Abstract-In a recent work, Coronel et al. initiated the study of the relation between the diversity-multiplexing tradeoff (DMT) performance of a multiuser multiple-input multiple-output (MU-MIMO) lattice code and the rate of the decay of the determinants of the code matrix as a function of the size of the signal constellation. In this note, we state a simple general upper bound on the decay function and study the promising code proposed by Badr & Belfiore in close detail. We derive a lower bound to its decay function based on a classical theorem due to Liouville. The resulting bound is applicable also to other codes with constructions based on algebraic number theory. Further, we study an example sequence of small determinants within the Badr-Belfiore code and derive a tighter upper bound to its decay function. The upper bound has certain conjectural asymptotic uncertainties, whence we also list the exact bound for several finite data rates.
I. BACKGROUND AND THE DECAY FUNCTIONAssume that we are to design a system for U simultaneously transmitting synchronized users, each transmitting with n t transmit antennas and, for simplicity so that we end up with square matrices, over U n t channel uses. We can describe each user's signals as n t × U n t complex matrices. A multiuser MIMO signal is then viewed as a U n t × U n t matrix obtained by using the signals of the individual users as blocks. So each user is occupying n t rows in this overall transmission matrix.Any study of DMT questions calls for a scalable set of finite signal constellations. For the sake of convenience most authors assume that these signal sets of individual users are carved out of a user specific lattice L j ⊂ M nt×Unt , j = 1, . . . , U .When studying DMT questions it is natural to assume that each user is maximally using the degrees of freedom available to him/her. Therefore, the lattices of the individual users should be of full rank n = 2U n 2 t , so that each user's signals consist of integral linear combinations of n user specific basis matrices. A natural scaling parameter is the range of the integer coefficients. We assume that the range is parameterized by a natural number N .