DRAM Scaling 1974 Build Stacked Capacitor Above Transistor

# Step 4 — Build the Stacked Capacitor: Rising Above the Transistor

## 1. Climbing Vertically into the Free Space Above the Gate

Rather than excavating deep into the fragile silicon substrate, the stacked capacitor builds the storage node vertically into the open space above the access transistor and field oxide. In Step 3, we examined the trench path, where depth into the substrate became the new surface-area budget at the cost of high-aspect-ratio reactive-ion etching and bulk punch-through risks. The stacked capacitor (STC) takes the diametrically opposite architectural approach: fabricate the access transistor on the silicon surface first, cover it with an inter-layer dielectric (ILD), open a storage contact via, and erect a multi-dimensional polysilicon electrode rising into the vertical dimension ($Z$-axis):

$$A_{\text{stacked}} \approx 2\pi (r_{\text{in}} + r_{\text{out}}) h_{\text{cyl}} + A_{\text{rim}} \approx 4\pi r \cdot h_{\text{cyl}}$$

where $h_{\text{cyl}}$ is the height of the cylindrical or crown electrode ($1.0\text{ to }2.0\ \mu\text{m}$), and $r_{\text{in}}, r_{\text{out}}$ are the inner and outer radii. By engineering hollow cylindrical cups, fins, or crowns, the stacked capacitor exposes both inner and outer sidewall surfaces to the capacitor dielectric and upper plate, doubling the effective vertical surface area compared to a solid pillar or a single trench cavity of identical diameter.

$$C_s = \frac{\varepsilon_{\text{di}} A_{\text{stacked}}}{t_{\text{di}}} \approx \frac{\varepsilon_{\text{di}} (4\pi r \cdot h_{\text{cyl}})}{t_{\text{di}}}$$

This architecture completely removes the storage node from the silicon crystal. Because the storage node is encapsulated within dielectric layers above the substrate, substrate leakage paths are physically broken, and minority carriers generated in the bulk by ionizing radiation cannot enter the capacitor. However, this vertical expansion creates a fierce new physical penalty: severe surface topography that destabilizes optical lithography and metal interconnect reliability.

Building the Stacked Capacitor: Crown/Cylinder Architecture elevating the storage node above the transistor to exploit both inner and outer sidewall surfaces STACKED CROWN CAPACITOR CROSS-SECTION Silicon Substrate Word Line ILD Oxide Layer Plug Inner Plate h ≈ 1.5 µm (Step Height) ← Inner Sidewall Outer Sidewall → THE DUAL-SURFACE ADVANTAGE Double Sidewall Area Utilization: Astack ≈ 2πrinh + 2πrouth ≈ 4πrh 2× more sidewall area than a solid post of same height Hemispherical Grained Silicon (HSG): Nanoscale textured poly adds microscopic bumps Multiplies effective area by another 1.8×–2.0× Electrical Decoupling from Bulk: ✓ Zero substrate punch-through or bulk leakage ✓ 100× improvement in alpha-particle soft error rate No deep silicon etching; standard thin-film deposition. THE PHYSICAL PAYOFF OF THE STACKED ARCHITECTURE ✓ Dual Sidewall Multiplication: Crown/cylinder geometry captures both inside and outside capacitor plates ✓ Substrate Immunity: Storage node is dielectrically isolated above the silicon, eliminating alpha-particle collection The stacked capacitor delivers immense electrical safety, but shifts the manufacturing burden to surface topography.

## 2. The Competing Constraint: Severe Topography and the Depth-of-Focus Crisis

While the stacked capacitor frees the memory cell from substrate electrical defects, rising $1.5\ \mu\text{m}$ above the wafer creates a severe topographic step that breaks the physics of optical lithography.

Topography vs. Immunity: The Two Faces of the Stacked Cell complete soft-error protection at the cost of catastrophic optical stepper defocus 1. ALPHA-PARTICLE SOFT-ERROR IMMUNITY Bulk Si Substrate α-Particle Ionization Track e⁻ / h⁺ pairs in bulk Stacked Capacitor Dielectric Barrier ✓ IMMUNE Physical Isolation from Substrate • Charge stored entirely in elevated polysilicon • Bulk ionization charges cannot cross the thick ILD • Retention time is stable against cosmic radiation 2. THE STEPPER DEPTH-OF-FOCUS CRISIS Memory Array (Tall) Periphery (Flat) Δh = 1.5 µm Focal Window (DOF < 0.8 µm) ✗ DEFOCUSED! Catastrophic Lithography Limit • Step height Δh exceeds optical stepper depth-of-focus • Cannot focus on array bitlines and peripheral logic at once • Metal lines thinning across cliffs cause bridging and opens The Architectural Tradeoff Defined: Stacked capacitors conquered substrate leakage and soft errors, but forced the industry to develop Chemical Mechanical Polishing (CMP) to survive.

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

Step 4 reveals the full symmetry of the 3D capacitor revolution:
1. The Opposing Answer to Step 3: Where Step 3 excavated downward into the silicon bulk ($A \approx \pi d \cdot d_{\text{etch}}$) to preserve surface planarity, Step 4 climbs upward ($A \approx 4\pi r \cdot h_{\text{cyl}}$) to preserve substrate electrical integrity. Neither alters Robert Dennard’s 1968 circuit equations ($\Delta V = \frac{C_s}{C_s + C_{bl}} V_{dd}$); both simply solve for $C_s$ in opposite directions along the vertical $Z$-axis.
2. Solving the Soft-Error Crisis: In the late 1970s, Intel and IBM discovered that trace radioactive thorium and uranium in ceramic chip packaging emitted alpha particles that corrupted DRAM data. The stacked capacitor emerged as the premier defense: by lifting the storage node above the bulk silicon, minority carriers generated in the substrate could not reach the stored bit.
3. Setting Up the Quantitative Comparison: By introducing dual-sidewall crown etching and hemispherical grained silicon (HSG), stacked capacitors demonstrated that vertical thin films could match the capacitance of deep trenches. But with height came a devastating optical depth-of-focus penalty.

Step 5 will place both architectures side-by-side on identical electrical terms—comparing aspect ratios, process complexity, and array compatibility to explain why the memory industry split into two fierce competing factions.

Take dram scaling 1974 build stacked capacitor above transistor further

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