The quick answer
Pick the alloy by what it has to seal against, not by habit. Kovar is for glass — specifically borosilicate hard glass — and it is the default for hermetic electronic packages and glass-to-metal seal feedthroughs. Alloy 42 is for ceramic and for semiconductor lead frames, where its slightly lower expansion matches alumina and silicon and its cobalt-free chemistry stamps and etches cheaply. Invar 36 is not a sealing alloy at all — its expansion is far below any glass or ceramic, and it is specified when the goal is dimensional stability across temperature.
The expensive mistake is treating the three as interchangeable "low-expansion alloys." They are not. A drawing that calls out Invar for a glass feedthrough, or Alloy 42 for a borosilicate seal, will machine and plate fine and then fail in the sealing furnace or in the first thermal cycle. If you already know Kovar is the right call, our Kovar CNC machining page covers hydrogen-fired stock, cutting parameters, and gold/nickel plating for hermetic packages.
The three alloys at a glance
| Property | Kovar | Alloy 42 | Invar 36 |
|---|---|---|---|
| Nickel | ~29% | ~42% | ~36% |
| Cobalt | ~17% | — | — |
| Iron | balance | balance | balance |
| Common spec | ASTM F15 | ASTM F30 | ASTM F1684 |
| UNS number | K94610 | K94100 | K93600 |
All three sit in the same nickel-iron family and share a counter-intuitive property: adding nickel to iron does not raise expansion the way alloying usually does — between roughly 30% and 50% nickel, the alloy's magnetostriction partly cancels normal thermal expansion. The exact nickel content (and, for Kovar, the cobalt addition) tunes where on that curve the alloy lands. That is the whole design space.
CTE — the number that decides everything
| Material | CTE (ppm/°C) | Seals / matches |
|---|---|---|
| Kovar | ~5.3 | Borosilicate hard glass (Corning 7052/7056, Schott 8245) |
| Alloy 42 | ~4.5 | Alumina ceramic, silicon, IC lead frames |
| Invar 36 | ~1.2 | Nothing sealed — dimensional stability use |
| Borosilicate glass | ~4.6–5.3 | Matched by Kovar |
| Alumina (96–99%) | ~6–7 | Approached by Alloy 42 |
| 304 stainless (reference) | ~17 | Not a controlled-expansion alloy |
A glass-to-metal seal works because the metal and the glass shrink together as the assembly cools from the sealing temperature. If the metal shrinks faster than the glass, the glass is left in tension — and glass is weak in tension, so it cracks, often days or weeks after assembly. If the metal shrinks slower, the glass is squeezed into compression, which it tolerates far better. This asymmetry is why a small CTE mismatch toward compression is survivable and a small mismatch toward tension is not, and why the alloy has to be matched to the specific glass rather than chosen from a generic "low-expansion" shelf.
Kovar — the borosilicate glass standard
Kovar was developed by Westinghouse in the 1930s specifically to match borosilicate hard glass for vacuum-tube envelopes, and nearly a century later it is still the default for hermetic electronic packages, microwave and RF feedthroughs, and high-reliability space and defense instrumentation. Its ~5.3 ppm/°C expansion tracks Corning 7052/7056 and Schott 8245-class glasses across the cool-down from the ~1000 °C sealing range.
The practical catch with Kovar is stock condition: for any glass-seal or plated part it must be hydrogen-fired before machining to decarburize the surface, otherwise free carbon outgasses into the glass seal and disrupts plating. Our Kovar machining page covers hydrogen-fired versus mill-finish stock, cutting parameters, and the gold-over-nickel plating typical of hermetic packages.
Alloy 42 — ceramic packages and lead frames
Alloy 42 (42% nickel, balance iron, no cobalt) sits a little below Kovar at roughly 4.5 ppm/°C. That target is aimed at alumina ceramic and silicon rather than glass, which is why Alloy 42 became the workhorse of semiconductor lead frames and ceramic-bodied packages — DIP frames, TO-can bases, and ceramic dual-in-line packages. Without cobalt it is cheaper than Kovar and easier to stamp and chemically etch into fine lead-frame geometry.
The trade-off is that Alloy 42 is a slightly poorer match to borosilicate glass than Kovar. For a package whose body is ceramic, Alloy 42 is correct; for one whose body is hard glass, Kovar is correct. The two are not substitutes — they are matched to different sealing partners.
Invar 36 — when you are not sealing to glass at all
Invar 36 is the odd one out. At ~1.2 ppm/°C its expansion is far below any sealing glass or ceramic, so it is never used for glass-to-metal seals. Instead it is specified when the requirement is simply to not move: optical benches and interferometer structures, semiconductor lithography stages, cryogenic vessels, and composite layup tooling whose CTE must match carbon-fiber laminate. If your part has to hold a dimension across a shop temperature swing rather than seal to a glass bead, Invar 36 — not Kovar — is the alloy. Our Invar 36 machining page covers the stress-relief protocol those applications require.
Machining all three
The three alloys machine in the same general family and share the nickel-iron headaches: moderate work-hardening, a tendency toward built-up edge, and poor thermal conductivity that concentrates heat at the tool tip. Kovar is the most forgiving — it behaves much like 410 stainless at ~28 HRC. Invar 36 is the most demanding because it work-hardens aggressively; a wrong feed rate glazes the surface and destroys the next pass. Alloy 42 sits between them.
For all three we run TiAlN-coated carbide, climb-cut exclusively, keep the tool moving to avoid dwelling in the cut, and flood-cool to clear heat. None of them is exotic to machine — the discipline is in consistent feed and never letting the surface work-harden. The harder problems are upstream (specifying the right stock condition) and downstream (plating and certification), not at the spindle.
Decision framework
Start from the sealing partner, not the alloy:
- Sealing to borosilicate hard glass — Kovar (ASTM F15), hydrogen-fired. Hermetic packages, glass-to-metal feedthroughs, vacuum tubes.
- Sealing to alumina ceramic, or making a lead frame — Alloy 42 (ASTM F30). Ceramic packages, DIP frames, TO bases.
- Not sealing — holding a dimension across temperature — Invar 36 (ASTM F1684). Optical, cryogenic, composite tooling.
- Operating continuously above ~450 °C — none of the three. Above the Curie point the controlled-expansion behavior is gone; the design needs a different approach.
- Need a non-magnetic part — none of the three. The whole family is ferromagnetic.
When a drawing reaches us with a generic "Kovar or equivalent" callout, we ask one question — what are you sealing to? — because that answer, not the alloy name, is what decides the part. If the callout and the sealing partner do not match, we flag it on the quote so the correction happens before metal is cut, not after the seal fails.