As mentioned in TFA, the most important factor for successfully joining a metal and a glass is to match their thermal expansion coefficients.
Most pure metals have a much greater thermal expansion than any glass, which will cause cracks.
In the nineteenth century, the first successful joinings of metal with glass were done using platinum, but that is obviously too expensive for normal applications.
Eventually a special alloy of iron-nickel-cobalt was developed, which is named kovar and whose thermal expansion is matched to that of a certain type of borosilicate glass.
The use of kovar was widespread in electronics, starting with the vacuum tubes and gas tubes, and then continuing with the first generations of transistors and integrated circuits, which used metal packages.
All the old transistors and operational amplifiers that were packaged in metal cans had pins and package bases made of kovar.
When kovar had to be joined with a different kind of glass than the type with which it is matched in thermal expansion, that glass was coated in one or more layers of different kinds of glasses, with that matched to kovar in contact with the metal and the intermediate layers having intermediate thermal expansion coefficients, interpolating between the bulk glass and kovar.
Kovar is not a good thermal or electrical conductor, which is why the modern power transistors that use plastic packages (e.g. TO-247) and copper bases and pins (which are plated with nickel or tin, to avoid corrosion) can easily dissipate much greater powers than the old transistors in TO-3 metal cans, which had the same size. On the other hand, the old transistors in metal packages were pretty much immune of environmental influences.
Ordinary incandescent bulbs must have similar sealing requirements, but they probably mostly rely on using a thin conductor that doesn't contract much when it cools. Also IIRC modern incandescent bulbs do not use a high vacuum but contain a low pressure inert gas so leakage would be slower if it occurs at all.
As you say, incandescent bulbs are less demanding because they do not use high vacuum, but they have the additional requirement that the pins that support the tungsten filament must resist to very high temperatures, because some heat is conducted through the filament into its support.
This is why the pins that support the filament are typically made of molybdenum. Molybdenum has a relatively low thermal expansion coefficient in comparison with most metals, so there are certain glass compositions that can match its TCE. The glass through which the pins pass is not of the same type as the bulb, which is made of cheaper glass, but it is of the type matched in TCE with molybdenum.
AFAIK, in in the US: Modern high-efficiency incandescent bulbs are exactly like the halogen bulbs of yore, but with the small quartz envelope wrapped in a familiar light-bulb-shaped shell, and with a base that matches. It's like a light bulb within a light bulb.
Except for all the ones that aren't modern or efficient. Common 40-Watt appliance bulbs, for instance: Those are still built using the old methods. They never changed. This strongly suggests that we never forgot how to seal metal wires into a glass bottle full of nothing.
But this article isn't about industrial processes. It's about rediscovering things at home, and that stands on its own merits. :)
Incandescent bulbs used dumet/platinite, which is an nickel-iron alloy like kovar except it's turned for a different CTE. This stuff isn't that expensive when mass produced - it's just those who can afford usually borosilicate can afford paying a premium for kovar.
A lot of thought went into selecting the correct metal and glass for the application.
Kovar[1] was typically used in commercial applications where tubes were constructed from hard (borosilicate) glass. In fact, there were special formulations of borosilicate, such as Corning 7052, and later 7073, which were designed to match with Kovar. So both the metal and the glass were designed to work together. This involves engineering the metal and glass such that they shrink at about the same rate from the glass's setting point (temperature where the glass's internal stresses start to align, but before the glass solidifies) down to room temperature.
An aside on how Kovar works, because it's neat: Kovar is ferromagnetic, and the mixture of metals changes the Curie Point - the temperature above which which a ferromagnetic material stops being magnetic due to the atoms being too energetic. The Curie Point isn't a single point, it's a region. Ferromagnetic materials' lattices actually expand as they become more magnetic - this is the Magnetovolume Effect. So by adjusting the ratios of materials, Westinghouse was able to balance the Magnetovolume Effect (materials wants to expand as it cools and regains its magnetism) with the natural lattice shrinking due to cooling, and create a region where the metal matches the shrink rate of glass.
Conversely, consumer-grade vacuum tubes, such as the ones in radios, guitar amplifiers, incandescent bulbs, and televisions, typically use cheaper soda-lime or lead-alkalai silicate glass[2]. This glass had completely different thermal expansion characteristics, so different materials for leads were required. For thin leads, what they typically[3] did is use a Dumet (42/58 Nickel/Iron) wire clad in a copper sheath and coated with borax. The bonded dumet-copper (about 80/20 by weight) expands at a compromise between the two, so it can be matched to the thermal expansion of glass. The borax aided in oxide control and bonding to the glass (this is copper's "red oxide" as mentioned in the article). But this format only works for thin wires, because as we accumulate surface area we start to have to worry about axial stress from the wire expanding along its axis. So for larger leads, a (more expensive, less conductive) one-piece alloy of 52/48 Nickel/Iron had to be used instead[4].
The anodes of CRTs used yet another alloy, designed for higher expansion volume, known as "Glass-Sealing 42-6", and standardized as ASTM F31. These are 42/6/52 Ni/Cr/Fe alloys.
Lastly, to bring it all back home, the glass matters as much as the metal, and the author of this article is using an exceptionally poor glass for vacuum tube work. It seems like they are using regular Pyrex, which has a much lower expansion coefficient than most vacuum tube glass, and in fact, most metals.
[1] - The generic term for Kovar is Fernico (from Iron-Nickel-Cobalt, Fe-Ni-Co). It was invented by Westinghouse in the 1930s. Other names for Kovar are: ASTM F-15, NILO K, Pernifer 2918, Rodar, and Dilvar P1.
[2] - An exception is tubes that experienced high temperatures that might melt normal glass - such as Xenon flash tubes. Another exception is metal vacuum tubes which had small glass borosilicate beads around each lead wire, bonded to both the wire and the surrounding metal. These were common in 1940s radios.
[4] - Interesting footnote: Platinum also works great as a soft-glass seal wire, if you have the $$$$. Dumet was originally marketed as "platinite" - a platinum substitute.
Most pure metals have a much greater thermal expansion than any glass, which will cause cracks.
In the nineteenth century, the first successful joinings of metal with glass were done using platinum, but that is obviously too expensive for normal applications.
Eventually a special alloy of iron-nickel-cobalt was developed, which is named kovar and whose thermal expansion is matched to that of a certain type of borosilicate glass.
The use of kovar was widespread in electronics, starting with the vacuum tubes and gas tubes, and then continuing with the first generations of transistors and integrated circuits, which used metal packages.
All the old transistors and operational amplifiers that were packaged in metal cans had pins and package bases made of kovar.
When kovar had to be joined with a different kind of glass than the type with which it is matched in thermal expansion, that glass was coated in one or more layers of different kinds of glasses, with that matched to kovar in contact with the metal and the intermediate layers having intermediate thermal expansion coefficients, interpolating between the bulk glass and kovar.
Kovar is not a good thermal or electrical conductor, which is why the modern power transistors that use plastic packages (e.g. TO-247) and copper bases and pins (which are plated with nickel or tin, to avoid corrosion) can easily dissipate much greater powers than the old transistors in TO-3 metal cans, which had the same size. On the other hand, the old transistors in metal packages were pretty much immune of environmental influences.