Mellon optical memory was an early form of computer memory invented at
Mellon Institute (today part of Carnegie Mellon University) in
1951. The device used a combination of photoemissive and
phosphorescent materials to produce a "light loop" between two
surfaces. The presence or lack of light, detected by a photocell,
represented a one or zero. Although promising, the system was rendered
obsolete with the introduction of core memory in the early 1950s. It
appears that the system was never used in production.
The main memory element of the Mellon device consisted of a very large
(television sized) square vacuum tube consisting of two slightly
separated flat glass plates. The inner side of one of the plates was
coated with a photoemissive material that released electrons when
struck by light. The inside of the other plate was coated with a
phosphorescent material, that would release light when struck by
The tube was charged with a high electrical voltage. When an external
source of light struck the photoemissive layer, it would release a
shower of electrons. The electrons would be pulled toward the positive
charge on the phosphorescent layer, traveling though the vacuum. When
they struck the phosphorescent layer, it would release a shower of
photons (light) travelling in all directions. Some of these photons
would travel back to the photoemissive layer, where they would cause a
second shower of electrons to be released. To ensure that the light
did not activate nearby areas of the photoemissive material, a baffle
was used inside the tube, dividing the device up into a grid of cells.
The process of electron emission causing photoemission in turn causing
electron emission is what provided the memory action. This process
would continue for a short time; the light emitted by the
phosphorescent layer was much smaller than the amount of energy
absorbed by it from the electrons, so the total amount of light in the
cell faded away at a rate determined by the characteristics of the
Overall the system was similar to the better-known Williams tube. The
Williams tube used the phosphorescent front of a single CRT to create
small spots of static electricity on a plate arranged in front of the
tube. However, the stability of these dots proved difficult to
maintain in the presence of external electrical signals, which were
common in computer settings. The Mellon system replaced the static
charges with light, which was much more resistant to external
Writing to the cell was accomplished by an external cathode ray tube
(CRT) arranged in front of the photoemissive side of the grid. Cells
were selected by using the deflection coils in the CRT to pull the
beam into position in front of the cell, lighting up the front of the
tube in that location. This initial pulse of light, focussed through a
lens, would set the cell to the "on" state. Due to the way the
photoemissive layer worked, focusing light on it again when it was
already "lit up" would overload the material, stopping electrons from
flowing out the other side into the interior of the cell. When the
external light was then removed, the cell was dark, turning it off.
Reading the cells was accomplished by a grid of photocells arranged
behind the phosphorescent layer, which emitted photons
omnidirectionally. This allowed the cells to be read from the back of
the device, as long as the phosphorescent layer was thin enough. To
form a complete memory the system was arranged to be regenerative,
with the output of the photocells being amplified and sent back into
the CRT to refresh the cells periodically.
Mellon Institute of Industrial Research: Computer Component
Fellowship #347, Quart. Rep. no. 3 (Apr.–July 1951) Sec. I–VI;
Quart. Rep. no. 5 (Oct.–Jan. 1952) Sec. I–V; Quart. Rep. no. 6
(Jan.–Apr. 1952) Sec. II, III, VI; Quart. Rep. no. 9 (Oct.–Jan.
1953) Section III.
^ Eckert, J. P. Jr. (1998-10-01). "A Survey of Digital Computer Memory
Systems". IEEE Ann. Hist. Comput. 20 (4): 15–28.
doi:10.1109/85.728227. ISSN 1058-6180.
Echert Jr., J.P., "A Survey of Digital Computer Memory Systems",
Proceedings of the IRE, October 1953. Reprinted in 'IEEE Annals of the
History of Computing, Vol.20