Tungsten raised to 2,500°C slowly evaporates, and this limits the life of 'bright' emitter filaments. A much lower temperature dramatically reduces thermionic emission. Fleming's original work was to investigate the blackening of the insides of lamps due to this evaporation, and the shadows cast by the filament supports.
Tungsten was typically worked by powder metallurgy methods, and it was found by Coolidge that the addition of a little thoria (thorium oxide) made the Tungsten more ductile. Langmuir found that filaments made this way produced excess electron emission over pure Tungsten and at could work at lower temperatures. This work resulted in emitting surfaces that were efficient at lower temperatures, the original dull emitters exploited this combination and were known as thoriated tungsten filaments. They operated at about 1,800°C.
As early as 1903 Wehnelt had found that grease contaminating a platinium filament produced enhanced emission. This work resulted in the knowledge that oxides of certain alkaline earth metals (calcium, barium etc.) produced emission equal to pure Tungsten but at considerably lower temperatures. The problem was devising a suitable technology to reliably exploit these facts.
The azide process, sometimes confused with the oxide-coating process, was developed on the Continent by Philips in 1924-5 and introduced to Britain when Philips acquired a controlling interest in Mullard in 1925-26. The first Philips-Mullard (PM series) azide valves to be announced (1925) were the PM3 and PM4. However, there were great technical problems at the Mullard works at this time and hearsay has it that all the early specimens were imported from the Philips factory in Eindhoven and relabelled as 'British Made' in order to appear to conform to the rules of the BVA cartel. The first British-made azide type to reach the market was the PM3 (late 1925?), followed by the PM4 which had a better specification and was harder to make. It is said that Mullards had so many thousands of below- spec valves that they bought up a small company (the Six-Sixty Valve Co.) to provide an outlet at the bottom end of the market.
Several methods were employed as the technology evolved. In the azide process a Tungsten (or possibly nickel) filament was first oxidised, or a copper layer on the filament was oxidised. The anode would be initially coated with barium azide. In manufacture, during vacuum pumping, the anode would be heated to vaporise the azide. At high temperature the vapour would decompose to form barium and nitrogen. The barium would react with the oxide on the filament to form a barium oxide layer on the Tungsten. The barium also deposited itself on the inner surface of the envelope where it acted as a getter to absorb the final gas molecules after the valve was sealed. One conspicuous feature of the Mullard form of the azide process was heavy internal blackening of the glass bulb (e.g. PM22). This was OK while competitors' valves were equally black (although some used a magnesium pre-getter to make the glass look silvery rather than black)
High efficiency emitters required an even and thin layer of oxide on the filament. The azide process was not capable of generating the uniformity required. Additionally the process could not confine the barium to just the required areas of the filament. In fact the azide process tended to splash barium and other material all over the place; glass insulators placed 'out of the way' had to be used instead of mica supports.
The azide process was not suited to the manufacture of indirectly-heated valves
Another technology, the paste coated cathode, replaced the azide process and became the standard process.