Channel Strain Engineering is the technique of introducing controlled mechanical stress into the transistor channel to modify the silicon crystal lattice and enhance carrier mobility — achieving 20-80% mobility improvement for electrons (nMOS) and 30-100% for holes (pMOS) through tensile or compressive strain, enabling 15-40% higher drive current at same gate length, and utilizing stress sources including strained epitaxial source/drain (eSi:C for nMOS, eSiGe for pMOS), stress liners (tensile SiN for nMOS, compressive SiN for pMOS), and substrate engineering to maintain performance scaling as transistors shrink below 10nm gate length.
Strain Fundamentals:
- Mobility Enhancement: strain modifies band structure; reduces effective mass; increases carrier mobility; tensile strain benefits electrons (nMOS); compressive strain benefits holes (pMOS)
- Strain Types: tensile strain (lattice stretched) increases electron mobility by 20-80%; compressive strain (lattice compressed) increases hole mobility by 30-100%
- Strain Magnitude: typical strain 0.5-2.0 GPa (0.5-2% lattice deformation); higher strain gives more mobility improvement; but reliability concerns above 2 GPa
- Strain Direction: uniaxial strain (along channel) most effective; biaxial strain (in-plane) also beneficial; triaxial strain (3D) less common
Strained Source/Drain Epitaxy:
- SiGe for pMOS: epitaxial Si₁₋ₓGeₓ with x=0.25-0.50 (25-50% Ge); larger Ge atoms create compressive strain in channel; 30-100% hole mobility improvement
- Si:C for nMOS: epitaxial Si with 0.5-2.0% carbon substitutional doping; smaller C atoms create tensile strain in channel; 20-50% electron mobility improvement
- Growth Process: selective epitaxial growth at 600-800°C; in-situ doping with B (pMOS) or P (nMOS); thickness 20-60nm; strain transfer to channel
- Strain Transfer: strain from S/D epitaxy transfers to channel through silicon lattice; effectiveness depends on S/D proximity to channel (5-20nm spacing)
Stress Liner Technology:
- Tensile SiN for nMOS: silicon nitride film with tensile stress (1-2 GPa); deposited over nMOS transistors; creates tensile strain in channel; 10-30% electron mobility improvement
- Compressive SiN for pMOS: silicon nitride film with compressive stress (1-2 GPa); deposited over pMOS transistors; creates compressive strain in channel; 15-40% hole mobility improvement
- Dual Stress Liner (DSL): separate liners for nMOS and pMOS; requires additional mask; optimizes strain for both transistor types
- Contact Etch Stop Layer (CESL): stress liner also serves as etch stop during contact formation; dual function; thickness 20-80nm
Strain Mechanisms:
- Lattice Mismatch: SiGe has 4% larger lattice constant than Si; creates compressive strain when grown on Si; Si:C has smaller lattice; creates tensile strain
- Stress Transfer: stress from S/D epitaxy or liner transfers to channel; magnitude depends on geometry, distance, and material properties
- Band Structure Modification: strain splits degenerate valleys in Si conduction band (nMOS) or valence band (pMOS); reduces effective mass; increases mobility
- Scattering Reduction: strain reduces phonon scattering; increases mean free path; further enhances mobility
Mobility Enhancement:
- nMOS Electron Mobility: unstrained Si: 400-500 cm²/V·s; with Si:C S/D: 500-700 cm²/V·s (25-40% improvement); with tensile liner: 550-750 cm²/V·s (30-50% improvement)
- pMOS Hole Mobility: unstrained Si: 150-200 cm²/V·s; with SiGe S/D: 250-400 cm²/V·s (60-100% improvement); with compressive liner: 200-300 cm²/V·s (30-50% improvement)
- Combined Effect: S/D strain + liner strain can be additive; total mobility improvement 50-150% possible; but diminishing returns above certain strain level
- Saturation Effects: mobility improvement saturates at high strain (>2 GPa) or high electric field; practical limit to strain engineering
Process Integration:
- S/D Recess Etch: etch Si in S/D regions; depth 20-60nm; creates cavity for epitaxial growth; critical dimension control ±2nm
- Selective Epitaxy: grow SiGe (pMOS) or Si:C (nMOS) in recessed regions; selective to Si; no growth on dielectric; temperature 600-800°C; growth rate 1-5 nm/min
- Stress Liner Deposition: plasma-enhanced CVD (PECVD) of SiN; control stress by deposition conditions (temperature, pressure, gas flow); thickness 20-80nm
- Dual Liner Process: deposit tensile liner; mask pMOS; etch nMOS liner; deposit compressive liner; mask nMOS; etch pMOS liner; 2 additional masks
Performance Impact:
- Drive Current: 15-40% higher Ion due to mobility enhancement; enables higher frequency or lower voltage at same performance
- Transconductance: 20-50% higher gm; improves analog circuit performance; better gain and bandwidth
- Saturation Velocity: strain increases saturation velocity by 10-20%; benefits short-channel devices; improves high-frequency performance
- Threshold Voltage: strain can shift Vt by ±20-50mV; must be compensated by work function or doping adjustment
Strain in FinFET:
- Fin Strain: strain in narrow fins (5-10nm width) differs from planar; quantum confinement affects strain; requires 3D strain modeling
- S/D Epitaxy: SiGe or Si:C grown on fin sidewalls; strain transfer to fin channel; effectiveness depends on fin width and height
- Stress Liner: liner wraps around fin; 3D stress distribution; more complex than planar; but still effective
- Strain Relaxation: narrow fins may partially relax strain; reduces effectiveness; requires optimization of fin geometry
Strain in GAA/Nanosheet:
- Nanosheet Strain: strain in suspended nanosheets (5-8nm thick, 20-40nm wide); different from bulk or fin; requires careful engineering
- S/D Epitaxy: SiGe or Si:C grown around nanosheet stack; strain transfer through nanosheet edges; effectiveness depends on sheet dimensions
- Strain Uniformity: achieving uniform strain across multiple stacked sheets challenging; top and bottom sheets may have different strain
- Inner Spacer Impact: inner spacers between sheets affect strain transfer; must be considered in strain engineering
Reliability Considerations:
- Defect Generation: high strain (>2 GPa) can generate dislocations or defects; reduces reliability; limits maximum strain
- Strain Relaxation: strain may relax over time at operating temperature; reduces mobility benefit; must be stable for 10 years
- Electromigration: strain affects electromigration in S/D and contacts; can improve or degrade depending on strain type; requires testing
- Hot Carrier Injection (HCI): strain affects HCI; higher mobility increases carrier energy; may degrade HCI reliability; trade-off
Design Implications:
- Mobility Models: SPICE models must include strain effects; mobility as function of strain; affects timing and power analysis
- Vt Compensation: strain-induced Vt shift must be compensated; work function or doping adjustment; maintains target Vt
- Layout Optimization: strain effectiveness depends on layout; S/D proximity, liner coverage; layout-dependent effects (LDE)
- Analog Design: higher gm from strain benefits analog circuits; better gain, bandwidth, and noise; enables lower power analog
Industry Implementation:
- Intel: pioneered strain engineering at 90nm node (2003); continued through 14nm, 10nm, 7nm; SiGe S/D for pMOS, Si:C for nMOS, dual stress liners
- TSMC: implemented strain at 65nm node; optimized for each node; N5 and N3 use advanced strain techniques; SiGe with 40-50% Ge content
- Samsung: similar strain techniques; 3nm GAA uses strain in nanosheet channels; optimized S/D epitaxy and stress liners
- imec: researching advanced strain techniques for future nodes; exploring alternative materials and geometries
Cost and Economics:
- Process Cost: strain engineering adds 5-10 mask layers; epitaxy, liner deposition, additional lithography; +10-15% wafer processing cost
- Performance Benefit: 15-40% drive current improvement justifies cost; enables frequency targets or power reduction
- Yield Impact: epitaxy defects and strain-induced defects can reduce yield; requires mature process; target >98% yield
- Alternative: without strain, would need smaller gate length for same performance; strain enables performance at larger gate length; reduces cost
Scaling Trends:
- 28nm-14nm Nodes: strain engineering mature; SiGe S/D with 25-35% Ge; dual stress liners; 30-60% mobility improvement
- 10nm-7nm Nodes: increased Ge content (35-45%); optimized liner stress; 40-80% mobility improvement; critical for FinFET performance
- 5nm-3nm Nodes: further optimization; 40-50% Ge; advanced liner techniques; strain in GAA nanosheets; 50-100% mobility improvement
- Future Nodes: approaching limits of strain engineering; >50% Ge difficult; alternative channel materials (Ge, III-V) may replace strained Si
Comparison with Alternative Approaches:
- vs Channel Material Change: strain is cheaper and more manufacturable than Ge or III-V channels; but lower mobility improvement; strain is near-term solution
- vs Gate Length Scaling: strain provides performance without gate length scaling; reduces short-channel effects; complementary to scaling
- vs Voltage Scaling: strain enables performance at lower voltage; reduces power; complementary to voltage scaling
- vs Multi-Vt: strain improves performance for all Vt options; complementary to multi-Vt design; both used together
Advanced Strain Techniques:
- Embedded SiGe Stressors: SiGe regions embedded in S/D; higher Ge content (60-80%); larger strain; but integration challenges
- Strain-Relaxed Buffer (SRB): grow relaxed SiGe layer; then grow strained Si on top; biaxial strain; used in some SOI processes
- Ge-on-Si: grow Ge channel on Si substrate; high hole mobility (1900 cm²/V·s); but high defect density; research phase
- III-V on Si: grow InGaAs or GaAs on Si; ultra-high electron mobility (>2000 cm²/V·s); but integration challenges; research phase
Future Outlook:
- Continued Optimization: strain engineering will continue at 2nm and 1nm nodes; incremental improvements; approaching fundamental limits
- Material Transition: beyond 1nm, may transition to Ge or III-V channels; strain engineering in new materials; different techniques required
- Heterogeneous Integration: combine strained Si (logic) with Ge (pMOS) and III-V (nMOS) on same chip; ultimate performance; integration challenges
- Quantum Effects: at <5nm dimensions, quantum confinement affects strain; requires quantum mechanical modeling; new physics
Channel Strain Engineering is the most successful mobility enhancement technique in CMOS history — by introducing controlled tensile or compressive stress through epitaxial source/drain and stress liners, strain engineering achieves 20-100% mobility improvement and 15-40% higher drive current, enabling continued performance scaling from 90nm to 3nm nodes and beyond while providing a manufacturable and cost-effective alternative to exotic channel materials, making it an indispensable tool for maintaining Moore's Law in the face of fundamental scaling limits.