Nobel Physics Prize honors optical fibers and CCD sensors

Fiber-optic cables are the world's information arteries. They carry internet traffic under the oceans, across cities, into offices, and, increasingly, into our homes. Today's fastest commercial cables operate at 10 terabytes per second—enough to transmit all 22 James Bond movies in 10 milliseconds. In the mid-1960s, while he was at Standard Telephones and Cables (STC) Ltd outside London, Kao and his small group of collaborators worked out the essential properties of the first generation of optical fibers for telecommunications.

Digital CCD cameras are more sensitive than their film-based predecessors and, because they produce intrinsically digital images, are far more convenient. Besides spawning a boom in personal photography, CCD sensors have revolutionized astronomy and are steadily improving diagnostic medicine. The images of nebulae, galaxies, and other celestial objects taken by the Hubble Space Telescope are not only famously gorgeous, they have also led to significant discoveries. Boyle and Smith invented the CCD sensor at Bell Labs in the late 1960s.

Higher capacity

The higher the frequency of an electromagnetic wave, the more information it can carry. By the 1960s, when Kao began his pioneering work, the UK's Post Office, AT&T, and other telecommunications companies were trying to develop optical communication. Signal generation was not the biggest hurdle. The recently invented laser, though still neither cheap nor convenient, could produce an information-carrying modulated signal.

Signal transmission, however, remained a formidable challenge. To be practical, a transmission conduit must retain information with as little loss as possible and be capable of going around corners. Some kind of transparent cable was needed. Flexible, dielectric waveguides made from plastic were already in use for microwave transmission. Could they be made to work in the optical waveband?

Kao and George Hockham, his colleague at STC, tackled the problem in a landmark 1966 paper.¹ Like others before them, they realized that a sufficiently narrow waveguide would transmit only one mode, thereby forestalling the smearing of information due to dispersion. Their key insight was to identify the role played by impurities, notably metal ions, which scatter and absorb the signal. In a perfectly pure material, the loss to Rayleigh scattering would be 1 dB/km. An impure material would suffice for practical transmission, provided loss was kept below 20 dB/km.

Of the materials Kao and Hockham tested, fused silica was the most promising. As one of Earth's most abundant materials, silica was certainly cheap, but whether it could be made into low-loss kilometers-long fibers was unclear. Although Kao's work at STC had been funded by the UK Post Office, he could not persuade the organization to finance the materials research needed to develop his idea. His overtures to AT&T also failed.

Corning Glass Works, however, did take up the challenge. By 1970, Corning's Robert Maurer and his coworkers had developed a method, based on chemical vapor deposition, to make long silica fibers from silicon tetrachloride. Doping the fibers with titanium reduced the loss to 17 dB/km, below the threshold Kao and Hockham had predicted.

A digital sensor

The impetus for Boyle and Smith's invention of the CCD came from two quite different Bell Labs projects: magnetic bubble memory and the silicon diode camera. In the 1960s Andrew Bobeck had the idea to store and retrieve data in magnetic domains or bubbles in a magnetorestrictive material. Applying a current would shift those domains underneath an array of read heads. No mechanical moving parts would be necessary.

Smith was working to develop a silicon diode camera to meet AT&T's goal of creating a telecommunications device that could transmit video as well as voice. In its underlying physics the silicon diode array is almost identical to a CCD. Photons impinging on a silicon substrate excite electrons across the material's bandgap to create a temporarily charged region. An image can be derived from those regions of photoelectric charge—provided the charge can be efficiently transferred from each individual diode.

That transfer step proved challenging. At a one-hour meeting on 16 October 1969, Boyle and Smith came up with a solution. In their proposed device, charge, like the domains of bubble memory, would reside not in individual lumps of material, but in a continuous piece. The pixels would be formed electronically by a grid of metallic electrodes that creates an array of positively charged potential wells for trapping photoelectrons.

After one exposure time, the image, consisting of the trapped charges (or absence of charges), could be extracted by applying a sequence of voltage pulses. The first pulse transfers the first row of charges to an analog-to-digital converter for readout and all the other rows of charges into the adjacent rows. Successive pulses send successive rows to the ADC until the entire image has been read out.

Within a week of their first meeting, Boyle, Smith, and their coworkers had made a prototype to test the all-important charge transfer concept. By April the following year, they published two papers in the Bell System Technical Journal: one describing the CCD concept,² the other describing a working device.³ That early device had an integration time of 16 s and could transfer all but 2% of the total photoelectric charge. Today's fastest CCD cameras can integrate images in a few 100 microseconds, and the most efficient can transfer all but 1 in 100 000 electrons.

Charles Day

1. C. K. Kao, G. A. Hockham, Proc. Inst. Electr. Eng. 113, 1151 (1966).
2. W. S. Boyle, G. E. Smith, Bell Syst. Tech. J. 49, 587 (1970).
3. G. F. Amelio, M. F. Thompsett, G. E. Smith, Bell Syst. Tech. J. 49, 593 (1970).