DRAM Scaling 1974 Measure Payoff Capacitance Process Complexity

# Step 9 — Measure the Payoff: Capacitance Recovered Per Unit of Process Complexity

## 1. The Cost-Accounting of 3D Charge Storage

Escaping the planar scaling crisis required adding masks, etch steps, and high-capital equipment; quantifying the true payoff requires measuring the capacitance recovered per unit of fabrication complexity. In Robert Dennard’s 1968 planar baseline, the storage capacitor was essentially free: a thin gate oxide grown alongside the transistor and covered by the same polysilicon layer ($N_{\text{mask}} \approx 6\text{ to }8$). But as minimum feature size shrank to $0.5\ \mu\text{m}$, planar capacitance collapsed below $3.5\ \text{fF}$, failing the sensing threshold ($\Delta V \ge 150\ \text{mV}$). Both the trench path (Steps 3, 6, 7) and the stacked path (Steps 4, 8) succeeded in recovering $C_s \ge 35\ \text{fF}$, but they did so by imposing heavy financial and process burdens on the wafer fabrication line.

To evaluate which architecture delivered superior manufacturing efficiency, semiconductor economists and fab directors defined the Storage Figure of Merit (FOM), quantifying capacitance recovered relative to mask layers and critical tool passes:

$$\text{FOM}_{\text{cap}} = \frac{C_s}{A_{\text{cell}} \cdot (N_{\text{mask}} + \alpha N_{\text{crit}})}$$

where $C_s$ is storage capacitance, $A_{\text{cell}}$ is cell area ($F^2$), $N_{\text{mask}}$ is total mask count, $N_{\text{crit}}$ is the number of high-capital critical tool steps (deep RIE, CMP, LPCVD, high-temperature anneals), and $\alpha \approx 1.2$ accounts for tool depreciation and scrap risk.

Process Cost Accounting: Planar vs. Trench vs. Stacked comparing mask additions, critical processing modules, and recovered capacitance 1968 PLANAR BASELINE Mask Count: 6–8 Masks Added 3D Steps: 0 Steps (Free) Capacitance at 0.5 µm: < 3.5 fF Verdict: Thermally Unviable • Low mask cost, but cell fails • Sensing margin collapses to zero DEEP TRENCH (ETCH DOWN) Mask Count: +4 Masks (11 Total) Critical Tool Steps: 8 Critical Steps Capacitance at 0.5 µm: 38 fF (11× Gain) Cost Profile: High Substrate Risk • Deep Si RIE + 3 poly recesses • Buried plate doping + collar oxide STACKED CROWN + CMP Mask Count: +5 Masks (12 Total) Critical Tool Steps: 10 Critical Steps Capacitance at 0.5 µm: 38 fF (11× Gain) Cost Profile: High Thin-Film Rework • Poly plug + crown etch + HSG • Thick ILD + CMP global planarization THE FINAL ECONOMIC ACCOUNTING ✓ Identical Electrical Gain: Both 3D architectures delivered 38 fF, rescuing the cell from the planar collapse ✓ Complexity Premium: Required +4 to +5 masks and ~10 critical tool passes, raising wafer fabrication cost by 35% to 45% The cost was severe, but it enabled a 1,000× increase in memory capacity over the subsequent decade.

## 2. The Economic Crossover Curve: Femtofarads per Manufacturing Dollar

Evaluating the marginal manufacturing cost across technology nodes reveals why the industrial balance shifted decisively over time.

The Economic Crossover: Cost per Femtofarad Across Nodes why trench led the 4M–16M era, and why stacked + CMP conquered 64M–1Gb and beyond NODE-BY-NODE ECONOMIC COMPETITION 4M – 16M DRAM Era (0.8 – 0.5 µm): Trench Wins • CMP tools were unproven, expensive ($2M/station) • Trench avoided CMP completely, delivering lower wafer cost 64M DRAM Era (0.35 µm): Parity • CMP matured for multi-level metal in logic and DRAM • HSG texturing added 2× area gain to stacked without height 256M – 1Gb Era (< 0.18 µm): Stacked Conquers • Trench aspect ratios exceeded 40:1 (etch-stop crisis) • Trench bulk punch-through forced expensive triple-wells ✓ Stacked thin-film rework gave strictly superior die yield FAB REWORK & SCRAP DYNAMICS Trench Scrap Penalty (Bulk Silicon): • A void or collar defect deep in bulk cannot be stripped • Whole wafer is scrapped ($1,500 loss) • In-line metrology cannot inspect bottom of 8 µm hole • Destructive SEM cross-sectioning only Stacked Rework Advantage (Thin Films): • Poly crowns and oxide ILD reside above the transistor • Litho/CMP misalignments can be wet-stripped and reworked • Optical scatterometry measures crown height non-destructively ✓ Saves expensive crystalline silicon substrates The Decisive Economic Factor: Stacked capacitors won the long-term industry battle because thin-film process steps above the wafer could be inspected, controlled, and reworked.

## 3. Place This Step in the Project's Own Lineage

Step 9 grounds the physical innovations of 3D scaling in manufacturing economics:
1. The Culmination of the Fork: Steps 2 through 8 followed the physical branching of the 3D capacitor revolution: etching down into the substrate versus building up above the gate. Step 9 performs the rigorous accounting: how many masks, how many tool passes, and what was the ultimate scrap risk? It demonstrates that both paths achieved the mandatory physical threshold ($C_s \ge 35\ \text{fF}$), but did so at a $+35\%\text{ to }+45\%$ wafer fabrication cost premium over planar processing.
2. The Reason Stacked Won the Endgame: Why did IBM and Infineon eventually abandon deep trenches in favor of stacked capacitors? Step 9 supplies the economic answer: as lithography pushed below $0.18\ \mu\text{m}$, trench aspect ratios crossed $40:1$ and bulk punch-through destroyed wafer yields. Stacked capacitors, bolstered by CMP and non-destructive thin-film inspection, yielded more functional bits per capital dollar.
3. Leading to the Series Retrospective: Having quantified the 3D charge-storage revolution from Dennard's 1968 circuit equations through deep etching, buried collars, and CMP planarization, we are ready for the final synthesis. Step 10 will assemble the entire 1974 narrative into an enduring retrospective on the triumph and limits of three-dimensional semiconductor scaling.

Take dram scaling 1974 measure payoff capacitance process complexity 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