Soliton Cellular Automata Using Rigged Configurations
Travis Scrimshaw
This is joint work with Xuan Liu (UMN).- states are an empty box (vacuum) or a ball;
- time evolution is given by moving balls from right to left and putting them in the next available position.
- Collections of adjacent balls, called solitons, move with speed according to their length/size.
- After interaction, called scattering, the solitons retain their sizes.
We can generalize the box-ball system even further by using a certain class of crystals in affine type called Kirillov–Reshetikhin (KR) crystals. The idea comes from row-to-row transfer matrices of solvable lattice models. We represent elements of the box-ball system by the type (or ) KR crystals : the set with crystal operators Note that the vacuum state corresponds to the element .
We also need the KR crystal , which are monomials in considered as commuting variables and the crystal operators act as above. Crystals also have a natural tensor product structure. Moreover, there is a unique crystal isomorphism called the combinatorial -matrix, that is given by if and only if Last, we require the local energy function, defined as as the maximal number of boxes the row can cover of .
For more background on type crystals, see Crystals for dummies by Mark Shimozono.
- Solitons of length are parameterized by crystals of type given by subtracting 1 from each entry.
- The scattering of two solitons corresponds to the combinatorial -matrix of type .
- The phase shift, the difference of positions if there was no interaction, is equal to , where is the length of the smaller soliton.
A rigged configuration is a sequence of partitions , one for each node of the classical Dynkin diagram, with integers attached to row called riggings.
There is a (conjectural) bijection between rigged configrations and tensor products of KR crystals. This bijection is recursively defined and very techincal (it comes from the Bethe ansatz of a related object in mathematical physics called Heisenberg spin chains), however, it linearizes the dynamics of SCA. Thus, it allows us to extract information about SCA and the time evolution.
Theorem: [Liu-S. 2017]
Assuming certain conjectures hold about the bijection , we show the following for certain classes of SCA:
- Time evolution is described by changing the riggings on by adding their corresponding row length.
- The lengths of the solitons corresponds to the rows of .
- Solitons of length are parameterized by crystals given by removing from the classical Dynkin diagram and is the (set of) adjacent node(s).
- The scattering of two solitons corresponds to the branched combinatorial -matrix.
- The phase shift is equal to , where is the length of the smaller soliton and is the Cartan matrix.
- Question: Is this the correct definition of soltions and length?
We think so.
- Open Problem: The description of the phase shift is more subtle when the length is not the number of factors. Moreover, our statement about the phase shift likely does not generalize. Make this statement precise and true for the general case.