# Strain propagation from surface to bulk in COMSOL

I am trying to simulate strain propagation from the surface into the bulk. I have a rectangular semiconductor block (~2 μm thick) on top of which metal gates (~25 nm thick) are deposited as seen in the picture below of xz-plane crosssection (zoomed near the gates). The x-axis is from left to right, the z-axis is from bottom to top and the y-axis is going into the picture.

I am fixing the bottom boundary with a constraint, other surfaces are free and all initial values for all domains are zero. Using Linear Elastic Material > Initial Stress and Strain (applied in the Metal Gate domains) I am putting a strain of $$\epsilon_{xx} = 10^{-2}$$ and $$\epsilon_{yy} = 10^{-2}$$. After a stationary study, I have plotted strain $$\epsilon_{xx}$$ in the same picture below (red is positive and blue is negative).

According to strain propagation, the positive surface strain under the contacts (and negative surface strain under the gaps) are supposed to propagate all the way until the bottom (with an exponential decay). However, in my simulation, the strain reverses at a certain depth. Even if I change my initial strain from $$10^{-2}$$ to $$10^{-3}$$, this inversion still occurs at that depth. This is not supposed to happen. • Do you have a reference that states that it would decay exponentially? – nicoguaro Jul 13 at 13:21
• Also, I don't think that the option you used is supposed to use for imposing boundary conditions and I wouldn't trust the results that much. As a final comment, the mesh does not seem to be fine enough to analyze strains and stresses. – nicoguaro Jul 13 at 13:22
• Please provide more details on your analysis: as far as I understand "initial strain" condition can be imposed on a (sub)-domain and not on an "interface", which I assume is the boundary between two subdomains. Moreover I would suggest to check the effect of your kinematic boundary condition (constraint) on the results. Maybe you should relax the fixed condition ($u_X = u_Y = u_Z = 0$) to a weaker one ($u_Z = 0$). Moreover is the material isotropic or not? Do results change with the Poisson ratio $\nu$? – Stefano M Jul 14 at 17:09
• Hi, Stefano, the initial strain is applied in the metal gate domains. Sorry about the mistake. – duncanidaho Jul 15 at 17:12
• Changing the value from $10^{-2}$ to $10^{-3}$ should give the same results, except for the magnitude since it is a linear analysis. – nicoguaro Jul 15 at 21:18