The human eye, once it acclimates to darkness, can detect bursts of light containing as few as a hundred photons. That ability suggests that rod cells—the retinal photoreceptors that specialize in night vision—can detect single quanta of light: Even after accounting for focusing, it’s unlikely that any two photons from so faint a flash would arrive at the same one of the roughly 120 million rods that line the inner eye.
Studies dating back to the 1940s have generated compelling, albeit circumstantial, evidence in favor of that proposition. Statistical analyses of human behavioral data suggest that only about 10% of the photons in a just-visible flash of light contribute to retinal stimulation.1 The rest are scattered within the eye or go unabsorbed by the retina’s photoreceptors, which makes it even less likely that a single rod would register multiple-photon hits. In other experiments, rods irradiated with few-photon pulses generated electrical signals of quantized amplitude.2 Presumably, the smallest signals correspond to single-photon stimulations. The next largest, roughly twice as big, are thought to herald two-photon stimulations, and so forth.
Researchers led by Leonid Krivitsky at the Agency for Science, Technology and Research in Singapore have now probed the limits of rod sensitivity in a more direct fashion—by firing photons one by one at rods plucked from the eyes of African clawed frogs.3 The figure shows the group’s experimental setup. Photons are delivered via an optical fiber to a rod held in a pipette, where electrodes record spikes in the cell’s transmembrane potential.
Similar experiments have been performed using lamps or LEDs dimmed to deliver pulses with a mean photon number close to one. Such pulses, however, are described by a Poisson distribution; although many of them may carry one photon, others will carry two or more, and a large share will carry none at all. One can’t say for sure how many photons contribute to a particular cell response and, as a result, one can’t directly calculate the cell’s single-photon detection efficiency.
To generate pulses containing precisely one photon, Krivitsky and company exploited a process known as parametric down conversion, in which a crystal converts one high-energy photon into a pair of low-energy ones. (See the article by Alan Migdall, Physics Today, January 1999, page 41.) In their particular implementation, the researchers shine a UV laser into a barium borate crystal, which occasionally spits out a pair of visible photons. One of the photons is steered to an avalanche photodiode, where its detection triggers an acousto-optical modulator to divert the second photon into the fiber that leads to the rod.
“The thing about an experiment like this,” says Krivitsky, “is that time works against you.” After harvesting a rod from its amphibian host, he explains, there’s just a one- to two-hour window in which to work before the cell loses viability. Meanwhile, to allow the cell adequate recovery time between spikes, pulses must be delivered at a rate of no more than about 10 per minute. In the pursuit of a statistically robust data set, every photon is precious. The optical fiber’s shape helps ensure that photons find their target: Its rounded tip acts as a lens that focuses outgoing light onto a spot 4 µm across, to match the diameter of the rod.
The team’s results, amassed with 10 rods harvested from 10 different specimens, indicate that an impinging photon has a nearly one in three chance of eliciting a spike. Previous experiments using conventional light pulses to stimulate rods of a related species, the cane toad, found detection rates of just 6%. The new estimate is roughly in line with the range predicted for human rods.
Indeed, rods appear to detect single photons almost as reliably as do many commercially available sensors, including some avalanche photodiodes. The rod response, however, is much more sluggish. Each spike is the product of a regulatory cascade that’s thought to start with the photoisomerization of a lone pigment molecule and culminate, nearly two seconds later, with the closing of several hundred ion channels. For Nicolas Gisin, a quantum optician at the University of Geneva, that raises an intriguing question: How is it that even in dim settings we can react to visual cues almost instantly? “As far as I know, there’s still no good answer,” he says. “But the information processing that’s involved must be remarkable.”