From: rusin@vesuvius.math.niu.edu (Dave Rusin) Subject: Re: first problem: more variables Date: 19 Aug 2000 05:42:53 GMT Newsgroups: sci.math Summary: [missing] In article , Zakir F. Seidov wrote: >Let (n=4): > >a=x+y+z+w, >b=x^2+y^2+z^2+w^2, >c=x^3+y^3+z^3+w^3, >d=x^4+y^4+z^4+w^4. > >Then: > >e=x^5+y^5+z^5+w^5= > (-a^5 + 10*a^3*b - 15*a*b^2 - 20*a^2*c + 20*b*c + 30*a*d)/24. > >May anyone derive general formula for n variables? You want the Newton formulae. Given a finite set of numbers x_i we may form the infinite sequence of the sums of their powers: p_n = \sum_i (x_i)^n . We may also compute their elementary symmetric functions, only finitely many of which are non-zero: expand \product_i (1 + x_i T) = \sum_n s_n T^n . (So s_0 = 1.) These two collections of numbers are related by the Newton formulae: 0= p_1 s_0 - 1 s_1 0= p_2 s_0 - p_1 s_1 + 2 s_2 0= p_3 s_0 - p_2 s_1 + p_1 s_2 - 3 s_3 0= p_4 s_0 - p_3 s_1 + p_2 s_2 - p_1 s_3 + 4 s_4 and so on. These formulae can be used in sequence to express the p_n as integer polynomials in the s_n. They may similarly be used to express the s_n as rational polynomials in the p_n (assuming we are working in characteristic zero). Since s_n = 0 for n > N (the count of the numbers x_i) we may combine these two sorts of computations to express each p_n as a rational polynomial in {p_1, ..., p_N}. These calculations lead to the formula shown for N=4; for N=5 I get 120*p6= 40*p3^2 + 90*p2*p4 - 15*p2^3 + 144*p5*p1 - 120*p2*p3*p1 + 45*p2^2*p1^2 - 90*p4*p1^2 + 40*p3*p1^3 - 15*p2*p1^4 + p1^6 Of course, there's no reason except exhaustion to stop here. By the way, once we know that N! * p_{N+1} is an integer linear combination of the appropriate monomials in the other p_n, we need only compute the coefficients; it suffices to pick sufficiently many sets of the {x_i} to generate enough linear equations relating these coefficients, then solve (using integer arithmetic!) dave