ChipFoundryServices
From CD-SEM & Spectroscopic Ellipsometry to Optical Scatterometry (OCD) & Defect Review

Metrology University

The quantum measurement science, optical diffraction physics, and statistical analytics of semiconductor nanometrology: critical dimension scanning electron microscopy (CD-SEM), spectroscopic ellipsometry (Ψ, Δ), optical critical dimension (OCD) scatterometry with RCWA, high-throughput multi-beam e-beam defect inspection, HAADF-STEM atomic imaging, and hybrid run-to-run APC yield engineering for sub-2nm logic.

7 Levels
Elementary to Fellow
21 Modules
Rigorous Curriculum
7 Sim Labs
Real-Time Engines
7 Diplomas
Industry Fellow Laureate
Academic Level 1 • Ages 6–10
The Nano-Detectives
Discover how scientists measure invisible structures smaller than light waves!
Module 1.1

If You Can't Measure It, You Can't Make It

Imagine building a Lego tower where every brick is smaller than a virus. If one brick is off by just the width of a single atom, the entire tower will collapse! In chip manufacturing, this is why 'Metrology'—the science of measurement—is king.

Before a wafer moves from etching to implantation or metallization, engineers must measure feature widths and layer thicknesses to verify everything matches the blueprint perfectly.

  • Metrology: The rigorous science and engineering of nanometer measurement.
  • Zero Guesswork: Every single nanometer must be verified before proceeding.
$$\text{Blueprint Specification } (10.0\,\text{nm}) \pm 0.2\,\text{nm} \iff \text{Fabrication Verification}$$
Module 1.2

Why Regular Microscopes Are Blind

Why can't engineers simply look through a standard optical microscope like in biology class? Because light waves are too fat! Visible light has a wavelength of about 500 nanometers.

Trying to see a 5-nanometer transistor with visible light is like trying to feel the teeth of a tiny wristwatch gear while wearing giant boxing gloves! The light waves simply wash around the transistor without reflecting an image.

  • Diffraction Limit: Visible light cannot resolve features smaller than $\approx 250\,\text{nm}$.
  • Electron Beams: Using high-speed electrons with picometer wavelengths to see individual atoms.
$$\lambda_{\text{light}} \approx 500\,\text{nm} \gg \text{Transistor Feature } (5\,\text{nm}) \implies \text{Optical Blindness}$$
Module 1.3

Scanning with Electrons

To see nanometer details, fabs use Scanning Electron Microscopes (SEMs). Instead of light, a gun shoots a focused beam of electrons down at the wafer.

When electrons hit the silicon, they knock out secondary electrons that fly into a detector. A computer counts these electrons to paint a crystal-clear, 3D picture of transistor lines on the monitor!

  • Electron Gun: Emitting a fine beam of electrons accelerated by high voltage.
  • Secondary Electrons: Low-energy electrons emitted from surface edges, highlighting sharp contours.
$$\lambda_{\text{de Broglie}} = \frac{h}{p} = \frac{h}{\sqrt{2 m_e q V_{\text{acc}}}} \approx 0.01-0.05\,\text{nm}$$
⚡ Interactive Laboratory L1
Electron de Broglie Wavelength & Resolution Calculator
Calculate the quantum de Broglie wavelength of electron beams under different accelerating voltages and compare against visible light.
Accelerating Voltage (V)1000V
Objective Aperture Angle (mrad)15mrad
Electron Beam Spot Size (nm)1.2nm
REAL-TIME SIMULATION TELEMETRY
Interactive physics simulator running client-side transfer models, carrier drift-diffusion kinetics, and boundary potential solvers.
Electron de Broglie Wavelength
-- pm
Diffraction Resolution Limit
-- nm
Equivalent Magnification
-- x
🎓 Level 1 Examination
Level 1 Conceptual Mastery Assessment
Why is semiconductor metrology essential throughout the fabrication process?
Why can standard optical microscopes using visible light NOT resolve modern 5nm transistors?
How does a Scanning Electron Microscope (SEM) form high-magnification images of silicon features?

Level 1 Completed: The Nano-Detectives Mastery Certificate

Conferred for mastery of Level 1 (Academic Level 1 • Ages 6–10) curriculum, simulation laboratory, and assessment evaluation.

Academic Level 2 • Ages 11–14
Critical Dimension (CD) & The CD-SEM
Master the measurement of gate widths, electron beam edge detection, and beam damage mitigation.
Module 2.1

What is Critical Dimension (CD)?

In every integrated circuit, there is one dimension that matters above all others: the smallest feature width that governs electrical performance. This is called the 'Critical Dimension' (CD), typically the physical gate length of the transistor.

If the gate CD is 1 nanometer too wide, the transistor switches too slowly and the processor fails its speed clock target. If the CD is 1 nanometer too narrow, excessive leakage current drains the battery! CD must be controlled within $\pm 0.3\,\text{nm}$ across millions of transistors.

  • Critical Dimension (CD): The minimum patterned feature width determining circuit performance.
  • CD Uniformity (CDU): Statistical 3-sigma variation across the 300mm wafer ($3\sigma < 0.5\,\text{nm}$).
  • Electrical Correlation: Drive current $I_{ ext{on}} \propto rac{1}{ ext{CD}}$ and leakage $I_{ ext{off}} \propto \exp(- ext{CD})$.
$$\text{CDU} = 3\sigma_{\text{CD}} \le 0.50\,\text{nm} \quad (\text{Cross-Wafer Specification})$$
Module 2.2

The CD-SEM Edge Detection Algorithm

A specialized tool called the CD-SEM (Critical Dimension SEM) is the fab's workhorse. As its electron beam sweeps horizontally across a line, the detector records secondary electron intensity, creating a characteristic waveform with sharp 'edge peaks'.

Why peaks at the edges? When electrons strike vertical sidewalls, more secondary electrons can escape into vacuum from the sharp corners than from flat floors. Edge detection algorithms (such as maximum slope or 50% threshold) determine the exact left and right line edges to calculate CD with picometer precision!

  • Edge Blooming Effect: Enhanced secondary electron emission at steep sidewall edges.
  • Threshold Algorithm: Locating the edge position at a fixed percentage (e.g. 50%) of the peak intensity.
  • Automated Recipe Execution: Tool locates test targets and measures 500 sites across a wafer in minutes.
$$\text{CD} = x_{\text{right\_edge}} - x_{\text{left\_edge}} \quad (\text{Extracted from Peak Waveform})$$
Module 2.3

Combating Electron Beam Shrinkage

Measuring photoresist patterns with an electron beam creates a tough dilemma: high-energy electrons break polymer bonds, causing the photoresist line to shrink during the very act of measurement! This is called 'e-beam shrinkage'.

To prevent corrupting the features it measures, CD-SEMs use ultra-low landing energies (300 to 500 eV) and rapid scanning techniques that deposit minimum electron dose while still acquiring clear signal waveforms.

  • Low Landing Energy: $V_{ ext{landing}} pprox 300-500\, ext{eV}$ minimizing penetration depth.
  • Dose Budget: Restricting total electron exposure to $< 10\,\text{electrons/\AA}^2$.
  • Shrinkage Compensation: Mathematical regression correcting for measured dimension changes.
$$\Delta\text{CD}_{\text{shrink}} = \Delta\text{CD}_{\infty}\left(1 - \exp\left(-\frac{\text{Dose}}{D_0}\right)\right)$$
⚡ Interactive Laboratory L2
CD-SEM Secondary Electron Waveform & Edge Extraction Lab
Simulate the line-scan secondary electron waveform across a gate line and calculate CD using the 50% threshold algorithm.
True Feature Physical CD (nm)22nm
SEM Beam Probe Size (nm)2.0nm
Edge Threshold Criterion (%)50%
REAL-TIME SIMULATION TELEMETRY
Interactive physics simulator running client-side transfer models, carrier drift-diffusion kinetics, and boundary potential solvers.
Extracted Measured CD
-- nm
Measurement Offset / Bias
-- nm
Waveform Signal-to-Noise
-- dB
🎓 Level 2 Examination
Level 2 Conceptual Mastery Assessment
What is meant by 'Critical Dimension' (CD) in semiconductor logic processing?
Why does secondary electron emission peak sharply at pattern edges in a CD-SEM line scan?
What phenomenon occurs when delicate organic photoresist patterns are repeatedly measured with high-energy electron beams?

Level 2 Completed: Critical Dimension (CD) & The CD-SEM Mastery Certificate

Conferred for mastery of Level 2 (Academic Level 2 • Ages 11–14) curriculum, simulation laboratory, and assessment evaluation.

Academic Level 3 • Ages 15–18
Spectroscopic Ellipsometry & Thin-Film Optics
Deconstruct polarized light reflection, determine complex refractive indices ($n, k$), and measure angstrom films.
Module 3.1

Polarized Light & The Ellipsometric Parameters (Ψ, Δ)

Thin dielectric films (such as gate oxides, nitride spacers, and ARC coatings) cannot be measured by SEM without slicing the wafer in half. Instead, fabs use Spectroscopic Ellipsometry (SE), a non-destructive optical technique with sub-Angstrom precision.

Linearly polarized light is bounced off the wafer at an oblique angle (e.g. 70°). Reflection from the film surfaces changes the light from linear to elliptical polarization. The ellipsometer measures two fundamental angles: Psi ($\Psi$), the change in amplitude ratio, and Delta ($\Delta$), the phase difference between p- and s-polarizations.

  • Amplitude Ratio ($\tan\Psi$): Ratio of p-polarized to s-polarized reflection coefficients $|r_p| / |r_s|$.
  • Phase Shift ($\Delta$): Phase difference $\delta_p - \delta_s$ between orthogonal electric field vectors.
  • Sub-Angstrom Sensitivity: $\Delta$ changes by several degrees for every 0.1 nm change in oxide thickness!
$$\rho = \frac{r_p}{r_s} = \tan(\Psi) \cdot e^{i \Delta} \quad (\text{Fundamental Ellipsometry Equation})$$
Module 3.2

The Complex Refractive Index: n and k

Every material interacts with electromagnetic radiation through its complex refractive index: $ ilde{n} = n + i k$, where $n$ is the refractive index (phase velocity) and $k$ is the extinction coefficient (optical absorption).

In transparent dielectrics ($SiO_2, Al_2O_3$), $k = 0$ across the visible spectrum, governed by the Cauchy dispersion model. In absorbing materials (amorphous carbon, polysilicon, metals), $k > 0$, requiring oscillator models (Tauc-Lorentz or Forouhi-Bloomer) to fit both thickness and material composition simultaneously.

  • Cauchy Dispersion: $n(\lambda) = A + \frac{B}{\lambda^2} + \frac{C}{\lambda^4}$, used for transparent dielectric films.
  • Extinction Coefficient ($k$): Governs optical absorption $\alpha = \frac{4\pi k}{\lambda}$.
  • Multi-Layer Stacks: Resolving up to 10 stacked layers simultaneously using broadband spectra (190 nm to 1,700 nm).
$$\tilde{n}(\lambda) = n(\lambda) + i k(\lambda), \quad I(z) = I_0 \cdot \exp\left(-\frac{4\pi k}{\lambda} z\right)$$
Module 3.3

Statistical Process Control (SPC) & Cpk Analytics

In high-volume manufacturing, metrology feeds directly into Statistical Process Control (SPC). Fab engineers track thousands of measurements in real time on control charts to detect process drifts before defects occur.

The ultimate benchmark of process capability is the $C_{pk}$ index, which quantifies how well the process distribution fits within upper and lower specification limits (USL/LSL). A world-class foundry requires $C_{pk} \ge 1.67$ ($5\sigma$ quality), meaning fewer than 1 defect per million operations!

  • Process Capability ($C_{pk}$): $C_{pk} = \min\left(\frac{\text{USL} - \mu}{3\sigma}, \frac{\mu - \text{LSL}}{3\sigma}\right)$.
  • World-Class Target: $C_{pk} \ge 1.67 \implies \text{Defect Rate } < 0.6\,\text{PPM}$.
  • Western Electric Rules: Statistical trend detectors flagging tool drifts in real time.
$$C_{pk} = \frac{\min(\text{USL} - \mu, \mu - \text{LSL})}{3\sigma} \ge 1.67 \quad (\text{Six-Sigma Foundry Standard})$$
⚡ Interactive Laboratory L3
Spectroscopic Ellipsometry Film Thickness & Cpk Solver
Calculate thin film thickness from phase shift Delta and evaluate fab line statistical capability (Cpk) against specification limits.
Ellipsometric Phase Delta Δ (deg)145deg
Film Refractive Index n1.46
Process Standard Deviation σ (nm)0.12nm
REAL-TIME SIMULATION TELEMETRY
Interactive physics simulator running client-side transfer models, carrier drift-diffusion kinetics, and boundary potential solvers.
Extracted Film Thickness
-- nm
Process Capability Index Cpk
--
Manufacturing Status
--
🎓 Level 3 Examination
Level 3 Conceptual Mastery Assessment
What physical parameter does Spectroscopic Ellipsometry measure upon polarized light reflection?
In the complex refractive index ñ = n + ik, what does the extinction coefficient k represent?
What Cpk score is required by premier semiconductor foundries to certify a process as 'World-Class Six-Sigma' capable?

Level 3 Completed: Spectroscopic Ellipsometry & Thin-Film Optics Mastery Certificate

Conferred for mastery of Level 3 (Academic Level 3 • Ages 15–18) curriculum, simulation laboratory, and assessment evaluation.

Academic Level 4 • Undergraduate
Optical Scatterometry (OCD) & Overlay Metrology
Decode diffraction signatures using Rigorous Coupled-Wave Analysis and solve the Edge Placement Error budget.
Module 4.1

Optical Critical Dimension (OCD) Scatterometry

While CD-SEMs measure top-down widths, transistors are complex 3D structures with fin heights, sidewall angles, undercut profiles, and footing. Scanning every 3D profile with SEM would require cross-sectioning and destroying the wafer!

Fabs invented Optical Critical Dimension (OCD) Scatterometry. Broadband polarized light illuminates periodic grating targets. The diffracted light signature contains rich optical information about the full 3D geometry. By solving Maxwell's equations using Rigorous Coupled-Wave Analysis (RCWA), OCD reconstructs top CD, middle CD, bottom CD, sidewall angle, and feature height simultaneously in milliseconds!

  • Non-Destructive 3D Profiling: Full profile reconstruction without cross-section cutting.
  • Rigorous Coupled-Wave Analysis (RCWA): Maxwell's equations solved in the spatial frequency domain.
  • Mueller Matrix Polarimetry: 16-element matrix capturing cross-polarization and 3D asymmetries.
$$\nabla \times (\nabla \times \mathbf{E}) - k_0^2 \varepsilon_r(x, z) \mathbf{E} = 0 \quad (\text{RCWA Maxwell Solver})$$
Module 4.2

Overlay Metrology: AIM vs Diffraction-Based Overlay (DBO)

Every chip consists of 60 to 80 patterned layers stacked on top of one another. If Layer 20 (contact vias) is misaligned with Layer 19 (gate electrodes) by more than 1.5 nanometers, circuits open or short circuit! Measuring this alignment is 'Overlay Metrology'.

Fabs evolved from image-based overlay (AIM, box-in-box targets viewed optically) to Diffraction-Based Overlay (DBO). DBO shines laser beams onto overlapping periodic gratings; any overlay misalignment breaks optical symmetry, producing an asymmetry in the +1st and -1st diffraction orders directly proportional to nanometer displacement!

  • Diffraction-Based Overlay (DBO): Measures optical intensity difference between $\pm 1 ext{st}$ diffraction orders.
  • Sub-Nanometer Precision: Overlay measurement uncertainty $\le 0.10\,\text{nm}$.
  • High-Speed On-The-Fly: Measures hundreds of targets per wafer in under 20 seconds.
$$\Delta I = I_{+1} - I_{-1} \propto \Delta x_{\text{overlay}} \implies \text{Overlay} = K \cdot \left(\frac{I_{+1} - I_{-1}}{I_{+1} + I_{-1}}\right)$$
Module 4.3

The Edge Placement Error (EPE) Vector Budget

In advanced nodes, device yield is no longer determined by lithography overlay alone. It is dictated by the total Edge Placement Error (EPE)—the absolute physical vector difference between intended and actual feature boundaries.

EPE combines four independent error distributions: lithography CD variation ($ ext{CDU}_{ ext{litho}}$), etch bias variation ($ ext{CDU}_{ ext{etch}}$), overlay error ($ ext{OVL}$), and Line Edge Roughness ($ ext{LER}$). Balancing the EPE budget within $\le 2.0\,\text{nm}$ is the primary challenge of modern scaling.

  • EPE Formula: $\text{EPE} = \frac{\Delta\text{CD}_{\text{litho}}}{2} + \frac{\Delta\text{CD}_{\text{etch}}}{2} + \text{Overlay} + \frac{\text{LWR}}{2}$.
  • Self-Aligned Integration: SADP/SAQP and self-aligned contacts (SAC) eliminating overlay terms.
  • Sub-3nm EPE Limit: Margin must remain strictly below half the dielectric spacer thickness.
$$\text{EPE}_{\text{total}} = \sqrt{\left(\frac{3\sigma_{\text{CD}}}{2}\right)^2 + (3\sigma_{\text{OVL}})^2 + \left(\frac{3\sigma_{\text{LWR}}}{2}\right)^2} \le \text{EPE}_{\text{budget}}$$
⚡ Interactive Laboratory L4
DBO Overlay Asymmetry & EPE Vector Budget Solver
Calculate physical overlay displacement from DBO diffraction asymmetry and evaluate the total Edge Placement Error (EPE) vector budget.
+1st Diffraction Order Intensity105a.u.
-1st Diffraction Order Intensity95a.u.
3-Sigma CD Variation (nm)1.2nm
Line Width Roughness LWR (nm)1.5nm
REAL-TIME SIMULATION TELEMETRY
Interactive physics simulator running client-side transfer models, carrier drift-diffusion kinetics, and boundary potential solvers.
Calculated Overlay Shift
-- nm
Total 3-Sigma EPE Budget
-- nm
Yield Margin (< 2.0 nm)
--
🎓 Level 4 Examination
Level 4 Conceptual Mastery Assessment
What primary advantage does Optical Critical Dimension (OCD) scatterometry offer over top-down CD-SEM?
How does Diffraction-Based Overlay (DBO) determine nanometer layer misalignment?
What is Edge Placement Error (EPE)?

Level 4 Completed: Optical Scatterometry (OCD) & Overlay Metrology Mastery Certificate

Conferred for mastery of Level 4 (Academic Level 4 • Undergraduate) curriculum, simulation laboratory, and assessment evaluation.

Academic Level 5 • Master's
Transmission Electron Microscopy (TEM/STEM) & HAADF Z-Contrast
Slice sub-30nm lamellae via Dual-Beam FIB and resolve atomic columns with aberration-corrected HAADF-STEM.
Module 5.1

Dual-Beam FIB Lamella Preparation

When fab engineers need definitive physical proof of a sub-2nm gate stack, they must view the structure in a Transmission Electron Microscope (TEM). But electrons cannot penetrate a 775-micron thick silicon wafer—the sample must be thinned to less than 30 nanometers!

This is achieved using a Dual-Beam FIB-SEM (Focused Ion Beam). A gallium ($Ga^+$) or argon/xenon plasma beam mills away silicon surrounding the target site, extracts a microscopic chunk with a nanomanipulator needle, and polishes it down to an ultra-thin 'lamella' just tens of atoms thick!

  • FIB Milling: Focused $Ga^+$ or $Xe^+$ ion beam performing micro-machining with nanometer precision.
  • Omniprobe Nanomanipulator: Microscopic tungsten needle picking and welding the sample to a TEM grid.
  • Low-kV Polishing: Final $500\,\text{eV}$ ion polish removing amorphous surface damage layers.
$$\text{Lamella Thickness } t_{\text{sample}} \le 20-30\,\text{nm} \quad (\text{Electron Transparency Regime})$$
Module 5.2

Aberration-Corrected STEM & HAADF Z-Contrast Imaging

In a Scanning Transmission Electron Microscope (STEM), a sub-Angstrom electron beam probe ($d_{ ext{probe}} < 0.08\,\text{nm}$) scans across the ultra-thin lamella. High-angle scattered electrons are collected by a High-Angle Annular Dark-Field (HAADF) detector.

HAADF scattering is governed by Rutherford scattering at high angles ($> 50\,\text{mrad}$). The image intensity is proportional to the square of the atomic number ($I \propto Z^{1.7-2.0}$). Heavy elements (like Hafnium $Z=72$ in gate dielectrics or Tungsten $Z=74$) glow brilliantly white, while silicon ($Z=14$) appears dark gray, resolving individual atomic columns!

  • HAADF Z-Contrast: Intensity scales as $I \propto Z^2$, providing immediate elemental contrast.
  • Spherical Aberration Corrector ($C_s$): Hexapole lens assemblies focusing electron probes below 0.5 Angstroms.
  • Atomic Column Resolution: Directly visualizing oxygen vacancies and interfacial dipole layers.
$$I_{\text{HAADF}} \propto \sigma_{\text{Rutherford}} \propto Z^{1.7-2.0} \quad (\text{Atomic Number Z-Contrast})$$
Module 5.3

Atomic Elemental Mapping: EDS & EELS Spectroscopy

Seeing atoms is only half the battle; knowing which element is which requires analytical spectroscopy inside the TEM.

Energy-Dispersive X-ray Spectroscopy (EDS) collects characteristic X-rays emitted when electrons fill core-level vacancies. Electron Energy-Loss Spectroscopy (EELS) measures the exact kinetic energy lost by transmitted electrons as they excite atomic inner-shell electrons, generating 2D chemical bonding maps with atomic spatial resolution!

  • EDS Mapping: Simultaneous detection of all elements from Boron to Uranium.
  • EELS Fine Structure: Resolving chemical oxidation states (e.g. $Ti^{3+}$ vs $Ti^{4+}$) and bandgap profiles.
  • Fast Silicon Drift Detectors (SDD): Large solid-angle detectors capturing atomic maps in seconds.
$$\Delta E_{\text{loss}} = E_0 - E_{\text{transmitted}} = E_{\text{binding}}(\text{Core Shell}) \quad (\text{EELS Core Loss})$$
⚡ Interactive Laboratory L5
HAADF-STEM Z-Contrast Intensity & Lamella Thickness Simulator
Calculate Rutherford Z-contrast scattering intensity ratios for gate stack materials (Si, HfO2, W, TiN) as a function of lamella thickness.
Lamella Sample Thickness (nm)25nm
STEM Accelerating Voltage (kV)200kV
HAADF Inner Collection Angle (mrad)60mrad
REAL-TIME SIMULATION TELEMETRY
Interactive physics simulator running client-side transfer models, carrier drift-diffusion kinetics, and boundary potential solvers.
HfO2 / Si Intensity Ratio
-- x
W / Si Intensity Ratio
-- x
Beam Broadening in Sample
-- nm
🎓 Level 5 Examination
Level 5 Conceptual Mastery Assessment
Why must a wafer sample be thinned into a 20-30 nm lamella using a Dual-Beam FIB before TEM imaging?
In HAADF-STEM imaging, why do heavy atoms like Hafnium (Z=72) and Tungsten (Z=74) appear much brighter than Silicon (Z=14)?
What analytical technique inside a STEM measures characteristic X-rays to identify elemental composition across a device cross-section?

Level 5 Completed: Transmission Electron Microscopy (TEM/STEM) & HAADF Z-Contrast Mastery Certificate

Conferred for mastery of Level 5 (Academic Level 5 • Master's) curriculum, simulation laboratory, and assessment evaluation.

Academic Level 6 • Ph.D.
Multi-Beam E-Beam Inspection & Machine Learning Defect Review
Harness 300+ parallel e-beams, detect buried electrical opens with voltage contrast, and classify defects with deep CNNs.
Module 6.1

The Throughput Bottleneck: Multi-Beam E-Beam Inspection

Optical darkfield wafer inspection tools scan an entire 300mm wafer in minutes, but cannot detect sub-10nm non-scattering defects. Single-beam e-beam inspection has nanometer resolution, but scanning a whole wafer with one beam would take over six months!

The breakthrough is Multi-Beam E-Beam Inspection (MBI). Using micro-electro-mechanical (MEMS) aperture arrays and magnetic lens splitting, modern systems split a primary electron source into an array of $334$ to over $1,000$ parallel beamlets, boosting inspection throughput by 1,000x to inspect whole wafers in hours!

  • MEMS Aperture Arrays: Splitting one gun into hundreds of coherent micro-beamlets.
  • Parallel Multi-Detector: Dedicated electron detectors for each individual beamlet.
  • High-Speed Defect Capture: Sensitivity to sub-5nm bridges, breaks, and missing contacts.
$$\text{Throughput} \propto N_{\text{beamlets}} \times \frac{I_{\text{beamlet}}}{\text{Pixel Size}^2} \quad (N \ge 334 \text{ beams})$$
Module 6.2

Voltage Contrast Inspection: Finding Invisible Opens & Shorts

Some of the most dangerous semiconductor defects are physically invisible: a contact plug that looks perfectly etched and filled, but fails to make electrical contact with the transistor drain underneath (a 'floating open').

Fabs detect this using Voltage Contrast (VC) inspection. The e-beam deposits a controlled surface electrical charge. Connected contact plugs bleed the charge harmlessly away to the grounded silicon substrate and appear bright. Floating contacts accumulate positive charge, which traps secondary electrons from escaping—causing the defect to glow dark!

  • Floating Open: Accumulates positive surface charge ($V_{ ext{surf}} > 0$), trapping electrons $ o$ Appears DARK.
  • Grounded Plug: Dissipates charge to substrate, secondary electrons escape $ o$ Appears BRIGHT.
  • Electrical Defect Detection: Finding 1 defective contact out of 100 billion without cutting the wafer.
$$V_{\text{surface}} = \frac{Q_{\text{accumulated}}}{C_{\text{contact}}} = \frac{(I_{\text{beam}} - I_{\text{SE}}) \cdot \tau_{\text{dwell}}}{C_{\text{plug}}}$$
Module 6.3

Deep Learning Automated Defect Classification (ADC)

Modern inspection tools capture millions of potential defect images per day. Human review is impossible. Foundries deploy Deep Convolutional Neural Networks (CNNs) for Automated Defect Classification (ADC).

The AI classifier ingests multi-channel SEM images, separates false-positive nuisance defects (grain noise, line roughness) from true yield-killer defects, and classifies killer defects into fine categories (bridge, pinch, CMP scratch, gate void) with $> 99\%$ accuracy at 100,000 images per hour!

  • Nuisance Filtering: Discarding 98% of non-killing surface noise detections.
  • Fine Classification: Identifying root-cause mechanism from morphology.
  • Real-Time Yield Feedback: Halting upstream etch or litho tools within minutes of defect excursions.
$$P(\text{Class}_k \mid \mathbf{I}_{\text{SEM}}) = \text{Softmax}\left(\mathbf{W} \cdot f_{\text{CNN}}(\mathbf{I}) + \mathbf{b}\right) > 0.99$$
⚡ Interactive Laboratory L6
Multi-Beam Inspection Throughput & Voltage Contrast Simulator
Model multi-beam e-beam wafer inspection throughput, calculate scan time vs single beam, and simulate voltage contrast charging potential.
Number of Parallel Beamlets334beams
Pixel Inspection Resolution (nm)8nm
Wafer Inspection Area (%)20%
REAL-TIME SIMULATION TELEMETRY
Interactive physics simulator running client-side transfer models, carrier drift-diffusion kinetics, and boundary potential solvers.
Multi-Beam Inspection Time
-- hours
Throughput Speedup Factor
-- x
Voltage Contrast Delta V
-- V
🎓 Level 6 Examination
Level 6 Conceptual Mastery Assessment
Why is Multi-Beam E-Beam Inspection (MBI) revolutionary for leading-edge semiconductor foundries?
How does Voltage Contrast (VC) inspection detect a contact via that has failed to touch the underlying source/drain diffusion (a floating open)?
What is the primary role of Deep Convolutional Neural Networks (CNNs) in automated defect review?

Level 6 Completed: Multi-Beam E-Beam Inspection & Machine Learning Defect Review Mastery Certificate

Conferred for mastery of Level 6 (Academic Level 6 • Ph.D.) curriculum, simulation laboratory, and assessment evaluation.

Academic Level 7 • Ph.D. & Technical Fellow
3D CFET Metrology Architecture & Hybrid Run-to-Run APC
Pioneer non-destructive buried channel profiling, CD-SAXS, and multi-sensor Bayesian APC control.
Module 7.1

The Buried 3D Metrology Crisis in CFET

In 3D Complementary FET (CFET) architectures, an nFET nanosheet stack is fabricated directly on top of a pFET nanosheet stack. While CD-SEM and OCD easily probe the top nFET, the bottom pFET channels are completely buried beneath 100 nm of active silicon, metal gates, and dielectric isolation!

Top-down CD-SEM is completely blind to buried nanosheet dimensions, footing, and inner spacer recesses. Measuring these hidden 3D structures non-destructively requires an entirely new metrology paradigm combining X-ray physics and multi-sensor Bayesian inversion.

  • Vertical 3D Stacking: nFET over pFET nanosheets separated by sub-15nm dielectric isolation.
  • Optical Extinction Limits: Visible and DUV light absorbed completely by upper silicon nanosheets.
  • Buried Interface Parameters: Requiring measurement of bottom sheet width, gate wrap-around, and inner spacer thickness.
$$\text{CFET Architecture: } \left[\text{Top nFET Nanosheets}\right] \Big/ \left[\text{Middle Dielectric Isolation}\right] \Big/ \left[\text{Buried pFET Nanosheets}\right]$$
Module 7.2

Critical Dimension Small-Angle X-ray Scattering (CD-SAXS)

To see right through top nanosheets into buried layers without cutting the wafer, fabs deployed in-fab Critical Dimension Small-Angle X-ray Scattering (CD-SAXS).

High-brightness X-rays ($\lambda = 0.05-0.15\,\text{nm}$) penetrate through 300mm silicon with near-zero absorption. By recording the small-angle transmission diffraction pattern across multiple incident tilt angles ($\pm 60^\circ$), CD-SAXS reconstructs the complete 3D electron density map, measuring buried pFET nanosheet dimensions with 0.1 nm accuracy!

  • Hard X-Ray Transmission: Complete penetration through dense metal gates and silicon stacks.
  • Reciprocal Space Inversion: Fourier transform mapping electron density distributions in 3D.
  • Atomic-Scale Precision: Unambiguous extraction of top vs buried nanosheet width differences.
$$q = \frac{4\pi}{\lambda}\sin(\theta/2), \quad I(\mathbf{q}) = \left|\int \rho_e(\mathbf{r}) e^{-i \mathbf{q}\cdot\mathbf{r}}\,d\mathbf{r}\right|^2 \quad (\text{CD-SAXS Diffraction})$$
Module 7.3

Hybrid Metrology Sensor Fusion & Run-to-Run (R2R) APC

No single metrology technique can measure everything. CD-SEM provides high lateral precision; OCD provides high throughput and vertical profiles; CD-SAXS measures buried layers; and AFM provides absolute reference heights.

The pinnacle of foundry engineering is Hybrid Metrology Sensor Fusion. Bayesian inference algorithms synthesize simultaneous data streams from OCD, CD-SEM, and CD-SAXS, breaking individual parameter correlations and feeding real-time offsets directly into lithography scanners and etch bias controllers in a closed-loop Run-to-Run (R2R) Advanced Process Control network.

  • Bayesian Sensor Fusion: Combining diverse physics models to break optical cross-correlations.
  • Closed-Loop R2R Control: Dynamic scanner dose, focus, and etch bias adjustments per wafer.
  • Angstrom-Era Zero Excursion: Preventing multi-million dollar scrap events across 100k wafer starts.
$$P(\mathbf{p} \mid \mathbf{y}_{\text{OCD}}, \mathbf{y}_{\text{SEM}}, \mathbf{y}_{\text{SAXS}}) \propto P(\mathbf{p}) \prod_{i} \exp\left(-\frac{1}{2}(\mathbf{y}_i - \mathbf{f}_i(\mathbf{p}))^T \mathbf{\Sigma}_i^{-1}(\mathbf{y}_i - \mathbf{f}_i(\mathbf{p}))\right)$$
⚡ Interactive Laboratory L7
CFET Buried Nanosheet CD-SAXS & Hybrid Fusion Solver
Simulate CD-SAXS diffraction intensity peaks and calculate hybrid metrology uncertainty reduction for buried CFET nanosheet channels.
Scattering Vector q (nm^-1)0.15nm^-1
Top nFET Nanosheet CD (nm)18nm
Buried pFET Nanosheet CD (nm)21nm
REAL-TIME SIMULATION TELEMETRY
Interactive physics simulator running client-side transfer models, carrier drift-diffusion kinetics, and boundary potential solvers.
SAXS Diffraction Peak Intensity
-- a.u.
Buried CD Offset (Bot - Top)
-- nm
Hybrid Fused Measurement Uncertainty
-- pm
🎓 Level 7 Examination
Level 7 Conceptual Mastery Assessment
Why is conventional CD-SEM incapable of measuring the bottom pFET nanosheet channel in a 3D CFET device?
How does Critical Dimension Small-Angle X-ray Scattering (CD-SAXS) measure buried nanostructures non-destructively?
What is the core principle of Hybrid Metrology Sensor Fusion in advanced run-to-run (R2R) process control?

Level 7 Completed: 3D CFET Metrology Architecture & Hybrid Run-to-Run APC Mastery Certificate

Conferred for mastery of Level 7 (Academic Level 7 • Ph.D. & Technical Fellow) curriculum, simulation laboratory, and assessment evaluation.

🏅
Distinguished Fellow in Nano-Metrology, OCD Scatterometry & Yield Analytics
Highest academic honor conferred by ChipFoundryServices OS for demonstrated mastery across all 7 curriculum tiers, interactive simulation laboratories, and verified examination standards.