As far as we know, only two types of waves can travel through the vacuum of space: electromagnetic and gravitational. Gravitational waves would be exceedingly hard to generate or detect, so any signals headed our way will probably lie somewhere in the electromagnetic spectrum—from X-rays at one extreme to frequencies lower than an ordinary AM radio at the other end. For years, the SETI community rejected the idea of looking for anything other than RF signals, even though they represent just a tiny portion of the electromagnetic spectrum [see illustration, “Casting a Wider Net”]. Limiting SETI to just radio waves is like losing a diamond ring on a football field and only searching for it at the 1-yard line.

A good argument can be made that the optical spectrum is a more likely place to find alien signals than are either RF or microwave frequencies. For one thing, it’s much easier to deal with noise at optical wavelengths. People who measure radio waves have to contend with interference from radar antennas, radio stations, and other terrestrial sources. The receiver itself also adds noise, which is why the detectors attached to advanced telescopes are typically cooled to near absolute zero. Yet there is always some residual thermal noise to contend with, and for certain wavelengths the cosmic microwave background—a ­vestige of the big bang—makes SETI searches difficult.

For optical observations, the only significant terrestrial source of interference is lightning, which is at worst a sporadic problem. In the early days, many investigators discounted optical SETI because they imagined that the sender’s star would be an overwhelming noise source. But they didn’t appreciate that it is actually quite easy to arrange a transmission that outshines whatever sun you’re circling: just use a pulsed laser rather than the continuous-wave type.

The development of laser communications systems for military satellites, submarines, and aircraft has proved that short bursts of light are far more efficient than continuous waves at carrying information over great distances. Each pulse has a high peak power, but most of the time the laser isn’t active, so the overall power consumption is low.

Presumably a distant intelligent civilization would have figured this out as well. With transmissions in brief bursts, each pulse could easily be 1000 times as bright as any nearby star in the receiving telescope’s field of view. The shorter the pulse, the less background light there is per pulse to compete with the signal. Reducing the pulse to nanosecond intervals makes the signal even more distinct, because there’s no source in nature that generates flashes that short.

The sender could vary the interval between pulses to convey information. Think of each interval as a roulette wheel. Each slot in the wheel represents a number: if the pulse fell in slot 36, it would be conveying the number 36; if it fell in slot 1, it would mean the number 1. The roulette wheel could have more or less than 36 slots, of course, and as the number of slots changes, so does the amount of information you can send per pulse. With just two slots, you could send just one bit per pulse; with 256 slots, you could send 8 bits; and with 1024 slots, you could send 10 bits. This technique, called pulse-position modulation, was used for many years in both radio and optical communications. So, if you were to detect a pulsed signal coming from space, the next step would be to analyze it for any repeating sequences.

Another reason to prefer optical methods over radio SETI is that it’s much easier to form a narrow beam of light. Remember, any message will have to travel many trillions of miles through space to reach us. If the sender were to broadcast a signal in all directions at once, the power needed would be prohibitively high, regardless of what wavelength was used.

George W. Swenson Jr., professor emeritus of astronomy and of electrical and computer engineering at the University of Illinois at Urbana-Champaign, has calculated that if a radio transmitter were 100 light-years away and projecting its energy omnidirectionally, it would require 5800 trillion watts to provide a detectable signal—an amount, Swenson points out, that is “more than 7000 times the total electricity-generating capacity of the U.S.”

SETI researchers therefore generally assume that the transmitter will point toward specific star systems and that the beam will be as tight as possible. The ratio of the wavelength being transmitted to the diameter of the antenna used is roughly proportional to the width of the beam. And the wavelength of visible light is six orders of magnitude smaller than that of microwaves, allowing the beam to be considerably narrower. So the physics of optics over RF wins out again.

Charles Townes—the coinventor of the laser—and Robert Schwartz first suggested the idea of searching for optical signals from extraterrestrials in 1961. Their paper, published in Nature, theorized that beings in a nearby star system, “some few or tens of light-years away,” could use laser (or maser) beams to communicate with us earthlings.

It took several more decades for optical SETI to catch on, in large part because laser technology wasn’t nearly as mature as radio technology. That’s no longer the case. Photodetectors today have quantum efficiencies of 40 percent or more; that is, for every 100 incoming photons, 40 are actually counted. The detectors are also much faster now, and picking up nanosecond pulses is no problem.

The new Harvard telescope takes advantage of these and other technological advances. Unlike a regular imaging telescope, Horowitz’s brainchild has no lens, nor is it designed to pluck pristine images of celestial bodies orbiting overhead. Its mission is a bit cruder. It uses a 1.8-meter-diameter mirror and a 0.9-meter secondary mirror to scoop up raw photons from the sky—much as a wooden barrel collects rainwater. For that reason, the telescope is more properly, though less glamorously, known as a photon bucket [see photos, “A Closer Look”].

Photon buckets are less expensive to build and operate than imaging telescopes, because all you’re really worried about is that the photons arrive at the detector. If a given photon ends up traveling a few extra millimeters because it hit an air pocket in the atmosphere, it doesn’t really matter. An advanced imaging telescope, by contrast, employs all kinds of sophisticated mechanisms to compensate for distortion in the incoming signal, so that it can accurately reproduce what it sees.

A key feature of the Harvard telescope is its use of ­multipixel photomultiplier tubes. Photomultipliers work by converting incoming photons into electrons, which then get amplified until the electrical signal can be distinguished from the noise. A multi­pixel photomultiplier divides the collection area up into tiny squares—64 per tube in this case—each acting like a separate detector. This setup lets you look at more star systems at a time.

The Harvard telescope breaks up the sky into 1.6- by 0.2‑degree patches, observing each patch for about 48 seconds before moving on to the next patch. A mere 48 seconds doesn’t sound like a long time, but remember that the associated electronics are sampling the data in nanosecond intervals. At that rate, the instrument should be able to cover the entire sky above the northern hemisphere in 150 nights of observation.

All those signals from all those pixels are then fed into 32 microprocessors, which were custom-designed for the project by Horowitz’s graduate student Andrew Howard. These PulseNet chips crank through the data—3.5 trillion bits per second—searching for a large spike in the photon count, which may indicate a possible light pulse from afar.

The telescope’s photodetectors are divided into two arrays, so that if one of them receives an interesting signal, you can check it against the second array. Ideally, you’d like to have an entirely separate telescope. In a previous search, Horowitz collaborated with David Wilkinson’s group at Princeton University, using a 1.5-meter telescope at Oak Ridge and a 0.9‑meter telescope at Princeton’s FitzRandolph Observatory. The paired instruments conducted a targeted—rather than an all-sky—optical search, examining more than 6000 stars over a six-year run. In its first three nights of observation, the all-sky search took in 200 times as many stars as the targeted search did during the entire experiment.