Temperature evolution due to a single...

Log Out | Topics | Search Moderators | Register | Edit Profile

Author Message   Dominic (dodo) New member Username: dodo Post Number: 2 Registered: 09-2005 Posted on Wednesday, October 05, 2005 - 05:38 am:    I am using FlexPDE 5.0 Professional for the study of heat conduction in silicon due to a single laser pulse (308 nm). I am interested in the temperature evolution in the material (silicon) and on the surface (max. surface temperature) taking into account that the optical and physical constants CHANGE with the temperature. The laser spot has Gaussian type distribution and about 18 ns pulse length. The spot size is 1mm. The optical penetration depth is 5nm. I already wrote a program to determine the temperature distribution on titanium by IR laser irradiation (Spot size: 15µm, Pulse length 10µs) à titan_orginal.pde. Titan_orginal Titan_orginal.pde (2.0 k) This works very well and I tried to use this as a model for my new problem. Unfortunately this does not work to good, as my computer calculates for hours, although I did not even include the temperature dependence of the constants. It seems that the program has difficulties to handle the dimensions 5nm (optical penetration) -- 1mm (spot size) as it takes very long to generate the mesh. Maybe the boundaries are also wrong. Si_XeCl Si_XeCl.pde (3.1 k) Could some please help me solve these two problems Thank you very much, Dominic   Robert G. Nelson (rgnelson) Moderator Username: rgnelson Post Number: 467 Registered: 06-2003 Posted on Thursday, October 06, 2005 - 01:02 am:    FlexPDE tries to create cells that are roughly equilateral. With your disparity of R/Z scales of 100000:1, this results in an enormous mesh, probably exceeding your RAM space and sending the problem into virtual memory. I suggest you use the techniques described in the Technical Note "Coordinate Scaling" to apply a scaling of perhaps 100 to the Z coordinate. Be sure to track the effects of the scaling through all parameters and sources.   Dominic (dodo) Junior Member Username: dodo Post Number: 3 Registered: 09-2005 Posted on Friday, October 14, 2005 - 08:03 am:    If I understand the technical note correctly I will have to describe my problem in Cartesian coordinates as the operator EXTRUSION may only be used in an 3D environment? Or is there a possibility to do this in cylinder coordinates?   Robert G. Nelson (rgnelson) Moderator Username: rgnelson Post Number: 470 Registered: 06-2003 Posted on Friday, October 14, 2005 - 07:19 pm:    The Technical Note describes the methods by which coordinates can be scaled. It does it in a 3D model for completeness. Algebra has a wonderful feature: you can change the names of coordinates to apply the scaling in any direction you want. You can then delete the coordinates that do not exist in your problem. The important thing is the principle that is applied. The easiest way to use scaling is to scale a chosen coordinate uniformly throughout the domain. In this mode, you don't have to worry about flux-matching at interfaces. The attached script shows scaling of the Z coordinate in your cylindrical problem. You should take it as an example of how to proceed, not as the final solution to your problem. Si_XeCl1.pde (3.2 k)   Dominic (dodo) Member Username: dodo Post Number: 4 Registered: 09-2005 Posted on Thursday, October 20, 2005 - 10:26 am:    thanks for the help!!!