From hydrolytically labile to hydrolytically stable Ru(II)-arene anticancer complexes with carbohydrate-derived co-ligands.

C. R. Acad. Sci. Paris, Ser. I 348 (2010) 487–490 Contents lists available at ScienceDirect C. R. Acad. Sci. Paris, Ser. I www.sciencedirect.com Algebra/Ordinary Differential Equations Differential ‘Galois’ extensions with new constants Extensions différentielles « galoisiennes » avec nouvelles constantes Lourdes Juan a , Andy R. Magid b a b Department of Mathematics, Texas Tech University, Lubbock, TX 79409, United States Department of Mathematics, University of Oklahoma, Norman, OK 73019, United States a r t i c l e i n f o a b s t r a c t Article history: Received 12 February 2009 Accepted after revision 6 April 2010 Presented by Bernard Malgrange Let F be a differential ﬁeld with algebraically closed ﬁeld of constants C and let E be a differential ﬁeld extension of F . The ﬁeld E is a differential Galois extension if it is generated over F by a full set of solutions of a linear homogeneous differential equation with coeﬃcients in F and if its ﬁeld of constants coincides with C . We study the differential ﬁeld extensions of F that satisfy the ﬁrst condition but not the second. © 2010 Académie des sciences. Published by Elsevier Masson SAS. All rights reserved. r é s u m é Soit F un corps différentiel dont le corps des constantes C est algébriquement clos et soit E ⊃ F une extension de corps différentiels. Le corps différentiel E est une extension galoisienne différentielle de F s’il est engendré sur F par une base de solutions d’une équation différentielle linéaire homogène à coeﬃcients dans F et si son corps des constantes est C . Nous étudions les extensions différentielles de F qui satisfont la première condition et non la seconde. © 2010 Académie des sciences. Published by Elsevier Masson SAS. All rights reserved. 1. Introduction Let F be a differential ﬁeld of characteristic zero. Let C denote the ﬁeld of constants of F (assumed algebraically closed). Let L = Y (n) + a1 Y (n−1) + · · · + a0 Y be a linear differential operator over F . A Picard–Vessiot, or differential Galois, extension E of F for L is a differential ﬁeld extension, also with ﬁeld of constants C (i.e. a no new constant extension), generated over F as a differential ﬁeld by n solutions of L = 0 linearly independent over C (i.e. a full set of solutions). The differential Galois group G ( E / F ) is the group of differential ﬁeld automorphisms of E ﬁxing F , and the resulting Galois correspondence includes the fact that the ﬁxed ﬁeld of E under G ( E / F ) is F . The same extension E is a Picard–Vessiot extension of F for many different operators L. One way to suppress the explicit reference to a speciﬁc operator is to note that a differential ﬁeld extension E ⊃ F is a Picard–Vessiot extension for some operator L provided that (1) There is a group G of differential automorphisms of E over F whose ﬁxed ﬁeld is F . (2) E is generated over F as a differential ﬁeld by a G-submodule V which is ﬁnite dimensional over C . (3) E is a no new constant extension of F . E-mail addresses: lourdes.juan@ttu.edu (L. Juan), amagid@math.ou.edu (A.R. Magid). 1631-073X/$ – see front matter doi:10.1016/j.crma.2010.04.004 © 2010 Académie des sciences. Published by Elsevier Masson SAS. All rights reserved. 488 L. Juan, A.R. Magid / C. R. Acad. Sci. Paris, Ser. I 348 (2010) 487–490 The operator L for which E is Picard–Vessiot is produced from V using Wronskian determinants, and then V has a full set of solutions of L as a C -basis (see [2]). There are natural situations, however, for example the case of certain derivations on coordinate rings of connected algebraic groups, which lead to extensions E ⊃ F which meet the ﬁrst two criteria but not the third. Such extensions are our subject here. We reformulate the situation in terms of rings generated by solutions. The ring F [ y i j : 1 ⩽ i ⩽ n, 0 ⩽ j ⩽ n − 1][ w −1 ], where w = det( y i j ), with derivation y i j = y i , j +1 , y i ,n−1 = −an−1 y i ,n−1 − · · · − a0 y i0 contains and is differentially generated by n solutions y i0 to the differential equation L = 0 whose Wronskian is a unit, and the same remains true for its homomorphic images. (This is why it is called a full universal solution algebra for L in [2].) If we take a maximal differential ideal of the full universal solution algebra and mod it out, the quotient ﬁeld of the resulting integral domain is a Picard–Vessiot extension of F for L: it has the additional property that its constants are C as well. It doesn’t matter which maximal differential ideal we choose: all are conjugate under the (differential) action of GLn (C ) extended from the linear action on the y i j [2]. And the stabilizer in GLn (C ) of the chosen maximal ideal, which acts naturally mod it and on the quotient ﬁeld, is the group of all differential automorphisms and the differential Galois group over F . We focus on what happens when a not necessarily maximal prime differential ideal P of the full universal solution algebra is selected, modded out, and we consider the resulting quotient ﬁeld extension E of F . When the prime ideal is not maximal, the constant ﬁeld K of E is a proper extension of F , and of course need not be algebraically closed. Our main result shows that nonetheless E is much like a Picard–Vessiot extension of the compositum ﬁeld F K . In particular, we ﬁnd an algebraic subgroup G of GLn ( K ) with E G = F K . Unless (0) is a maximal differential ideal of the full universal solution algebra, the latter will always have non-maximal prime differential ideals to which our results apply. A typical situation to which our results apply is the following: Let H be a connected algebraic group over C , let D 1 , . . . , D m be a basis of Lie( H ), let bi ∈ F and consider the derivation D = D F ⊗ 1 + b i ⊗ D i on F [ H ] = F ⊗ C [ H ]. We show that there is an operator L as above which has a full set of solutions in F ( H ) (the quotient ﬁeld of F [ H ]) with invertible Wronskian, so that the algebra they differentially generate over F ﬁts into our theory above. The appropriate consequences are drawn. We will preserve throughout the notational conventions of this introduction. Another useful reference on the Picard– Vessiot theory is [3]. 2. Modulo non-maximal primes Let P be a prime differential ideal of the full universal solution algebra F [ y i j : 1 ⩽ i ⩽ n, 0 ⩽ j ⩽ n − 1][ w −1 ] for L = Y (n) + a1 Y (n−1) + · · · + a0 Y , let R F denote F [ y i j ][ w −1 ], and let E denote the quotient ﬁeld of R F / P . Let K denote the ﬁeld of constants of E. We choose an algebraic closure E of E. Then K , the algebraic closure of K in E, is also algebraically closed and is the ﬁeld of constants of the composita E K and F K . To see that K is indeed the ﬁeld of constants, we use the fact that E K ⊇ E is an algebraic extension, so that any constant must be algebraic over the constants K of E. Lemma 1. E K is a Picard–Vessiot extension of F K for L. Proof. E K is generated over F K by a full set of solutions, with non-zero Wronskian of L = 0, so the only issue is new constants. As we just noted, the constants of E K are K which is also the constants of F K . 2 Let E 0 be a differential subﬁeld of E containing F and K , for example F K . Then Lemma 1 also implies that E K is a Picard–Vessiot extension of E 0 K . Let zi j be the image of y i j in E K . Let R E 0 K = E 0 K ⊗ F R F and consider the homomorphism R E 0 K → E K induced from y i j → zi j . Let M be its kernel. Note that R E 0 K is also a full universal solution algebra for L over E 0 K . If M were not maximal then as above the quotient ﬁeld E K of R E 0 K / M would contain a new constant, a contradiction. So M is a maximal differential ideal. Let R E 0 = E 0 ⊗ F R F and let M 0 be the kernel of the homomorphism R E 0 → E induced from y i j → zi j . We have R E 0 K ⊇ R E 0 and M ∩ R E 0 = M 0 . We note that K ∩ E = K , since anything algebraic over constants is a constant. It follows that K ⊗ K E is an integral domain, and hence so is K ⊗ K E 0 . Since these are integral domains algebraic over ﬁelds, they are themselves ﬁelds. Thus E K = K ⊗ K E and E 0 K = K ⊗ K E 0 . Lemma 2. E 0 K ∩ E = E 0 . Proof. From the above equalities, the assertion is the obvious one that ( K ⊗ K E 0 ) ∩ ( K ⊗ K E ) = K ⊗ K E 0 in K ⊗ K E. These observations allow us to conclude that the ideal M is induced (we use the above notation): 2 L. Juan, A.R. Magid / C. R. Acad. Sci. Paris, Ser. I 348 (2010) 487–490 489 Lemma 3. K ⊗ K R E 0 / M 0 is isomorphic to R E 0 K / M as K -algebra. In particular: (1) M 0 is a maximal differential ideal. (2) M = R E 0 K M 0 as ideals. (3) M = K M 0 as K -vector spaces. Proof. By the discussion preceding the lemma, the subring S = K ⊗ K R E 0 / M 0 of K ⊗ K E can be regarded as a subring of E K . Viewed in that way, it is the E 0 K algebra generated by zi j and the inverse Wronskian, and that algebra is R E 0 K / M. This implies the main assertion of the lemma and the others are direct consequences. 2 The rings R E 0 and R E 0 K are generated by the vector spaces V K = i , j K y i j and V K = i , j K y i j (and the inverse Wron- skians) over their coeﬃcient ﬁelds. Differential actions of the groups GLn ( K ) and GLn ( K ) on R F K and R F K come from their actions on these vector spaces [2]. The inclusion GLn ( K ) ⊆ GLn ( K ) is compatible with the inclusion V K ⊆ V K . So we can regard GLn ( K ) as acting on R E 0 K , and the restriction of that action to the subring R E 0 is the given action. The stabilizer GLn ( K ) M of the ideal M is an algebraic subgroup of GLn ( K ); in fact, it is the differential Galois group G ( E K / E 0 K ) [2]. Lemma 4. The stabilizer GLn ( K ) M 0 of M 0 is Zariski dense in GLn ( K ) M . Proof. We have by Lemma 3 that M = K M 0 . If M 0 were a ﬁnite dimensional K -vector space, the result would be obvious. Since the actions here are rational, we can reduce immediately to the ﬁnite dimensional case and conclude the same result. 2 We are now ready for the main result. Theorem 1. Let G ( E / E 0 ) = GLn ( K ) M 0 . Then E G ( E / E 0 ) = E 0 . Proof. Because G ( E / E 0 ) is Zariski dense in G ( E K / E 0 K ) by Lemma 4, and E K ⊃ E 0 K is Picard–Vessiot by Lemma 1 we have ( E K )G ( E / E 0 ) = E 0 K . It follows that E G ( E / E 0 ) ⊆ E 0 K . So it suﬃces to show that E ∩ E 0 K = E 0 . This is just Lemma 2. 2 3. Derivations of group coordinate rings Warning: the symbol E is used in this section differently than above. Let H be a connected linear algebraic group over C , let D 1 , . . . , D m be a basis of Lie( H ), let bi ∈ F and consider the derivation D = D F ⊗ 1 + 1 ⊗ D i on F [ H ] = F ⊗ C [ H ]. We regard H as acting on F [ H ] by left translations (h · f ( g ) = f ( gh)) so that D is H -equivariant. Both D and the H -action extend to the quotient ﬁeld F ( H ) and commute with each other. Let W be some ﬁnite dimensional C -subspace which is H -stable and generates C [ H ] as a C -algebra. Then the following properties hold for the differential ﬁeld extension E = F ( H ) ⊃ F : (1) The group H is a group of differential automorphisms of E over F such that E H = F . (2) There is a ﬁnite dimensional H -stable, C -vector space W ⊂ E such that E = F W is differentially generated over F by W . As remarked in the introduction, if additionally the constants of E were those of F , E would be a Picard–Vessiot extension of F . We do not make that assumption here; hence any ﬁeld extension meeting the above two criteria is called a pre-Picard–Vessiot (brieﬂy pPV) extension of F . For example, if X is an irreducible H -torsor and D a derivation on F ( X ) induced by an element of the corresponding twisted Lie algebra (see [1] for details) then F ( X ) is a pPV extension of F . Let K be the ﬁeld of constants of an arbitrary pPV extension E. Let V = K W , where W is as deﬁned above except that we can replace C [ H ] by the coordinate ring C [ X ] of an irreducible H -torsor, and let z1 , . . . , zm be a K -basis of W . Note that the Wronskian w ( z1 , . . . , zm ) is non-zero by construction. H acts on K W , although not K -linearly in general. Nonetheless, for h ∈ H we have a matrix α (h) ∈ GLm ( K ) such that h h = ( z1 , . . . , zm )α (h). z1 , . . . , zm We can differentiate both sides of this equation and obtain h z1 h = z1 , . . . , zm , . . . , zm α (h) using the fact that K is the ﬁeld of constants. Repeated differentiation shows that a similar formula holds for higher derivatives as well. 490 L. Juan, A.R. Magid / C. R. Acad. Sci. Paris, Ser. I 348 (2010) 487–490 Now suppose Y is a differential indeterminate over F , and consider the Wronskian determinant w (Y , z1 , . . . , zn ) in E {Y }. When we expand this determinant along the ﬁrst column, the various minors that occur have rows of the form (i ) (i ) (z1 , . . . , zm ). These rows transform under h ∈ H as above via multiplication by α (h), and hence the minors transform via multiplication by det(α (h)). This applies to the coeﬃcient w ( z1 , . . . , zm ) of Y (m) as well. It then follows that the coeﬃcients of L = w ( z1 , . . . , zm )−1 w (Y , z1 , . . . , zn ) = Y (m) + a1 Y (m−1) + · · · + am Y (0) are invariant under any h ∈ H and hence lie in F . (This is an adaptation of the argument in [2].) We thus have a homomorphism from the full universal solution algebra R F = F [ y i j ][ w −1 ] over F for L to E by y i0 → zi whose kernel is a prime ideal P . Let E 1 denote the quotient ﬁeld of its image. By construction, F is a subﬁeld of E 1 , and G(E /F K1) E = E 1 K . The constants of E 1 are K 1 = K ∩ E 1 . By Theorem 1, we have F K 1 = E 1 1 , where G ( E 1 / F K 1 ) is a subgroup G(E1/F K1) of GLm ( K 1 ). Thus the extension E ⊃ F breaks into the subextensions E = E 1 K ⊃ E 1 ⊃ E 1 extensions on the ends are by constants and that in the middle is by a group. = F K 1 ⊃ F , where the References [1] L. Juan, A. Ledet, Equivariant vector ﬁelds on non-trivial SOn -torsors and differential Galois theory, Journal of Algebra 316 (2007) 735–745. [2] A. Magid, Lectures on Differential Galois Theory, University Lecture Series, vol. 7, American Mathematical Society, Providence RI, 1997 (second printing with corrections). [3] M. van der Put, M. Singer, Differential Galois Theory, Springer-Verlag, New York, 2003.