Neighbor identification is the most computationally intensive step in particle based simulations. To contain its cost, a common approach consists in using a regular grid to sort particles according to the cell they belong to. Then, neighbor search only needs to test the particles contained in a constant number of cells. During the simulation, a usually small amount of particles are moving between consecutive steps. Taking into account this temporal coherency to save on the maintenance cost of the acceleration data structure is difficult as it usually triggers costly dynamics memory allocations or data moves. In this paper we propose to rely on a Packed Memory Array (PMA) to efficiently keep particles sorted according to their cell index. The PMA maintains gaps in the particle array that enable to keep particle sorted with O(log2(n)) amortized data moves. We further improve the original PMA data structure to support efficient batch data moves. Experiments show that the PMA can outperform a compact sorted array for up to 50% element moves.