Pure-crystal-silicon detector has ultralow dark current

Although the field of silicon photonics is progressing rapidly, developing photodetectors remains a major challenge. In most cases, their operating wavelength lies near 1.55μm. That is because silicon is a low-loss material and this matches the operating wavelength of wideband telecommunications systems. Low loss, however, is not suited to detector applications because silicon does not absorb light at 1.55μm. Thus, we need to alter or replace silicon to detect light at this wavelength. Such approaches are ongoing in the silicon-photonics community. For example, germanium (which does absorb light at 1.55μm) can be integrated onto silicon by heteroepitaxy^1,2 or absorption centers that are intentionally introduced into silicon using ion implantation.³ Both approaches have exhibited high efficiency in detector operations. However, such additional processes complicate fabrication and result in high dark current as the crystal quality is degraded.

We propose another approach, using two-photon absorption (TPA) in pure silicon, for light detection. It is widely known that pure-crystal silicon can absorb light at 1.55μm through TPA, but its absorption efficiency is very low since it is based on a third-order nonlinearity. We used a p-i-n diode (with a lightly doped near-intrinsic semiconductor region between the p- and n-type regions) embedded in a high-quality (Q)-factor silicon photonic-crystal nanocavity to greatly enhance TPA efficiency. The result is a high detection efficiency (~20%) with ultralow dark current (15pA) for all-silicon photodetectors in the 1.55μm band.⁴

Figure 1 shows our device structure. We formed⁵ a nanocavity with a mode volume of 0.15μm³ at the center of a line defect in a silicon photonic-crystal slab by slight adjustment (3--9nm) of the air-hole position inside the red hexagon. A p-i-n junction surrounding the cavity was formed by a conventional process.⁶ The most important technological achievement was that the final device shows a surprisingly high Q value. The loaded (with external coupling) and unloaded Q (no external coupling) were 4.3×10⁵ and 8.4×10⁵, respectively, even after fabrication of the p-i-n junction with electrical pads, accomplished through careful device design.

Figure 1. Schematic of an all-silicon photodetector with a photonic-crystal nanocavity embedded in a p-i-n junction with electrical pads. Insulator-region and p-i-n-junction widths: w_i = 8.72 and w_w = 8.4μm. Triangular air-hole photonic crystal period: 420nm. Slab thickness: 204nm. Measured loaded Q: 4.3 ×10⁵.

Such high Q and ultrasmall mode volume may greatly enhance the TPA process. Indeed, this device shows high detection efficiency at 1.55μm. Figure 2 summarizes its photocurrent response as a function of the input power at a bias voltage of −3V. Red and black circles represent the response of devices with and without a cavity, respectively. Deviation from a linear curve corresponds to TPA onset. The figure clearly shows that TPA dominates the photocurrent of our device, which is greatly enhanced with respect to the reference device without a cavity. The upper green dashed curve represents perfect detection (a photon is converted to an electron), which we denote as a quantum efficiency (QE) of 100%. Our device performance corresponds to a QE of approximately 10%, which means that ~20% of photons are converted to electrons (since two photons generate one electron through TPA). If we take the cavity-waveguide-coupling efficiency into account, 44% of photons are absorbed. This efficiency is comparable to that for conventional linear-absorption photodetectors. We achieved this high conversion efficiency for TPA at a micro-Watt optical input level, which is primarily caused by the high Q and small mode volume of the photonic-crystal cavity.

Figure 2. Photocurrent response of an all-silicon photodetector. Red and black solid symbols correspond to experimental data for a sample with a cavity and a reference without one, respectively. The blue curve was obtained by theoretical simulation assuming two-photon absorption in silicon. The upper green dashed line represents the theoretical upper limit of the quantum efficiency (QE). A/W: Ampere per Watt.

Another important feature of our device is its dark-current performance. The device showed that a dark current of 15pA has a bias voltage of −3V. This is two to four orders of magnitude smaller than for other detectors based on silicon because we use only pure-crystal silicon.

Our work has demonstrated that we can realize an efficient photodetector for the 1.55μm band consisting of pure-crystal silicon. We have shown that we can boost TPA at the micro-Watt level comparable to linear absorption, which may open up novel functions. It is also worth noting that our device is very small (8.4 × 8.72 × 0.2μm³) and easily integrated within a chip. The small size also results in a very small capacitance (~6aF), indicating potentially faster operation (although the speed of our device is currently limited below 1Gbps because of unoptimized electrical-contact formation).

As a next step, we aim to integrate these nanoscale photoreceivers with other photonic devices (such as switches, modulators, and splitters that are also based on photonic crystals) within a single chip. We believe that this will enable photonic network-on-chip applications in the future. In addition, this very efficient two-photon detection scheme may open up new possibilities in integrated quantum-communication systems.

This work is supported in part by the Core Research for Evolutionary Science and Technology program of the Japan Science and Technology Agency.