First Year Engineering is coming soon

We're still writing and reviewing the notes for every subject, and we'll open it when every subject is ready. Meanwhile you can practise aptitude, ask Aria about any subject, or see the branches that are ready.

VLSI technology and IC fabrication

VLSI technology and IC fabrication: from silicon wafer to CMOS chips—crystal growth, oxidation, photolithography, doping, deposition, etching, interconnect, isolation, design rules, yield, testing, and packaging.

Drafted with Aria, reviewed by the AiCanCode.org team · Updated . How we write and check our notes. Spotted an error? Use Give Feedback at the bottom of the page.

Why it matters

VLSI (Very Large Scale Integration) technology is how modern microprocessors, memories, sensors, and controllers are physically built on silicon. Understanding IC fabrication helps engineers read datasheets correctly, choose suitable components, appreciate process limitations, and communicate with chip vendors and PCB designers.

Key ideas

  • CMOS (Complementary Metal-Oxide-Semiconductor): A technology using both n-channel MOSFETs (NMOS) and p-channel MOSFETs (PMOS) with complementary switching, yielding very low static power. CMOS dominates modern digital ICs.
  • Moore’s law (empirical observation): “The number of transistors on an economically optimal IC doubles approximately every 18–24 months.” It is not a physical law; it describes a historical trend in integration.
  • Wafer: A thin circular slice of monocrystalline silicon used as the substrate for ICs. Common diameters: 200 mm and 300 mm (industrial standard today).
  • Czochralski crystal growth: A process to grow large single-crystal silicon ingots by dipping a seed crystal into molten silicon and slowly pulling while rotating; then the ingot is sliced into wafers.
  • Cleanroom: A controlled environment with very low particle counts, temperature and humidity control. ISO Class 5 (formerly Class 100) and better are typical for critical lithography.
  • Unit process: A single fabrication step (oxidation, lithography, etch, etc.). Process flow is the ordered combination yielding devices.
  • Thermal oxidation: Growing silicon dioxide (SiO2) on silicon by exposing it to oxygen or steam at high temperature (e.g., 900–1100 °C). SiO2 is used as an insulator and mask.
  • Photolithography: Pattern transfer using light. Steps: wafer clean → photoresist coat (spin coat) → soft bake → align mask → expose (UV) → develop → hard bake. Positive resist: exposed regions become soluble; negative resist: exposed regions become insoluble.
  • Etching: Removing material using chemicals (wet etch) or plasma (dry/RIE). Isotropic etch removes equally in all directions; anisotropic etch removes mainly vertically (better for fine features).
  • Doping: Introducing impurity atoms (donors or acceptors) to control carrier concentration. Methods: diffusion (high-temperature drive-in governed by Fick’s laws) and ion implantation (accelerated ions shot into silicon), typically followed by annealing to repair damage and activate dopants.
  • Epitaxy (EPI): Depositing a crystalline silicon layer on a crystalline substrate to form lightly doped, high-quality layers over heavily doped sublayers.
  • Thin-film deposition: Adding layers such as polysilicon, SiO2, Si3N4, metals. Techniques: PVD (Physical Vapor Deposition, e.g., sputtering), CVD (Chemical Vapor Deposition; LPCVD at low pressure, PECVD with plasma at lower temperature).
  • Isolation: Electrically separating devices. Historically LOCOS (Local Oxidation of Silicon); modern processes use STI (Shallow Trench Isolation): etch trenches and fill with oxide, then planarize by CMP (Chemical Mechanical Planarization).
  • Well formation: Creating p-well and n-well regions in a p-type or n-type substrate to host NMOS and PMOS respectively; twin-well and triple-well options for better control and noise isolation.
  • Gate stack: Thin gate dielectric (e.g., SiO2 or high‑k material) and a gate electrode (polysilicon or metal). Advanced nodes use high‑k/metal gate stacks.
  • Interconnect and dielectric: Metal layers (Al, Cu) separated by interlayer dielectrics (SiO2 or low‑k dielectrics). Barrier/liner materials (e.g., TiN, TaN) prevent copper diffusion. Multiple metal layers are connected by vias.
  • Design rules: Geometric constraints (minimum width, spacing, enclosure) that ensure manufacturability. Scalable rules often use λ (lambda) units, where feature sizes are multiples of λ.
  • Yield: Fraction of good dies per wafer. It depends on defect density and die area. A common simple model is the Poisson yield model: Y = exp(−D0·A), where D0 is defect density and A is die area.
  • Testing and packaging: After fabrication, wafers undergo wafer-probe testing; good dies are packaged (DIP, QFP, QFN, BGA, CSP, or flip‑chip), then final test and burn-in as needed.

Exact statements and typical derivations

  • Fick’s first law (diffusion): “The diffusion flux J is proportional to the concentration gradient: J = −D·(dC/dx), assuming steady state and constant diffusion coefficient D.”
  • Fick’s second law: “∂C/∂t = D·(∂²C/∂x²), assuming constant D.” Solution forms (erfc/Gaussian) are used for constant-source and limited-source diffusion profiles; universities often expect you to state these forms conceptually even if not deriving the full PDE solution.
  • Rayleigh resolution criterion (optics): “The minimum resolvable half-pitch (critical dimension) is R = 0.61·λ/NA,” assuming incoherent imaging and circular aperture. Lithography also uses the empirical form CDmin = k1·λ/NA.
  • Sheet resistance (Rs): “The resistance of a thin film per square is Rs = ρ/t,” assuming uniform film thickness t and bulk resistivity ρ. Device resistors implemented in polysilicon or diffusion are often calculated by counting squares.
  • Oxide capacitance per unit area (Cox): “Cox = εox/tox,” assuming parallel-plate approximation, uniform thickness tox, and dielectric permittivity εox = ε0·εr.
  • Poisson yield model: “Yield Y = exp(−D0·A),” assuming random, independent defects across area A with mean density D0.

What to draw and label (exam diagrams)

  • Cross-section of an NMOS (or CMOS) fabrication sequence:
    1. Start with p-type substrate. 2) Grow pad oxide and nitride (for STI). 3) Trench etch and oxide fill (STI), CMP planarization. 4) Well implantation (n-well/p-well). 5) Gate oxide growth. 6) Polysilicon deposition and patterning (gate). 7) Lightly doped drain (LDD) implants. 8) Sidewall spacer formation. 9) Source/drain heavy implants. 10) Silicidation on source/drain/gate. 11) Interlayer dielectric deposition. 12) Contact/via etch and fill. 13) Metal interconnect layers. Label substrate type, wells, STI oxide, gate oxide with thickness, polysilicon gate, spacers, S/D regions, contacts, metal1.
  • Top view (layout) of a CMOS inverter: show pMOS in n-well and nMOS in p-substrate, shared VDD and GND rails, polysilicon gate crossing active diffusion to form both transistors, and metal1 interconnects; label active regions, poly, contact cuts, well taps, and minimum width/spacing according to λ rules.
  • Photolithography stepper: lamp/laser source, condenser, reticle (mask), projection lens with numerical aperture NA, wafer on stage; indicate reduction factor (e.g., 4×).

Formulas

  • Rs = ρ/t — Sheet resistance; Rs (Ω/□), ρ is resistivity (Ω·m), t is film thickness (m). Uniform thin films.
  • R = Rs·(L/W) — Resistor from sheet resistance; R (Ω), Rs (Ω/□), L (m), W (m). Layout with L/W squares.
  • Cox = εox/tox — Gate-oxide capacitance per unit area; Cox (F/m²), εox (F/m), tox (m). Parallel-plate approximation.
  • C = Cox·A — MOS capacitor/gate capacitance; C (F), Cox (F/m²), A area (m²). Uniform oxide thickness.
  • τ ≈ R·C — RC delay estimate; τ (s), R (Ω), C (F). First-order delay of an RC network.
  • L = √(D·t) — Diffusion length scale; L (m), D (m²/s), t (s). Order-of-magnitude junction depth for limited cases.
  • J = −D·(dC/dx) — Fick’s first law; J (atoms·m⁻²·s⁻¹), D (m²/s), C (atoms·m⁻³), x (m). Steady state.
  • Y = exp(−D0·A) — Poisson yield; Y (dimensionless), D0 defect density (defects/m²), A die area (m²). Random, independent defects.

Worked examples

Example 1: Gate capacitance and RC delay of a small MOS gate Given: SiO2 gate oxide thickness tox = 10 nm, εr(SiO2) = 3.9, ε0 = 8.854×10⁻¹² F/m, transistor gate area A = W·L with W = 1.0 µm and L = 0.10 µm. Polysilicon gate sheet resistance Rs = 20 Ω/□. Gate length to first contact has effective L/W = 3 squares.

  1. Formula: Cox = εox/tox where εox = ε0·εr.
  2. Compute εox = 8.854×10⁻¹² F/m × 3.9 = 3.453×10⁻¹¹ F/m.
  3. Compute Cox = 3.453×10⁻¹¹ F/m ÷ 10×10⁻⁹ m = 3.453×10⁻³ F/m².
  4. Gate area A = 1.0 µm × 0.10 µm = (1.0×10⁻⁶ m) × (1.0×10⁻⁷ m) = 1.0×10⁻¹³ m².
  5. Formula: C = Cox·A.
  6. Compute C = 3.453×10⁻³ F/m² × 1.0×10⁻¹³ m² = 3.453×10⁻¹⁶ F = 0.3453 fF.
  7. Gate poly resistance to contact: R = Rs·(L/W) = 20 Ω/□ × 3 = 60 Ω.
  8. RC delay estimate: τ ≈ R·C = 60 Ω × 0.3453×10⁻¹⁵ F = 2.07×10⁻¹⁴ s = 20.7 fs. Answer: Gate capacitance ≈ 0.345 fF, RC delay to gate contact ≈ 2.07×10⁻¹⁴ s (idealized; interconnect and load dominate in real circuits).

Example 2: Optical resolution in photolithography Given: KrF lithography with wavelength λ = 248 nm, numerical aperture NA = 0.60. Use Rayleigh’s criterion R = 0.61·λ/NA.

  1. Formula: R = 0.61·λ/NA.
  2. Convert units: λ = 248 nm = 248×10⁻⁹ m.
  3. Compute R = 0.61 × 248×10⁻⁹ m ÷ 0.60 = (151.28×10⁻⁹ m) ÷ 0.60 = 252.1×10⁻⁹ m. Answer: Minimum resolvable half-pitch R ≈ 252 nm. Feature sizes smaller than this require higher NA, shorter λ, or lower k1 via process enhancements.

Example 3: Yield estimate using the Poisson model Given: Defect density D0 = 0.5 defects/cm², die area A = 100 mm². Wafer diameter = 300 mm. Estimate the yield and approximate number of good dies per wafer, ignoring edge losses for simplicity.

  1. Convert die area: 100 mm² = 1.0 cm².
  2. Formula: Y = exp(−D0·A).
  3. Compute Y = exp(−0.5 × 1.0) = exp(−0.5) ≈ 0.607.
  4. Wafer area: A_wafer = π·(D/2)² = π × (300 mm/2)² = π × (150 mm)² = π × 22500 mm² ≈ 70685.8 mm².
  5. Approximate dies per wafer (ideal) = A_wafer / A_die = 70685.8 mm² / 100 mm² ≈ 706.9 ≈ 706 dies (integer placement and edges ignored).
  6. Good dies ≈ dies per wafer × Y ≈ 706 × 0.607 ≈ 428.5. Answer: Yield ≈ 60.7%, good dies per 300 mm wafer ≈ 429 (idealized; real numbers are lower due to edge/partial-die losses and non-Poisson defects).

Model long-answer (theory): CMOS process flow (short note)

  • Start with a p-type silicon substrate, grow pad oxide and silicon nitride. Pattern nitride and etch shallow trenches. Fill with oxide and planarize (STI) to isolate devices. Perform well implants to create n-well for pMOS and adjust threshold with channel implants. Grow thin gate oxide thermally. Deposit and pattern polysilicon to form gates. Perform lightly doped drain (LDD) implants, deposit spacer oxide and etch-back to form sidewall spacers. Perform heavy source/drain implants. Anneal to activate dopants and repair damage. Silicide formation on source/drain/gate lowers contact resistance. Deposit interlayer dielectric, open contacts, deposit and pattern metal1. Repeat for higher metal layers separated by dielectrics and connected by vias. Finish with passivation, open bond pads, wafer-probe test, dice, and package.

Common mistakes

  • Using bulk resistivity directly for thin-film resistors instead of converting to sheet resistance (Ω/□).
  • Applying Rayleigh resolution without checking NA or mixing λ units (nm vs m).
  • Assuming 100% yield or ignoring the exponential nature in Y = exp(−D0·A).
  • Drawing CMOS layout without well taps/contacts to VDD (n-well) and GND (p-substrate), which are essential to avoid latch-up.
  • Confusing LOCOS with STI; STI uses trench etch and fill, not local oxidation encroachment (“bird’s beak”).

For your exam

  • Derivations/expressions: Fick’s laws (statement and meaning), Rayleigh resolution, Cox and RC delay, Poisson yield Y = exp(−D0·A), Rs = ρ/t and R = Rs·(L/W).
  • Diagrams: Cross-section sequence of CMOS fabrication; photolithography setup; CMOS inverter layout with labels.
  • Numericals: Calculate Cox and MOS capacitance; compute resistor value from sheet resistance and geometry; estimate lithography resolution for given λ and NA; estimate yield for given D0 and die area. Full-marks answers show: correct formula, unit conversions in SI, step-by-step substitution, clear final boxed value with units, and mention of assumptions (e.g., uniform oxide, ignoring edge losses).

Quick check

  1. State the difference between diffusion and ion implantation in doping.
  2. Give the formula for sheet resistance and explain each symbol.
  3. What is the empirical relation connecting minimum critical dimension, wavelength, numerical aperture, and process factor in lithography? Answers: 1) Diffusion drives dopants into silicon at high temperature based on concentration gradients (Fick’s laws); ion implantation shoots accelerated ions into silicon to a controlled depth and requires annealing to activate and repair damage. 2) Rs = ρ/t, where Rs (Ω/□) is sheet resistance, ρ (Ω·m) is resistivity of the film, t (m) is film thickness; assumes uniform film. 3) CDmin = k1·λ/NA, where k1 is a process-dependent constant, λ (m) is exposure wavelength, and NA is numerical aperture.

Finished this topic? Mark it so your progress, study plan and readiness keep up.

Stuck on something here?