This 2D counterpart looks at the physics side-on, as a cross-section through the silicon die, and models the charge-collection mechanism itself rather than re-drawing the 3D grid flat. An ion track deposits charge along its path; instead of a single geometric falloff, that charge splits into two physically distinct components exactly as real device physics does: a prompt/drift component collected almost instantly by the struck cell's electric-field funnel, and a slower diffusion component that spreads outward as a genuine two-dimensional Gaussian charge cloud (ambipolar diffusion) and can reach neighboring cells.
Diffusion Green's function: n(r,t) = Q0/(4πDt)·exp(−r²/4Dt)
Charge available to a pad of area A at distance d peaks at t* = d²/(4D):
Q_peak(d) = Q0·A·e⁻¹ / (π·d²) — independent of D, an exact inverse-square law
Struck-cell charge: Q_struck = f_prompt·Q0 + Q_peak(d_min)
Upset condition: Q_cell > Q_crit → bit flips
The charge-vs-distance chart on the right plots this closed-form Q(d) live against the current LET and Qcrit, so you can see directly where the crossing point sits — the same quantity the strike engine tests every time a particle lands.
- LET — sets the total charge Q0 deposited along the ion's track (same silicon rule of thumb as the 3D twin: ≈10.3 fC per MeV·cm²/mg per μm of path).
- Qcrit — the memory cell's critical charge; smaller, lower-voltage cells hold less and upset more easily.
- Shielding — aluminum attenuates the strike probability exponentially with thickness, exactly as before.
- At high LET the diffusion tail can exceed Qcrit at more than one neighboring cell from a single strike — a genuine multi-cell upset, driven here by the 1/d² diffusion law rather than a fixed decay length.
Real-world relevance: this drift-then-diffusion picture (sometimes called the "funneling" model) is the actual mechanism radiation-effects engineers simulate with device-level TCAD tools to predict a chip's real-world upset rate in orbit.