Some applications of model theory to differential Galois theory
Transcription
Some applications of model theory to differential Galois theory
Some applications of model theory to differential Galois theory Javier Moreno Institut Camille Jordan Université Claude Bernard Lyon 1 Model Theory in Kirishima March 4th, 2010 Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 1 / 25 Differential Galois Theory The goal of Differential Galois Theory is to study differential equations by algebraic means. We would like to have methods, mimicking the ones from classical Galois theory, to evaluate if the integral of a given function (or the solution of a differential equation) can be expressed in terms of “elementary functions”. Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 2 / 25 Differential Algebra Definition Let R be an arbitrary (commutative) ring. A derivation on R is an additive function ∂ : R → R such that ∂(xy ) = x∂(y ) + y ∂(x) for any x, y ∈ R. A ring (field) R equipped with a derivation ∂ is what we call a differential ring (field). By the constants of R, denoted CR , we mean the set of elements in R where ∂ vanish. Remark Notions of ideals (prime, maximal), localizations, &c. can be defined and developed. Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 3 / 25 Differential Algebra Definition Let R be an arbitrary (commutative) ring. A derivation on R is an additive function ∂ : R → R such that ∂(xy ) = x∂(y ) + y ∂(x) for any x, y ∈ R. A ring (field) R equipped with a derivation ∂ is what we call a differential ring (field). By the constants of R, denoted CR , we mean the set of elements in R where ∂ vanish. Remark Notions of ideals (prime, maximal), localizations, &c. can be defined and developed. Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 3 / 25 Classic Galois Theory (Ch. 0) Let us recall the way (algebraic) Galois theory works. Take, for convenience, a polynomial P(X ) ∈ F [X ] of degree n with no repeated roots. The idea is to associate to this polynomial a certain (finite) group and extract properties about the solution of the polynomial from the structural properties of the group. Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 4 / 25 Classic Galois Theory (Ch. 0) Let us recall the way (algebraic) Galois theory works. Take, for convenience, a polynomial P(X ) ∈ F [X ] of degree n with no repeated roots. The idea is to associate to this polynomial a certain (finite) group and extract properties about the solution of the polynomial from the structural properties of the group. Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 4 / 25 Classic Galois Theory (Ch. 0) Let 1 ], i6=j (Xi − Xj ) S = F [X1 , . . . , Xn , Q and let I be the ideal generated by the P(Xi ) in S. Let M > I maximal in S. Define the splitting field of P over F as K = S/M. And let G = Aut(K /F ). We call this the Galois group of P. Q Since S contains i6=j (Xi − Xj ) then the images of Xi in K are distinct roots of P. K is unique up to isomorphism and we have Galois correspondence. We can prove, for instance, that a P is solvable by radicals if G is a solvable group. Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 5 / 25 Classic Galois Theory (Ch. 0) Let 1 ], i6=j (Xi − Xj ) S = F [X1 , . . . , Xn , Q and let I be the ideal generated by the P(Xi ) in S. Let M > I maximal in S. Define the splitting field of P over F as K = S/M. And let G = Aut(K /F ). We call this the Galois group of P. Q Since S contains i6=j (Xi − Xj ) then the images of Xi in K are distinct roots of P. K is unique up to isomorphism and we have Galois correspondence. We can prove, for instance, that a P is solvable by radicals if G is a solvable group. Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 5 / 25 Classic Galois Theory (Ch. 0) Let 1 ], i6=j (Xi − Xj ) S = F [X1 , . . . , Xn , Q and let I be the ideal generated by the P(Xi ) in S. Let M > I maximal in S. Define the splitting field of P over F as K = S/M. And let G = Aut(K /F ). We call this the Galois group of P. Q Since S contains i6=j (Xi − Xj ) then the images of Xi in K are distinct roots of P. K is unique up to isomorphism and we have Galois correspondence. We can prove, for instance, that a P is solvable by radicals if G is a solvable group. Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 5 / 25 Classic Galois Theory (Ch. 0) Let 1 ], i6=j (Xi − Xj ) S = F [X1 , . . . , Xn , Q and let I be the ideal generated by the P(Xi ) in S. Let M > I maximal in S. Define the splitting field of P over F as K = S/M. And let G = Aut(K /F ). We call this the Galois group of P. Q Since S contains i6=j (Xi − Xj ) then the images of Xi in K are distinct roots of P. K is unique up to isomorphism and we have Galois correspondence. We can prove, for instance, that a P is solvable by radicals if G is a solvable group. Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 5 / 25 Classic Galois Theory (Ch. 0) Let 1 ], i6=j (Xi − Xj ) S = F [X1 , . . . , Xn , Q and let I be the ideal generated by the P(Xi ) in S. Let M > I maximal in S. Define the splitting field of P over F as K = S/M. And let G = Aut(K /F ). We call this the Galois group of P. Q Since S contains i6=j (Xi − Xj ) then the images of Xi in K are distinct roots of P. K is unique up to isomorphism and we have Galois correspondence. We can prove, for instance, that a P is solvable by radicals if G is a solvable group. Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 5 / 25 Linear differential Galois theory (Picard-Vessiot) Let y 0 = Ay be a linear differential equation of order n (= dim(A)) with coefficients in F , a differential field with CF algebraically closed. Easy Fact The set of solutions of this equation in L > F an extension of F is a CF -vector space of dimension at most n. We would like to obtain a corresponding splitting field for this equation (i.e. with solution space of full (n) dimension). Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 6 / 25 Linear differential Galois theory (Picard-Vessiot) Let y 0 = Ay be a linear differential equation of order n (= dim(A)) with coefficients in F , a differential field with CF algebraically closed. Easy Fact The set of solutions of this equation in L > F an extension of F is a CF -vector space of dimension at most n. We would like to obtain a corresponding splitting field for this equation (i.e. with solution space of full (n) dimension). Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 6 / 25 Linear differential Galois theory (Picard-Vessiot) Let y 0 = Ay be a linear differential equation of order n (= dim(A)) with coefficients in F , a differential field with CF algebraically closed. Easy Fact The set of solutions of this equation in L > F an extension of F is a CF -vector space of dimension at most n. We would like to obtain a corresponding splitting field for this equation (i.e. with solution space of full (n) dimension). Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 6 / 25 Linear differential Galois theory (Picard-Vessiot) Let S = F [Y1,1 , . . . , Yn,n , 1 ] det(Y ) and define a derivation on S setting Y 0 = AY . Let M be a maximal differential ideal of S. Define the Picard-Vessiot ring of the given equation with respect to F as R = S/M. In this case R is not necessarily a field, but it can be proven that it is an integral domain. Let K = Q(R) and let G = Aut∂ (K /F ). This will be the Galois group for our equation. Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 7 / 25 Linear differential Galois theory (Picard-Vessiot) Let S = F [Y1,1 , . . . , Yn,n , 1 ] det(Y ) and define a derivation on S setting Y 0 = AY . Let M be a maximal differential ideal of S. Define the Picard-Vessiot ring of the given equation with respect to F as R = S/M. In this case R is not necessarily a field, but it can be proven that it is an integral domain. Let K = Q(R) and let G = Aut∂ (K /F ). This will be the Galois group for our equation. Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 7 / 25 Linear differential Galois theory (Picard-Vessiot) Some properties of P-V extensions If K /F is a P-V extension for y 0 = Ay , then: 1 K is unique up to isomorphism. 2 K = F (Z ) for some Z ∈ GLn (K ) such that Z 0 = AZ , and 3 CK = CF . Moreover! G < GLn (CF ) is a linear algebraic group (i.e. a zero set in GLn (CF ) of a system of polynomials over CF ). Properties (2) and (3) above define the P-V extension for the given differential equation. Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 8 / 25 Linear differential Galois theory (Picard-Vessiot) Some properties of P-V extensions If K /F is a P-V extension for y 0 = Ay , then: 1 K is unique up to isomorphism. 2 K = F (Z ) for some Z ∈ GLn (K ) such that Z 0 = AZ , and 3 CK = CF . Moreover! G < GLn (CF ) is a linear algebraic group (i.e. a zero set in GLn (CF ) of a system of polynomials over CF ). Properties (2) and (3) above define the P-V extension for the given differential equation. Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 8 / 25 Linear differential Galois theory (Picard-Vessiot) Some properties of P-V extensions If K /F is a P-V extension for y 0 = Ay , then: 1 K is unique up to isomorphism. 2 K = F (Z ) for some Z ∈ GLn (K ) such that Z 0 = AZ , and 3 CK = CF . Moreover! G < GLn (CF ) is a linear algebraic group (i.e. a zero set in GLn (CF ) of a system of polynomials over CF ). Properties (2) and (3) above define the P-V extension for the given differential equation. Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 8 / 25 Linear differential Galois theory (Picard-Vessiot) Galois correspondence There is a correspondence between Zariski-closed subgroups H ⊂ G and intermediate differential subfields between F and K . Solutions of differential equations (One example) If G is a solvable group then the corresponding equation can be solved by a finite sequence of field extensions F = F0 < F1 · · · < Fl = K such that Fi+1 = Fi (ti ) where either 1 ti0 ∈ Fi (integral), or 2 ti0 /ti ∈ Ki (exponential). Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 9 / 25 Linear differential Galois theory (Picard-Vessiot) Galois correspondence There is a correspondence between Zariski-closed subgroups H ⊂ G and intermediate differential subfields between F and K . Solutions of differential equations (One example) If G is a solvable group then the corresponding equation can be solved by a finite sequence of field extensions F = F0 < F1 · · · < Fl = K such that Fi+1 = Fi (ti ) where either 1 ti0 ∈ Fi (integral), or 2 ti0 /ti ∈ Ki (exponential). Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 9 / 25 Model Theory, anywhere? (Robinson, 1950’s!) There is a model companion of the theory of differential fields (ch. 0): Differentially Closed Fields (a.k.a. DCF0 .) DCF0 is stable (it is actually totally transcendental (Blum)). It has QE and elimination of imaginaries. Also, dcl(A) = hAi. By stability, the field of constants is a pure algebraically closed field. Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 10 / 25 Model Theory, anywhere? (Robinson, 1950’s!) There is a model companion of the theory of differential fields (ch. 0): Differentially Closed Fields (a.k.a. DCF0 .) DCF0 is stable (it is actually totally transcendental (Blum)). It has QE and elimination of imaginaries. Also, dcl(A) = hAi. By stability, the field of constants is a pure algebraically closed field. Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 10 / 25 Model Theory, anywhere? (Robinson, 1950’s!) There is a model companion of the theory of differential fields (ch. 0): Differentially Closed Fields (a.k.a. DCF0 .) DCF0 is stable (it is actually totally transcendental (Blum)). It has QE and elimination of imaginaries. Also, dcl(A) = hAi. By stability, the field of constants is a pure algebraically closed field. Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 10 / 25 Non-linear differential Galois theory (Kolchin) Let U be a big saturated model of DCF0 . Strongly normal extensions An extension K/F of differential fields is strongly normal if: 1 CF = CK algebraically closed. 2 Something like: K = F (Z ) for some Z ∈ GLn (K ) such that Z 0 = AZ . Non-linear Galois theory! Let K /F an strongly normal extension of differential fields, then: Gal(K /F ) = Aut∂ (K /F ) is isomorphic to the CF rational points of an algebraic group over CF . Galois correspondence, &c. Q: Equations? A: Logarithmic differential equations!. Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 11 / 25 Non-linear differential Galois theory (Kolchin) Let U be a big saturated model of DCF0 . Strongly normal extensions An extension K/F of differential fields is strongly normal if: 1 2 CF = CK algebraically closed. K = F (a0 , . . . , am ) and, for any embedding σ : K → U over F , we have σ(K ) ⊆ K hCU i. Non-linear Galois theory! Let K /F an strongly normal extension of differential fields, then: Gal(K /F ) = Aut∂ (K /F ) is isomorphic to the CF rational points of an algebraic group over CF . Galois correspondence, &c. Q: Equations? A: Logarithmic differential equations!. Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 11 / 25 Non-linear differential Galois theory (Kolchin) Let U be a big saturated model of DCF0 . Strongly normal extensions An extension K/F of differential fields is strongly normal if: 1 2 CF = CK algebraically closed. K = F (a0 , . . . , am ) and, for any embedding σ : K → U over F , we have σ(K ) ⊆ K hCU i. Non-linear Galois theory! Let K /F an strongly normal extension of differential fields, then: Gal(K /F ) = Aut∂ (K /F ) is isomorphic to the CF rational points of an algebraic group over CF . Galois correspondence, &c. Q: Equations? A: Logarithmic differential equations!. Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 11 / 25 Non-linear differential Galois theory (Kolchin) Let U be a big saturated model of DCF0 . Strongly normal extensions An extension K/F of differential fields is strongly normal if: 1 2 CF = CK algebraically closed. K = F (a0 , . . . , am ) and, for any embedding σ : K → U over F , we have σ(K ) ⊆ K hCU i. Non-linear Galois theory! Let K /F an strongly normal extension of differential fields, then: Gal(K /F ) = Aut∂ (K /F ) is isomorphic to the CF rational points of an algebraic group over CF . Galois correspondence, &c. Q: Equations? A: Logarithmic differential equations!. Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 11 / 25 Non-linear differential Galois theory (Kolchin) Let U be a big saturated model of DCF0 . Strongly normal extensions An extension K/F of differential fields is strongly normal if: 1 2 CF = CK algebraically closed. K = F (a0 , . . . , am ) and, for any embedding σ : K → U over F , we have σ(K ) ⊆ K hCU i. Non-linear Galois theory! Let K /F an strongly normal extension of differential fields, then: Gal(K /F ) = Aut∂ (K /F ) is isomorphic to the CF rational points of an algebraic group over CF . Galois correspondence, &c. Q: Equations? A: Logarithmic differential equations!. Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 11 / 25 Internality and definable automorphism groups In the early 80’s, Hrushovski and Zilber isolated conditions to guarantee that certain automorphism groups were definable (this, in essence, makes them visible to model theory). Internality Consider p and q possibly partial types over A. We say that p is q-internal (over A) if there is a set B containing A such that, for every realization a of p, there is a tuple b of realizations of q such that a ∈ dcl(Bb). Fundamental system of solutions Given p q-internal, we say that a tuple a of realizations of p is a fundamental system of solutions of p relative to q if there exists u(·, ·), an A-definable function, such that for any b realizing p, we have that b = u(a, c) for some tuple c of realisations of q. (Conditions?) Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 12 / 25 Internality and definable automorphism groups In the early 80’s, Hrushovski and Zilber isolated conditions to guarantee that certain automorphism groups were definable (this, in essence, makes them visible to model theory). Internality Consider p and q possibly partial types over A. We say that p is q-internal (over A) if there is a set B containing A such that, for every realization a of p, there is a tuple b of realizations of q such that a ∈ dcl(Bb). Fundamental system of solutions Given p q-internal, we say that a tuple a of realizations of p is a fundamental system of solutions of p relative to q if there exists u(·, ·), an A-definable function, such that for any b realizing p, we have that b = u(a, c) for some tuple c of realisations of q. (Conditions?) Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 12 / 25 Internality and definable automorphism groups In the early 80’s, Hrushovski and Zilber isolated conditions to guarantee that certain automorphism groups were definable (this, in essence, makes them visible to model theory). Internality Consider p and q possibly partial types over A. We say that p is q-internal (over A) if there is a set B containing A such that, for every realization a of p, there is a tuple b of realizations of q such that a ∈ dcl(Bb). Fundamental system of solutions Given p q-internal, we say that a tuple a of realizations of p is a fundamental system of solutions of p relative to q if there exists u(·, ·), an A-definable function, such that for any b realizing p, we have that b = u(a, c) for some tuple c of realisations of q. (Conditions?) Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 12 / 25 Internality and definable automorphism groups Binding group theorem Let T be a stable theory and U a big saturated model. Suppose that p and q are over A and p is q-internal. Suppose also that there is a fundamental system of solutions of p relative to q. Then the group of automorphisms of U that fixes q(U) and A pointwise induces a (type-)definable group of automorphisms on p(U) Actually... Stability is not really necessary. Something like this can be proven in a completely general setting. (Hrushovski 2001) Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 13 / 25 Internality and definable automorphism groups Binding group theorem Let T be a stable theory and U a big saturated model. Suppose that p and q are over A and p is q-internal. Suppose also that there is a fundamental system of solutions of p relative to q. Then the group of automorphisms of U that fixes q(U) and A pointwise induces a (type-)definable group of automorphisms on p(U) Actually... Stability is not really necessary. Something like this can be proven in a completely general setting. (Hrushovski 2001) Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 13 / 25 Poizat’s beautiful observation Kolchin’s differential Galois Theory could be remade as an application of internality in DCF0 ! Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 14 / 25 Poizat’s beautiful observation Strongly normal extensions An extension K/F of differential fields is strongly normal if: 1 2 CF = CK algebraically closed. K = F (a0 , . . . , am ) and, for any embedding σ : K → U over F , we have σ(K ) ⊆ K hCU i. Corollary (of sorts) Kolchin’s differential Galois theory In addition Motivation for elimination of imaginaries and key question regarding definable groups in algebraic closed fields!! Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 15 / 25 Poizat’s beautiful observation Strongly normal extensions An extension K/F of differential fields is strongly normal if: 1 2 CF = CK algebraically closed. K = F (a) and for any a 0 ≡F a, we have that a 0 ∈ dcl(FaCU ). (Thus, tp(a/F ) is CU -internal!) Corollary (of sorts) Kolchin’s differential Galois theory In addition Motivation for elimination of imaginaries and key question regarding definable groups in algebraic closed fields!! Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 15 / 25 Poizat’s beautiful observation Strongly normal extensions An extension K/F of differential fields is strongly normal if: 1 2 CF = CK algebraically closed. K = F (a) and for any a 0 ≡F a, we have that a 0 ∈ dcl(FaCU ). (Thus, tp(a/F ) is CU -internal!) Corollary (of sorts) Kolchin’s differential Galois theory In addition Motivation for elimination of imaginaries and key question regarding definable groups in algebraic closed fields!! Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 15 / 25 Poizat’s beautiful observation Strongly normal extensions An extension K/F of differential fields is strongly normal if: 1 2 CF = CK algebraically closed. K = F (a) and for any a 0 ≡F a, we have that a 0 ∈ dcl(FaCU ). (Thus, tp(a/F ) is CU -internal!) Corollary (of sorts) Kolchin’s differential Galois theory In addition Motivation for elimination of imaginaries and key question regarding definable groups in algebraic closed fields!! Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 15 / 25 Positive characteristic differential Galois theory? Immediate problem Direct translation does not work. Ordinary differential constants are too weak in characteristic p > 0: any extension creates new constants. Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 16 / 25 Possible solution: Iterative derivations Differential iterative rings (Hasse, Schmidt (1937!)) An ID-Ring is a ring R equipped with a sequence D = (∂i : R → R)i<ω of additive endomorphisms such that: P (Derivation) ∂0 = IdR and ∂m (ab) = k +l=m ∂k (a)∂l (b); (Iterativity) ∂m ∂n = m+n n ∂m+n ; Given (R, D), define CR , the ring (field) of constants, as the set of elements that are killed by the ∂i (i ≥ 1). Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 17 / 25 Possible solution: Iterative derivations Differential iterative rings (Hasse, Schmidt (1937!)) An ID-Ring is a ring R equipped with a sequence D = (∂i : R → R)i<ω of additive endomorphisms such that: P (Derivation) ∂0 = IdR and ∂m (ab) = k +l=m ∂k (a)∂l (b); (Iterativity) ∂m ∂n = m+n n ∂m+n ; Given (R, D), define CR , the ring (field) of constants, as the set of elements that are killed by the ∂i (i ≥ 1). Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 17 / 25 Model theory pays back its debts Iterative differential fields are well behaved The theory of iterative differential fields of characteristic p has a stable, non superstable model companion, SCHp , with elimination of imaginaries and quantifier elimination. Actually SCHp is just another presentation of the theory of separably closed fields of characteristic p and imperfection degree 1 (SCFp,1 ). (Remember Soha Sin’s talk on tuesday!) Precedent Pillay (2002), following Hrushovski, found a presentation of linear iterative differential Galois theory using the model theory of SCHp . Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 18 / 25 Model theory pays back its debts Iterative differential fields are well behaved The theory of iterative differential fields of characteristic p has a stable, non superstable model companion, SCHp , with elimination of imaginaries and quantifier elimination. Actually SCHp is just another presentation of the theory of separably closed fields of characteristic p and imperfection degree 1 (SCFp,1 ). (Remember Soha Sin’s talk on tuesday!) Precedent Pillay (2002), following Hrushovski, found a presentation of linear iterative differential Galois theory using the model theory of SCHp . Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 18 / 25 Model theory pays back its debts Iterative differential fields are well behaved The theory of iterative differential fields of characteristic p has a stable, non superstable model companion, SCHp , with elimination of imaginaries and quantifier elimination. Actually SCHp is just another presentation of the theory of separably closed fields of characteristic p and imperfection degree 1 (SCFp,1 ). (Remember Soha Sin’s talk on tuesday!) Precedent Pillay (2002), following Hrushovski, found a presentation of linear iterative differential Galois theory using the model theory of SCHp . Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 18 / 25 Iterative strongly normal extensions Let U be a large and highly saturated model of SCHp and let C = CU . Iterative Strongly Normal Extensions Let (F , D) < (K , D) an extension of definably closed ID-fields. We say that K /F is a iterative strongly normal extension if: 1 CF = CK and CF is algebraically closed. 2 K = dcl(Fa) (=Closing Fa under derivations and then under pth roots). 3 For any σ : K ,→ U embedding over F , σ(K ) ⊆ K hCi 4 acl(F ) ∩ K = dcl(Fd) for some d = (d1 , · · · , dm ). Let Gal(K /F ) be Aut(K /F ). Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 19 / 25 Iterative strongly normal extensions Let U be a large and highly saturated model of SCHp and let C = CU . Iterative Strongly Normal Extensions Let (F , D) < (K , D) an extension of definably closed ID-fields. We say that K /F is a iterative strongly normal extension if: 1 CF = CK and CF is algebraically closed. 2 K = dcl(Fa) (=Closing Fa under derivations and then under pth roots). 3 For any σ : K ,→ U embedding over F , σ(K ) ⊆ K hCi 4 acl(F ) ∩ K = dcl(Fd) for some d = (d1 , · · · , dm ). Let Gal(K /F ) be Aut(K /F ). Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 19 / 25 Iterative strongly normal extensions Let U be a large and highly saturated model of SCHp and let C = CU . Iterative Strongly Normal Extensions Let (F , D) < (K , D) an extension of definably closed ID-fields. We say that K /F is a iterative strongly normal extension if: 1 CF = CK and CF is algebraically closed. 2 K = dcl(Fa) (=Closing Fa under derivations and then under pth roots). 3 For any σ : K ,→ U embedding over F , σ(K ) ⊆ K hCi 4 acl(F ) ∩ K = dcl(Fd) for some d = (d1 , · · · , dm ). Let Gal(K /F ) be Aut(K /F ). Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 19 / 25 Iterative strongly normal extensions Let U be a large and highly saturated model of SCHp and let C = CU . Iterative Strongly Normal Extensions Let (F , D) < (K , D) an extension of definably closed ID-fields. We say that K /F is a iterative strongly normal extension if: 1 CF = CK and CF is algebraically closed. 2 K = dcl(Fa) (=Closing Fa under derivations and then under pth roots). 3 For any σ : K ,→ U embedding over F , σ(K ) ⊆ K hCi 4 acl(F ) ∩ K = dcl(Fd) for some d = (d1 , · · · , dm ). Let Gal(K /F ) be Aut(K /F ). Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 19 / 25 A good Galois Group and a Galois correspondence Some results If K /F is iterative strongly normal with dcl(Fa) = K , then 1 Gal(K /F ) is isomorphic to the CF -rational points of G, an algebraic group defined over CF , and 2 There is a Galois correspondence between algebraic subgroups of G(CF ) and definably closed ID-subfields of K which contain F . 3 There are equations associated to these extensions: iterative logarithmic differential equations. Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 20 / 25 A good Galois Group and a Galois correspondence Some results If K /F is iterative strongly normal with dcl(Fa) = K , then 1 Gal(K /F ) is isomorphic to the CF -rational points of G, an algebraic group defined over CF , and 2 There is a Galois correspondence between algebraic subgroups of G(CF ) and definably closed ID-subfields of K which contain F . 3 There are equations associated to these extensions: iterative logarithmic differential equations. Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 20 / 25 A good Galois Group and a Galois correspondence Some results If K /F is iterative strongly normal with dcl(Fa) = K , then 1 Gal(K /F ) is isomorphic to the CF -rational points of G, an algebraic group defined over CF , and 2 There is a Galois correspondence between algebraic subgroups of G(CF ) and definably closed ID-subfields of K which contain F . 3 There are equations associated to these extensions: iterative logarithmic differential equations. Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 20 / 25 But wait, there is more! Generalised strongly normal theory: What if we substitute the constants by an arbitrary formula? (Explored in characteristic zero by Pillay, Marker, Süer, &c.) The Galois group is finite-dimensional differential definable. (Ch. p?) Difference Galois theory: Automorphisms instead of derivatives? (Kameski (“partial automorphism” (preserving an almost arbitrary set of formulas), Singer, Chatzidakis, &c.) Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 21 / 25 But wait, there is more! Generalised strongly normal theory: What if we substitute the constants by an arbitrary formula? (Explored in characteristic zero by Pillay, Marker, Süer, &c.) The Galois group is finite-dimensional differential definable. (Ch. p?) Difference Galois theory: Automorphisms instead of derivatives? (Kameski (“partial automorphism” (preserving an almost arbitrary set of formulas), Singer, Chatzidakis, &c.) Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 21 / 25 Arigato Gozaimasu! Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 22 / 25 Arigato Gozaimasu! Unless... Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 22 / 25 One application I dream of: Galois theory of p-adic differential equations Let Cp be the completion of the algebraic closure of the field of p-adic numbers. (|a| = p−vp (a) ) Consider Cp (z) with | · |Gauss , the Gauss norm, defined by |am z m + · · · + a0 |Gauss = max |ai |. Denote by F the completion of Cp (z) with respect to the Gauss df is continuous norm. The differentiation on Cp (z) given by f 7→ dz with respect to the Gauss norm and thus extends uniquely to ∂F a differentiation of F. By a p-adic differential equation we mean a differential equation on the differential field (F, ∂F ). Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 23 / 25 One application I dream of: Galois theory of p-adic differential equations Let Cp be the completion of the algebraic closure of the field of p-adic numbers. (|a| = p−vp (a) ) Consider Cp (z) with | · |Gauss , the Gauss norm, defined by |am z m + · · · + a0 |Gauss = max |ai |. Denote by F the completion of Cp (z) with respect to the Gauss df is continuous norm. The differentiation on Cp (z) given by f 7→ dz with respect to the Gauss norm and thus extends uniquely to ∂F a differentiation of F. By a p-adic differential equation we mean a differential equation on the differential field (F, ∂F ). Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 23 / 25 One application I dream of: Galois theory of p-adic differential equations Let Cp be the completion of the algebraic closure of the field of p-adic numbers. (|a| = p−vp (a) ) Consider Cp (z) with | · |Gauss , the Gauss norm, defined by |am z m + · · · + a0 |Gauss = max |ai |. Denote by F the completion of Cp (z) with respect to the Gauss df is continuous norm. The differentiation on Cp (z) given by f 7→ dz with respect to the Gauss norm and thus extends uniquely to ∂F a differentiation of F. By a p-adic differential equation we mean a differential equation on the differential field (F, ∂F ). Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 23 / 25 One application I dream of: Galois theory of p-adic differential equations Let Cp be the completion of the algebraic closure of the field of p-adic numbers. (|a| = p−vp (a) ) Consider Cp (z) with | · |Gauss , the Gauss norm, defined by |am z m + · · · + a0 |Gauss = max |ai |. Denote by F the completion of Cp (z) with respect to the Gauss df is continuous norm. The differentiation on Cp (z) given by f 7→ dz with respect to the Gauss norm and thus extends uniquely to ∂F a differentiation of F. By a p-adic differential equation we mean a differential equation on the differential field (F, ∂F ). Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 23 / 25 One application I dream of: Galois theory of p-adic differential equations Fun facts 1 The field of constants of F is Cp . 2 Its residue field is Fp (z) and, more importantly, 3 1 n The reduction of ( n! ∂F )n≥0 to the residue field is (∂(n) )n≥0 , the m−n standard iterative derivative on Fp (z) given by ∂(n) z m = m . n z Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 24 / 25 One application I dream of: Galois theory of p-adic differential equations Fun facts 1 The field of constants of F is Cp . 2 Its residue field is Fp (z) and, more importantly, 3 1 n The reduction of ( n! ∂F )n≥0 to the residue field is (∂(n) )n≥0 , the m−n standard iterative derivative on Fp (z) given by ∂(n) z m = m . n z Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 24 / 25 One application I dream of: Galois theory of p-adic differential equations Fun facts 1 The field of constants of F is Cp . 2 Its residue field is Fp (z) and, more importantly, 3 1 n The reduction of ( n! ∂F )n≥0 to the residue field is (∂(n) )n≥0 , the m−n standard iterative derivative on Fp (z) given by ∂(n) z m = m . n z Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 24 / 25 One application I dream of: Galois theory of p-adic differential equations Natural questions 1 What logarithmic differential equations on (F, ∂F ) become iterative logarithmic differential equations on (Fp (z), ∂(n) ) after reduction? 2 In the cases where this happens, what is the relationship between the two Galois groups? Key problem What is the right model theoretical setting to study structures like (F, ∂F ) (i.e. differential valued fields of mixed characteristic and the iterative differential structure in the residue)? Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 25 / 25 One application I dream of: Galois theory of p-adic differential equations Natural questions 1 What logarithmic differential equations on (F, ∂F ) become iterative logarithmic differential equations on (Fp (z), ∂(n) ) after reduction? 2 In the cases where this happens, what is the relationship between the two Galois groups? Key problem What is the right model theoretical setting to study structures like (F, ∂F ) (i.e. differential valued fields of mixed characteristic and the iterative differential structure in the residue)? Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 25 / 25 One application I dream of: Galois theory of p-adic differential equations Natural questions 1 What logarithmic differential equations on (F, ∂F ) become iterative logarithmic differential equations on (Fp (z), ∂(n) ) after reduction? 2 In the cases where this happens, what is the relationship between the two Galois groups? Key problem What is the right model theoretical setting to study structures like (F, ∂F ) (i.e. differential valued fields of mixed characteristic and the iterative differential structure in the residue)? Javier Moreno (UCJ-Lyon 1) Mod. Th. & Dif. Gal. Th. Kirishima, March 4th, 2010 25 / 25