Silicon Gate 1967 Pattern Polysilicon Gate Channel Length

# Step 3 — Pattern the Polysilicon Gate, Defining the Channel by the Gate’s Own Edge: The Single Lithographic Step That Sets Channel Length

## 1. Why Channel Length Is Now Set by a Single Feature Width, Not a Mask Alignment

In the 1962 metal-gate process, channel length was the distance between two independently diffused p-n junctions, and the gate had to be patterned afterward with deliberate overlap margin to ensure it covered that channel under worst-case mask misalignment. This step eliminates that entire geometric chain: after depositing blanket polysilicon over the thin gate oxide, a single photolithographic mask patterns the polysilicon into isolated gate stripes, and the etched width of that polysilicon stripe directly defines the future channel length of the transistor. The chemical etchant cuts through the polycrystalline silicon layer with high selectivity, stopping cleanly on the underlying silicon dioxide ($SiO_2$) without pitting the substrate. Wherever the polysilicon remains, it shields the underlying silicon crystal; wherever it is removed, the bare oxide is exposed to receive dopants in the subsequent diffusion furnace.

$$L_{\text{channel}} = W_{\text{gate}} - 2 x_{\text{lateral}} = (W_{\text{mask}} - 2 \Delta x_{\text{etch}}) - 2 x_{\text{lateral}}$$

where $W_{\text{mask}}$ is the drawn width of the gate line on the photomask, $\Delta x_{\text{etch}}$ is the controlled lateral undercut from the isotropic or semi-anisotropic silicon etch, and $x_{\text{lateral}}$ is the small sideways diffusion of dopant under the gate during the subsequent anneal. Crucially, there is no mask-alignment term $\Delta_{\text{align}}$ in this equation. In 1962, the drawn gate width had to satisfy $W_{\text{gate,1962}} \ge L_{\text{diff}} + 2 \Delta_{\text{align}}$ to guarantee that mechanical alignment error between two separate masks did not leave the channel uncovered; here, the gate's own physical edge determines where the channel starts and ends, collapsing channel definition into a single lithographic event.

Patterning the Polysilicon Gate: High Selectivity Over Thin Oxide chemical etching dissolves unmasked polysilicon and stops dead on gate dielectric PHOTORESIST MASK OVER BLANKET POLY Single-crystal silicon wafer Continuous Gate Oxide (SiO2 ~100 nm) Blanket Polysilicon Layer (~500 nm) Patterned Photoresist Single mask defines gate width Wmask Ready for chemical etching ETCHED POLYSILICON GATE ELECTRODE Single-crystal silicon wafer Source window Drain window Polysilicon Gate Wgate → Defines Lchannel High Etch Selectivity (Poly-Si vs SiO2 > 25:1) Oxide preserved; gate edges set future dopant boundary Lchannel = Wgate - 2xlateral = (Wmask - 2Δxetch) - 2xlateral channel length is governed strictly by gate lithography and etch, with zero inter-mask alignment error

## 2. Real Diagram: Single-Mask Channel Definition vs. Two-Mask Misalignment

The fundamental architectural difference between 1962 and 1967 is captured in how each process establishes the physical boundary of the conducting channel.

Two Mask Alignments vs. One Self-Determining Gate Line why 1962 had to oversize the gate, and why 1967 Step 3 scales directly with lithography 1962: CHANNEL SET BY DIFFUSION SPACING Source Drain Ldiff Aluminum Gate Gate Must Exceed Ldiff by 2Δalign Independent masking forces oversized gate 1967 STEP 3: CHANNEL SET BY GATE STRIPE Polysilicon Gate Future Channel Wgate ≡ Channel Boundary Zero Overlap Margin Needed Gate edges will block the next diffusion step 1962: Alignment tolerance Δalign imposed a floor on parasitic gate overlap capacitance Cgd 1967 Step 3: Gate width directly sets channel length — scaling lithography directly shrinks channel length

## 3. Resolving the Fundamental Geometric Constraint of 1962

In 1962's metal-gate MOSFET series, scaling channel length was constrained not by how narrow a line the optical lithography could resolve, but by how accurately two different masking layers could be registered to one another. Even if a photolithographic tool could print a 3 µm line, the fab could not build a 3 µm channel: if mask alignment uncertainty $\Delta_{\text{align}}$ was 2 µm, the drawn channel and gate had to incorporate safety margins to prevent catastrophic misalignment. If the aluminum gate shifted slightly to the left, it failed to bridge the gap to the drain, creating an open circuit; if it was drawn large enough to guarantee overlap, the resulting parasitic capacitance destroyed switching frequency.

Step 3 resolves that dilemma permanently. By patterning the polysilicon gate *before* the source and drain exist, the physical boundaries of the channel become identical to the physical boundaries of the etched polysilicon stripe.

The consequences for semiconductor scaling are profound:
1. Direct Scaling: Channel length $L$ now scales directly with lithographic feature resolution ($W_{\text{gate}}$), rather than being throttled by tool alignment precision.
2. Selective Chemical Etch: The selective chemical etch carves away polysilicon while leaving the 100 nm gate dielectric intact, preserving surface cleanliness for the subsequent dopant introduction.
3. Automatic Shadowing: The patterned polysilicon gate is now positioned to act as an impermeable physical barrier against incoming dopant flux, setting the stage for the self-aligned diffusion in Step 4.

Take silicon gate 1967 pattern polysilicon gate channel length further

Ask the copilot about this term, or have our engineers assess it against your process.

Ask CFSGPT about this → Request a technical assessment → Open the CFS app