the Poincar´e lemma (see Theorem 3.4 in [12]). Proposition 2.11. Let V∨ ⊂ Hom(∧2V,V) be the natural inclusion of the dual space of V given by contracting with one factor. When dimM ≥ 3, a coframe ω is conformally closed if and only if σω takes values in V∨. Although Cartan’s method is applicable to the equivalence problem for arbitrary cone structures, its actual implementation can be challenging, depending onthe type of the cone structure. For isotrivial cone structures, however, this becomes simple: Definition 2.12. Let C ⊂ PT(M) be a Z-isotrivial cone structure for a projective variety Z ⊂ PV. A coframe ω : T(M) → M ×V is adapted to the cone structure if Pω(C) = Z. Proposition 2.13. An isotrivial cone structure is locally flat if and only if after restricting to an open subset, it admits a conformally closed adapted coframe. Since an isotrivial cone structure always admits an adapted coframe, we can use Propo- sition 2.11 and Proposition 2.13 to check the local flatness of an isotrivial cone structure. One difficulty here is that there may be several different adapted coframes, so we need to choose the right one. Different choices of adapted coframes are related by the linear automorphism group of the fiber. Let us elaborate this point. b For a projective variety Z ⊂ PV, let Z ⊂ V be its homogeneous cone. Denote by b b b Aut(Z) ⊂ GL(V) the linear automorphism group of Z and by aut(Z) ⊂ gl(V) its Lie b b algebra. SinceZ ⊂ V isacone, theLiealgebraaut(Z)alwayscontainsthescalarsC. When b aut(Z) = C,aZ-isotrivialconestructurehasauniqueadaptedcoframeuptomultiplication byfunctions. Consequently, themethodof Proposition2.13essentially determines thelocal b flatness of Z-isotrivial cone structures when aut(Z) = C. b b When aut(Z) 6= C, however, compositions with Aut(Z)-valued functions give rise to many different choices of adapted coframes for a Z-isotrivial cone structure. In this case, Proposition 2.13 is not decisive and we have to consider the problem of choosing the right coframe. This leads to the equivalence problem for G-structures where G corresponds to b the group Aut(Z) ⊂ GL(V). The general theory of G-structures has been developed by manymathematicians. Inparticular, fortheG-structuresmodeledonHermitiansymmetric