SNSPD as multi-photon detector

The Duke­–Ohio State team, using a standard SNSPD from Quantum Opus, developed a signal-processing technique for counting multiple photons using a single-photon detector. [Image: Duke University]

Single-photon detectors have become a core technology in areas ranging from quantum communication to exquisitely precise sensors and imaging systems. But what do you do when you want to count more than one photon? Two papers recently published in Optica offer different solutions to the quandary.

In one study, researchers at Duke University, Ohio State University and the company Quantum Opus unveil a technique for tweaking a widely used lab technology to detect and resolve up to four separate photons in a single optical pulse (Optica, doi: 10.1364/OPTICA.4.001534). And a separate paper, from Dartmouth University, describes a highly sensitive sensor for imaging applications that can pick out and resolve individual photons at room temperature, and that’s compatible with the CMOS technology used for conventional image sensors (Optica, doi: 10.1364/OPTICA.4.001474).

A lab staple

The Duke/Ohio State/Quantum Opus work, led by OSA Fellows Daniel Gauthier at Ohio State and Jungsang Kim at Duke, involved superconducting nanowire single-photon detectors (SNSPDs), devices that have become a staple item in quantum-technology labs. These detectors, which operate at cryogenic temperatures of only around 4 K, consist of a tiny loop of superconducting wire that experiences a brief loss of superconductivity when a stray photon wanders by, which in turn can be read out as an electrical signal.

SNSPDs have been a boon in the quantum lab, with extremely high detection efficiencies (approaching 100 percent) and with low “dark count” rates—false positive signals of single photons in the absence of light. In one sense, though, the “single-photon detector” part of SNSPD is a bit of a misnomer; the devices can detect the presence of at least one photon, but can’t resolve the actual number of photons in a wave packet that might contain more than one.

Resolving photon number is a problem that crops up increasingly often in quantum-optics experiments; as a result, a number of labs have looked into ways to nudge SNSPDs into a mode that can resolve the number of photons in a pulse. Some of those approaches have had success in certain applications. But they’ve also often degraded the performance metrics of SNSPDs, and led to slower operation and higher dark counts.

Rethinking the signal

To get past this problem, the team led by Gauthier and Kim opted not to invent a new SNSPD, but to find a way to pull more information out of the signal from existing devices. By modeling the resistance of the nanowire in a conventional SNSPD from Quantum Opus as a function of photon number, the researchers found that the varying resistances led to varying rise times of the electrical signal. Those differing rise times could in principle be mapped to the number of photons in a given optical pulse, once the signal passed through a sufficiently high-bandwidth amplifier.

The team tested the scheme in the lab by pumping wave packets from a 1550-nm laser, attenuated to have a specific mean number of photons per packet, into a length of optical fiber tied to the SNSPD, and then reading out the signal from a cryogenic high-bandwidth amplifier. They found that they could resolve counts of up to four photons in a given wave packet.

One advantage of the scheme, according to the researchers, is that it does not require any changes to the SNSPDs that labs may already be using—all that’s needed is the addition of a cryogenically cooled, high-bandwidth amp for the read-out circuit at the end. The team is working on boosting the resolving power of the method to ten or more photons. And the researchers see immediate potential use for the technique, even in its current state, in labs working on emerging areas such as quantum encryption and quantum information.

A room-temperature counter

CMOS QIS demo

The Dartmouth researchers offered a demonstration of their quanta image sensor that involved building up a complete image (right-hand side) by stacking eight individual frames of continuously acquired binary images. [Image: Jiaju Ma]

In another study, the Dartmouth lab of OSA Fellow Eric Fossum—who in early 2017 was a co-recipient of the prestigious Queen Elizabeth Prize in engineering for his pioneering work on CMOS image sensors—looked at the problem of multi-photon detection outside of the quantum lab and its cryogenic capabilities, in room-temperature sensing and imaging applications.

Commercially available room-temperature photon counters, such as single-photon avalanche diodes (SPADs), commonly rely on electron avalanche multiplication to get a readable signal from single photons. That, in turn, requires high operating voltages, which don’t square well with advanced CMOS technology and, thus, limit the size and resolution of such devices. SPADs and other avalanche devices are also susceptible to high dark counts and can have relatively low quantum efficiencies and slow readout speeds.

A collection of “jots”

Fossum and his team sought an alternative approach that could leverage CMOS technology and avoid the pitfalls of avalanche-based amplification. The result is an instrument that the group calls a quanta image sensor (QIS).

The QIS consists of billions of tiny, specialized pixels, called “jots” by the Dartmouth team. The jots, with a pitch size of 1.1 microns, collect individual photoelectrons and accumulate them over some integration period, returning a single- or multi-bit value depending on the number of electrons collected. (The photoelectrons themselves are generated as incident photons hit a silicon surface.) The key to counting individual photons is scanning the sensor at an extremely high rate (a thousand frames per second). The team also gave special attention to adjustments to minimize the instrument’s read noise—electrical noise at a key capacitor point that, if too high, would swamp the delicate signal from individual photons.

The team tested the approach in a two-layer CMOS device, with one wafer including the jots and the other layer containing readout and control circuits. They were able to clearly see the signatures of multiple individual photoelectrons, corresponding to high-probability photon counts, with “ultra-low” read-noise and dark-count levels, and with quantum efficiencies ranging from 70 to 80 percent. They also put the technology through its paces with a high-speed imaging demo in which a grayscale image of a tiger (above) was built up at a fast scanning rate, with power consumption of only 17 mW.

The Dartmouth researchers conclude that the QIS is “qualified for high-speed photon-counting imaging with high spatial resolution,” and expects that the technology “will be widely adopted in scientific and space imaging, life science, security, automotive, and other applications in the near future.” To that end, Fossum and two of his co-authors, Jiaju Ma and Saleh Masoodian, have founded a startup firm, Gigajot Technology, to commercialize the sensors.