Silicon Gate 1967 Polysilicon Gate Resistance RC Delay

# Step 6 — Confront a New Problem: Polysilicon’s Own Resistance Slows the Gate Signal Itself: The Distributed RC Delay of a Semiconductor Gate

## 1. The Cost of Replacing a True Metal with a Doped Semiconductor

While replacing aluminum with polysilicon eliminated the catastrophic overlap capacitance that crippled the 1962 metal-gate process, it immediately introduced a brand-new physical vulnerability: electrical resistance. Aluminum is a true metal with a bulk resistivity of $\rho_{\text{Al}} \approx 2.7\ \mu\Omega\cdot\text{cm}$, giving a standard half-micron film an almost negligible sheet resistance of $R_{s,\text{Al}} \approx 0.05\ \Omega/\Box$—an essentially perfect equipotential conductor. Polycrystalline silicon, by contrast, is a semiconductor whose crystal structure is fractured into microscopic grains separated by disordered grain boundaries that trap free charge carriers and scatter conducting electrons. Even when doped during the source/drain furnace cycle, polysilicon exhibits a sheet resistance of $R_{s,\text{poly}} \approx 20\text{--}50\ \Omega/\Box$—between $400\times$ and $1000\times$ higher than aluminum. Because the gate electrode sits directly on top of the thin dielectric layer to form a distributed capacitor, this high sheet resistance transforms the gate from a fast equipotential node into a distributed RC transmission line that delays and distorts the incoming switching signal.

$$\tau_{\text{distributed}} = \frac{1}{2} R_{\text{line}} C_{\text{line}} = \frac{1}{2} \left(R_s \frac{L_{\text{wire}}}{W_{\text{gate}}}\right) \left(C_{\text{ox}} W_{\text{gate}} L_{\text{wire}}\right) = \frac{1}{2} R_s C_{\text{ox}} L_{\text{wire}}^2$$

where $R_s$ is the sheet resistance of the gate layer, $C_{\text{ox}} = \varepsilon_{\text{ox}} / t_{\text{ox}}$ is the specific gate oxide capacitance, and $L_{\text{wire}}$ is the physical length of the gate run. Notice that the signal propagation delay $\tau_{\text{distributed}}$ scales with the square of the gate line length ($L_{\text{wire}}^2$). For a single, compact transistor where $L_{\text{wire}}$ is only a few microns, this distributed RC delay is tiny (picoseconds), easily overwhelmed by the massive speed dividend won by removing Miller capacitance in Step 5. But for long gates running across wide multi-finger output stages or interconnect runs spanning a memory array, the high resistance of polysilicon causes severe signal attenuation, where the far end of the transistor turns on long after the near end has switched.

Distributed RC Delay Along a Polysilicon Gate Line distributed resistance and oxide capacitance create signal delay scaling with L² NEAR END (Driven Contact) Sharp input step (trise ≈ 0) DISTRIBUTED RC TRANSMISSION LINE ΔR ΔR ΔR Distributed Gate Oxide Capacitance ΔC FAR END (Degraded) Slew degradation & latency PHYSICAL STRIPE: RESISTANCE ACCUMULATES ALONG THE GATE Metal Polysilicon Gate Line (Rs ≈ 30 Ω/□) — Length Lwire Continuous Thin Gate Oxide (Forms Grounded Shunt Capacitance Across Entire Length) Silicon Substrate τpropagation ≈ 0.5 · Rs · Cox · Lwire² — quadratic length penalty ✓ Local transistors switch instantly because individual gate width is tiny Long wordlines and cross-chip interconnect suffer severe RC slowdown unless strapped with metal

## 2. Real Diagram: The Two-Regime Boundary of Silicon-Gate Scaling

Whether polysilicon gate resistance helps or hurts total circuit speed depends entirely on whether the device is dominated by localized transistor switching or by distributed interconnect wiring.

The Architectural Trade-off: Local Gate Win vs. Global Wire Loss quantifying where self-alignment triumphs and where polysilicon resistance imposes a design constraint REGIME 1: LOCAL TRANSISTOR GATE (L < 50 µm) Short Gate Line (Individual Logic Inverter): W = 20 µm, L = 5 µm Gate RC Time Constant: τ < 15 ps Miller Capacitance Reduction: >11× NET RESULT: 4× TO 5× SPEED INCREASE Gate resistance is negligible compared to the driver resistance The elimination of overlap capacitance completely dominates ✓ UNQUESTIONED VICTORY REGIME 2: LONG RUN / WORDLINE (L > 500 µm) Long Array Bus Line (Memory Wordline): Lwire = 1000 µm, W = 5 µm (200 squares) Total Line Resistance: Rline = 6,000 Ω Distributed Delay: τ ≈ 0.5 R C > 5.2 ns NET RESULT: CRIPPLING PROPAGATION DELAY High series resistance attenuates switching edge Cannot route clocks or long buses purely in polysilicon ⚠ NEW ARCHITECTURAL BOTTLENECK The Material Trade-off: Polysilicon solves the transistor, but complicates the wiring. This friction forces two historical solutions: intentional heavy doping (Step 7) and two-level metal/poly interconnect hierarchies.

## 3. The New Engineering Reality Born From Self-Alignment

Every major breakthrough in semiconductor processing solves a primary limitation while exposing a secondary one. The 1962 metal-gate process was bottlenecked by a geometric flaw: because the gate had to be aligned after the source and drain, overlap margin was mandatory, and the resulting feedback capacitance throttled device frequency.

Steps 1 through 5 eliminated that geometric flaw, proving that self-alignment collapses overlap capacitance by $11\times$. But in substituting polysilicon for aluminum, the process traded a capacitive problem for a resistive problem:
1. The Equipotential Illusion Shattered: In 1962, designers treated the aluminum gate electrode as an instantaneous equipotential surface ($R \approx 0$). In 1967, the polysilicon gate must be treated as a physical distributed resistor.
2. Quadratic Interconnect Penalty: Because distributed RC delay scales as $L_{\text{wire}}^2$, polysilicon is exceptional as a local gate electrode but terrible as long-distance chip wiring. If an entire integrated circuit attempted to route all interconnections in polysilicon, signal propagation across the die would stall.
3. The Mandate for Doping Control: This realization directly motivates Step 7: rather than relying purely on accidental dopant absorption from the source/drain furnace, the polysilicon gate must be intentionally and heavily doped to its solid solubility limit to push its sheet resistance down from hundreds of ohms per square to the lowest value physics permits.

Step 6 documents the discovery that self-alignment had not eliminated parasitics entirely; it had traded an unmanageable geometric capacitance for a manageable semiconductor resistance—defining the exact problem Step 7 must solve.

Take silicon gate 1967 polysilicon gate resistance rc delay 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