Understanding Thermal Resistance in Liquid-Cooled Power Electronics

In modern power conversion, managing heat is just as critical as managing current. Every watt that passes through a semiconductor eventually becomes heat, and that heat must go somewhere. In high power inverter and converter systems - especially those using silicon carbide (SiC) or gallium nitride (GaN) devices - the ability to remove heat efficiently defines the boundary between high performance and premature failure.

Wide bandgap (WBG) technologies have changed the design landscape. Devices now switch faster, operate at higher voltages, and deliver unprecedented power densities. Yet those same strengths bring new challenges: high local heat flux, tight packaging, and higher junction temperature limits that demand more sophisticated cooling.

At the center of this challenge lies a deceptively simple metric: thermal resistance, expressed in degrees Celsius per watt (°C/W). It quantifies how much temperature rise occurs for each watt of heat flow and is as fundamental to thermal design as resistance is to electrical design.

This article provides a detailed technical discussion of thermal resistance in liquid cooled power electronics. It explains how to calculate it, what typical values mean, how it scales with flow rate and material choice, and what designers must consider for coolant temperatures of 65°C, 85°C, and 105°C.

1. Fundamentals of Thermal Resistance

1.1 Definition

Thermal resistance (Rθ) describes the relationship between temperature rise and power dissipation:

Rθ = ΔT / Q

where ΔT is the temperature difference between two reference points in °C and Q is heat flow in watts. A system with Rθ = 0.02 °C/W will rise 20°C when dissipating 1000 W. Lower values are better - they indicate a smaller temperature rise for a given load.

1.2 Why It Matters

Semiconductor performance and lifetime depend directly on temperature. In SiC MOSFETs, on resistance, switching losses, and avalanche ruggedness all degrade as junction temperature rises. Every 10°C increase roughly halves solder joint or bond wire life under cyclic stress.

Thermal resistance governs this relationship: Tjunction = Tcoolant,in + Q × Rθ,total. If Rθ,total is high, junction temperature climbs rapidly and reliability suffers. If it is low, the same system can handle more current or operate at higher ambient temperature with equal safety margin.

2. The Thermal Resistance Network

Heat moves through a sequence of physical layers, each with its own resistance:

Junction → Case/Baseplate → Thermal Interface → Cold Plate → Coolant → Heat Exchanger → Ambient

The total resistance is the sum: Rθ,total = Rθ(j–c) + Rθ(c–tim) + Rθ(tim–plate) + Rθ(plate–fluid) + Rθ(fluid–loop)

Some segments are fixed by device construction. Others - interface, cold plate, and coolant loop - are controllable through design. Those segments determine whether a system meets its temperature targets.

3. Segment by Segment Analysis

3.1 Junction to Case

This resistance reflects the conduction path from the semiconductor junction through the die attach, substrate, and baseplate.

Device Type	Rθ(j–c) (°C/W per die)
Discrete SiC MOSFET	0.2 to 0.6
Power Module (400–800 A)	0.03 to 0.08

This value is set by the manufacturer and defines the starting point for all further calculations.

3.2 Case to Plate (Thermal Interface)

Between the module and cold plate lies one of the most misunderstood bottlenecks: the thermal interface material (TIM). The conduction through this layer follows: Rθ,tim = t / (k × A) where t is thickness, k is thermal conductivity, and A is contact area.

Example: a 100 µm grease layer with k = 3 W/m·K over 25 cm² yields Rθ = 0.013 °C/W. This single layer can consume half the total thermal budget if applied thickly or unevenly. Metallic bonds - sintered silver, solder, or braze - reduce this by an order of magnitude to 0.001–0.003 °C/W.

3.3 Cold Plate Conduction

For a 3 mm aluminum plate (k = 205 W/m·K) and 25 cm² area: Rθ = 0.003 / (205 × 0.0025) = 0.0058 °C/W. Replacing aluminum with copper (k = 385 W/m·K) cuts this roughly in half. When heat sources are small compared to plate area, add spreading resistance - this can add 20–40% to the conduction term.

3.4 Convective Transfer to Coolant

The convective component is where liquid cooling earns its reputation: Rθ,conv = 1 / (h × Awet)

Flow Type	h (W/m²·K)
Laminar channel flow	500 to 2,000
Turbulent (moderate Re)	4,000 to 8,000
Micro channel or jet impingement	10,000 to 30,000

The Dittus–Boelter correlation estimates h for turbulent internal flow: Nu = 0.023 × Re^0.8 × Pr^0.4. For a 5 L/min water glycol stream through a 60 × 100 mm plate, h approaches 9,000–10,000 W/m²·K - good enough for cold plate resistances below 0.01 °C/W.

3.5 Coolant Temperature Rise

As the fluid absorbs heat, it warms along the flow path: ΔTfluid = P / (ṁ × cp). For 3 kW load at 5 L/min and cp = 3900 J/kg·K, ΔT ≈ 9.3°C. The outlet is roughly 9°C warmer than the inlet, so the device never sees ideal coolant temperature everywhere across its footprint.

4. Calculating Effective Thermal Resistance

Suppose a 3 kW inverter leg must maintain a maximum case temperature of 90°C with 65°C coolant inlet:

Rθ,target = (90 − 65) / 3000 = 0.0083 °C/W

At 85°C coolant and 120°C case: Rθ = 0.0117 °C/W. At 105°C coolant and 150°C case: Rθ = 0.015 °C/W. Higher coolant temperature relaxes the numeric target but simultaneously raises material stress and aging rates - high temperature systems require more robust TIMs, seals, and plate materials to maintain the same reliability.

5. Realistic Performance Targets

Application	Coolant (°C)	Rθ surface→fluid (°C/W)	Characteristics
Commercial	≤ 65	0.005 to 0.010	Aluminum cold plate, 5–10 LPM, glycol water
Extended Temp	≤ 85	0.008 to 0.012	Copper or Cu-Al hybrid, 8–12 LPM, high temp seals
MIL Grade	≤ 105	0.010 to 0.015	Copper nickel alloy, metallic TIM, 10–15 LPM

6. Flow Rate, Pressure Drop, and Pump Power

Increasing flow rate lowers Rθ,conv roughly as Q^-0.8 but increases pressure drop as Q². Pump power follows: Ppump = ΔP × Vdot / η. Doubling flow can cut thermal resistance by ~30% but can quadruple pressure drop. At 5 L/min and 0.2 bar, Ppump ≈ 17 W. At 10 L/min and 0.8 bar, Ppump ≈ 68 W. The optimal design minimizes total energy loss, not just Rθ.

7. Coolant Properties and Material Compatibility

Coolant	k (W/m·K)	cp (J/kg·K)	Typical Use
Water + 50% glycol	0.4 to 0.6	3,800 to 4,200	Standard industrial
PAO oil	0.13	2,000	High temp dielectric
Fluorinated (Novec)	0.06 to 0.08	1,100 to 1,400	Dielectric, low conductivity

Switching from water glycol to a dielectric can quadruple Rθ,conv if geometry and flow remain constant. Above 100°C, fluorocarbon (FKM or FFKM) or metal seals are preferred. Copper nickel or stainless structures are preferred for 105°C service.

8. Reliability Considerations

8.1 Thermal Cycling

Repeated power cycling generates expansion mismatches between die, substrate, and baseplate. Lower thermal resistance reduces the amplitude of each temperature swing, extending solder and wire bond life.

8.2 TIM Degradation

Greases can pump out and pads can dry. Both increase Rθ over time. Select materials with low volatility and mechanical stability at maximum case temperature. Phase change TIMs or metallic bonds provide consistent long-term performance.

8.3 Corrosion and Contamination

Galvanic couples between copper and aluminum accelerate corrosion in mixed metal systems. Maintain inhibitor chemistry and use deionized, filtered coolant.

9. Advanced Cooling Technologies

9.1 Pin Fin and Micro Channel Designs

Arrays of pins or narrow channels increase surface area and turbulence. Micro channels (0.1–0.5 mm) can reach h > 20,000 W/m²·K, achieving 0.002–0.005 °C/W. They require fine filtration and careful manufacturing but offer exceptional density for systems near 50 kW/L.

9.2 Jet Impingement Cooling

Coolant jets striking the heated surface directly disrupt boundary layers, creating local h up to 30,000 W/m²·K. Jet arrays are common in high performance computing and are now migrating into traction inverter research.

9.3 Additive Manufactured Cold Plates

3D printed metal plates enable internal lattice or spiral geometries that maximize uniformity and minimize mass. These can integrate flow manifolds, structural ribs, and mounting bosses directly into the design.

9.4 Embedded Cooling

Emerging modules embed coolant channels within substrates or busbars, removing interface layers entirely. The result is near-zero spreading resistance and higher allowable current density without sacrificing reliability.

10. Validation and Characterization

Steady state tests apply known power and record surface to coolant temperature rise. Transient thermal impedance (Zth) pulses power input and extracts time constant response to characterize internal layers. Flow and pressure mapping measures differential pressure to confirm channel uniformity. A rise greater than 15% in measured Rθ typically signals interface fatigue or corrosion onset.

11. Temperature Class Scenarios

Commercial (≤ 65°C Coolant)

Target Rθ ≈ 0.007 °C/W at 3 kW. Aluminum plate, standard glycol water, silicone-based TIM, 5–10 LPM. Focus: low cost, manufacturability, weight reduction.

Extended Temperature (≤ 85°C Coolant)

Target Rθ ≈ 0.010 °C/W at 3 kW. Copper or Cu-Al hybrid plate, graphite or phase change TIM, high temperature EPDM seals, 8–12 LPM. Focus: higher reliability under sustained high coolant temperature and vibration.

MIL Grade (≤ 105°C Coolant)

Target Rθ ≈ 0.015 °C/W at 3 kW. Copper nickel alloy or stainless cold plate, metallic bond TIM, FFKM or metal seals, 10–15 LPM at 0.8–1.0 bar. Focus: long-term stability, zero leak tolerance, compatibility with 105°C continuous coolant.

12. Practical Design Tradeoffs

Parameter Change	Improves	Tradeoff
Higher flow rate	Convective heat transfer	Pumping losses, noise, cavitation
Copper vs. aluminum	Conductivity and spreading	Weight and cost
Thinner TIM	Interface resistance	Surface flatness tolerance
Micro channels	High h, low Rθ	Clogging risk, filtration cost
Dual sided cooling	Overall Rθ	Complexity and packaging
Higher coolant temperature	Smaller radiator	Material stress, lower margin

13. Best Practice Guidelines

1. Define the thermal budget early. Determine maximum junction temperature, total power loss, and coolant inlet temperature. Derive a numeric Rθ target and allocate portions to each layer.
2. Model the full stack. Use multiphysics tools to combine electrical, thermal, and fluid dynamics analysis.
3. Prioritize surface flatness. Target ≤ 25 µm flatness and ≤ 0.8 µm Ra roughness for consistent TIM thickness. Uneven contact can increase interface resistance by 30% or more.
4. Select TIMs deliberately. For 105°C coolant applications, consider metallic TIMs or sintered silver to prevent pump-out and dry-out.
5. Validate under real thermal loads. Confirm measured gradients align with simulation within ±10%.
6. Manage flow distribution. Use CFD to design manifolds that maintain uniform Reynolds number across all channels.
7. Include margin for aging. Include at least 25% margin below maximum allowable Rθ to accommodate drift over lifetime.

14. Example Calculation

Consider a 3-phase SiC inverter leg dissipating 3 kW total losses with 85°C coolant inlet. Assume:

• Rθ(j–c) = 0.05 °C/W
• Rθ(c–tim) = 0.003 °C/W
• Rθ(plate–fluid) = 0.010 °C/W
• Rθ(fluid–loop) = 0.002 °C/W

Total Rθ,total = 0.065 °C/W. Temperature rise = 3000 W × 0.065 = 195°C. Junction = 85 + 195 = 280°C - unacceptable. To achieve a safe 150°C junction: Rθ,req = (150 − 85) / 3000 = 0.0217 °C/W. Achieving that requires at least a 3× improvement through reduced interface thickness, a higher-performance cold plate, and increased flow rate.

15. Conclusion

Thermal resistance defines the invisible boundary between electrical performance and real-world reliability. In power electronics - especially SiC-based systems operating at 400–800 V and hundreds of amps per phase - it determines whether theoretical power density can be realized in practice. Liquid cooling provides unmatched capability, but success depends on understanding every part of the thermal path from semiconductor to coolant.

At SiCore Research, thermal, electrical, and mechanical disciplines are integrated from the outset of every program. Quantitative thermal resistance budgeting, verified through analysis and test, forms the foundation for our scalable, high-density SiC platforms capable of operating reliably across 65°C, 85°C, and 105°C coolant regimes.

References

1. Advanced Cooling for Power Electronics. Electronics Cooling Magazine. 2017.
2. Cooling of Power Electronics. Mersen Technical Brochure. 2025.
3. Liquid Cooling Design Guide. Advanced Thermal Solutions. 2024.
4. IEEE Components, Packaging and Manufacturing Technology Society. Thermal Design in Power Electronics Packaging. Tutorial Notes. 2020.
5. T. H. Nguyen et al. Experimental Investigation of Jet Impingement Cooling for SiC Power Modules. Applied Thermal Engineering. 2022.

---

© 2025 SiCore Research LLC  ·  Michigan  ·  Partner Inquiry →