# Robinson–Schensted–Knuth correspondence

In mathematics, the **Robinson–Schensted–Knuth correspondence**, also referred to as the **RSK correspondence** or **RSK algorithm**, is a combinatorial bijection between matrices *A* with non-negative integer entries and pairs (*P*,*Q*) of semistandard Young tableaux of equal shape, whose size equals the sum of the entries of *A*. More precisely the weight of *P* is given by the column sums of *A*, and the weight of *Q* by its row sums.
It is a generalization of the Robinson–Schensted correspondence, in the sense that taking *A* to be a permutation matrix, the pair (*P*,*Q*) will be the pair of standard tableaux associated to the permutation under the Robinson–Schensted correspondence.

The Robinson–Schensted–Knuth correspondence extends many of the remarkable properties of the Robinson–Schensted correspondence, notably its symmetry: transposition of the matrix *A* results in interchange of the tableaux *P*,*Q*.

## The Robinson–Schensted–Knuth correspondence

### Introduction

The Robinson–Schensted correspondence is a bijective mapping between permutations and pairs of standard Young tableaux, both having the same shape. This bijection can be constructed using an algorithm called Schensted insertion, starting with an empty tableau and successively inserting the values *σ*_{1},…,*σ*_{n} of the permutation *σ* at the numbers 1,2,…*n*; these form the second line when *σ* is given in two-line notation:

.

The first standard tableau *P* is the result of successive insertions; the other standard tableau *Q* records the successive shapes of the intermediate tableaux during the construction of *P*.

The Schensted insertion easily generalizes to the case where σ has repeated entries; in that case the correspondence will produce a semistandard tableau *P* rather than a standard tableau, but *Q* will still be a standard tableau. The definition of the RSK correspondence reestablishes symmetry between the *P* and *Q* tableaux by producing a semistandard tableau for *Q* as well.

### Two-line arrays

The *two-line array* (or *generalized permutation*) *w*_{A} corresponding to a matrix *A* is defined[1] as

in which for any pair (*i*,*j*) that indexes an entry *A*_{i,j} of *A*, there are *A*_{i,j} columns equal to , and all columns are in lexicographic order, which means that

- , and
- if and then .

#### Example

The two-line array corresponding to

is

### Definition of the correspondence

By applying the Schensted insertion algorithm to the bottom line of this two-line array, one obtains a pair consisting of a semistandard tableau *P* and a standard tableau *Q*_{0}, where the latter can be turned into a semistandard tableau *Q* by replacing each entry *b* of *Q*_{0} by the *b*-th entry of the top line of *w*_{A}. One thus obtains a bijection from matrices *A* to ordered pairs,[2] (*P*,*Q*) of semistandard Young tableaux of the same shape, in which the set of entries of *P* is that of the second line of *w*_{A}, and the set of entries of *Q* is that of the first line of *w*_{A}. The number of entries *j* in *P* is therefore equal to the sum of the entries in column *j* of *A*, and the number of entries *i* in *Q* is equal to the sum of the entries in row *i* of *A*.

#### Example

In the above example, the result of applying the Schensted insertion to successively insert 1,3,3,2,2,1,2 into an initially empty tableau results in a tableau *P*, and an additional standard tableau *Q*_{0} recoding the successive shapes, given by

and after replacing the entries 1,2,3,4,5,6,7 in *Q*_{0} successively by 1,1,1,2,2,3,3 one obtains the pair of semistandard tableaux

#### Direct definition of the RSK correspondence

The above definition uses the Schensted algorithm, which produces a standard recording tableau *Q*_{0}, and modifies it to take into account the first line of the two-line array and produce a semistandard recording tableau; this makes the relation to the Robinson–Schensted correspondence evident. It is natural however to simplify the construction by modifying the shape recording part of the algorithm to directly take into account the first line of the two-line array; it is in this form that the algorithm for the RSK correspondence is usually described. This simply means that after every Schensted insertion step, the tableau *Q* is extended by adding, as entry of the new square, the *b*-th entry *i*_{b} of the first line of *w*_{A}, where *b* is the current size of the tableaux. That this always produces a semistandard tableau follows from the property (first observed by Knuth[2]) that for successive insertions with an identical value in the first line of *w*_{A}, each successive square added to the shape is in a column strictly to the right of the previous one.

Here is a detailed example of this construction of both semistandard tableaux. Corresponding to a matrix

one has the two-line array

The following table shows the construction of both tableaux for this example

Inserted pair | ||||||||

P |
||||||||

Q |

## Combinatorial properties of the RSK correspondence

### The case of permutation matrices

If is a permutation matrix then RSK outputs standard Young Tableaux (SYT), of the same shape . Conversely, if are SYT having the same shape , then the corresponding matrix is a permutation matrix. As a result of this property by simply comparing the cardinalities of the two sets on the two sides of the bijective mapping we get the following corollary:

**Corollary 1**: For each we have
where means varies over all partitions of and is the number of standard Young tableaux of shape .

### Symmetry

Let be a matrix with non-negative entries. Suppose the RSK algorithm maps to then the RSK algorithm maps to , where is the transpose of .[1]

In particular for the case of permutation matrices, one recovers the symmetry of the Robinson–Schensted correspondence:[3]

**Theorem 2**: If the permutation corresponds to a triple , then the inverse permutation, , corresponds to .

This leads to the following relation between the number of involutions on with the number of tableaux that can be formed from (An *involution* is a permutation that is its own inverse):[3]

**Corollary 2**: The number of tableaux that can be formed from is equal to the number of involutions on .

*Proof*: If is an involution corresponding to , then corresponds to ; hence . Conversely, if is any permutation corresponding to , then also corresponds to ; hence . So there is a one-one correspondence between involutions and tableaux

The number of involutions on is given by the recurrence:

Where . By solving this recurrence we can get the number of involutions on ,

### Symmetric matrices

Let and let the RSK algorithm map the matrix to the pair , where is an SSYT of shape .[1] Let where the and . Then the map establishes a bijection between symmetric matrices with row() and SSYT's of type .

## Applications of the RSK correspondence

### Cauchy's identity

The Robinson–Schensted–Knuth correspondence provides a direct bijective proof of the following celebrated identity for symmetric functions:

where are Schur functions.

### Kostka numbers

Fix partitions , then

where and denote the Kostka numbers and is the number of matrices , with non-negative elements, with row() and column() .

## References

- Stanley, Richard P. (1999).
*Enumerative Combinatorics*.**2**. New York: Cambridge University Press. pp. 316–380. ISBN 0-521-55309-1. - Knuth, Donald E. (1970). "Permutations, matrices, and generalized Young tableaux".
*Pacific Journal of Mathematics*.**34**(3): 709–727. doi:10.2140/pjm.1970.34.709. MR 0272654. - Knuth, Donald E. (1973).
*The Art of Computer Programming, Vol. 3: Sorting and Searching*. London: Addison–Wesley. pp. 54–58. ISBN 0-201-03803-X.

- Brualdi, Richard A. (2006).
*Combinatorial matrix classes*. Encyclopedia of Mathematics and Its Applications.**108**. Cambridge: Cambridge University Press. pp. 135–162. ISBN 0-521-86565-4. Zbl 1106.05001.