Skip to main content
added 501 characters in body
Source Link

For a polynomial or a formal power series $F(x_1,\ldots,x_N)$ in $x_1,\ldots,x_N$, let $$ {\rm Sym} [F(x_1,\ldots,x_N)]=\frac{1}{N!}\sum_{\sigma\in \mathfrak{S}_N} F(x_{\sigma(1)},\ldots,x_{\sigma(N)}) $$ denote the (normalized) symmetrization, as an average over permutations in the symmetric group $\mathfrak{S}_N$.

For an integer partition $\lambda=(1^{a_1}2^{a_2}\cdots)$, with $a_1$ parts equal to 1, $a_2$ parts equal to 2, etc., the monomial symetric function $m_{\lambda}$ satisfies $$ {\rm Sym}[x_1^{\lambda_1}\cdots x_N^{\lambda_N}]=\frac{a_1!a_2!\cdots}{N!}\ m_{\lambda}(x_1,\ldots,x_N) $$ because the overcounting factor or size of the stabilizer of the monomial being symmetrized is $a_1! a_2!\cdots$. By the Orbit-Stabilizer Theorem, the number of terms in $m_{\lambda}$ is $$ L_{r,\lambda}=\frac{N!}{a_1!a_2!\cdots}\ . $$ As a result, the wanted series is $$ S(x_1,\ldots,x_N)={\rm Sym}[U(x_1,\ldots,x_N)] $$ with $$ U(x_1,\ldots,x_N)=\sum_{r\ge 0}\sum_{\lambda \vdash r} x_1^{\lambda_1}x_2^{\lambda_2}\cdots x_N^{\lambda_N} $$ Changing variables to $u_1=\lambda_1-\lambda_2$, $u_2=\lambda_2-\lambda_3$,..., $u_{N-1}=\lambda_{N-1}-\lambda_N$, $u_N=\lambda_N$, we get $$ U(x_1,\ldots,x_N)=\sum_{u_1,\ldots,u_N\ge 0}x_1^{u_1+\cdots+u_N} x_2^{u_2+\cdots+u_N}\cdots x_{N-1}^{u_{N-1}+u_{N}} x_N^{u_N} $$ This immediately gives $$ S(x_1,\ldots,x_N)={\rm Sym}\left[\frac{1}{(1-x_1)(1-x_1 x_2)\cdots(1-x_1\cdots x_N)}\right] $$


Edit:

Figuring out the result as a single fraction may perhaps be interesting. There are related formulas where, in the numerator of $U$, one has a product over descents of products of the $x$'s up to that descent. This would be the trivial or antichain poset case of formula (6.3) in the article "Linear extension sums as valuations of cones" by Boussicault, Féray, Lascoux, and Reiner.

For a polynomial or a formal power series $F(x_1,\ldots,x_N)$ in $x_1,\ldots,x_N$, let $$ {\rm Sym} [F(x_1,\ldots,x_N)]=\frac{1}{N!}\sum_{\sigma\in \mathfrak{S}_N} F(x_{\sigma(1)},\ldots,x_{\sigma(N)}) $$ denote the (normalized) symmetrization, as an average over permutations in the symmetric group $\mathfrak{S}_N$.

For an integer partition $\lambda=(1^{a_1}2^{a_2}\cdots)$, with $a_1$ parts equal to 1, $a_2$ parts equal to 2, etc., the monomial symetric function $m_{\lambda}$ satisfies $$ {\rm Sym}[x_1^{\lambda_1}\cdots x_N^{\lambda_N}]=\frac{a_1!a_2!\cdots}{N!}\ m_{\lambda}(x_1,\ldots,x_N) $$ because the overcounting factor or size of the stabilizer of the monomial being symmetrized is $a_1! a_2!\cdots$. By the Orbit-Stabilizer Theorem, the number of terms in $m_{\lambda}$ is $$ L_{r,\lambda}=\frac{N!}{a_1!a_2!\cdots}\ . $$ As a result, the wanted series is $$ S(x_1,\ldots,x_N)={\rm Sym}[U(x_1,\ldots,x_N)] $$ with $$ U(x_1,\ldots,x_N)=\sum_{r\ge 0}\sum_{\lambda \vdash r} x_1^{\lambda_1}x_2^{\lambda_2}\cdots x_N^{\lambda_N} $$ Changing variables to $u_1=\lambda_1-\lambda_2$, $u_2=\lambda_2-\lambda_3$,..., $u_{N-1}=\lambda_{N-1}-\lambda_N$, $u_N=\lambda_N$, we get $$ U(x_1,\ldots,x_N)=\sum_{u_1,\ldots,u_N\ge 0}x_1^{u_1+\cdots+u_N} x_2^{u_2+\cdots+u_N}\cdots x_{N-1}^{u_{N-1}+u_{N}} x_N^{u_N} $$ This immediately gives $$ S(x_1,\ldots,x_N)={\rm Sym}\left[\frac{1}{(1-x_1)(1-x_1 x_2)\cdots(1-x_1\cdots x_N)}\right] $$

For a polynomial or a formal power series $F(x_1,\ldots,x_N)$ in $x_1,\ldots,x_N$, let $$ {\rm Sym} [F(x_1,\ldots,x_N)]=\frac{1}{N!}\sum_{\sigma\in \mathfrak{S}_N} F(x_{\sigma(1)},\ldots,x_{\sigma(N)}) $$ denote the (normalized) symmetrization, as an average over permutations in the symmetric group $\mathfrak{S}_N$.

For an integer partition $\lambda=(1^{a_1}2^{a_2}\cdots)$, with $a_1$ parts equal to 1, $a_2$ parts equal to 2, etc., the monomial symetric function $m_{\lambda}$ satisfies $$ {\rm Sym}[x_1^{\lambda_1}\cdots x_N^{\lambda_N}]=\frac{a_1!a_2!\cdots}{N!}\ m_{\lambda}(x_1,\ldots,x_N) $$ because the overcounting factor or size of the stabilizer of the monomial being symmetrized is $a_1! a_2!\cdots$. By the Orbit-Stabilizer Theorem, the number of terms in $m_{\lambda}$ is $$ L_{r,\lambda}=\frac{N!}{a_1!a_2!\cdots}\ . $$ As a result, the wanted series is $$ S(x_1,\ldots,x_N)={\rm Sym}[U(x_1,\ldots,x_N)] $$ with $$ U(x_1,\ldots,x_N)=\sum_{r\ge 0}\sum_{\lambda \vdash r} x_1^{\lambda_1}x_2^{\lambda_2}\cdots x_N^{\lambda_N} $$ Changing variables to $u_1=\lambda_1-\lambda_2$, $u_2=\lambda_2-\lambda_3$,..., $u_{N-1}=\lambda_{N-1}-\lambda_N$, $u_N=\lambda_N$, we get $$ U(x_1,\ldots,x_N)=\sum_{u_1,\ldots,u_N\ge 0}x_1^{u_1+\cdots+u_N} x_2^{u_2+\cdots+u_N}\cdots x_{N-1}^{u_{N-1}+u_{N}} x_N^{u_N} $$ This immediately gives $$ S(x_1,\ldots,x_N)={\rm Sym}\left[\frac{1}{(1-x_1)(1-x_1 x_2)\cdots(1-x_1\cdots x_N)}\right] $$


Edit:

Figuring out the result as a single fraction may perhaps be interesting. There are related formulas where, in the numerator of $U$, one has a product over descents of products of the $x$'s up to that descent. This would be the trivial or antichain poset case of formula (6.3) in the article "Linear extension sums as valuations of cones" by Boussicault, Féray, Lascoux, and Reiner.

Source Link

For a polynomial or a formal power series $F(x_1,\ldots,x_N)$ in $x_1,\ldots,x_N$, let $$ {\rm Sym} [F(x_1,\ldots,x_N)]=\frac{1}{N!}\sum_{\sigma\in \mathfrak{S}_N} F(x_{\sigma(1)},\ldots,x_{\sigma(N)}) $$ denote the (normalized) symmetrization, as an average over permutations in the symmetric group $\mathfrak{S}_N$.

For an integer partition $\lambda=(1^{a_1}2^{a_2}\cdots)$, with $a_1$ parts equal to 1, $a_2$ parts equal to 2, etc., the monomial symetric function $m_{\lambda}$ satisfies $$ {\rm Sym}[x_1^{\lambda_1}\cdots x_N^{\lambda_N}]=\frac{a_1!a_2!\cdots}{N!}\ m_{\lambda}(x_1,\ldots,x_N) $$ because the overcounting factor or size of the stabilizer of the monomial being symmetrized is $a_1! a_2!\cdots$. By the Orbit-Stabilizer Theorem, the number of terms in $m_{\lambda}$ is $$ L_{r,\lambda}=\frac{N!}{a_1!a_2!\cdots}\ . $$ As a result, the wanted series is $$ S(x_1,\ldots,x_N)={\rm Sym}[U(x_1,\ldots,x_N)] $$ with $$ U(x_1,\ldots,x_N)=\sum_{r\ge 0}\sum_{\lambda \vdash r} x_1^{\lambda_1}x_2^{\lambda_2}\cdots x_N^{\lambda_N} $$ Changing variables to $u_1=\lambda_1-\lambda_2$, $u_2=\lambda_2-\lambda_3$,..., $u_{N-1}=\lambda_{N-1}-\lambda_N$, $u_N=\lambda_N$, we get $$ U(x_1,\ldots,x_N)=\sum_{u_1,\ldots,u_N\ge 0}x_1^{u_1+\cdots+u_N} x_2^{u_2+\cdots+u_N}\cdots x_{N-1}^{u_{N-1}+u_{N}} x_N^{u_N} $$ This immediately gives $$ S(x_1,\ldots,x_N)={\rm Sym}\left[\frac{1}{(1-x_1)(1-x_1 x_2)\cdots(1-x_1\cdots x_N)}\right] $$