Remarks on sum rules in the heavy quark limit of QCD

We underline a problem existing in the heavy quark limit of QCD concerning the rates of semileptonic B decays into P-wave $D_J(j)$ mesons, where $j = {1 \over 2}$ (wide states) or $j = {3 \over 2}$ (narrow states). The leading order sum rules of Bjorken and Uraltsev suggest $\Gamma [ \bar{B} \to D_{0,1} ({1 \over 2}) \ell \nu ] \ll \Gamma [ \bar{B} \to D_{1,2} ({3 \over 2}) \ell \nu ]$, in contradiction with experiment. The same trend follows also from a sum rule for the subleading $1/m_Q$ curent matrix element correction $\xi_3(1)$. The problem is made explicit in relativistic quarks models \`a la Bakamjian and Thomas, that give a transparent physical interpretation of the latter as due, not to a $L \cdot S$ force, but to the Wigner rotation of the light quark spin. We point out moreover that the selection rule for decay constants of $j = {3 \over 2}$ states, $f_{3/2} = 0$, predicts, assuming the model of factorization, the opposite hierarchy $\Gamma [ \bar{B} \to \bar{D}_{s_{1,2}} ({3 \over 2}) D^{(*)} ] \ll \Gamma [ \bar{B} \to \bar{D}_{s_{0,1}} ({1 \over 2}) D^{(*)} ]$.