Q2IC Technology

As discussed in relation to another question in this examination, Fick's second law dC/dt = D(d^2C/dx^2) admits two standard closed-form analytic solutions under two different, commonly encountered boundary conditions. For the predeposition case, where the wafer surface is held at a constant concentration Cs throughout the diffusion time (an effectively infinite dopant source continuously replenishing the surface), the solution is the complementary error function profile, C(x,t) = Cserfc(x/(2sqrt(Dt))), with total dose Q(t) = (2/sqrt(pi))Cssqrt(Dt) increasing with the square root of time as more total dopant enters the wafer. For the drive-in case, where a fixed, finite total dose Q (deposited during a preceding short predeposition or ion implantation step) diffuses further into the wafer with no additional dopant source, the solution is the Gaussian profile, C(x,t) = (Q/sqrt(piDt))exp(-x^2/(4Dt)), where the surface concentration Cs(t) = Q/sqrt(piDt) now decreases over time as the same fixed total dose spreads to ever-greater depth.

Correction to the Simple Theory

Both of these classical analytic solutions rest on the simplifying assumption that the diffusion coefficient D is constant, independent of the local dopant concentration and position within the wafer; this assumption holds reasonably well at low to moderate doping concentrations, well below the intrinsic carrier concentration ni at the diffusion temperature. However, at high doping concentrations (approaching or exceeding ni at the diffusion temperature, commonly the case for heavily doped source, drain, or emitter regions), the diffusivity becomes strongly concentration-dependent, because the diffusion of most common dopants in silicon is mediated by charged point defects (vacancies or self-interstitials) whose local equilibrium concentration itself depends on the local Fermi level, which in turn depends on the local net dopant concentration; at high concentrations, this coupling causes the effective diffusivity to increase substantially above its low-concentration value, and the resulting experimentally measured profile deviates significantly from the ideal erfc or Gaussian shape, typically showing a much flatter, more box-like profile near the surface followed by a steeper fall-off near the diffusion front, rather than the smooth, gradually decreasing shape predicted by the constant-D solutions. This correction requires either a full numerical solution of Fick's equation with a concentration-dependent D(C) extracted from separate experimental measurements, or the use of tabulated or graphical correction factors developed specifically for each common dopant species, in place of the simple closed-form analytic solutions, whenever accurate profile prediction is required at high doping concentrations.

Range Theory of Ion Implantation

When an energetic ion is implanted into a target such as silicon, it loses energy through a combination of two distinct stopping mechanisms as it travels through the crystal before finally coming to rest: nuclear stopping, in which the incoming ion undergoes elastic collisions with the nuclei of target atoms, transferring energy and momentum in discrete, relatively large increments and causing significant lattice displacement damage along the ion's path (dominant at lower ion energies and for heavier ions), and electronic stopping, in which the ion loses energy through a large number of small, quasi-continuous inelastic interactions with the electron cloud of the target material, analogous to a viscous drag force (dominant at higher ion energies). Because both stopping mechanisms, and particularly nuclear stopping, involve a statistically random sequence of individual collision events, ions of identical initial energy implanted along the same nominal direction do not all come to rest at exactly the same depth, but rather at a range of depths described by a statistical distribution, commonly approximated as Gaussian, characterized by a mean projected range Rp (the average total depth of penetration measured along the original beam direction) and a standard deviation delta-Rp (the projected straggle, describing the spread of the depth distribution around this mean value), along with a lateral straggle describing spread perpendicular to the beam direction, relevant for lateral doping profile control under a mask edge. The resulting as-implanted dopant concentration profile is therefore well-approximated by a Gaussian distribution centered at depth Rp (rather than at the surface, as would be the case for a purely diffused profile), N(x) = Npexp(-(x-Rp)^2/(2delta-Rp^2)), where Np is the peak concentration at the mean projected range, related to the total implanted dose Q through Np = Q/(sqrt(2pi)delta-Rp). Both Rp and delta-Rp depend on the implant energy, the mass of the implanted ion species relative to the target atom mass, and the target material, and are typically obtained from tabulated simulation data (such as from the widely used SRIM/TRIM Monte Carlo simulation software) or from analytic range theory calculations based on the Lindhard-Scharff-Schiott (LSS) stopping power theory, which provides the fundamental physical framework relating ion energy, mass, and target properties to the resulting nuclear and electronic stopping powers and hence to the resulting range and straggle values.

It is also worth further emphasizing that the Lindhard-Scharff-Schiott range theory framework, while foundational, is in modern practice supplemented or entirely replaced by detailed Monte Carlo ion transport simulations (such as the widely used SRIM and TRIM codes), which individually track thousands or millions of simulated ion trajectories through the target material, each undergoing a statistically sampled sequence of nuclear and electronic stopping events, to build up an accurate, non-Gaussian (and hence more physically realistic) implant profile directly from first-principles collision physics rather than relying on the simplified analytic Gaussian approximation, particularly important for accurately predicting channeling effects (anomalously deep penetration along certain crystallographic directions where the ion beam aligns with open channels between atomic planes) that the simple LSS range theory does not capture but that Monte Carlo simulation, given an appropriately detailed crystal structure model, can reproduce.

It is also worth noting that both the corrected (concentration-dependent) diffusion theory and the ion implantation range theory discussed in this question ultimately serve the same overarching purpose in device fabrication: accurately predicting the final, as-processed dopant concentration profile so that the resulting electrical device characteristics can be reliably designed and simulated before committing to an expensive physical fabrication run, meaning modern process technology computer-aided design (TCAD) tools incorporate both the concentration-dependent diffusion corrections and detailed ion implantation range and straggle models (often derived from or validated against Monte Carlo simulation) within a single, unified process simulation framework used throughout the semiconductor industry to predict device behavior directly from a specified process recipe.