[math-fun] Possibly naive question about algebraic numbers
Suppose we have not one integer polynomial in one complex variable, but n > 1 integer polynomials P_1, . . . P_n in n complex variables z_1, . . . , z_n. Set Z = (z_1, . . ., z_n) and P(Z) = (P_1(Z), . . ., P_n(Z)) in C^n, for any z_1, . . ., z_n in C. Suppose (*) Z_0 in C^n satisfies P(Z_0) = 0. ------------------------------------------------------------------- QUESTIONS: 1. Does (*) imply that the complex coordinates of Z_0 are all algebraic numbers ??? 2, If the answer to this is "not necessarily", what if we add the condition (**) det(JP(Z_0)) is non-zero, where JP is the n x n Jacobian matrix (df_j/dz_k (Z_0)), 1 <= i, j <= n ??? 3. If still no, then what kind of numbers do we get, anyway, esp. if we assume both (*) and (**) -- supposing we toss in all complex coordinates of all such "algebraic vectors" Z_0 for all n >= 1 ??? Or for a fixed n ??? Or for all n <= N, for some fixed N ??? Just curious. --Dan
Certainly (*) is not sufficient. If all the polynomials are divisible by (z_1 - z_2), then any Z_0 with z_1 = z_2 will satisfy it. Franklin T. Adams-Watters -----Original Message----- From: dasimov@earthlink.net Suppose we have not one integer polynomial in one complex variable, but n > 1 integer polynomials P_1, . . . P_n in n complex variables z_1, . . . , z_n. Set Z = (z_1, . . ., z_n) and P(Z) = (P_1(Z), . . ., P_n(Z)) in C^n, for any z_1, . . ., z_n in C. Suppose (*) Z_0 in C^n satisfies P(Z_0) = 0. ------------------------------------------------------------------- QUESTIONS: 1. Does (*) imply that the complex coordinates of Z_0 are all algebraic numbers ??? 2, If the answer to this is "not necessarily", what if we add the condition (**) det(JP(Z_0)) is non-zero, where JP is the n x n Jacobian matrix (df_j/dz_k (Z_0)), 1 <= i, j <= n ??? 3. If still no, then what kind of numbers do we get, anyway, esp. if we assume both (*) and (**) -- supposing we toss in all complex coordinates of all such "algebraic vectors" Z_0 for all n >= 1 ??? Or for a fixed n ??? Or for all n <= N, for some fixed N ??? Just curious. --Dan ________________________________________________________________________ Check Out the new free AIM(R) Mail -- 2 GB of storage and industry-leading spam and email virus protection.
On Tuesday 29 August 2006 01:59, Daniel Asimov wrote:
Suppose we have not one integer polynomial in one complex variable, but n > 1 integer polynomials P_1, . . . P_n in n complex variables z_1, . . . , z_n.
Set Z = (z_1, . . ., z_n) and P(Z) = (P_1(Z), . . ., P_n(Z)) in C^n, for any z_1, . . ., z_n in C.
Suppose
(*) Z_0 in C^n satisfies P(Z_0) = 0. -------------------------------------------------------------------
QUESTIONS:
1. Does (*) imply that the complex coordinates of Z_0 are all algebraic numbers ???
Obviously not as it stands. P1(z1,z2) = z1^2 - z2^2 P2(z1,z2) = z1^3 - z2^3 Z = (pi,pi)
2, If the answer to this is "not necessarily", what if we add the condition
(**) det(JP(Z_0)) is non-zero,
where JP is the n x n Jacobian matrix (df_j/dz_k (Z_0)), 1 <= i, j <= n ???
The coordinates must then be algebraic. Here's a proof, which is probably unnecessarily clumsy although its basic idea is very simple: if there's a non-algebraic root then you can tweak it a little and still have a root. Let K be the field generated by the coordinates of Z, and let A be any field homomorphism from K to any subfield of C. Then P(A(Z)) = A(P)(Z) = P(Z) = 0, where A(P) means "P with A applied to its coefficients". Suppose K isn't algebraic. Then it's generated by some (finite, but nonempty) set of independent transcendentals y1, ..., yk, plus perhaps something x algebraic over y1,...,yk; a homomorphism A : K -> C is defined by choosing A(y1), ..., A(yk), A(x) such that q(y1,...,yk,x) = 0 where q is the minimal polynomial for x. Since q, being a minimal polynomial, is irreducible, we have dq/dx(y1,...,yk,x) nonzero. Hence, by the implicit function theorem, there's a neighbourhood of x where we can write x as a smooth function of y1,...,yk. So now choose A(y1), ..., A(yk) to be arbitrarily close to y1,...,yk; and define A(x) according to that smooth function (so it's also close to x). Then A(Z) is arbitrarily close to Z. In other words, there are points where P(_)=0 arbitrarily close to Z. But this is impossible, since the Jacobian of P is nonzero at Z. Contradiction! -- g
participants (3)
-
Daniel Asimov -
franktaw@netscape.net -
Gareth McCaughan