A lattice calculation of the nucleon's spin-dependent structure function g_2 revisited

Our previous calculation of the spin-dependent structure function g_2 is revisited. The interest in this structure function is to a great extent motivated by the fact that it receives contributions from twist-two as well as from twist-three operators already in leading order of 1/Q^2 thus offering the unique possibility of directly assessing higher-twist effects. In our former calculation the lattice operators were renormalized perturbatively and mixing with lower-dimensional operators was ignored. However, the twist-three operator which gives rise to the matrix element d_2 mixes non-perturbatively with an operator of lower dimension. Taking this effect into account leads to a considerably smaller value of d_2, which is consistent with the experimental data.