Let $T$ be a standard Young tableaux on $[n]$. Denote the RSK algorithm $\text{RSK}(w)=(P(T),Q(T))$ for $w\in\mathfrak{S}_n$, where $P(T)$ is the Schencted insertion tableaux.

For $1\leq i\leq j\leq n$. Let $T_{[i,j]}$ be the skew SYT by restricting $T$ to the segament $[i,j]$. For a skew shape $Y$, define the rectification of Y, $\text{Rect}(Y)$ to be applying jeu de taquin on $Y$ to obtain a standard shape. See section 2.1 of this paper, in which $\text{Rect}$ is denoted as $\text{std}$.

It is well known that

For $w\in \mathfrak{S}_n$, $T\in \text{SYT}_n$. If $P(w)=T$, then $$\text{Rect}(T_{[i,j]})=P(w_{[i,j]})$$ for all $[i,j]\subseteq [n]$, where $w_{[i,j]}$ means restricting the permutation to the subalphabet $[i,j]$, e.g. $126534_{[2,5]}=2534$.

My question is: is there a $K$-theoretic analog of this property, in terms of Hecke insertion, $K$-jeu-de-taquin of increasing tableaux?

Specifically, define $K$-rectification by replacing jdt with $K$-jdt, and denote $K$-$P(w)$ the Hecke-insertion tableau of the word $w$. We say a Tableau is unique-rectification-target if it's unique in its $K$-Knuth class. (Definition 3.15 of [BS13], see also section 4 of [PP14])

Let $T$ be an increasing Tableau (of alphabet $[n]$) and $Y=T_{[1,i]}$ such that $Y$ is an SYT and $1\cdots i\notin T\backslash Y$ (so that there is no ambiguity). If $K\text{-Rect}(T\backslash Y)$ is a URT, then is it always true that: $$K\text{-Rect}(T\backslash Y)=K\text{-}P(w_{[i+1,n]})$$? where $w$ is the row-reading word of $T$, or even all words such that $K\text{-}P(w)=T$.

In general, is it true that $$\{K\text{-Rect}(T\backslash Y)\}=\{K\text{-}P(w_{[i+1,n]}):K\text{-}P(w)=T\}$$

For example, Let $T=\begin{matrix}1&2&4\\3&5&6\\4&6&9 \end{matrix},Y=\begin{matrix}1&2\\3&\end{matrix}$. We have $$K\text{-Rect}\left(\begin{matrix}*&*&4\\*&5&6\\4&6&9 \end{matrix} \right)=\begin{matrix}4&5&6\\5&9&\\6&& \end{matrix}$$ The row reading word of $T$ is $w=\mathfrak{row}(T)=469356124$, and $$K\text{-}P(w_{[4,9]})=K\text{-}P(469564)= \begin{matrix}4&5&6\\5&9&\\6&& \end{matrix}$$

Thanks !

  • $\begingroup$ Nice question ;) $\endgroup$ Jun 16 '19 at 21:40
  • 1
    $\begingroup$ But one issue is that $K$-rectification is not necessarily uniquely defined because $K$-jdt lacks the same confluence properties as usual jdt. $\endgroup$ Jun 16 '19 at 21:46
  • $\begingroup$ Perhaps there is a nice interplay with K-theoretic crystals? de.arxiv.org/pdf/1904.09674.pdf RSK and crystal operators in the classical case are closely related... $\endgroup$ Jun 17 '19 at 9:19
  • $\begingroup$ @PerAlexandersson Thanks! $\endgroup$ Jun 17 '19 at 20:17
  • $\begingroup$ @SamHopkins Actually I was very wrong. $K$-Rect is not uniquely defined. See the edit of my question. $\endgroup$ Jun 18 '19 at 1:16

The statement is not true for increasing tableau.

$$K\text{-rect}\left(\begin{matrix}*&*&1&2\\*&1&3&4\\2&3\end{matrix}\right)=\begin{matrix}1&3&2\\2&4\end{matrix}$$ $$K\text{-}P(2313412)=\begin{matrix}1&2&4\\2&3\end{matrix}$$ I think the reason is that in the standard setting, jdt preserves Knuth equivalence and Knuth equivalence $\iff$Schensted insertion equivalence. But in $K$-theoretic version, $K$-jdt preserves $K$-knuth equivalence but $K$-Knuth equivalence if (but not only if) Hecke insertion equivalence.


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.