You may try to perform an isochron annealing of the first layer before depositing the second one. Pores, dislocations, point defects, and impurities create a non-uniform distribution of electrical potential and surface stress across the surface of the first layer, which changes eventually the kinetics of the second layer growth.

Your second layer thickness is near 2000 Angstrom, that is near 400 atomic layers stock one over another. With non-uniform electric potential and the surface stress distribution, local vertical and lateral growth velocities across the surface can be visibly different. It's just my God's feeling that a reasonable isochron annealing of the first layer might help.

It will be helpful to have more details. Is this process done on a small chip or in large wafers? How much is your tolerance for misalignment and what degree of misalignment do you get with your experiments?

It will be ideal to include local alignment marks (X, Y and rotation) during your first e-beam step to achieve better alignment in the second step. You can look into this paper if interested ( Biomicrofluidics 8, 016501 (2014); https://doi.org/10.1063/1.4861435).

Hope it helps!

I faced similar issues but after some trials I had optimized and never faced it again..so it is a matter of practice and optimization..do you have sufficiently large markers at the magnification you are doing 3 point alignment? I realized we cannot do 3 point alignment at the same magnification as the writing (we used to do 3 point alignment @ 3-6KX while write @ 80K that's why we needed lager markers)..also they are many other small issues you have to keep in mind ..