Quantum principal bundles on projective bases

The purpose of this paper is to propose a sheaf theoretic approach to the theory of quantum principal bundles over non affine bases. We study the quantization of principal bundles G ->G/P, where G is a semisimple group and P a parabolic subgroup.


Introduction
A quantum principal bundle is usually described as an algebra extension B ⊂ A, with A the "total space" algebra on which coacts a quantum group, and B the "base space" subalgebra of coinvariant elements.Local triviality is encompassed in the notion of locally cleft extension.
In the commutative setting, this picture proves to be extremely effective when the base space M is affine, that is, when the algebra B is containing all of the information to reconstruct the base space.For a projective base, however, the coinvariant ring B consists of just the constants, so it is not the object of interest anymore.
In this paper we take a very general point of view on the definition of quantum principal bundle (see Definition 2.3), so that we can accomodate the affine setting mentioned above, but also the case of projective base, together with a preferred projective embedding.In our definition a quantum principal bundle is a locally cleft sheaf of H comodule algebras for a given Hopf algebra H.In the commutative setting, when the base is affine the algebra of global sections (regular functions on the total space) is an Hopf-Galois extension; when the base is a projective variety our notion still makes sense and it actually gives the correct point of view to proceed to the quantization.
The definition is tested on an important special case, that when M is the quotient of a semisimple group G and a parabolic subgroup P .In this case, in fact, M = G/P is projective, and we can effectively substitute the coinvariant ring B with the homogeneous coordinate ring Õ(G/P ) of G/P with respect to a chosen projective embedding, corresponding to a line bundle L. The line bundle L can be recovered more algebraically via a character χ of P ; the corresponding sections are the semi-coinvariant elements of O(G) with respect to χ and generate the homogeneous coordinate ring Õ(G/P ) of G/P .In this case the locally cleft sheaf of H = O(P )-comodule algebras, denoted F , gives the subsheaf of coinvariants F co O(P ) that is the structure sheaf O G/P of G/P .The relation between this latter and the homogeneous coordinate ring Õ(G/P ) is then as usual by considering projective localizations (zero degree subalgebras of the localizations) of Õ(G/P ).
Similarly, in the quantum case, as in [9,17] we obtain the quantum homogeneous coordinate ring Õq (G/P ) as the O q (P )-semi-coinvariant elements of the quantum group O q (G), the quantization of the semisimple group G. Assuming Ore conditions for localizations, we then proceed to obtain from Õq (G/P ) and O q (G) a suitable sheaf F of O q (P )-comodule algebras, which will be the quantum principal bundle over the quantum space obtained through Õq (G/P ).More explicitly, the coinvariant subsheaf F co Oq(P ) will be the quantum structure sheaf associated with the (noncommutative) projective localizations of Õq (G/P ).
The quantization of the flag variety G/P and its noncommutative geomety has recently attracted a lot of attention.The theory, also following the remarkable classification of differential calculi over irreducible quantum flag manifolds in [22,23], has been conspicuously developed in the past years, see for example [10,24,25,30,31,11].In particular, the study of quantum projective space as a quantum homogeneous space has proven fruitful, however, it has mainly concerned quantum projective space as the base space of a quantum principal U(N − 1)-bundle with quantum SU(N) total space, i.e., a study not in the projective context.Indeed, despite the progress on quantum principal bundles [5,3,6,20], the projective setting, describing quantum versions of principal bundles G → G/P , with P parabolic, is yet to be fully understood.The aim of this paper is to provide a key step in this direction, together with an appropriate setting for a future differential calculus on such quantizations.
We summarize the main results by explaining the organization of the paper.
In Section 2 we recall basic notions in Hopf-Galois extensions, including the inspiring sheaf approach of [34,7].We then present our sheaf theoretic definition of quantum principal bundle.We also provide the example of SL 2 (C)/P both in the classical and in the quantum setting.This serves also as motivation and preparation for the general theory we develop in later sections.
In Section 3 we discuss quantum homogenous projective varieties, mainly following [9, §2].Starting from a quantum section d ∈ O q (G), quantum version of the lift to O(G) of the character χ of P defining the line bundle L giving the projective embedding of G/P , we construct the homogeneous ring Õq (G/P ).
In Section 4, we develop a general theory for quantum principal bundles on homogeneous projective varieties.We construct the sheaf F of O q (P )comodule algebras on the quantum projective variety Õq (G/P ) by local data, that is by considering suitable projective localizations of Õq (G/P ), obtained via a corresponding quantum section d ∈ O q (G).As shown in Theorem 4.8, if this sheaf is locally cleft we have a quantum principal bundle.
In Section 5, we exemplify the construction of Section 4 in the case of quantum projective space.We prove that quantum projective space is the base space of a canonical quantum principal bundle with total space O q (SL n ) and structure group O q (P ) (quantum parabolic subgroup of O q (SL n )).
In Section 6, we apply and further develop the results in [1] and show that 2-cocycle deformations (twists) of quantum principal bundles give new quantum principal bundles.We construct three classes of quantum principal bundles on quantum projective spaces.The first two are locally cleft but not locally trivial.The total spaces are not Hopf algebras hence they are not quantum principal bundles on quantum homogenous projective space as in the construction presented in Theorem 4.8.The second and third class are on multiparametric quantum projective space, the third class being also an example of the construction in Theorem 4.8, with total space the multiparametric special linear quantum group.

Quantum Principal bundles
In the category of locally compact Hausdorff topological spaces, a principal bundle is a bundle E → M, with compatibility requirements regarding the P -space structure, for a given topological group P .These requirements can be effectively summarized by asking that the map is a homeomorphism, with M = E/P .We can dualize this picture by replacing spaces with their function algebras, that is we replace E with A = C(E), M with B = C(M) and P with H = C(P ).The notion of principal bundle is then replaced by that of faithfully flat Hopf-Galois extension.The Hopf-Galois property is the freeness of the P -action, and amounts to the requirement that the pull-back of the above map, called canonical map, is a bijection.The faithfully flat property, or equivalently, the equivariant projectivity conditions correspond to the principality of the action (see e.g.[6]).
In the affine algebraic category we can proceed and give the same definitions, where in place of C(E), C(M) and C(P ) we take the coordinate rings of E, M and P .In fact, the contravariant functor associating to affine varieties their coordinate ring is an equivalence of categories (see [21,Proposition 2.6, §II] for more details).
However, when we turn to examine the case of projective varieties, since the above mentioned equivalence of categories does not hold anymore as stated, but becomes more involved, we need to take a different approach to the theory of principal bundles, introducing the sheaves of functions on our geometric objects.As it turns out, this approach, despite its apparent complication and abstraction is very suitable for quantization.

The Classical description
We start with a description of the classical setting.Definition 2.1.Let E and M be topological spaces, P a topological group and ℘ : E −→ M a continuous function.We say that (E, M, ℘, P ) is a Pprincipal bundle (or principal bundle for short) with total space E and base M, if the following conditions hold 1. ℘ is surjective.
2. P acts freely from the right on E.

P acts transitively on the fiber
4. E is locally trivial over M, i.e. there is an open covering M = ∪U i and homeomorphisms σ i : We can speak of algebraic, analytic or smooth P -principal bundles, we just take the objects and the morphism of Def.2.1 in the appropriate categories.Notice that ℘ is open.
In [34] Pflaum gives a sheaf theoretic characterization of principal bundles, in the category of locally compact topological spaces, which is very suitable for noncommutative geometry.
In the algebraic category, over a field k, we can give another characterization of principal bundles, closely related to Pflaum's one.For the basic definitions regarding algebraic groups we refer e.g. to [2,§II], for Hopf algebras e.g. to [28], [4, Part VII §5 ].
Proposition 2.2.Let ℘ : E −→ M be a surjective morphism of algebraic varieties, and O E , O M the structural sheaves of E and M respectively.Let F be the sheaf on M defined by F (U) = O E (℘ −1 (U)).Let P be an affine algebraic group, H the associated Hopf algebra.Then E −→ M is a principal bundle if and only if

comodule algebras;
• There exists an open covering {U i } of M such that we have the following algebra isomorphisms where We notice that condition (1) establishes M ≃ E/P ; we will identify M and E/P , so that correspondingly (2) gives the local triviality, the transitive action of P on the fiber and the freeness of the P action on E. We leave the details of this characterization to the reader, it will be a small variation of the argument given in [34].

The Quantum description
We now proceed and extend this point of view in order to give the definition of quantum principal bundle: it is based on [34] (see also Proposition 2.2) and also on [7], but it is more general since it encompasses the possibility for the base manifold to be projective.Furthermore, we take our category to be algebraic.
We will work with algebras (not necessarily commutative) over a field k of characteristic 0, or the ring of Laurent polynomials k q = k[q, q −1 ], q an indeterminate.All algebras will be unital and morphisms preserve the unit.In particular we will work with H-comodule algebras (A, δ), where δ denotes the Hopf algebra coaction (frequently omitted).Hopf algebras will be with bijective antipode.Definition 2.3.Let (H, ∆, ε, S) be a Hopf algebra and A be an H-comodule algebra with coaction δ : The extension A of the algebra B is called H-Hopf-Galois (or simply Hopf-Galois) if the map If E −→ M is a P -principal bundle and E, M and P are affine algebraic varieties or differentiable manifolds, then the algebra of functions (algebraic or differential) on E and P correspond respectively to the algebras A and H satisfying Definition 2.3.The algebra B is the algebra of functions on the base manifold M (see e.g.[6], [1] for details).
Example 2.4.Let B be an algebra with trivial right H-coaction, i.e., δ(b) = b ⊗ 1 for all b ∈ B. Consider H as an H-comodule algebra with the coaction given by the coproduct ∆.Then A := B ⊗ H is a right H-comodule algebra (with the usual tensor product algebra and right H-comodule structure).We have We denote as usual by ℓ * j the convolution product of two linear maps j : we can require j : H → A to be a right H-comodule map where H has H-comodule structure given by ∆, i.e., δ • j = (j ⊗ id) • ∆.Definition 2.5.Let H be a Hopf algebra and A an H-comodule algebra.The algebra extension A co H ⊂ A is called a cleft extension if there is a right H-comodule map j : H → A, called cleaving map, that is convolution invertible.
An extension A co H ⊂ A is called a trivial extension if there is an Hcomodule algebra map j : H → A.
Since 1 H is a grouplike element j(1 H )j −1 (1 H ) = 1 A , so that j(1 H ) is an invertible element in A co H . Hence a cleaving map can always be normalised to j(1 H ) = 1 A .We will always consider normalized cleaving maps.
Remark 2.7.A trivial extension A co H ⊂ A is automatically a cleft extension.In fact, since an H-comodule algebra map j : H → A maps the unit of H in that of A, its convolution inverse is j −1 = j • S. Furthermore, the H-comodule algebra map j : The extension B ⊂ B⊗H of Example 2.4 is an example of trivial extension (with j(h) = 1 B ⊗ h, for all h ∈ H).
Theorem 2.8.Let A be an H-comodule algebra (with base ring a field k), then A co H ⊂ A is a cleft extension if and only if A co H ⊂ A is an Hopf-Galois extension and there is an H-comodule and left Here B ⊗ H is an H-comodule with H-coaction id ⊗ ∆.For later use we recall that the relation betweeen a cleaving map j : H → A and the left The notion of cleft extension is the noncommutative generalization of that of trivial principal bundle.The next observation sharpens the relation between trivial Hopf-Galois extensions, trivial principal bundles and cleft extensions.
Observation 2.9.If j : H → A is an H-comodule algebra map, then we have an action of H on B = A co H given by h ⊲ b = j(h 1 ) b j −1 (h 2 ) = j(h 1 ) b j(S(h 2 )), for all h ∈ H, b ∈ B. We can therefore consider the smashed product algebra B ♯ H, that is the H-comodule B ⊗H with product structure central the smashed product is the usual tensor product of algebras.In particular, in the affine case, we immediately recover that a P -principal bundle E → E/P is trivial if and only if O(E/P ) ⊗ O(P ) ≃ O(E) as O(P )-comodule algebras.
In the more general case of an extension that is nontrivial but cleft, the map j : H → A is not an H-comodule algebra map, and the 2-cocycle measures this failure.In general the map h ⊗ b → j(h 1 ) b j −1 (h 2 ) is not an action of H on B. In this cleft case we can still induce via the isomorphism θ : B ⊗ H → A an algebra structure on B ⊗ H, this corresponds to a crossed product B ♯ τ H (see e.g.[28,Proposition 7.2.3]).
We want to present a notion of quantum principal bundle that is more general than that of Hopf-Galois extension presented in Def.2.3, and which can accomodate also the case where M is an algebraic variety, which is not affine.To this end, we consider a sheaf theoretic description of quantum principal bundles.We start by introducing the notion of quantum ringed space.
The locally cleft property is equivalent to the existence of a projective cleaving map that is a collection of cleaving maps j i : H −→ F (U i ).
Remark 2.12.A sheaf of Hopf-Galois extentions is locally cleft if it so as a sheaf of H-comodule algebras.A locally cleft sheaf F of Hopf-Galois extensions is in particular a quantum principal bundle on the quantum ringed space (M, F coH ).
Moreover, a sheaf F of H-comodule algebras, such that the extension O M (M) = F (M) coH ⊂ F (M) is Hopf-Galois, is equivalent to a sheaf of Hopf-Galois extensions, indeed, as observed in [7], the property of being Hopf-Galois restricts locally.Therefore, a quantum principal bundle F has the property O M (M) = F (M) coH ⊂ F (M) is Hopf-Galois, if and only if it is a locally cleft sheaf of Hopf-Galois extensions.
Let us see a simple example, in the commutative setting, that we will generalize to the noncommutative setting and generic dimensions.
Example 2. 13.Let E = SL 2 (C) and consider the principal bundle ℘ : SL 2 (C) −→ SL 2 (C)/P ≃ P 1 (C), where P is the upper Borel in SL 2 (C), i.e., the subgroup of all matrices with vanishing entry (1,2).Let A = O(SL 2 ) be the algebra of regular functions on the complex special linear group SL 2 (C).We explicitly have where C[a, b, c, d] denotes the commutative algebra over C freely generated by the symbols a, b, c, d, while (ad − bc − 1) denotes the ideal generated by the element ad − bc − 1, that implements the determinant relation.
Let O(P ) be the algebra of functions on ( The coinvariants B = A co O(P ) of this coaction are just the constants, indeed the coinvariant are functions on the base space P 1 (C), and the only regular functions on all projective space are the constants (Liouville theorem).We see that the extension A co O(P ) ⊂ A is not Hopf-Galois, and that this is due to the lack of regular functions on the base space of the P -principal bundle Nevertheless, we can define an O(P )-principal bundle structure according to Definition 2.11.To this aim, we first consider an affine open cover of the total space and then we project it to the base.
Let {V 1 , V 2 } be the open cover of SL 2 (C) where V i consists of those matrices in SL 2 (C) with entry (i, 1) not equal to zero.Define U i = ℘(V i ) and observe that {U 1 , U 2 } is an open cover of P 1 (C) since ℘ is an open map.The algebras of functions on the opens V 1 and V 2 are the localizations

The coaction in (3) uniquely extends to coactions δ
Notice that they are the coordinate rings of the affine algebraic varieties Next we consider on P 1 (C) the topology {∅, is a rough topology, but sufficient to describe the principal bundle on P 1 (C)).We then define the ringed space (P 1 (C), O P 1 (C) ) with sheaf of regular functions O P 1 (C) given by and with O P 1 (C) (∅) being the one element algebra over C, terminal object in the category of algebras.It is easy to verify that the restriction morphism r 12,2 : B 2 → B 12 , w → z −1 , with all other ones being given by the obvious inclusions (but for the empty set where we have the canonical projections), indeed define the sheaf of regular functions on P 1 (C).
Finally we define the sheaf F of O(P )-comodule algebras and F (∅) := {0} (the one element algebra) with the obvious restriction morphisms.
We now show that all properties required by Def.2.11 are satisfied.

Indeed by construction O(U
) is a trivial extension (and hence a cleft extension) because the map j 1 : O(P ) → A 1 defined on the generators by and extended as algebra morphism to all O(P ) is well defined and easily seen to be an O(P )-comodule morphism (recall Example 2.14.We discuss the quantum deformation of the previous example.Consider the algebra A q that is the algebra C q a, b, c, d freely generated (over C q = C[q, q −1 ], q an indeterminate that may be specialized to a complex number) by the symbols a, b, c, d, modulo the ideal I M generated by the q-commutation relations (or Manin relations, cf.Def.5.1), and modulo the ideal (ad − q −1 bc − 1) generated by the determinant relation.
In short: Let us similarly define Let U i be a cover of M = SL 2 (C)/P as in Example 2.13.In analogy with the classical case we define , the noncommutative localizations in the elements a and c respectively.The coinvariants are given by and the ringed space (P 1 (C), O q P 1 (C) ) can be then easily constructed in analogy with the commutative case: with the nontrivial restriction map given by r q 12,2 : B q 2 → B q 12 , v → u −1 that is again well defined since on U 12 one has uv = 1 = vu.The natural candidate is again a sheaf of O q (P )-comodule algebras on P 1 (C); note in particular that A q 12 is well defined since the localization we choose satisfies the Ore condition (see [35]).As in the previous section we define the cleaving maps j i : O q (P ) −→ A q,i , i = 1, 2 on the generators as: We observe that j 1 extends to an algebra map to all A q,1 : and similarly for j 2 .The comodule property of j 1 (and similarly for j 2 ) is then easily checked on the generators: We can then conclude that A q i are trivial O q (P )-extensions of B q i .
We will study a generalization of the above example in Section 5.In that more general setting we will use the following proposition (see e. O q (P )-comodule algebras (here r q 12,i : A q i → A q 12 are the obvious restriction maps).

Quantum homogenous projective varieties
A homogenous projective variety can be realized as quotient of affine algebraic groups G, P .Its homogenous coordinate ring Õ(G/P ) with respect to a chosen projective embedding, when corresponding to a very ample line bundle L, is obtained via a section of L; this is a a given element t ∈ O(G).A quantum homogenous projective variety Õq (G/P ) can be similarly characterized via a quantum section d ∈ O q (G).We review this construction due to [9], see also [17], adapting, for the reader's convenience, the main definitions and results to the present setting that differs from the first reference setting (there the accent was on Poisson geometry and Quantum Duality principle).

Projective embeddings of homogeneous spaces
If G is a semisimple algebraic group, P a parabolic subgroup, the quotient G/P is a projective variety and the projection G −→ G/P is a principal bundle (see Definition 2.1).G/P is an homogeneous space for the G-action and just an homogeneous variety for the P -action, which is not transitive.
We now recall how a character of P determines a projective embedding of G/P and its coordinate ring Õ(G/P ).Given a representation ρ of P on some vector space V , we can construct a vector bundle associated to it, namely The space of global sections of this bundle is identified with the induced module (see, e.g., [21] for more details) In particular, for χ : P −→ k * a character of P , i.e. a one dimensional representation of P on L ≃ k , we can consider L n := G × P L ⊗n and define Assume L is very ample, i.e. it is generated by a set of global sections f 0 , f 1 , . . ., f N ∈ Õ(G/P ) 1 ; so that the algebra Õ(G/P ) is graded and generated in degree 1 (by the f i 's).Then Õ(G/P ) is the homogeneous coordinate ring of the projective variety G P with respect to the embedding given via the global sections of L (see [14], p. 176).
Observation 3.1.While O G/P denotes the structure sheaf of G/P , so that O G/P (G/P ) is the space of global sections, that is k since G/P is a projective variety, Õ(G/P ) denotes the homogeneous coordinate ring of G/P .
We want to reformulate this classical construction in purely Hopf algebraic terms.The character χ is a group-like element in the coalgebra O(P ) .The same holds for all powers χ n (n ∈ N ).As the χ n 's are group-like, if they are pairwise different they also are linearly independent, which ensures that the sum n∈N Õ(G/P ) n , inside O(G), is a direct one.Moreover, once the embedding is given, each summand Õ(G/P ) n can be described in purely Hopf algebraic terms as where Õ(G/P ) is the homogeneous coordinate ring generated by the global sections of L, i.e. generated by Õ(G/P ) 1 .
Vice-versa, given t ∈ O(G) satisfying (5), (6), if Õ(G/P ) as defined in (7), ( 8) is generated in degree 1, namely by Õ(G/P ) 1 , then Õ(G/P ) is the homogeneous coordinate ring of the projective variety G/P associated with the projective embedding of G/P given by the very ample line bundle L = G× P k, the P -action on the ground field k being induced by π(t).
Notice that while S(χ) = π(t) is group-like, t has an "almost grouplike property", given by (5).We call an element t ∈ O(G) satisfying ( 5), (6) a classical section because t ∈ Õ(G/P ) 1 .The line bundle L and the homogenous coordinate ring Õ(G/P ) depend only on π(t), not on the lift t.
Remark 3.3.We point out that Õ(G/P ) is a unital subalgebra as well as a (left) coideal of O(G); the latter property reflects the fact that G P is a (left) G-space.Thus, the restriction of the comultiplication of O(G) , namely is a left coaction of O(G) on Õ(G/P ), which structures Õ(G/P ) into an O(G)-comodule algebra.Moreover Õ(G/P ) is graded and the coaction ∆ Õ(G/P ) is also graded with respect to the trivial grading on O(G) , so that each Õ(G/P ) n is indeed a coideal of O(G) as well.

Quantum homogeneous projective varieties and quantum sections
We quickly recall some definitions of quantum deformations and quantum groups, establishing our notation.We define quantum homogeneous spaces and then turn to the quantization of the picture described in the previous section.
Definition 3.4.By quantization of O(G), we mean a Hopf algebra O q (G) over the ground ring k q := k[q, q −1 ], where q is an indeterminate, such that: 1. the specialization of O q (G) at q = 1 , that is O q (G)/(q −1) O q (G) , is isomorphic to O(G) as an Hopf algebra; 2. O q (G) is torsion-free, as a k q -module; We also call O q (G) a quantum deformation of G , or for short, quantum group.We also say that the is the coordinate ring of an affine variety M, we further say that O q (M) is a quantization of M. If Õ(M) is the homogeneous coordinate ring of a projective variety, with respect to a given projective embedding, we say that Õq (M) is a quantization of M provided it is graded and the quantization preserves the homogeneous components.
We next define quantum homogeneous varieties, in this case M = G/P .Definition 3.5.Let G/P be a homogeneous space with respect to the action of an algebraic group G.If G/P is affine we say that its quantization O q (G/P ) is a quantum homogeneous variety (space) if O q (G/P ) is a subalgebra of O q (G) and an O q (G)-comodule algebra.If G/P is projective and Õ(G/P ) is its homogeneous coordinate ring with respect to a given projective embedding, then we ask its quantization Õq (G/P ) to be a O q (G)-comodule subalgebra of O q (G).We furtherly ask the algebra Õq (G/P ) to be graded and the O q (G)-coaction to preserve the grading.In this case we call Õq (G/P ) a quantum homogeneous projective variety.
Let O q (G) be a quantum group and O q (P ) a quantum subgroup (quotient Hopf algebra), quantizations respectively of G and P as above.Since from Proposition 3.2 a classical section t defines a line bundle on G/P and a projective embedding, we study a quantum projective embedding by quantizing this classical section.Definition 3.6.A quantum section of the line bundle L on G P associated with the classical section t, is an element where π : O q (G) −→ O q (P ) := O q (G)/I q (P ), I q (P ) ⊂ O q (G) being a Hopf ideal, quantization of the Hopf ideal I(P ) defining P .

Define now:
Õq (G/P ) := n∈N Õq (G/P ) n , where We recall a result from [9].where we consider O q (G) with the trivial grading.
3. As algebra Õq (G/P ) is a subalgebra of O q (G).
Hence Õq (G/P ) is a quantum homogeneous projective variety.
From now on we assume that Õq (G/P ) is generated in degree one, namely by Õq (G/P ) 1 .The quantum Grassmannian and flag are examples of this construction and they are both generated in degree one.
Example 3.8.Let us consider the case G = SL n (C) and P the maximal parabolic subgroup of G: The quotient G/P is the Grassmannian Gr of r spaces into the n dimensional vector space C n .It is a projective variety and it can be embedded, via the Plücker embedding, into the projective space P N (C) where N = n r .This embedding corresponds to the character: The coordinate ring O(Gr) of Gr, with respect to the Plücker embedding, is realized as the graded subring of O(SL n ) generated by the determinants d I of the minors obtained by taking (distinct) rows I = (i 1 , . . ., i r ) and columns 1, . . ., r.In fact one can readily check that d = det(a ij ) 1≤i,j≤r is a classical section and, denoting by π : O(SL n ) −→ O(P ) the natural projection dual to the inclusion P ⊂ SL n , that In [15] the quantum Grassmannian O q (Gr) is defined as the graded subring of O q (SL n ) generated by all of the quantum determinants D I of the minors obtained by taking (distinct) rows I = (i 1 , . . ., i r ) and columns 1, . . ., r.
It is a quantum deformation of O(Gr) and a quantum homogeneous projective space for the quantum group O q (SL n ), (see [15,17] for more details).Again one can readily check that d = D 1...r is a quantum section and that where O q (P ) = O q (G)/I q (P ) is the quantum subgroup of O q (G) defined by the Hopf I q (P ) = (a ij ) generated by the elements a ij for r + 1 ≤ i ≤ n and 1 ≤ j ≤ r, and π : O q (G) −→ O q (P ).
4 Quantum Principal bundles from parabolic quotients G/P In the previous section we have seen how to construct a quantum homogenous projective variety Õq (G/P ) given a quantum section d ∈ O q (G).We here show how quantum sections lead to quantum principal bundles over quantum homogeneous projective varieties.

Sheaves of comodule algebras
Let as before G be a semisimple algebraic group, P a parabolic subgroup.
We start with a classical observation recalling the construction of a (finite) basis {t i } i∈I of the module of global sections of the very ample line bundle L → G/P associated with a classical section t ∈ O(G).We also construct the corresponding open cover {V i } i∈I of G. Observation 4.1.Recalling Proposition 3.2, we consider an element in t ∈ O(G) satisfying ( 5) and ( 6) and defining a very ample line bundle L → G/P , with t ∈ Õ(G/P ) 1 ⊂ O(G) that is now a section of L. Let ∆(t) = t (1) ⊗ t (2) = i∈I t i ⊗ t i be its coproduct and notice that the elements t i can be chosen to be linearly independent.We now show that {t i } i∈I is a basis of Õ(G/P ) 1 , the module of global section of L, hence the t i 's generate Õ(G/P ) as a (graded) algebra.Indeed, by the Borel-Weyl-Bott theorem, Õ(G/P ) 1 is an irreducible G module (corresponding to the infinitesimal weight uniquely associated to χ).By the very definition of ∆, the G-action on t is given by, for all g, x ∈ G: Since Õ(G/P ) 1 is irreducible, for any f ∈ Õ(G/P ) 1 there exists a g ∈ G, such that f = g • t and consequently f is a linear combination of the t i 's by (10).Hence the t i 's form a basis of Õ(G/P ) 1 .Furthermore, a covering of G is given by {V i } i∈I , where the open sets V i are defined by the non vanishing of the corresponding t i ∈ O(G).This is so because the line bundle L defines a projective embedding of G/P , hence there are no common zeros for its global sections.
Based on the previous observation we have the following important property of the quantum homogeneous projective variety Õq (G/P ).Lemma 4.2.Let d be a quantum section, and Then the d i 's can be chosen so to form a basis of Õq (G/P ) 1 as k q free module, hence of Õq (G/P ) as graded algebra.
Proof.The fact that the d i 's belong to Õq (G/P ) 1 is non trivial, but it is an immediate consequence of Proposition 3.10 in [9].The property that they generate Õq (G/P ) 1 as k q free module is a consequence of the same property being true in the classical setting (see Observation 4.1) and comes through the application of Proposition 1.1 in [19] followed by Lemma 3.10 in [18].The last property immediately follows from the assumption that Õq (G/P ) is generated by Õq (G/P ) 1 .
We assume that S i := {d r i , r ∈ Z ≥0 } is Ore in order to consider localizations of O q (G) and hence define a sheaf.We furtherly assume that S i is Ore in the graded subalgebra O q (G/P ) of O q (G).We can then define: the Ore extension of O q (G) with respect to the multiplicatively closed set given by: where with an abuse of notation we write π(d) −1 for the antipode of π(d) in O q (P ).
Proof.Notice that O q (G) is an O q (P )-comodule algebra with coaction ∆ ) by the universality of the Ore construction we have our definition of δ i .
Assume now we can form iterated Ore extensions: ) independently from the order, i.e. we assume to have a natural isomorphism between O q (V i ∩ V j ) and O q (V j ∩ V i ).This is in general a very restrictive hypothesis, neverthless we will see it is verified in some interesting examples in the next section.
We also define: as the natural morphism obtained from the Ore extension.
Setting as usual V I = ∩ i∈I V i we immediately have the following proposition (cf.Proposition 4.3).Proposition 4.4.O q (V I ) is an O q (P )-right comodule algebra and the morphisms r IJ are O q (P )-right comodule algebra morphisms.
Let us now consider the opens U I := ℘(V I ), obtained via the projection ℘ : G −→ G/P .We have the following.
Proposition 4.5.The assignment: with the restriction maps r IJ : O q (V I ) → O q (V J ), defines a sheaf of O q (P )comodule algebras on G/P = ∪ i∈I U i , and more in general on M := ∪ i∈J U i ⊂ G/P , where I ⊂ I and I ⊂ J ⊂ I respectively.
Proof.The opens U I with I ⊂ I (and the empty set) form a basis B for a topology on G/P .Recalling Proposition 2.15 we just have to show that the assignment U I → F (U I ) := O q (V I ), with the restriction maps r IJ , defines a B-sheaf of O q (P )-comodule algebras.Since restrictions morphisms are actually algebra inclusions, using the existence of iterated Ore extension and their compatibility this is straighforwardly seen to be a B-sheaf of algebras and of O q (P )-comodule algebras.
The sheaf on the more general open submanifold M = U i∈J U i is simply obtained by considering the opens U I with I ⊂ J ⊂ I.

Quantum principal bundles on quantum homogeneous spaces
In the previous section we have constructed a sheaf of comodule algebras F on M ⊂ G/P .We now want to define a quantum ringed space structure on the topological space M as in Definition 2.10 and show that F is a quantum principal bundle on it.Notice that M coincides with G/P if J = I, while for J I, i.e. for a proper subset of the set of indices I of the open cover {V i } i∈I of G, we have that M is a proper open subset of G/P .

By Observation 4.1 we know that {U
Because of our (graded) Ore hypothesis, this is also the subalgebra of elements of degree zero inside Õq (G/P )S −1 i and, for this reason, it is called the (noncommutative) projective localization of Õq (G/P ) at S i .Proposition 4.6.Let the notation be as above.The assignment Proof.According to Proposition 2.15 it is enough to check that our assignment is a B-sheaf for the basis associated with the opens {U i }, but this is immediate by our hypothesis on the existence of iterated Ore extension and their compatibility.Proposition 4.7.Let the notation be as above.Then F (U i ) co Oq(P ) = O M (U i ), i.e. it is the subring in F (U i ) generated by the elements d j d −1 i .Proof.By our definition of coaction δ i (see (12)) ) for a suitable r.Notice that: Hence zd r i ∈ Õq (G/P ) r , which, by Lemma 4.2, is generated by the d j 's: zd r i = λ j i ...jr ∈kq λ j i ...jr d j 1 . . .d jr .
Therefore we have: We now proceed by induction on r.The case r = 0 is clear.For generic r, since d i satisfies the Ore condition: .
By induction we obtain: We conclude summarizing the main results we have obtained.
Theorem 4.8.Let G be a semisimple algebraic group and P a parabolic subgroup, let the quantum group O q (G) and the quantum subgroup O q (P ) := O q (G)/I q (P ) be the quantizations of the coordinate rings O(G) and O(P ).
Let d be a quantum section (see Definition 3.6), denote with {d i } i∈I a choice of linearly independent elements in the coproduct ∆(d) = i∈I d i ⊗ d i , and assume they generate the homogenous coordinate ring Õq (G/P ) (see Lemma 4.2).Assume furtherly that O q (V i ) := O q (G)S −1 i , S i = {d r i , r ∈ Z ≥0 } is Ore and that subsequent localizations do not depend on the order (see (13)).Then: In particular, for M = G/P (J = I), the sheaf O G/P is the projective localization of the homogeneous coordinate ring Õq (G/P ).
2. The assignment: U I → F (U I ) := O q (V I ) defines a sheaf F of O q (P )comodule algebras on the quantum ringed space M = ∪ i∈J U i ⊂ G/P .
If the sheaf F is locally cleft (see Definition 2.11) then F is a quantum principal bundle.

Examples
In this section we apply the general theory we have developped and present quantum principal bundles over quantum projective spaces.We hence sharpen the notion of quantum projective space as quantum homogenous space.In this section the ground field is k = C.

Quantum deformations of function algebras
We start with an important example of quantum group and its quantum homogeneous varieties.For more details see [26] and [15].
Definition 5.1.We define the quantum matrices as the C q algebra O q (M n ): where i, j = 1, . . .n and I M is the ideal of the Manin relations: The quantum matrix algebra O q (M n ) is a bialgebra, with comultiplication and counit given by: We define the quantum general linear group to be the algebra where det q is the quantum determinant: where ℓ(σ) is the length of the permutation σ (see [33] for more details on quantum determinants).We define the quantum special linear group to be the algebra O q (GL n ) and O q (SL n ) are Hopf algebras and quantum deformations respectively of the general linear and the special linear groups.

Quantum principal bundles on quantum Projective spaces
We consider the special case of a maximal parabolic subgroup P of G = SL n (C) of the form: In this case G/P ≃ P n−1 (C) is the complex projective space, and Õ(P n−1 ) is the corresponding free graded ring with n generators.Its quantization Õq (P n−1 ) is well known and, for example, it is constructed in detail in [15] (see Theorem 5.4 for r = 1), see also [11].Õq (P n−1 ) is the subring of O q (SL n ) generated by the elements x i = a i1 , i ∈ I = {1, ...n}.We can immediately give a presentation: We reinterpret this construction within the present framework, first showing that Õq (P n−1 ) is a quantum homogeneous projective space according to Definition 3.5 and then constructing, along Theorem 4.8, an O q (P )-principal bundle on the ringed space obtained via projective localizations of Õq (P n−1 ).
Let O q (G) = O q (SL n ) be the quantum special linear group of Definition 5.1, and define the quantum parabolic subgroup where I q (P ) = (a α1 ) is the Hopf ideal generated by a α1 , α ∈ {2, . . .n}.We use coordinates p ij for the images of the generators a ij under π : Furthermore, from the coproduct ∆(a 11 ) = i∈I a 1i ⊗ a i1 we choose the linearly independent elements Hence, by Lemma 4.2, the elements d i span Õq (SL n /P ) 1 , as defined in (9).The quantum homogeneous projective variety Õq (SL n /P ) is generated in degree one, cf.Example 3.8, and one can see immediately that Õq (SL n /P ) coincides with Õq (P n−1 ), as defined in (17).
We now structure Õq (P n−1 ) as a quantum ringed space and construct a sheaf of locally trivial O q (P )-comodule algebras, i.e., a quantum principal bundle on the quantum projective space Õq (P n−1 ), where O q (P ) is the quantum parabolic subgroup of O(SL n ) defined in (18).
Let us consider the two classical open covers of the topological spaces SL n (C) and P n−1 (C) respectively: where a 0 ij denote the generators of O(SL n ) and similarly x 0 i those of Õ(P n−1 ), i, j = 1, . . ., n. Evidently, ℘( is , does not depend on the order of the Ore extensions.
Proof.See [36, pp. 4 and 5].Notice that a i1 is a quantum minor of order 1 and two such minors q-commute, hence their product forms an Ore set.
As a corollary of Theorem 4.8 we then immediately obtain Proposition 5.3.Let the notation be as in the previous section.The assignment: defines a sheaf of O q (P )-comodule algebras on SL n (C)/P .Furthermore, F (U i ) co Oq(P ) is generated by a i1 a −1 11 , F co Oq(P ) equals the projective localization of Õq (P n−1 ) and (SL n (C)/P, F co Oq(P ) ) is a quantum ringed space.
We now show that F is a quantum principal bundle on the quantum ringed space (SL n (C)/P, F co Oq(P ) ).The only property to be checked is the locally cleft condition (cf.Definition 2.5).We actually show the stronger local triviality condition, i.e., the collection of maps j i : O q (P ) → F (U i ) are O q (P )-comodule algebra maps, hence, in particular, are cleaving maps (cf.Remark 2.7).
We first study the map 11 ], i, j = 1, ...n; since a 11 is invertible we have the matrix factorization (20) where α, β = 2, . . .n, and D kl ij = a ik a jl − q −1 a il a jk , with i < j and k < l, denotes the quantum determinant of the 2 × 2 quantum matrix obtained by taking rows i, j and columns k, l.
In the commutative case this factorization corresponds to the trivialization the total space of SL n (C) → SL n (C)/P (cf.eq. ( 19)).In the quantum case we similarly have that F (U 1 ) co Oq(P ) ⊂ F (U 1 ) is a trivial Hopf-Galois extension.Recalling Remark 2.7 and Observation 2.9, we shall see it is the smashed product where the generators The properties of j 1 : O q (P ) → F (U 1 ) follow from the properties of an associated lift J 1 that maps into the localization O q (M n )[a − 1  11 ] of the quantum matrix algebra defined in (15).Lemma 5.4.Let O q (p ij ) denote the quantum matrix algebra with generators p ij = p 11 , p 1β , p αβ and p α1 = 0; α, β = 2, . . ., n.We have a well defined algebra map 1  11 ], that on the generators reads [16], the following commutation relations in O q (M n ) among quantum determinants and generators of the algebra of quantum matrices: . Also, by Theorem 7.3 in [16], the indeterminates u αβ := D 1β 1α satisfy the Manin relations as in Definition 5.1, where we replace a αβ with u αβ .In order to show that J 1 is an algebra map, we have to show it is well defined.First, we easily compute the commutation relations 11 , that imply that the a −1 11 D 1β 1α 's satisfy the Manin relations among themselves.Next, we need to check that the commutation relations between a 1γ , γ = 2, . . .n, and a −1 11 D 1β 1α are of the Manin kind.If γ > β, we have: because a 1β and D 1β 1α commute.If γ < β, we need to check the commutation: We leave this calculation as an exercise.
Lemma 5.5.Let the notation be as above.Let det q (p ij ) and det q (a ij ) denote respectively the quantum determinants in O q (p ij ) and Proof.In the factorization (20), define: for i, j = 1, . . ., n, α, β = 2, . . ., n.Since c ij = J 1 (p ij ), by Lemma 5.4, they form a quantum matrix and our claim amounts to det q (a ij ) = det q (c ij ).
We start by noticing that b ij and c kl satisfy the following commutation relations: We also notice the obvious facts: We proceed with a direct calculation of det q (a ij ) using a ij = k b ik c kj .Recall the quantum Laplace expansion along the first column (see [33] pg 47): where A(r, 1) is the quantum determinant obtained from (a ij ) by removing the r-th row and first column, where u means that we omit the term u.
Let us look at the term b where k 2 , . . ., k n = 1, . . ., n.We want to reorder it, and we claim that: By (23) b 2k 2 = 0 if and only if k 2 = 1, 2. So we have to reorder c k 2 σ(2) b 3k 3 only for k 2 < 3, hence, by (22), we have that they commute.The rest follows by repeated application of this argument.
Therefore, we can write the first term in (24) as: where C[k 2 , . . ., k n |2, . . ., n] is the quantum determinant in the indeterminates c ij obtained by taking rows (k 2 , . . ., k n ) (in this order) and columns (2, . . ., n).Notice that, by (23), the sum over the index k t runs only on the values k t = 1 and t.We rewrite the first term in (24) as: Let us now look at the second term in (24).Reasoning as before, we have: However, we notice that here it must be k 1 = 1, otherwise b 1k 1 = 0, hence this forces k t = t for all t > 1.So we can write: because by (22) we have b t1 c 11 = qc 11 b t1 .
If we substitute expressions ( 26) and ( 28) in ( 24) and simplify we remain with just one term: 11 ] defined on the generators as: 11 ]/(det q (a ij ) − 1) and O q (P ) = O q (SL n )/I q (P ) = O q (p ij )/(det q (p ij ) − 1) as algebras.Because of the previous lemma, j 1 : O q (P ) → O q (SL n )[a − 1  11 ] is well defined; in fact it is the algebra map 11 ] induced on the quotients.We next show that j 1 is an O q (P )-comodule morphism, i.e., δ 1 • j 1 = (j 1 ⊗ id) • ∆ P , where ∆ P is the comultiplication in O q (P ) and δ 1 is the O q (P ) coaction on F (U 1 ) = O(V 1 ) as defined in Proposition 4.3.Since j 1 is an algebra map, it is enough to check the comodule property on the generators.Let us look at the case of p αβ , the case p 1j being an easy calculation.On the one hand, using the coproduct formula for quantum minors (see e.g.[16]) On the other hand: We now extend the previous proposition in order to define the O q (P )comodule algebra maps j k : O q (P ) → F (U k ) = O q (SL n )[a −1 α1 ], (k = 1, . . .n) thus proving the triviality of the Hopf-Galois extensions F (U k ) co Oq(P ) ⊂ F (U k ).
Reasoning as before, for each fixed value of k, we consider the factorization of quantum matrices (a ij ) similar to (20): where β = 2, . . .n.This suggests to exchange row k with row 1 in order to identify the last matrix with the matrix of generators p 11 p 1β 0 p αβ of O q (P ).
Proof.This is a direct check similar to Proposition 5. Then we are left to check the commutation relations of a kγ with −qa −1 k1 D 1β αk , a −1 k1 D 1j ki and a −1 k1 D 1β kα .There are nine of these, depending on the combinations k > α, k = α, k < α with γ > β, γ = β, γ < β.These indeed correspond to the commutation relations between p 1γ and p αβ .
We conclude that j k is a well defined algebra map because in O q (SL n )[a −1 k1 ] we have j k (p 11 )j k (det q (p αβ )) = 1, consistently with the last of the defining relations of the algebra O q (P ): p 11 det q (p αβ ) = 1.This is obtained with the same argument as in Lemma 5.5.
Since j k is an algebra map it is an O q (P )-comodule map provided the comodule property δ 1 • j 1 = (j 1 ⊗ id) • ∆ P holds on the generators.It is straighforward to see that this is indeed the case on p 1j .Let's compute the case p αβ with α > k (the other cases being similar): Remark 5.8.Recalling Remark 2.7 and Observation 2.9, as corollary of the above proposition we have where it is easy to check that the smashed product is nontrivial (i.e., different from the tensor product).Theorem 5.9.Let the notation be as in the previous section.The assignment: defines a quantum principal bundle on the quantum ringed space (SL n (C)/P, F co Oq(P ) ), with structure sheaf F co Oq(P ) = O SLn/P given by projective localizations of the quantum homogeneous projective space Õq (P n−1 ) = Õq (SL n /P ).
Proof.After Proposition 5.3 we only need to prove the locally cleft property.This is a direct consequence of Proposition 5.7 and Remark 2.7.
Remark 5.10.Notice that our construction, and in particular Theorem 5.9, holds also when we take q ∈ C, that is, we specialize the indeterminate q to a complex value.
6 Quantum principal bundles from twists In this section we obtain new quantum principal bundles via 2-cocycle deformations.In particular we provide examples that are locally cleft from examples that are locally trivial.We here consider the ground ring to be a field, hence specialize q ∈ k.As in [1] we consider 2-cocycle (twist) deformations based on the "structure group" Hopf algebra H and also on an "external symmetry" Hopf algebra K, i.e. a Hopf algebra coacting on the quantum principal bundle, the coaction being compatible with that of H (in the commutative case K is associated with automorphisms of the bundle, possibly nontrivial on the base).

Deformations from twists of H
Let γ : H ⊗ H → k be a 2-cocycle of the Hopf algebra H, denote by γ −1 : H ⊗ H → k its convolution inverse and by H γ the new Hopf algebra that has the same costructures of H and new product • γ and antipode obtained by twisting the ones of H via γ.Explicitly the product reads, for all h, h ′ ∈ H, h ).We also denote with Γ the functor from the category of right H-comodule algebras to that of right H γcomodule algebras ) ).Since H and H γ have the same costructures, A γ is a right H γ -comodule algebra using the same comodule structure map as for A. The functor Γ is the identity on morphisms.Theorem 6.1.Let γ be a 2-cocycle of the Hopf algebra H and Γ the corresponding functor of comodule algebras.The sheaf F is an H-principal bundle (quantum principal bundle) over the ringed space Proof.If F is a sheaf of H-comodule algebras over M then Γ • F is easily seen to be a sheaf of H γ -comodule algebras over M (locality and the gluing property immediately follow recalling that Γ is the identity on objects).Vice versa, since the convolution inverse γ −1 is a 2-cocycle of H γ , and (H Let {U i } be a covering of M with F (U i ) coH = O M (U i ) and such that F is locally cleft.Since H γ and H have the same coproduct we have Remark 6.2.We further observe that if the H-principal bundle F is locally trivial with respect to a covering {U i }, i.e., the cleft extensions F (U i ) coH ⊂ F (U i ) are trivial extensions, so that F (U i ) ≃ F (U i ) coH ♯H (cf.Observation 2.9), then this is no more the case for the twisted H γ -principal bundle Γ • F because the extensions F (U i ) coHγ ⊂ F (U i ) γ are cleft but nontrivial.Indeed, as follows from [29, Theorem 5.2], F (U i ) γ ≃ F (U i ) coH ♯ γ −1 H γ , where ♯ γ −1 denotes the crossed product given by the 2-cocyle γ −1 of H γ .

Deformations from twists of K
Let now K be another Hopf algebra and F be a sheaf over the ringed space (M, O M ) of (K, H)-bicomodule algebras, i.e. right H-comodule algebras and left K-comodule algebras with left and right coactions commuting: (ρ ⊗ id) • δ = (id ⊗ δ) • ρ.Since k is a field, K is free as a k-module and F coH : U → F (U) coH is a subsheaf of K-comodule algebras (because F (U) coH are K-subcomodule algebras, cf.[1, Proposition 3.12]).
A twist σ of K gives the functor Σ from left K-comodule algebras A to left K σ -comodule algebras Σ(A) ≡ σ A, where the new product is given by a σ • a ′ = σ(a (−1) ⊗ a ′ (−1) )a (0) a ′ (0) (the comodule structure maps of A and σ A being the same).The functor Σ is the identity on morphisms.As in Theorem 6.1, composition of this functor with the sheaf F of (K, H)-bicomodule algebras gives the sheaf Σ • F of (K σ , H)-bicomodule algebras.Theorem 6.3.Let the sheaf F of (K, H)-bicomodule algebras over the ringed space (M, F coH ) be an H-principal bundle.If the H-comodule (H, ∆) has a compatible K-comodule structure, so that it is a (K, H)-bicomodule and the cleaving maps j i : H → F (U i ), relative to a covering {U i } of M, are (K, H)bicomodule maps, then the sheaf Σ • F of (K σ , H)-bicomodule algebras over the ringed space (M, Σ • F coH ) is an H-principal bundle.
Proof.Since the sheaf F coH of K-comodule algebras is a subsheaf of the sheaf F of K-comodule algebras the sheaf Σ•F coH of K σ -comodule algebras is well defined.Since the Σ functor is the identity on objects Σ • F coH = (Σ • F ) coH as K σ -comodule algebras.
We are left to show that the sheaf Σ • F is locally cleft.From Theorem 2.8, for each open U i we have the local trivialization that is an isomorphism of left F (U i ) coH -modules and right H-comodules.Since j i is also a left K-module map and F (U i ) is a K-comodule algebra we easily have that ϑ i is also a left K-comodule map.
Applying the functor Σ to the F (U i ) coH -module and (K, H)-bicomodule isomorphism ϑ i we obtain the isomorphism of left σ F (U i ) coH -modules and (K σ , H)-bicomodules . Using the (K σ , H)-bicomodule isomorphism (we suppress the pedices of ϕ ℓ for simplicity) where σ H := Σ(H) is just the (K, H)-bicomodule H now seen as a (K σ , H)bicomodule, we obtain the left σ F (U i ) coH -module and (K σ , H)-bicomodule isomorphism Forgetting the K σ -comodule structure and recalling that as H-comodules σ H = H, and that as tensor products of H-comodules we have σ ⊗ = ⊗, this isomorphism becomes an σ F (U i ) coH -module and H-comodule

Examples
We twist the quantum principal bundle F on the quantum ringed space (SL n (C)/P, F co Oq(P ) ) of Theorem 5.9 and obtain three new quantum principal bundles: Γ • F , Σ • F and Γ • Σ • F ; the first on the locally ringed space associated with the homogeneous ring of quantum projective space Õq (P n−1 ), the other two on its multiparametric deformation Õq,γ (P n−1 ).
Deformations from twists of H = O q (P ).The (n − 1)-dimensional torus T n−1 is a subgroup of SL n (C) and correspondingly we have that the Hopf algebra O(T n−1 ) (the group Hopf algebra over C of the free abelian group generated by n − 1 elements) is a quotient of O q (SL n ).It is useful to present O(T n−1 ) as the algebra over C generated by the n elements t i , i = 1, . . .n and their inverses t −1 i modulo the ideal generated by the relation t 1 t 2 . . .t n = 1.The Hopf algebra structure is fixed by requiring t i to be group like.The Hopf algebra projection O q (SL n ) pr −→ O(T n−1 ) on the generators is given by We consider the exponential 2-cocycle γ on O(T n−1 ) defined on the generators t i by γ (t j ⊗ t k ) = γ jk with γ jk = exp iπθ jk ; θ jk = −θ kj ∈ R is defined by γ(a ij ⊗ a kl ) = δ ij δ kl γ il , and (32) for all a, b, c ∈ O q (SL n ).Twist deformation via this 2-cocycle of the quantum group O q (SL n ) gives the multiparametric special linear quantum group studied e.g. in [32].The torus Hopf algebra O(T n−1 ) is also a quotient of the parabolic quantum group O q (P ) defined in (18).Correspondingly the 2-cocycle γ on O(T n−1 ). is pulled back to a 2-cocycle, still denoted γ, on O q (P ) providing its multiparametric deformation O q,γ (P ).
We now apply Theorem 6.1 to the O q (P )-principal bundle F on the quantum ringed space (SL n (C)/P, F co Oq(P ) ) of Theorem 5.9 and obtain the O q,γ (P )-principal bundle Γ • F on (SL n (C)/P, F co Oq(P ) ).Furthermore, Remark 6.2 implies that while the O q (P )-principal bundle F is locally trivial on We now show that this O q,γ (P )-principal bundle Γ • Σ • F is an example of the construction of Theorem 4.8.This is so because the (graded) algebras O q (SL n ), O q (P ), O q (SL n /P ) and their localizations are left and right (graded) O(T n−1 )-comodule algebras.
We first observe that the total space (global sections) of Γ • Σ • F is the multiparametric quantum group with O q,γ (P ) that is a quantum subgroup.Indeed we can pull back the twist (31) on K = O(T n−1 ) to the twist (33) on O q (SL n ).Then (Γ•Σ)(O q (SL n )) is the twist of O q (SL n ) as a left O q (SL n )-comodule algebra and with the same twist (33) as a right O q (SL n )-comodule algebra, hence it is the twisting of O q (SL n ) as a Hopf algebra, giving the Hopf algebra O q,γ (SL n ).Similarly we have (Γ • Σ)(O q (P )) = O q,γ (P ) .
In order to show that O q,γ (P ) is a quantum subgroup of O q,γ (SL n ) recall that the deformation ( 34) is induced from a left and right action of the Hopf algebra O(T n−1 ) and notice that the ideal I q (P ) = (a α1 ) ⊂ O q (SL n ) is a left and right O(T n−1 )-subcomodule algebra.Its twist deformation I q,γ (P ) := (Σ • Γ)(I q (P )) is an ideal in O q,γ (SL n ).It is furthermore a Hopf ideal since so was I q (P ) in O q (SL n ), and because twisting does not affect the costructures and twisting via the exponential 2-cocycle (31) does not affect the antipode as a linear map.We can then consider the quotient Hopf algebra O q,γ (SL n )/I q,γ (P ), this is easily seen to be the multiparametric quantum group in (35).We next twist Õq (P n−1 ) = Õq (SL n /P ) seen as left K = O(T n−1 )-comodule algebra (and a trivial right O(T n−1 )-comodule algebra).The twist is grade preserving and therefore Õq,γ (P n−1 ) := (Σ • Γ)( Õq (P n−1 )) is a graded algebra.It is generated by the quantum section d = a 11 ∈ Õq,γ (SL n ) and the corresponding d i = a i1 obtained from the coproduct (that equals that of O q (SL n )).Indeed monomials in d i , respectively contructed with the product of Õq (P n−1 ) and of Õq,γ (P n−1 ), differ by a phase and hence span the same Cmodule Õq,γ (P n−1 ).Explicitly Õq,γ (P n−1 ) is the subalgebra generated by the elements x i := d i = a i1 ∈ O q,γ (SL n ), i.e. it is the multiparametric quantum homogeneous projective space Õq,γ (P n−1 ) = C q x 1 , . . .x n /(x i x j − q −1 γ 2 ij x j x i , i < j) .
We now observe that O q (SL n )S −1 i is canonically an O(T n−1 )-bicomodule algebra.We twist it to (Σ • Γ)(O q (SL n )S −1 i ) and denote by γ • γ the corresponding product (notice that γ • γ restricted to the sub O(T n−1 )-bicomodule O q (SL n ) is the Hopf algebra twist of the product of O q (SL n )).Due to γ(t −1 i ⊗ t i ) = 1 = γ(t i ⊗ t −1 i ) (cf. ( 31) and ( 32)), we have i .This shows that the inverse d −1 i of d i in O q (SL n ) is also the inverse in O q,γ (SL n ).
g. [12, §1.1]), Proposition 2.15. 1.Let B be a basis for a topology T on M. Then a B-sheaf of H-comodule algebras F (that is a sheaf defined for the open sets in B with gluing conditions) extends to a unique sheaf of H-comodules on M. 2. If {U i } is an open cover of M, then the empty set and finite intersections U i 1 ∩ • • • ∩ U ir form a basis for a topology on M. Remark 2.16.In Example 2.14, with {U i } open cover of P 1 (C), the B-sheaf is the restriction of F to B = {∅, U 12 , U 1 , U 2 }, and F (P 1 (C)) is recovered as the pull-back

( 4 )
with π : O(G) −→ O(P ) the standard projection, S the antipode of O(P ) .Lifting S(χ) ∈ O(P ) to an element t ∈ O(G) we have the following proposition.Proposition 3.2.Let P be a parabolic subgroup of a semisimple algebraic group G and denote by π : O(G) −→ O(P ) the natural projection dual to the inclusion P ⊂ G.If G/P is embedded into some projective space via some very ample line bundle L then there exists an element t ∈ O(G) such that

4 .
Recalling the commutation relations of the p ij 's (cf.proof of Proposition 5.4), and those between quantum minors in[16], we have: i) the a kj among themselves have the same commutation relations as the p 1j 's.ii) a k1 commutes with D 1β αk , D 1β 1k , D 1β kα .iii) The −qa −1 k1 D 1β αk 's, satisfy the same Manin relations among themselves as the p αβ 's; similarly for the a −1 k1 D 1β 1k 's and the a −1 k1 D 1β kα 's.iv) The mixed commutation relations: of −qa −1 k1 D 1β αk with a −1 k1 D 1β 1k and with a −1 k1 D 1β kα , and of a −1 k1 D 1β 1k with a −1 k1 D 1β kα , also satisfy the same Manin relations as those of the corresponding p αβ 's.
Ch. 3).We now define the key notion of quantum principal bundle by extending to the quantum case what we established in Proposition 2.2.Definition 2.11.Let (M, O M ) be a ringed space and H a Hopf algebra.We say that a sheaf of H-comodule algebras F is an H-principal bundle or quantum principal bundle over (M, O M ) if there exists an open covering {U i } of M such that: