| name | jet-impingement-cooling |
| description | Jet impingement cooling for electronics and high-flux applications — confined single-phase liquid, submerged jets, micro-jet arrays, two-phase (boiling) impingement, CHF (critical heat flux), thermal resistance networks, Intel/IBM chip cooling, datacenter liquid cooling, IEC 62368. |
| metadata | {"priority":7,"promptSignals":{"phrases":["jet impingement cooling electronics","chip cooling impingement","submerged jet cooling","liquid jet cooling","two-phase impingement","high flux cooling jet"],"minScore":3}} |
Jet Impingement Cooling for Electronics — Complete Skill
Overview and Heat Flux Regimes
Heat flux hierarchy (achievable with various cooling methods):
| Cooling Method | Max q'' [W/cm²] |
|---|
| Natural convection (air) | 0.05–0.1 |
| Forced air convection | 0.5–5 |
| Cold plate (liquid) | 10–50 |
| Single-phase liquid jet impingement | 20–500 |
| Two-phase (boiling) jet impingement | 100–2,000 |
| Spray cooling | 200–1,000 |
| Microchannels | 500–3,000 |
Jet impingement advantage: direct jet on chip surface → eliminates chip-to-lid and TIM resistances
Limitation: fluidic connection directly to die; hermetic sealing required; deionized water handling
Single-Phase Liquid Jet Impingement
Heat Transfer Correlations
Submerged jet (fluid submerged in same liquid — no air entrainment):
Nu = f(Re, Pr, H/D, r/D)
Zuckerman-Lior (2006) review — recommended correlations:
For laminar (Re < 1,500): Nu_stag = 0.765 × Re^0.5 × Pr^0.4
For turbulent (Re > 3,000): Nu_stag = 0.455 × Re^0.63 × Pr^0.36
Circular micro-jet (D < 1 mm) — turbulent:
Nu_stag = 0.6 × Re^0.5 × Pr^0.4 × (D/d_nozzle)^0.0 [correlation for micro-orifices]
Water at 25°C key properties:
ρ = 997 kg/m³; μ = 0.89×10⁻³ Pa·s; k = 0.605 W/(m·K); Pr = 6.1; c_p = 4,180 J/(kg·K)
Performance at High Heat Flux
h achievable (single-phase water):
V_j = 5 m/s, D = 1 mm → Re = 997 × 5 × 0.001 / (0.89×10⁻³) = 5,600 (turbulent)
Nu_stag = 0.455 × 5,600^0.63 × 6.1^0.36 = 0.455 × 262 × 2.05 = 244
h = 244 × 0.605 / 0.001 = 147,620 W/(m²·K) = 14.8 W/(cm²·K)
Maximum q'' (single-phase, ΔT_max = 40°C chip-to-coolant):
q''_max = h × ΔT = 147,620 × 40 = 5.9 × 10⁶ W/m² = 590 W/cm²
Micro-Jet Arrays
Silicon micro-fabricated jets (MEMS):
D = 50–250 μm; pitch = 200–1,000 μm; etched in Si wafer bonded to chip
Effective h = 50,000–200,000 W/(m²·K); q'' = 100–800 W/cm²
Calculation for micro-jet array:
Mass flux per jet: G = ρ × V_j × π/4 × D² [kg/s per jet]
Array heat load: Q = N_jets × ṁ_jet × c_p × ΔT_coolant [thermal balance]
Pump pressure: ΔP ≈ 0.5 × ρ × V_j² × (1 + K_inlet + K_outlet) + 4 × f × L/D [Darcy-Weisbach for orifice array]
Exit flow velocity in confined array:
V_exit = N_jets × A_jet × V_j / A_exit_channel [must check Re_channel < critical to avoid back-flow]
Two-Phase (Boiling) Jet Impingement
Boiling Heat Transfer Regimes
Pool boiling reference (Nukiyama curve):
- Natural convection (q'' < 1 W/cm²)
- Nucleate boiling (1–100 W/cm²): bubbles nucleate at surface; h increases dramatically
- Critical Heat Flux (CHF): q''_CHF ≈ 100–300 W/cm² for saturated water at 1 atm
- Transition boiling: unstable; avoid
- Film boiling: vapor blanket; very low h; surface overheating → burnout
Jet impingement enhances CHF:
q''_CHF,jet / q''_CHF,pool ≈ 1 + 0.26 × We^0.25 × (1 - 0.9 × x_exit) [x_exit = vapor quality at exit]
Jet impingement CHF = 200–1,000 W/cm² (water) vs. 100 W/cm² (pool)
Kandlikar CHF correlation for jets:
q''_CHF = 0.149 × h_fg × ρ_g × [σ × g × (ρ_f - ρ_g) / ρ_g²]^0.25 × (1 + 0.0022 × Re^0.8 × Pr^0.4)
Weber number (jet impingement):
We = ρ_f × V_j² × D / σ [ratio of inertia to surface tension; We > 1 for impingement to overcome surface tension]
Subcooled Flow Boiling
Chen correlation (subcooled flow boiling):
h_tp = h_macro + h_micro [two-component model]
h_macro = F × h_Dittus-Boelter [F = enhancement factor from Chisholm]
h_micro = S × h_Cooper_pool_boiling [S = suppression factor; Cooper pool boiling correlation]
Critical heat flux in impingement:
q''_CHF_impingement ≈ 1.5–3× q''_CHF_pool (at same conditions)
Subcooling enhances CHF: q''_CHF increases ~10% per 10°C subcooling (water)
Working fluid selection (two-phase):
Water: highest CHF; corrosive to metals without treatment; dielectric issue (use DI water)
FC-72 (3M Fluorinert): dielectric; safe for immersion; q''_CHF ≈ 15 W/cm² (much lower than water)
R-134a: refrigerant; phase change at low pressure; COP benefits; hermetic system
Novec 7100 (3M): dielectric; low GWP; q''_CHF ≈ 20–30 W/cm²
Thermal Resistance Network
Chip to coolant resistance (single-phase):
R_total = R_chip + R_TIM + R_lid + R_TIM2 + R_lid-to-jet [traditional with lid]
Direct die cooling (no lid):
R_total = R_chip_junction + R_chip_bulk + R_jet_surface [much lower]
Junction to ambient:
T_j = T_coolant_in + q × (R_total + R_coolant_rise)
R_coolant_rise = 1 / (ṁ_total × c_p) [°C/W; temperature rise of coolant as it picks up heat]
Chip thermal resistance:
R_jc = (T_junction - T_case) / q [°C/W; from datasheet; typically 0.1–2.0 °C/W for server CPUs]
With direct impingement (no TIM or lid):
T_junction = T_coolant + q × (R_jc + R_spreading + 1/h_avg/A_chip)
A_chip = chip area [m²]; h_avg from jet correlation
Target thermal resistance: data center GPUs: R_ja ≤ 0.2 °C/W; server CPUs: R_ja ≤ 0.1 °C/W (at 300 W TDP)
System Design
Flow and Pressure Requirements
Single 10 mm × 10 mm chip, q'' = 200 W/cm² = 2×10⁶ W/m²:
Q_chip = 200 W × A_chip = 200 × 10⁻⁴ = 2 W → wrong unit: Q = q'' × A = 2×10⁶ × (0.01)² = 200 W
With ΔT_coolant = 15°C: ṁ = Q / (c_p × ΔT) = 200 / (4180 × 15) = 0.0032 kg/s = 3.2 g/s per chip
Array of 20 chips (rack server):
ṁ_total = 20 × 3.2 = 64 g/s = 0.064 L/s = 3.8 L/min
Pump selection:
ΔP ≈ 50–200 kPa (jet array + manifold + piping)
Pump power: P = ΔP × Q_flow / η = 100,000 × 0.064×10⁻³ / 0.6 = 10.7 W (per server module)
Manifold Design
Inlet manifold: ensures uniform flow to all jets; pressure drop < 10% of jet pressure drop for ± 5% flow uniformity
Exit manifold: collects heated coolant; avoid vapor trap in two-phase systems
Header sizing:
V_header < V_jet / 5 (ensure low header velocity → good distribution)
A_header = N_jets × A_jet × V_j / V_header_max
Sealing and Reliability
Leakage concerns: direct die cooling → catastrophic failure on leak
Double O-ring seal: primary + secondary; leak detection between seals
Material: EPDM (water compatible), Viton (dielectric fluid), PTFE encapsulated
Deionized water system:
Resistivity: > 1 MΩ·cm (prevent electrolytic corrosion)
pH control: 6.5–8.0
Dissolved oxygen: < 1 ppb (prevents copper corrosion in cold plates)
Ion exchange resin: inline filter; regenerate annually
Standards and Design Codes
| Standard | Scope |
|---|
| IEC 62368-1 | Audio/video, IT, communication equipment safety — liquid cooling |
| JEDEC JESD15-3 | System level ESD and liquid cooling compatibility |
| ASHRAE TC9.9 | Liquid cooling for data centers |
| IPC-SM-785 | Guidelines for accelerated life testing (thermal) |
| SEMI F57 | Semiconductor grade DI water purity requirements |
| ASTM D1193 | Reagent grade water specifications |
Output
Provide: target heat flux q'' [W/cm²] and chip area [mm²], coolant type (water/FC-72/Novec) and inlet temperature [°C], jet diameter D [mm] and jet velocity V_j [m/s], Reynolds number Re and stagnation Nusselt number Nu_stag, heat transfer coefficient h [W/(m²·K)], chip-to-coolant thermal resistance R [°C/W], junction temperature T_j [°C] at full TDP, mode (single-phase/two-phase), CHF margin q''_CHF / q''_applied (target > 2×), total flow rate ṁ [g/s] and ΔT_coolant [°C], pump pressure [kPa] and pump power [W], micro-jet array parameters (D [μm], pitch [μm], N_jets), sealing specification, coolant purity requirements (DI water MΩ·cm), and applicable standard (ASHRAE TC9.9, JEDEC JESD15-3, IEC 62368-1).
