I'm sorry but it is not perfectly clear to me what issue you are facing. It sounds like you are trying to make the initial concentration (presumably of moisture) in the wood higher than that of the surrounding air, and you expect a gradual reduction in moisture as it diffuses into the surrounding air. This is of course to be expected. But it sounds like you find that at the beginning of the transient, the calcuklated concentration in the wood is lower than that of the surrouding air, which is physically impossible. I created a small test case to try to duplicate your observation (please see the listing at the end of this reply which you can copy into a file and read into MAPDL with the /INPUT command). My test case seems to work fine... I am so far unable to reproduce the problem. However, I have seen similar non-physical calculated results before.What I saw happens in thermal conduction transient analyses too. Sometimes called thermal undershoot or spurious oscillations, the calculated concentration (or temperature) is seen to oscillate - exhibit 2nd order system behavior. Thermal and diffusion systems are first order and do not physically respond in this way. The calculated oscillations are often caused by suddenly changing boundary conditions and/or very small time steps. Applying loads more gradually and/or using larger initial time steps can help alleviate this. Also, it sometimes helps to use a diagonalized diffusion damping matrix (set KEYOPT(1)=1 on the ET command) or create a mesh without midside nodes (use the MSHMID,2 command prior to creating the mesh so that the elements are linear rather than quadratic). I should say, though, that I've never seen a case of this that is as severe as what your first image implies, so I'm not sure if my suggestions will help or if they are even relevant. Hopefully my small test case (below) will help.