Physics and models
Established physics, with its limits on the page.
Greater Today TCAD solves 2D drift-diffusion with the open-source DEVSIM engine. Here is what it models, what it measures, and — just as important — what it does not model yet.
Devices
What you can simulate today.
| Template | Device | Notes |
|---|---|---|
| gan_hemt | AlGaN/GaN HEMT | Polarisation-induced 2DEG at the AlGaN/GaN interface; GaN uses an electron quasi-Fermi-potential formulation that stays robust in high-voltage off-state. |
| si_nmos | Silicon NMOS | Electron and hole densities solved directly. |
| custom | Your own 2D device | Rectangular regions, boundary contacts, uniform and Gaussian doping, and DC sweeps written as an input deck. |
Analyses
Sweeps
| op | Operating point at a fixed bias |
| idvg | Transfer characteristic, one curve per drain bias |
| idvd | Output characteristic |
| offstate | Off-state field, used for the breakdown estimate |
Metrics
Extracted figures
| vth_lin · vth_cc | Threshold voltage, linear and constant-current |
| gm_max | Peak transconductance |
| ss | Subthreshold swing |
| ion · ioff · on_off_ratio | On and off current and their ratio |
| ron | On-resistance |
| ns_gate · ns_access | 2DEG sheet density under the gate and in the access region |
| vbr_est | Breakdown estimate (see limits below) |
The solver
Drift-diffusion, solved properly.
Poisson's equation and the electron and hole continuity equations are solved together on a 2D mesh with Newton's method, using the open-source DEVSIM engine.
- Heterojunctions via a band-corrected Scharfetter–Gummel discretisation
- Caughey–Thomas velocity saturation on a constant low-field mobility
- Shockley–Read–Hall recombination
- Deep acceptor traps for semi-insulating GaN buffers
- Ohmic, Schottky and MOS gate contacts
- AlGaN/GaN polarisation as a fixed sheet charge,
pol(x)
Materials
Eleven built-in materials, and your own.
All material parameters live in one library. Copy any entry to make your own; each run stores a snapshot of the materials it used.
| Material | Type | Typical use |
|---|---|---|
| Si | Semiconductor | MOSFET channels and substrates |
| Ge | Semiconductor | High-mobility channels |
| GaAs | Semiconductor | III–V devices |
| 4H-SiC | Semiconductor | Wide-bandgap power devices |
| GaN | Semiconductor | HEMT channels and buffers |
| AlN | Semiconductor | Nucleation and spacer layers |
| AlGaN | Alloy | HEMT barriers, with mole fraction comp= |
| SiO₂ | Insulator | Gate oxides |
| Si₃N₄ | Insulator | Passivation and gate dielectrics |
| Al₂O₃ | Insulator | High-k gate dielectrics, MIS-HEMTs |
| HfO₂ | Insulator | High-k gate dielectrics |
A custom semiconductor needs permittivity, electron affinity, bandgap, effective densities of states, electron mobility and saturation velocity; other transport properties have documented defaults. Design searches use Bayesian optimisation (TPE).
Known model limits
What the models don't do yet.
We publish these because a device engineer needs them to judge any number the tool produces.
- Classical 2DEG. No Schrödinger–Poisson quantisation, so the carrier peak sits at the interface.
- No self-heating or contact resistance. This is why the simulated HEMT peak current, around 2.2 A/mm, is above typical measured values. It is the first item for calibration.
- Breakdown is an estimate.
vbr_estis the drain bias at which the electron impact-ionisation integral along the channel reaches 1, post-processed from the off-state field. Avalanche carriers are not fed back, and field plates are not modelled yet. - Simple mobility. No surface-roughness or doping-dependent mobility yet.
- Compensated HEMT surface. Polarisation charge at the AlGaN surface is assumed compensated by surface donors.
- Static buffer traps. Traps follow the conduction band only — no transients or current collapse.
- An open convergence issue. The full NMOS example sweep fails to converge near −0.4 V. Completed points are kept and the run is marked partial.
Working on GaN or silicon devices?
We're opening the private beta to a small number of labs and teams. Tell us what you simulate.