Pigment Holds Promise for 3-D Machine Vision, Optical Computing

Lawrence Normie In Israel

JERUSALEM -- A saltwater-borne bacterium may provide the key to the development of parallel optical computers and real-time 3-D robotic vision. Bacteriorhodopsin, the pigment protein derived from the bacterium, possesses unique electro-optical properties that make it useful in bistable optical switches capable of storing negative and positive values -- a crucial element for image processing operations such as deconvolution and edge detection.
The breakthrough came from Aaron Lewis at the Division of Applied Physics and Center for Neural Computation at the Hebrew Uni- versity. Lewis and his research team demonstrated the ability to represent negative values and differences of numbers by measuring only light intensity. He explained that, until now, optical systems relying on light intensity measurement recorded only positive values, because there is no direct way to represent negative values without the use of stable coherent light sources and phase detectors.

Two prominent bands
Differences in light intensity impressed on pixels fabricated from the pigment directly produce a corresponding electrical output through a phenomenon known as gradient of protons. The resulting current is conducive to the integration of the optical processing elements with conventional electronic computers and even on silicon chips. The pigment exhibits two well-separated absorption bands at purple (412 nm) and yellow (570 nm) wavelengths.
In the prototype system, yellow represents negative value and purple means positive values; equal concentrations of the two states represent zero.
Each of these light frequencies produces a photoelectric response where one is the negative of the other. In Lewis' words, "excitatory" and "inhibitory" characteristics impressed on the system can be measured electronically.
Typical outputs range between several microvolts to millivolts. A 0.5-µm pixel composed of 100 monolayers of pigment (each about 45 Å thick) actively generates a maximum photocurrent of about 1 µA.

Imaging functions
The prototype device has been configured to calculate difference of Gaussian functions. These functions are used to perform deconvolution and edge detection in image recognition and enhancement algorithms in robotic vision and microscopy.
Although femtosecond switching speeds will eventually be possible for parallel computation, Lewis said the transition speed from purple to yellow states is on the order of 100 µs, which is adequately fast for computational schemes.
The erasable property of the system also gives rise to the possibility of very high density, superfast optical memory and mass storage devices. For example, as an optical serial information storage device using a writing spot smaller than the wavelength of light, it might be possible to impress the whole Encyclopaedia Brittanica on a square centimeter of the protein film.
The technology has reached the laboratory prototype stage in small systems, but Lewis said it will take substantial funding to bring it to the industrial prototype stage.
"If the system is to be used with an electronic computer," he said, "you need to be able to microfabricate it on a chip, together with the ancillary electronics. Ordinary methods used for silicon would destroy the polymers."