Diffused Base Transistor Base Predeposition

# Diffused-Base Transistor Base Predeposition: Constant-Source Dose Control Before Drive-In

Base predeposition is the dopant-loading step that gives a diffused-base transistor process a measured inventory of base dopant before that inventory is redistributed deeper by drive-in. The wafer is exposed to a controlled dopant source at elevated temperature. While the source remains active, the near-surface concentration is maintained near a chemistry- and temperature-dependent boundary value, and dopant diffuses into the semiconductor. The result is a shallow, high-concentration profile with a total dose that can be monitored and then carried into a separate source-free drive-in.

Predeposition does not independently “set dose without affecting depth.” During the step, time and temperature affect both dose and penetration. Its manufacturing advantage is sequential control: after the source is removed, the integrated dopant inventory is approximately conserved during an ideal drive-in while its profile broadens and its peak falls. The process engineer can therefore qualify dopant loading first and redistribute it second, rather than asking one furnace exposure to satisfy surface concentration, dose, junction depth, and profile shape simultaneously.

## Constant-source diffusion model

For a semi-infinite solid initially free of the introduced dopant, a fixed surface concentration $C_s$ and constant diffusivity $D$ give the classical solution

The integrated dose per unit area is

These relations expose the control structure. Surface concentration is set by equilibrium and delivery chemistry only while the boundary remains established. Dose grows with the square root of diffusivity and time under the ideal assumptions. The metallurgical junction during predeposition is the depth at which the introduced base dopant compensates the collector background, not the point where its concentration becomes zero.

Diffusivity is strongly temperature-dependent and is commonly represented over a qualified range by

Consequently, a small furnace-temperature error can move dose and penetration more than an equal fractional timing error. Ramp and cooldown also contribute thermal budget. The wafer experiences an integral over $D[T(t)]$, not a rectangular recipe defined only by displayed temperature and dwell.

The analytic solution is a model, not an automatic process truth. Real behavior can include concentration-dependent diffusion, clustering, precipitation, electrically inactive dopant, segregation at an oxide or interface, surface reaction limits, finite source capacity, gas-phase depletion, lateral diffusion, crystal defects, and moving boundaries. A source may fail to establish constant $C_s$ or may produce a deposited glass whose later behavior differs from direct in-diffusion. Models must be calibrated to the material, dopant species, orientation, ambient, and concentration range actually used.

## What the furnace must control

A base-predeposition operation couples source preparation, transport, surface chemistry, and solid-state diffusion. The controlled chain is:

1. establish a clean, reproducible semiconductor surface or patterned diffusion window;
2. stabilize furnace temperature and carrier/reactant flows before product exposure;
3. introduce the dopant source without a transient overshoot or depleted front;
4. maintain uniform source activity and temperature across the load;
5. terminate the source exposure reproducibly;
6. purge and cool or transfer without adding an uncontrolled dose;
7. remove or condition any source glass and measure the loaded wafer before drive-in.

Gas residence time and boat loading can generate front-to-back signatures. Furnace-tube wall history can store and release dopant. Source temperature, bubbler state, vapor delivery, carrier flow, pressure, wafer spacing, and exhaust conductance affect boundary conditions. Temperature gradients change $D$ exponentially. Surface oxide, contamination, damage, orientation, and native surface chemistry affect incorporation. A stable furnace setpoint does not prove a stable dose.

Open-tube, closed-tube, deposited-source, solid-source, liquid-source, and gaseous-source arrangements impose different failure modes. The engineering abstraction is the same: prove the surface boundary and its uniformity, then prove the integrated electrically active dose. Historical recipes and modern diffusion systems should not be interchanged by source name alone.

## Dose, activity, and sheet resistance are different quantities

Chemical dose counts dopant atoms per unit area. Electrically active dose counts atoms occupying sites and contributing carriers under the measurement condition. Sheet resistance weights the conductivity profile,

or the corresponding hole expression. Mobility depends on concentration, temperature, compensation, and scattering, so sheet resistance is not simply the inverse of total chemical dose. Two profiles can have similar $R_s$ but different surface concentration, junction depth, activation, or high-concentration tails.

A four-point probe on a blanket wafer or monitor structure is valuable because it is fast and nondestructive, but it must be correlated with the intended profile. Spreading-resistance profiling, electrochemical capacitance-voltage profiling, SIMS, junction staining, bevel methods, or electrical test structures may be used depending on era and required information. Each has different spatial resolution, calibration, destructiveness, and sensitivity to chemical versus active dopant.

For a transistor base, the final quantity of interest is not the predeposition sheet resistance in isolation. It is the post-drive-in base profile, its compensation by the later emitter diffusion, the residual active base width, junction curvature, surface leakage, and device gain/breakdown distribution. The predeposition monitor is an intermediate predictor whose correlation must survive the remaining thermal history.

## Failure modes and discriminating evidence

The response should follow the mechanism. A source-delivery transient is corrected in delivery or conditioning. A furnace-zone fault is maintained and recalibrated. A surface-preparation signature belongs upstream. Adjusting time to make the lot-average sheet resistance look correct can hide a boat gradient and worsen the tails that govern device yield.

## Qualification and control plan

A robust qualification crosses source condition, temperature, time, load size, wafer slot, surface preparation, and relevant material lots. Temperature and time splits should be interpreted through accumulated thermal budget, not as independent linear knobs. Replicate center points and randomize experiments where practical so furnace drift does not masquerade as a factor effect.

The control plan records source lot and condition, tube and boat identity, prior conditioning, recipe version, gas/pressure traces, zone temperatures, ramp and cooldown, wafer slot, surface state, monitor-wafer locations, metrology recipe, and downstream drive-in genealogy. Control charts should distinguish within-wafer, wafer-to-wafer, slot-to-slot, run-to-run, and tube-to-tube terms.

Release limits require both gauge capability and downstream correlation. A narrow sheet-resistance distribution is not sufficient if its relationship to final base width drifts with source chemistry or activation. Reference wafers and destructive profile checks anchor the fast monitor. Measurement-system analysis should include probe condition, edge exclusion, wafer temperature, film removal or surface preparation, pattern effects, and reference calibration.

Automated correction should be bounded and slow relative to fault detection. Run-to-run adjustment may compensate a stable, identified drift after source and furnace health are established. It should not normalize heater failure, source exhaustion, contamination, tube-memory transients, or a broken metrology correlation. Hold rules and rollback criteria belong in the recipe-control design.

The core manufacturing insight is that predeposition creates a measured dopant inventory, not a finished base junction. Its value comes from handing a known, traceable dose to drive-in. Art-level control therefore connects source physics and furnace history to chemical dose, electrical activity, profile shape, and ultimately the base-width distribution after the later emitter diffusion.