Diffused Base Transistor Base Drive-in

# Diffused-Base Transistor Base Drive-In: Limited-Source Redistribution and Junction-Depth Control

Base drive-in is the source-free thermal step that redistributes the dopant inventory loaded during base predeposition into a deeper, broader base profile. Ideally, no additional dopant enters the wafer. The integrated dose remains approximately fixed while the surface concentration falls, the profile spreads, and the collector-side base junction moves inward. This is the step that converts a shallow dopant inventory into the depth margin needed for a later, shallower emitter diffusion.

The controlled quantity is not furnace time by itself. Junction depth follows the accumulated diffusivity over the complete wafer temperature history, and the final transistor base width depends on the difference between this driven base junction and the emitter junction formed afterward. A useful process therefore preserves the predeposition dose, controls the full ramp–dwell–cool thermal budget, limits surface loss and contamination, and verifies a depth-resolved profile before treating the wafer as ready for emitter formation.

## Limited-source diffusion model

For an ideal instantaneous plane dose $Q$ diffusing into a semi-infinite solid with no further source, the one-dimensional solution is

When temperature changes with time, the natural process variable is the accumulated diffusivity

so the constant-$D$ product $Dt$ is replaced by $\Theta$. The profile becomes broader as $\Theta$ increases, while its peak $Q/\sqrt{\pi\Theta}$ decreases. The area under the profile remains $Q$ only if dopant is neither added nor lost and remains represented by the same diffusivity model.

For a uniform collector background $N_C$, an ideal junction satisfies $C(x_j,\Theta)=N_C$. Rearrangement gives

This relation contains an important nuance: junction depth is not simply $2\sqrt{\Theta}$. The peak concentration also falls as the profile spreads, so the logarithmic threshold term changes. If the peak approaches the background, a meaningful compensated junction disappears. Dose, background doping, activation, and thermal budget jointly determine the crossing.

The plane-source Gaussian is a useful limiting model, not the exact starting condition after a finite predeposition. A rigorous prediction propagates the measured or calibrated predeposition profile through the later temperature history. Concentration-dependent diffusivity, point-defect coupling, electric-field effects, clustering, precipitation, segregation, evaporation, oxide/interface exchange, and incomplete activation can change both shape and electrical behavior.

## Full thermal history matters

The wafer accumulates diffusion during heat-up, stabilization, dwell, transfer, and cooldown. Because $D(T)$ is strongly Arrhenius-dependent, the hottest part of a ramp can contribute appreciably even when it is not counted as recipe dwell. Two furnaces with equal nominal temperature and time can produce different $\Theta$ because of ramp rate, load recovery, overshoot, wafer position, thermocouple calibration, and cooling practice.

The process record should retain a wafer-relevant temperature trajectory or a qualified proxy, not only the controller setpoint. Zone sensors observe the furnace, not every wafer. Monitor wafers, calibrated thermal mapping, load-position studies, and profile measurements establish the relationship between equipment state and material result.

Ambient affects more than heat transfer. An inert ambient may reduce oxidation but does not guarantee zero surface loss. An oxidizing drive-in can grow a protective oxide and alter surface boundary conditions, while oxidation-generated point defects may change dopant diffusion depending on species and material system. Chlorine-bearing cleans or ambients, moisture, residual source material, tube wall state, and prior runs can influence contamination and redistribution. “Source-free” means no intentional continuing dopant supply; it does not mean the surface and ambient are physically irrelevant.

## Profile handoff to emitter diffusion

Drive-in creates the collector-side base junction. The later emitter predeposition and drive-in introduce the opposite dopant from the same surface to a shallower depth. The electrical base is the region where the base dopant remains dominant between the emitter-base and base-collector compensation points.

In a simplified geometry,

but the electrically relevant neutral base width differs from the metallurgical separation because depletion regions extend into the base under bias. Curved junctions, lateral diffusion, surface fields, compensation gradients, lifetime, mobility, and contact geometry also influence device behavior. A depth pair alone cannot guarantee gain, transit time, punch-through margin, leakage, or breakdown.

Base drive-in must therefore leave margin for all subsequent thermal steps. Emitter diffusion, oxidation, contact anneals, and other furnace operations continue moving the base profile. The target at this stage is a precursor profile whose predicted final state meets the device window—not necessarily the final junction depth measured immediately after drive-in.

## Metrology and release criteria

No single measurement fully describes the driven base. Sheet resistance integrates conductivity and is sensitive to activation and mobility, but different profiles can share the same value. Junction staining or bevel methods can estimate a compensation depth but are destructive and technique-dependent. Spreading resistance, electrochemical capacitance-voltage, SIMS, and electrical structures provide different combinations of chemical concentration, carrier concentration, depth resolution, and calibration.

A practical control strategy uses fast blanket-wafer or monitor measurements anchored periodically by depth-resolved reference methods. The metrology recipe must specify surface preparation, bevel or crater geometry if used, calibration standards, spatial sampling, edge exclusion, and the assumed background profile. Gauge repeatability and tool matching should be separated from real furnace variation.

Release criteria should include more than a mean depth:

• retained dose or a qualified proxy for it;
• surface and peak concentration within the activation/solubility window;
• junction depth and profile slope against the collector background;
• within-wafer, wafer-to-wafer, boat-slot, run, and furnace matching;
• absence of surface haze, particles, slip, oxide defects, contamination, or abnormal film growth;
• remaining thermal-budget margin for emitter diffusion and later anneals;
• demonstrated correlation to final base width and electrical distributions.

## Failure signatures

Corrections must follow cause. A stable mean-depth drift with a healthy furnace and verified incoming dose may support a bounded thermal adjustment. A zone fault, thermocouple drift, contamination event, dose-loss mechanism, or broken metrology correlation requires hold and maintenance or process investigation. Adjusting dwell to compensate every profile anomaly converts diagnosable faults into hidden device tails.

## Qualification and manufacturing control

Qualification crosses incoming dose/profile, collector background, peak temperature, ramp, dwell, cooldown, ambient, load size, boat position, furnace identity, oxide condition, and relevant prior tube history. Replication and randomization separate recipe effects from temporal drift. Profile simulation should be calibrated on measured distributions, not one nominal wafer, and should carry uncertainty into the predicted emitter/base-collector separation.

Production genealogy links the predeposition lot and measurement, drive-in furnace and recipe version, full trace data, wafer slot, oxide or cap state, profile metrology, subsequent emitter operations, and final electrical results. Control charts should preserve spatial and hierarchical structure rather than pooling every wafer and slot into a single standard deviation.

Run-to-run control belongs behind equipment-health checks, profile-model validity limits, actuator bounds, and rollback rules. Virtual estimates of $\Theta$ or $x_j$ can improve sampling decisions, but they do not replace physical anchors when thermocouples drift, source carryover occurs, or diffusion physics changes with concentration and interface condition.

The central principle is that drive-in is a profile transformation, not a depth-only anneal. It spends thermal budget to exchange peak concentration for penetration while trying to retain dose and material quality. The correct output is a verified precursor base profile with enough downstream margin to form the emitter and still meet gain, speed, leakage, and breakdown requirements.