 Solution

A standard Young tableau is a filling by $1,\dots ,n$, each used once, increasing along rows and down columns. A semistandard Young tableau is a filling by positive integers, weakly increasing along rows and strictly increasing down columns. Its content $\mu$ means that the entry $i$ occurs ${\mu }_i$ times. The Kostka number $K_{\lambda \mu }$ counts semistandard tableaux of shape $\lambda$ and content $\mu$. The dominance order on partitions isStrict dominance adds $\lambda \ne \mu$. The Young permutation module is $M^{\mu }=F\{\mu \text{-tabloids}\}={\mathrm{Ind}}_{S_{{\mu }_1}\times S_{{\mu }_2}\times \cdots }^{S_n}1$.

Let $D^{\lambda }$ be a composition factor of $M^{\mu }$. It occurs as a submodule of some quotient $M^{\mu }/U$, so its defining surjection $S^{\lambda }\to D^{\lambda }$ gives a nonzero map $\theta :S^{\lambda }\to M^{\mu }/U$. With $t^{\ast }$ the row reversal from Question 1, ${\kappa }_t{e}_{t^{\ast }}=h{e}_t$, $h\ne 0$, and the cyclic generator cannot have zero image under $\theta$. Therefore ${\kappa }_t$ acts nontrivially on $M^{\mu }/U$, and hence on $M^{\mu }$. The column-cancellation argument proves $\lambda ⊵\mu$.

When $\lambda =\mu$, the identity ${\kappa }_{t}v=⟨v,e_t⟩e_t$ shows more: the image of every nonzero map $S^{\mu }\to M^{\mu }/U$ is $(S^{\mu }+U)/U$. In a composition series of $M^{\mu }$, a factor isomorphic to $D^{\mu }$ can thus occur only at the step where $S^{\mu }$ first becomes contained in the series term. Once contained, every later such map would have zero image. Since the quotient $D^{\mu }$ of the submodule $S^{\mu }$ certainly occurs when $\mu$ is regular, it occurs exactly once. If $\mu$ is not regular, no simple is indexed by $\mu$. Consequently every other factor has a strictly dominating regular label. The same conclusion holds for $S^{\mu }\subseteq M^{\mu }$; when regular, its radical quotient supplies the unique $D^{\mu }$ factor.

The integral standard-basis theorem says that the standard polytabloids form a basis over every field. Thus $\dim S_F^{\mu }=f^{\mu }$ is independent of characteristic. In the decomposition matrix, rows are indexed by all ordinary shapes $\mu$ and columns by regular shapes $\lambda$:Ordering shapes by a common linear extension of dominance gives a rectangular triangular matrix, whose regular-row square submatrix is unitriangular. The standard basis also gives $f^{\mu }={\sum }_{\lambda }{d}_{\mu \lambda }\dim D^{\lambda }$.

Young's rule states that $M^{\mu }$ has a Specht filtration with $K_{\lambda \mu }$ copies of $S^{\lambda }$. In characteristic zero this is a direct-sum decomposition into irreducibles. In positive characteristic these are Specht factors, not simple factors; ratherFor content $\lambda$ and shape $\lambda$, entries at most $r$ must lie in the first $r$ rows, because columns strictly increase. Their number is exactly the size of those rows, so successive rows must be filled by their row number. Hence $K_{\lambda \lambda }=1$. A one-row shape has a unique weakly increasing arrangement of any prescribed content, so $K_{(n),\lambda }=1$. Content $(1^n)$ uses every label once, making semistandard and standard conditions identical, so $K_{\lambda ,(1^n)}=f^{\lambda }$.

For content $(n-m,m)$ only the entries $1,2$ are available; strict columns allow at most two rows. In shape $(n-j,j)$ every bottom entry is $2$ and every entry above it is $1$. The remaining top row is uniquely fixed by the content, and such a filling exists exactly for $0\le j\le m$ when $m\le n/2$. Young's rule therefore giveswhere multiplication denotes induction from the indicated Young subgroup, not pointwise multiplication of characters of the same group. The dimension of this permutation module is $\left(\genfrac{}{}{0ex}{}{n}{m}\right)$. Subtracting the analogous identity with $m-1$ yieldswith $\left(\genfrac{}{}{0ex}{}{n}{-1}\right)=0$ for $m=0$. This dimension holds over every field.