What Is Thermal Conductivity?
Thermal conductivity is the property that tells you how easily heat moves through a material when there is a temperature difference across it. The symbol is k, or sometimes λ, and the standard unit is W/(m·K). If a material has a high k value, heat flows through it fast. If k is low, heat moves slowly.
I've spent fifteen years in building insulation and heat-sink design, and every time someone throws around "thermal conductivity" without context I have to bite my tongue. Datasheet numbers are measured under perfect lab conditions. Real parts in real systems almost never see those exact conditions.
Lab value vs real part
Lab thermal conductivity is measured on a flat, clean, defect-free sample with perfect contact pressure and no air gaps. In the field you have surface roughness, oxide layers, contact pressure variation, and often a thin air film that kills effective conductivity. A copper block might show 400 W/(m·K) in the textbook, but bolt two copper heat sinks together with 2 MPa pressure and you're lucky to see 50 000 W/(m²·K) at the interface.
Metals
Pure copper at room temperature is around 401 W/(m·K). Oxygen-free C10100 is a few points higher, commercial C11000 a few points lower. Aluminum 6061 sits at 167–180 W/(m·K) depending on temper. 6063 is usually 200–210 because it's extruded and the grain structure lines up better. Pure silver beats everything at ~430, but nobody uses it outside of a few RF cavities.
Non-metals that surprise people
Diamond (CVD or single-crystal) runs 1000–2200 W/(m·K). We use it for laser-diode submounts and GaN HEMT gates when money is no object. Graphite in-plane (pyrolytic or graphene sheets) can hit 1500–2000 along the plane, drops to 6 perpendicular. That anisotropy is why you have to be careful how you orient graphite heat spreaders in phones.
Ceramics and fillers
Aluminum nitride (AlN) is 170–220 W/(m·K) in production grades. Beryllium oxide was 250–300 but got banned in most places. Boron nitride (h-BN) platelets give you 300–600 in-plane when aligned in a polymer, maybe 30 through-plane. Plain alumina (Al2O3) is only 30–35, yet 90 % of LED boards still use it because it's cheap and dielectric.
Polymers and greases
Unfilled epoxy or polyurethane is 0.2–0.3 W/(m·K). Load it with 70 vol% alumina or boron nitride and you can push 2–4 W/(m·K). Thermal pads you buy off Digi-Key are usually 1–8 W/(m·K) bulk, but the real bottleneck is the contact resistance at each face. Phase-change materials start at 3–5, liquid metal (gallium) interfaces can get you above 30 if you can live with the mess.
Typical values I actually use when I size heat sinks
| Material | Bulk k (W/m·K) @ 25 °C | Notes most people forget |
|---|---|---|
| Copper C10200 | 401 | Falls to ~380 at 100 °C |
| Aluminum 6061-T6 | 167 | |
| Aluminum 6063-T6 | 201 | |
| CVD diamond | 1800–2200 | Only for tiny areas |
| Pyrolytic graphite (in-plane) | 1500–1700 | 6–10 out-of-plane |
| AlN ceramic | 170–220 | |
| Silver | 429 | Rarely used |
| Thermal grease (high-end) | 8–14 | Interface dominates anyway |
| Arctic Silver 5 (old reference) | ~8.5 | Still the benchmark people quote |
| Indium foil (53 °C melt) | 82 | Soft, conforms perfectly |
Bottom line: the bulk k number is only half the story. Contact resistance, surface finish, mounting pressure, and temperature dependence usually decide whether your CPU throttles or your IGBT lives.
I always derate lab numbers by at least 30 % on first-pass designs, then measure the actual junction temperature on the bench. Anything else is asking for field returns.