Author |
Message |
Abhishek (amotayed)
New member Username: amotayed
Post Number: 1 Registered: 01-2007
| Posted on Sunday, February 04, 2007 - 08:40 pm: | |
Dear Dr. Nelson, I am trying to solve the standard Poisson's equation for an oxide semiconductor interface. I am having some problem in assigning proper boundary conditions at the semiconductor-oxide interface. How to assign the continuity of normal component of D at the interface? I wrote the program but the solution is incorrect. I would appreciate if you could suggest me something or direct me towards any example. Thanks Abhishek
|
Robert G. Nelson (rgnelson)
Moderator Username: rgnelson
Post Number: 757 Registered: 06-2003
| Posted on Monday, February 05, 2007 - 03:26 pm: | |
What platform and version are you running? When I run this problem, I get a floating-point overflow because the iteration is diverging. The problem seems to be your source term. If I replace it with something simple, the solution converges. As to the question of interior boundary conditions, you don't have to do anything. If you re-read the documentation concerning Natural Boundary conditions, you will see that when you write div(eps*grad(u)) the default interior boundary condition is the continuity of the normal component of eps*grad(u), as required by Maxwell's equations.
|
Louis Gerrer (imep)
New member Username: imep
Post Number: 1 Registered: 01-2008
| Posted on Tuesday, January 22, 2008 - 09:22 am: | |
Hello, I'm a new user, solving the same kind of problem in 2D. I've got a problem with calculating the flow of the current density through a line (because of the problem dimension, the surface between oxyde and semiconductor is just a line) ; the command LINE_INTEGRAL does not match so I use a brutal approximation : Is=integral(Jx*ustep(L/20-x))/(L/20) . I'm sure their is a better way to do this ; could you please enligth my darkness ??? imep
|
Robert G. Nelson (rgnelson)
Moderator Username: rgnelson
Post Number: 1048 Registered: 06-2003
| Posted on Tuesday, January 22, 2008 - 02:04 pm: | |
You'll have to be a little more complete in your description. I can't figure out what you are talking about.
|
Louis Gerrer (imep)
New member Username: imep
Post Number: 2 Registered: 01-2008
| Posted on Wednesday, January 23, 2008 - 10:25 am: | |
Thank you for your answer, I've understood our mistake : we though we couldn't use line_integral because we ignored the possibility of naming separatly each segment of a domain!!! this was a really stupid question, sorry for wasting your time ; next time I'll ask something intersting... |