In 1911 William Coolidge at GE in the States developed a method of drawing tungsten wire (Tungsram used what appears to be a sintering process) and a bulb based on this was introduced. The higher the temperature the better the efficiency. Previous experimenters had used platinum (m.p. 2014 °K) but as well as having a lower melting point it is too expensive for commercial use. Carbon has a higher sublimation (solid to vapour) temperature of 3915 °K but it difficult to produce a robust filament. Tungsten at 3695 °K turned out to be the best compromise.
Hot filaments have a definite life. The tungsten gradually evaporates and when the filament thins in a particular spot it get hotter there, evaporates faster and soon breaks.
Scanning electron microscope image of blown filament
The evaporated metal darkens the glass and reduces the output. Filing the bulb with an inert gas cuts down the evaporation rate by supplying some back pressure. Another improvement was the adoption of a coiled coil filament which again reduces the evaporation: thin filaments have a high ratio of surface area to volume. Thick filaments, for example low voltage ones, have a lower ratio which means they can evaporate for longer before failure. A coiled coil behaves more as a filament of the outer coil diameter and also lasts longer.
There is a definite relationship between filament temperature, efficiency and life. Efficiency varies with applied voltage as V 1.8 but lifetime varies as V −16 . So designing for a 5% lower voltage gives a brighter light and increased efficiency (10%) but much reduced lifetime (