Hints, Tips and Solutions

 

Volume 12, Number 2, February 2002

Q. When I simulate my process flow the simulated profile is always lower than the SRP measurements. What could be causing this ?

A. In many simulated process flows the surface oxide is etched completely off after which there may be some diffusion cycles. When ATHENA performs a diffusion step it will model the diffusion of dopants not only into the silicon but also across region boundaries for instance from silicon to the gas above the silicon surface. So if no oxide is present on the surface of the silicon, dopant loss may occur into the gas. To illustrate a 1e14 Boron implant @ 50 KeV is performed into silicon and is annealed at 1150C for 60 minutes. Figure 1 shows the resulting Boron profile when an oxide layer is deposited just before the diffusion and when the silicon surface is exposed. Clearly significant dopant may be lost if no oxide is present. It is important therefore to check in each input deck that a surface oxide is present before any diffusion step is performed.


Figure 1. The Boron profile that results when a surface PAD oxide is
present compared to the result when it is absent. Significant dopant is
lost to the GAS above the silicon surface if no PAD oxide is present.

 

If it is desired to modify this dopant loss the user may adjust the segregation coefficients with the command

IMPURITY I.BORON SILICON /GAS SEG.0=1 SEG.E=0.0 TRN.0=27.9 TRN.E=2.48

 

Q. The structure I am simulating inside ATHENA is very large and the number of mesh points has become excessive. Is there anything I can do to get around this limitation ?

A. A possible solution does exist with the application of the Silvaco tool DevEdit. This tool allows a user to load one ATHENA structure and to then JOIN a second ATHENA structure to it. Therefore if the ATHENA simulation is very large the ATHENA input deck may be broken into two parts. The first may simulate only one half the length of the device whilst the second, which has an identical depth, will simulate the rest. By dividing the simulation domain into two discrete ATHENA input decks it is possible to maintain a dense mesh without hitting problems with the number of nodes. To illustrate, the DeckBuild example mos1ex01_0.str structure was loaded into DevEdit and by use of the "File --> Join" option was then joined to another file (in this case it was also mos1ex01_0.str). When remeshed the final structure is shown in Figure 2.


Figure 2: DEVEDIT was used to combine two ATHENA structures
together to form one single structure which was then remeshed.

 

 

Call for Questions

If you have hints, tips, solutions or questions to contribute, please contact our Applications and Support Department
Phone: (408) 567-1000 Fax: (408) 496-6080
email: support@silvaco.com