Advances in time-of-flight and time-correlated single-photon-counting devices

Single-photon avalanche diodes (SPADs) are semiconductor devices with picosecond time resolution that are used to detect low-intensity signals. They are used in several applications, including fluorescence life-time imaging, positron emission tomography, and 3D imaging. SPADs were developed as an alternative to photon multiplier tube and microchannel plate technologies to provide better cost-performance ratios than the older devices, and to give less complex mechanics and biases.¹

Originally, SPADs were built using specially designed silicon-based processes and were capable of both high temporal and spatial resolutions. SPADS have recently been redesigned and manufactured in a slightly adapted CMOS process to provide further optimization and higher readout circuit complexity. This has significantly increased their popularity in the fields of spectroscopy and imaging.

We have developed arrays based on SPAD smart pixels for advanced time-of-flight (TOF) and time-correlated single-photon-counting applications. These arrays are fabricated using a cost-effective 0.35μm CMOS technology. Our work is conducted as part of the European Commission's MiSPiA (Microelectronic Single-Photon 3D Imaging Arrays for low-light high-speed Safety and Security Applications) project.

Our new SPAD device consists of a p-n junction (interface between two types of semiconductor material), based on a highly doped p⁺ layer diffused within a deep n-well (see Figure 1) that is reverse-biased beyond its breakdown voltage (the minimum voltage that causes part of an insulator to become electrically conductive).² This photosensitive p-n junction thus operates in Geiger mode, where an electron-hole pair generated by a single impinging photon ignites a charge carrier avalanche. This event is counted, for example, by a 10 bit digital counter that is incorporated into each SPAD pixel.

Figure 1. Schematic cross section of an integrated CMOS single-photon-counting avalanche diode (SPAD) device.²HV: High-voltage. p, n: Semiconductor materials.

Our new technology eliminates problems with avalanche photodiodes that are related to the strong dependence of the amplification factor on temperature or biasing. The deep n-well isolates the photoactive and avalanche regions of each SPAD from the surrounding electronics and neighboring pixels. In addition, electrical breakdown, which is essential for SPAD operation, must be confined to the active area of each SPAD. We achieve this by keeping the electrical field, located beneath the p⁺-layer-based anode (see Figure 1), higher in the direction perpendicular to the wafer surface than in any other part of the SPAD structure at all times. For this reason, a low-doped p guard-ring surrounds the SPAD active area, and a special n implantation is incorporated to increase the electrical field in the SPAD active region (as shown in Figure 1). The n implantation is also used to adjust the breakdown voltage of the device.

The large (30μm active area diameter) CMOS detector we use has a very low noise level: 12 counts per second (cps) at room temperature and 5V of excess bias over the breakdown voltage. As shown in Figure 2, the noise rises to just 100cps for 50μm active area diameter SPADs, and remains below 100,000cps for those devices with 500μm active area diameters under the same operating conditions.³

Figure 2. Dark counting rate of SPADs with different diameter photoactive regions.³cps: Counts per second.

The SPADs we have developed yield high photon-detection efficiencies in a wide wavelength range, between 250nm (UV) and 1000nm (near-IR). This strong absorption is due to the modified stoichiometry of the silicon nitride-based passivation layer that is commonly used in CMOS processes. The analog front-end electronics in our system quickly sense and quench the avalanches. This creates an almost negligible after-pulsing effect if a 20ns time delay is incorporated after each event. The in-pixel 10 bit time-to-digital converter provides 312ps resolution and 320ns full-scale range. This equates to a single shot with 10cm spatial resolution within a 50m depth range for a TOF system.⁴ Our arrays range in size from 32 × 32 to 32 × 64 pixels, with areas of 100 and 150μm². We also use microlenses on top of the SPAD arrays (see Figure 3) to solve the photon fill-factor problem, where for a 30μm SPAD, only 3% of its area is photoactive.

Figure 3. Microphotograph of the 32 ×64 pixel CMOS SPAD imager with microlenses on top.

The in-pixel 10 bit memory and output buffers of our smart pixel devices make them a viable part of advanced single-photon imager arrays that can be used for 3D depth ranging in safety and security applications and for 2D fluorescence lifetime decays in biomedical imaging. We are currently working to develop back-side illuminated SPAD arrays with fill factors of about 70%.