A major motivation for the AKLT model was the Majumdar–Ghosh chain. Because two out of every set of three neighboring spins in a Majumdar–Ghosh ground state are paired into a singlet, or valence bond, the three spins together can never be found to be in a spin 3/2 state. In fact, the Majumdar–Ghosh Hamiltonian is nothing but the sum of all projectors of three neighboring spins onto a 3/2 state.
The main insight of the AKLT paper was that this construction could be generalized to obtain exactly solvable models for spin sizes other than 1/2. Just as one end of a valence bond is a spin 1/2, the ends of two valence bonds can be combined into a spin 1, three into a spin 3/2, etc.
Definition
Affleck et al. were interested in constructing a one-dimensional state with a valence bond between every pair of sites. Because this leads to two spin 1/2s for every site, the result must be the wavefunction of a spin 1 system.
For every adjacent pair of the spin 1s, two of the four constituent spin 1/2s are stuck in a total spin zero state. Therefore, each pair of spin 1s is forbidden from being in a combined spin 2 state.
By writing this condition as a sum of projectors that favor the spin 2 state of pairs of spin 1s, AKLT arrived at the following Hamiltonian
up to a constant,
where the are spin-1 operators, and the local 2-point projector that favors the spin 2 state of an adjacent pair of spins.
This Hamiltonian is similar to the spin 1, one-dimensional quantumHeisenberg spin model but has an additional "biquadratic" spin interaction term.
Ground state
By construction, the ground state of the AKLT Hamiltonian is the valence bond solid with a single valence bond connecting every neighboring pair of sites. Pictorially, this may be represented as
Here the solid points represent spin 1/2s which are put into singlet states. The lines connecting the spin 1/2s are the valence bonds indicating the pattern of singlets. The ovals are projection operators which "tie" together two spin 1/2s into a single spin 1, projecting out the spin 0 or singlet subspace and keeping only the spin 1 or triplet subspace. The symbols "+", "0" and "−" label the standard spin 1 basis states (eigenstates of the operator).[10]
Spin 1/2 edge states
For the case of spins arranged in a ring (periodic boundary conditions) the AKLT construction yields a unique ground state. But for the case of an open chain, the first and
last spin 1 have only a single neighbor, leaving one of their constituent spin 1/2s unpaired. As a result, the ends of the chain behave like free spin 1/2 moments even though
the system consists of spin 1s only.
The spin 1/2 edge states of the AKLT chain can be observed in a few different ways. For short chains, the edge states mix into a singlet or a triplet giving either a unique ground state or a three-fold multiplet of ground states. For longer chains, the edge states decouple exponentially quickly as a function of chain length leading to a ground state manifold that is four-fold degenerate.[11] By using a numerical method such as DMRG to measure the local magnetization along the chain, it is also possible to see the edge states directly and to show that they can be removed by placing actual spin 1/2s at the ends.[12] It has even proved possible to detect the spin 1/2 edge states in measurements of a quasi-1D magnetic compound containing a small amount of impurities whose role is to break the chains into finite segments.[13] In 2021, a direct spectroscopic signature of spin 1/2 edge states was found in isolated quantum spin chains built out of triangulene, a spin 1 polycyclic aromatic hydrocarbon.[14]
Matrix product state representation
The simplicity of the AKLT ground state allows it to be represented in compact form as a matrix product state.
This is a wavefunction of the form
Here the As are a set of three matrices labeled by and the trace comes from assuming periodic boundary conditions.
The AKLT ground state wavefunction corresponds to the choice:[10]