In previous direct measurements of nuclear heating for beryllium induced with DT neutrons, it was pointed out that the calculation withJENDL-3.2 underestimated the measured one by 25%. However, reasons of the discrepancy have not been understood clearly. Recently, we measured the a-particle emission double-differential cross section for beryllium and found that the evaluation of the (9)Be(n.2n+2 alpha) reaction in nuclear data libraries have some problems. We examined KERMA factors for beryllium deduced with three latest nuclear data libraries: JENDL-3.3, ENDF/B-VII.0 and JEFF-3.1. The partial KERMA factors for (9)Be(n,2n + 2 alpha) reaction channel at incident neutron energy of 14.2 MeV deduced from these libraries were compared with a new partial KERMA factor calculated based on our experimental model. The partial KERMA factor from JENDL-3.3 was smaller by 20% than our experiment-based one. The reason of the discrepancy in the previous nuclear heating measurement comes from the smaller partial KERMA factor in JENDL-3.3, which is caused by significant underestimation of higher energy part of the a-particle emission DDX at forward emission angles. (C) 2008 Elsevier B.V. All rights reserved.