EPILOGUE

Final Passage: Halley’s Comet 1986

THE COMPUTATIONS for the 1986 return of Halley’s comet began shortly after Gertrude Blanch retired from scientific life in 1967. Though she was not the last professional human computer, her departure coincided with the final days of many computing offices. The National Bureau of Standards, now identified as the National Institute of Standards and Technology, closed its Computational Laboratory and reassigned the few remaining veterans of the Mathematical Tables Project to other divisions. The American Nautical Almanac moved from punched card equipment to electronic computers. Observatories were either acquiring their own small computers or purchasing the services of larger machines. A few businesses, such as insurance firms and petroleum refiners, retained small staffs of calculating assistants, but these, too, were being replaced with IBM 370s, DEC PDP-8s, Burroughs B-6500s, and other computers that were powered by electricity and sported numerical names.

With the 1986 return, astronomers returned to the problem of testing Newton’s theory of gravitation in order that they might reduce Andrew Crommelin’s 1910 discrepancy of two days, sixteen hours, and forty-eight minutes. The basic principles of Isaac Newton’s gravitational theory were never questioned, but researchers hoped that some slight modification might produce a more accurate prediction of the comet’s perihelion. One organization prepared ephemerides with a gravitational force that was slightly weaker than the one identified by Newton. Another team postulated the existence of one more giant outer planet, which they identified as “Planet X,” and tried to find an orbit for the planet that would account for Crommelin’s missing two and a half days. A final group hypothesized that the discrepancy was created by the comet itself. They suggested that the nucleus of the comet acted like a weak rocket engine because of a phenomenon known as outgassing. As the comet approached the sun, its surface was warmed by the light until it boiled away as vapor. This vapor created the tail and produced a gentle thrust that slowed the comet’s approach to the sun and sped its retreat into the distant spheres of the solar system.¹

The ephemerides for the 1986 return, the fourth return since the death of Edmund Halley, were the first prepared in the United States. They were done at the Jet Propulsion Laboratory, an American research center that had flourished during the space race of the 1960s. The laboratory designed unmanned spacecraft and managed scientific space missions. It occupied a patch of flat land next to a Los Angeles freeway and resembled a bland contemporary office building. Like an observatory or an almanac office, the laboratory had a large central room, but this room was filled with displays and control panels rather than with telescopes or human computers. From this room, laboratory scientists controlled spacecraft in orbit around the earth, sitting on the face of the moon, and speeding across the solar system toward the planet Mars.

The calculations for the fourth return were prepared in a back office of the Jet Propulsion Laboratory by a young researcher named Donald Yeomans. Yeomans was a new addition to the laboratory and had recently finished a doctorate in astronomy. He prepared a mathematical model of the comet’s orbit that combined the methods of Andrew Crommelin with an analysis of outgassing. Instead of preparing a computing plan, Yeomans created a computer program, a list of instructions for an electronic computer. The program was written in a language called FORTRAN IV, which was then popular among scientists and engineers. Programming was more exacting than planning, as electronic computers were less forgiving than their human counterparts. Unlike a good human computer, who could correct errors on computing sheets, an electronic computer would follow the instructions blindly, executing each operation even if the action made no sense.

Yeomans submitted his program to a central computer at the laboratory, a UNIVAC 1108. In the years that followed the Second World War, the name “UNIVAC” had been nearly synonymous with the term “electronic digital computer.” The moniker had been invented by J. Presper Eckert and John Mauchly, the designers of the original ENIAC machine, as the name for their first commercial computer.² Eckert and Mauchly had stopped designing computers in 1958, but the name UNIVAC survived as the brand name for machines sold by the Sperry Rand Corporation.³ The UNIVAC 1108 was considerably faster than a room of human computers, but it was shared by the entire staff of the Jet Propulsion Laboratory. “We were lucky to get one run per day,” Yeomans recalled. “I had to come in nights and weekends to get the necessary turn around time.”⁴

Yeomans needed repeated calculations in order to remove the 1910 discrepancy. He computed old orbits of the comet, including those that ended in 1682, 1758, and 1835. He compared his predictions of those early returns with the actual observations, adjusted the values in his mathematical model, and repeated the process. He finished the computation in 1977, reported that the next perihelion of Halley’s comet should occur on February 9, 1986, at 15:50 (Universal time), and claimed that his result was accurate to within six hours.⁵ He revised his prediction in 1982, after an earthbound telescope caught the first images of the comet. The actual perihelion occurred at 10:48, five hours and two minutes before Yeomans’s original prediction.⁶

46. Comet Halley, 1986, from Giotto Spacecraft

Even though the fourth anticipated return of Comet Halley had been well publicized, the event quickly lost its hold upon the public consciousness in the spring of 1986. The comet had skimmed low in the skies of the Northern Hemisphere and had not presented a dramatic picture to most observers. The professional astronomers clung to Halley’s comet for only a little longer. They had observations to compare, theories to analyze, papers to write. During the next few years, they demonstrated how well modern astronomers could compute the paths of these irregular visitors. In 1994, they followed the frightening Levy-Shoemaker comet, which slammed into Jupiter and thereby demonstrated the role that the outer planets have played in protecting the Earth. Three years later, astronomers computed the orbit of the stunning Hale-Bopp comet, which had a nucleus several orders of magnitude bigger than the core of Halley’s comet and left a tail that stretched across the night sky.

As the astronomers put Halley’s comet behind them, the scientists at the Jet Propulsion Laboratory discarded their old tools of computation. They removed the central UNIVAC 1108 and placed a small computer on the desk of each researcher. FORTRAN IV was replaced by more sophisticated languages and was eventually succeeded by programs for symbolic mathematics. These programs could handle many of the tasks that had once been undertaken by the mathematicians on planning committees, such as the manipulation of mathematical expressions, the operations of algebra, and even the solution of many problems in calculus. Not only could they calculate numerical answers, they could help a scientist derive a concise mathematical theory of a physical phenomenon. Had the comet reached its perihelion in 1996 instead of 1986, it would have been predicted by a much more refined computing process. “If I had access to the more sophisticated algorithms and code that we now employ on [electronic computers],” wrote Yeomans, “[then my] Halley prediction would have been even more accurate than the realized [value].”⁷

The scientists who convene every seventy-five years to compute the perihelion of Halley’s comet have had trouble looking both forward and backward. They might project the next perihelion with some improvement in accuracy, but having computed the signs of the skies, they generally fail to foresee the signs of the times. Edmund Halley, who struggled to produce even a crude estimate of the first return of his comet, did not anticipate either the work of Clairaut, Lepaute, and Lalande or de Prony’s radical division of labor or the factory system of computation. By contrast, Charles Babbage overestimated the demand for computation. Alone among those who witnessed the 1835 return, Babbage understood the nature and potential of computing machinery. He would have been surprised to learn that one century later, the largest computing organization was a work relief project that resembled an office of the 1790s. Andrew Crommelin, L. J. Comrie, and the others who saw the 1910 return had grand visions for computation, but they would not have appreciated that the computing machine would replace the mathematical table and that all their labor to correct table entries would eventually be forgotten.

In a like manner, the scientists have freely taken of the historical data and old computations, but they overlooked the people who handled the calculations: Nicole-Reine Étable de la Brière Lepaute of the Palais Luxembourg, George Airy’s boy computers at the Greenwich Observatory, Oswald Veblen’s staff at Aberdeen, the WPA workers of Gertrude Blanch, and even the table makers of the Handbook of Mathematical Functions. It seems likely that the scientists of the 2061 return of Halley’s comet will shrink Yeomans’s discrepancy to a few seconds, as they will have detailed observations of the comet through its entire orbit and computing tools beyond what we can now imagine.⁸ It seems equally likely that those same scientists will know little of the daily lives of the computational workers who are so common in our age, the computer programmers, the network managers, and the Web designers. The generation of 2061 may be surprised when an elder of our time explains that she once worked with electronic computers, that she took pride in her skill, that she had been pleased to be part of a scientific endeavor, and that she had once opened a book and studied a subject called calculus.