Author |
Message |
Shu-Ting Cheng (coolmouse)
Member Username: coolmouse
Post Number: 6 Registered: 03-2008
| Posted on Friday, May 02, 2008 - 08:59 am: | |
Hello, I have problem with moving boundary problem. The situation is there is one layer of SiO2 over one layer of Si,and on the top of SiO2 is the costant flux of O2 which will build more SiO2 so that the boundary of SiO2 will move foward but Si backward. I want the bondary move when the interface concentration is high enough(like conc=3),so I use Vb=IF conc>eqconc THEN 3 else 0 for the velocity of Ym,but it didn't move anyway. Could you please help me if my scrpt is something wrong?
|
Robert G. Nelson (rgnelson)
Moderator Username: rgnelson
Post Number: 1115 Registered: 06-2003
| Posted on Friday, May 02, 2008 - 03:12 pm: | |
1. You have the wrong sign on your Conc boundary condition, so the values go negative. The switch point is therefore never reached. (Natural BC definitions correspond in sign to the second order terms when moved to the left side of the equation.) 2. You have an instantaneous jump in the velocity BC. When Conc=0.009999999999, the boundary is stationary. When Conc is 0.01000000001, the velocity jumps instantaneously to 3. Small oscillations in the solution approximation will cause the motion to start and stop. This is not an acceptable definition. (See "Discontinuities" in the Help Index). Use a velocity definition that changes smoothly through the switch point. Vb = IF conc>eqconc THEN 3*(conc-eqconc)^2 ELSE 0 will do the job at early time, but will eventually smash the interior boundary into the stationary top. 3. You have specified a boundary motion in the interior, but a zero velocity at the boundary. This is another discontinuity, requiring an infinite shear at the side walls. Fix the sidewall BCs to allow motion. nobc(Ym) and Natural(Vm)=0. See attached
|
Shu-Ting Cheng (coolmouse)
Member Username: coolmouse
Post Number: 7 Registered: 03-2008
| Posted on Monday, May 05, 2008 - 02:48 am: | |
Thank you very much!!But I still have some questions about your revise... (1) Did you mean that the sidewall must strech to fit the interior motion? (2)What's the difference betwen VELOCITY(Ym)=0 VALUE(Vm)=0 and NOBC(Ym) NATURAL(Vm)=0 ? Is NOBC the space derivaive? Does it mean d(Ym)/dy ,the strain, equal to 0 ? Then what does d(Vm)/dy mean? (3) What's the difference with for t=1 by 0.01 to 1 and for cycle=1 ? You didn't define the "cycle". |
Shu-Ting Cheng (coolmouse)
Member Username: coolmouse
Post Number: 8 Registered: 03-2008
| Posted on Monday, May 05, 2008 - 03:16 am: | |
I have found out what cycle means. Sorry for this stupid question.... |
Robert G. Nelson (rgnelson)
Moderator Username: rgnelson
Post Number: 1118 Registered: 06-2003
| Posted on Tuesday, May 06, 2008 - 02:20 pm: | |
1) Yes. You can't have a zero velocity on the boundary and a fixed nonzero velocity immediately adjacent to the boundary on the interior. This requires an infinite shear of the material at the boundary. I presume the molecules on the boundary will behave the same as the molecules immediately adjacent to them. If you want to hold the boundary velocities to zero, you cannot assign a fixed velocity to the interior material. 2) VELOCITY sets a time derivative. VALUE sets a value. Ym and Vm are separate variables. Each must be assigned a boundary condition. The VALUE of Vm is equal to the time derivative (i.e. VELOCITY) of Ym. NOBC means "assign NO boundary condition". NATURAL(Vm) means the outward normal component of grad(Vm), because your equation is div(grad(Vm))=0 (see Natural in the Help Index). Setting it to zero imposes a reflective boundary condition. |