This theorem stays that any symmetric polynomial can be expressed as a polynomial of elementary polynomials. So let's suppose I have a symmetric polynomial $f(x_1,x_2,...,x_n)$ in $R[\mathbb{X}]$. I can find a unique polynomial $g(s_1,s_2,...,s_n)$ in $R[\mathbb{X}]$ so I can express f in term of g, where $s_1,s_2,...,s_n$ are elementary polynomials. My question to you is: Is there a algorithm/set of stepts to determine g? Thanks.


Let $c x_1^{a_1} x_2^{a_2} \dots x_n^{a_n}$ be the lexicographically largest monomial of $f$, that is there are no monomials with strictly larger exponent of $x_1$, and no monomials with $x_1$ exponent ${a_1}$ that have a higher power of $x_2$, and so on. We'll think of this as being the leading term of the polynomial.

Now the key thing to notice is that $e_n^{a_n}e_{n-1}^{(a_{n-1}-{a_n})} \dots e_1^{(a_1-a_2)}$ contains the monomial $x_1^{a_1} x_2^{a_2} \dots x_n^{a_n}$ with coefficient $1$ and all other monomials it contains are smaller lexicographically.

Now the point is you can consider the leading term of $f - c e_n^{a_n}e_{n-1}^{(a_{n-1}-{a_n})} \dots e_1^{(a_1-a_2)}$, and so on. Each step reduces the leading term (in lexicographic order), so this process must eventually terminate, at which point you have written $f$ in terms of the elementary symmetric functions.

  • $\begingroup$ Ok, you kinda lost me at the last paragraph. Can you please show me how this work on a example. Let's say I have the following polynomial: $$ P(x_1,x_2,x_3) = x_1^4+x_2^4+x_3^4-2x_1^2x_2^2-2x_2^2x_3^2-2x_1^2x_3^2 $$. Thanks a lot for your help. $\endgroup$ – Raducu Mihai Mar 9 '16 at 22:35
  • $\begingroup$ The lexicographically largest term is $x_1^4$. So according to the algorithm described you should subtract off $e_1^4 = (x_1+x_2+x_3)^4$. The point being that what remains, while having more terms, will have a smaller leading term namely $-4x_1^3x_2$. Then we apply the same operation again to this polynomial $P - e_1^4$, this time adding $4e_2e_1^2$ to cancel the leading term. Next we would cancel the leading term $P - e_1^4 + 4e_2e_1^2$, and so on. Eventually we are left with $P - g(e_1,e_2,e_3)$ having no leading term and hence $P - g(e_1,e_2,e_3) = 0$ as desired. $\endgroup$ – Nate Mar 10 '16 at 15:14

You can use divided difference operators to express the polynomial in terms of Schur polynomials, then you can apply the second Jacobi-Trudi formula, which expresses a Schur polynomial as the determinant of a matrix whose entries are elementary symmetric polynomials. Computationally this is rather inefficient, but it is nonrecursive, which is sometimes advantageous.

  • $\begingroup$ Hey, thank you for your answare. Can you please show me how this would work on a real example. See the polynomial in the comment section for @Nate 's answare. $\endgroup$ – Raducu Mihai Mar 10 '16 at 15:10
  • $\begingroup$ @Raducu I don't have a computer algebra system here, but my hand calculations resulted in $\partial^{4123}(p)=3$, $\partial^{5423}(p)=-1$, $\partial^{3423}(p)=-2$, so $p=3s_{(31)}-s_{(211)}-2s_{(22)}$, and you can go to the Wikipedia article on the Jacobi-Trudi formula and substitute what it gives you. $\endgroup$ – Matt Samuel Mar 10 '16 at 16:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.