Introduction

Alamode (A LAyered Model Development Environment) is a program which solves systems of partial differential equations (PDEs) that are described using a dial-an-operator paradigm. Its purpose is to allow quick prototyping of PDE-based models for thermal diffusion in semiconductors, and also provides a framework for evaluating new numerical techniques to discretize and solve these equations. The emphasis in the user interface is to provide for complete specification of the equations and boundary conditions comprising the model, while eventually providing additional layers of access to the discretization and numerical techniques used to solve the equations.

Unlike SUPREM, which is designed to be a process simulator, Alamode is only a PDE solver. This difference is reflected strongly in the user interface. Whereas SUPREM has reasonable defaults for models and parameters, Alamode requires the complete specification of all terms, parameters, and boundary/interface constraints for a model as applicable to the specific device structure being simulated. While this may initially seem excessive, this enables more flexibility in model description than SUPREM allows and ensures that there are no hidden features of any model.

Everything described in this manual is subject to change, and backward compatibility to prior versions will be attempted, but is not guaranteed. Currently available features will probably not be deleted, but they may be significantly changed. User feedback is important to help mold future changes. Send suggestions to alamode@gloworm.Stanford.EDU.

This manual corresponds to release 97.06.18 of Alamode. The most recent version of the documentation is available on the WWW from http://WWW-TCAD.Stanford.EDU/tcad/programs/alamode.html.

The first widely distributed release was 96.11.07.

Between 96.08.19 and 97.08.19:

• PMM support.
• KPM support.
• Fixes to the profile function and command.

Between 96.11.07 and 97.06.18:

• Added profile function and command.
• Added box function.
• Added erfc function.
• Added sqrt function.
• Added fermi_nnp function.
• Added min function.
• Added max function.
• Added arcsinh function.
• TCL variables should now be accepted in all slots that used to require fields or functions.
• Nodal weighting for triangle elements is based on the intersection of the voronoi cell with the element instead of a row-sum lumped mass matrix. This seems to be a better estimate of the area for nodal lumping and preserves symmetry in meshes derived from a tessellation of rectangles.
• Implemented a primitive estimate for the initial time step (-estimateFirstTimeStep).
• Added ``field -linkVariableAtPoint'' which links a TCL variable to a field value nearest a given point. This is useful for monitoring concentrations at a specific point.
• Replaced the quasiFermiEField operator with quasiNeutralEField. In addition to being incorrect from a model perspective, the quasiFermiEField usually had slow nonlinear convergence. Another alternative for electric field enhancements is to add a Poisson equation to your model to solve for and use the diffusion operator with the diffusivity concentration product in the diffusivity slot (i.e. ).
• Changed the default diffusion operator to a more efficient geometric diffusion operator. This may produce slightly different results than older releases, but they are just as accurate. The old diffusion operator is still available as ``cdiffusion''.
• Timestepping error control, update norm convergence, and threshold detection/reporting criteria and are now specified for each field instead of for each model.
• Restructured the code to support nodal assembly and eliminate duplicated evaluation during assembly. The result has been speedups ranging from 2 to 20.
• Error-driven timestep estimation has been added to the generalized trapezoid family of integration algorithms. This has not been well-tested.