1. Introduction. In a recent paper [8], Imamoḡlu and Kohnen have studied the mth power of the Riemann theta function ϑ in relation with the number r m (n) of representations of a positive integer n as a sum of m integral squares. Their result is interesting, since, for each m, the computation of r m (n) does not require any pre-knowledge of r m (n ′ ) for n ′ < n. One of the main tools used in this proof was that ϑ m has highest order of vanishing at one cusp or, better, that a translate of ϑ has highest order of vanishing at the cusp ∞; subsequently Kohnen and the second author extended the result to the integral representations of the lattice D + m (see [9]). In this paper we want to treat the problem of theta series with the highest order of vanishing at the cusp ∞. When the level is a power of 2, these theta series are the mth powers of a certain theta function with characteristic, related to the quadratic form 2 k 1 m or, in the language of lattices, to the lattice √ 2 k Z m . Instead if the level is a power of 3, these theta series are the (m/2)th powers of a theta series associated to the 2-dimensional root lattice A 2 with characteristic. These modular forms have also many representations as theta series related to different lattices.Let L be an integral lattice of rank m. We denote by (·, ·) its associated scalar product, and we say as usual that L is even if (x, x) ≡ 0 mod 2. We define the dual lattice by