The Thermal Inkjet

Ichiro Endo's laboratory at Canon in 1977 had a syringe of ink on the bench when a hot soldering iron touched it and the ink squirted. John Vaught reached the same idea at Hewlett-Packard in 1979 from the other direction, thinking about percolators. Both had noticed that if you heat a liquid fast enough you do not get boiling in the ordinary sense - you get a single explosive bubble, and a bubble in a confined channel is a piston. A thermal inkjet nozzle is therefore a resistor with no moving part anywhere near it. Pulse a forty-micrometre-square thin film for a couple of microseconds and the water touching it passes about three hundred degrees, the homogeneous nucleation limit where it can no longer remain liquid at any pressure. The layer that flashes is a fifth of a micrometre thick and weighs a third of a nanogram. That is the whole engine. What limits it is not the boiling. Getting the ink back is slow: capillary refill takes tens of microseconds against the couple that the bubble takes, so a thermal nozzle fires at tens of kilohertz while its heater could go far faster. The answer to a faster printer was therefore never a faster nozzle - it was more of them, photolithographed side by side by the thousand. Compared with a piezo head it is worse in almost every measurable way: a hundred times the energy per drop, an ink that must be mostly water and nearly free of solids, and a head that erodes and fouls in months. It won anyway, because it is a silicon chip. Adding nozzles is free the way adding transistors is free, and a head that wears out gets sold with the ink as a cartridge - which turns a precision component into something the customer replaces themselves.