Realising high-performance sensors with heterogeneous integration

Manufacturing InGaAs photodetectors directly on CMOS silicon revolutionises shortwave infrared sensors for consumer markets.

BY BEI SHI AND JONATHAN KLAMKIN FROM AELUMA

In quite a number of applications there is much demand for scalable, low-cost, high-performance sensors. Such sensors are wanted for autonomous systems, robotics, defence and aerospace technologies, artificial intelligence (AI), mobile phones, smart watches, and augmented reality/virtual reality (AR/VR).

Within the sensor portfolio, those based on silicon CMOS dominate detection in the visible as well as in the near-infrared at wavelengths up to around 940 nm. However, silicon sensors fail to cover the vast majority of the short-wave infrared (SWIR) domain, which spans 900 nm to 1700 nm. That’s a significant impediment, because sensors operating within this spectral band enjoy a number of valuable benefits, including: lower solar interference, enabling better outdoor operation; access to highly mature components, including lasers, for illumination-detection scenarios such as light detection and ranging (lidar), which is used in 3D imaging and autonomous systems; performance in the dark, enabling night vision; the possibility to detect in inclement weather conditions, such as fog and rain; and ‘eye-safe’ wavelengths, which allow for a much higher illumination power in the presence of people – this enables higher-resolution, longer-range imaging for facial identification, lidar, and other 3D imaging applications.

Figure 1. (a) Sensor response comparison for silicon and InGaAs detectors. (b) Solar irradiance and maximum permissible exposure of eyes to the lidar emitter power from visible to SWIR. Red vertical bars indicate typical lidar operation wavelength bands.

Silicon versus InGaAs
When operating at 940 nm, a wavelength commonly used in mobile devices, silicon detectors benefit from a low solar interference. However, they are held back by a very low detection efficiency. Here a more attractive alternative is InGaAs, which as well as offering a much higher efficiency at 940 nm, is capable of operating in eye-safe wavelength bands. One of the great strengths of InGaAs sensors is that they can cover a much broader spectrum than their silicon counterparts, spanning from the visible to as far as 2600 nm – it’s an extensive response that allows them to be used for enhanced night vision and thermal imaging. Within the spectral response of InGaAs detectors sit bands for consumer lidar and 3D imaging at 905 nm to 940 nm, 1130 nm, 1350 nm to 1400 nm, and 1550 nm.

Note that while the first of these, the 905 nm to 940 nm band, is also supported by mature silicon detectors, such as silicon single-photon avalanche detectors (SPADs), it is not eye safe, which limits the maximum detectable object distance. That makes detectors based on the family of InGaAs materials compelling candidates, because they deliver high efficiency detection at longer, eye-safe wavelengths. It should be noted that although InGaAs detectors are not as well-known as those based on silicon, they are proven in many formats, including p-i-n photodiodes, linear-mode avalanche photodiodes (APDs), and Geiger-mode APDs that are otherwise known as SPADs. (A comparison of the sensor response of silicon detectors and InGaAs SWIR detectors is shown in Figure 1, and a comparison of photodetector technologies for the near infrared and SWIR sensing applications is provided in Figure 2).

Figure 2. Technology comparison regarding sensor manufacturing scalability, cost and performance. Aeluma is focused on large-wafer InGaAs-on-silicon photodetector manufacturing.

Performance, scale, cost and applicability
Although the quantum efficiency – defined as the fraction of incident light that contributes to a photocurrent signal – is low for silicon detectors in the commonly used 905 nm to 940 nm region, that’s not stopped this technology from being widely adopted for consumer applications, including mobile devices and automotive lidar. Behind this adoption are cost considerations. Traditional InGaAs detectors are manufactured on InP substrates, which are typically confined to 2- to 4-inches in diameter, and are expensive and fragile. The latter weakness is significant, because it poses handling challenges during manufacturing processes.

Yet despite this fragility, InP is widely used for manufacturing semiconductor lasers operating at SWIR wavelengths for telecommunication applications. Part of the reason for this appeal is that these InP lasers are far smaller than photodetector arrays used for 3D imaging and lidar. This substantial increase in size is an important consideration, because InP technology does not scale to larger substrate sizes, due to its high cost and fragility. But from a performance perspective, InGaAs-on-InP is by far the better candidate for SWIR imaging and lidar applications.

Driven by the demand for high-performance sensors in consumer applications, alternative technologies are generating much attention for SWIR detectors. Contenders piquing much interest include those based on the germanium-on-silicon and colloidal quantum dot systems. Under development, both technologies provide a path for scaling, but they are somewhat unproven, especially for the common 1550 nm SWIR wavelength. What’s more, both candidates only provide a limited performance. For wavelengths beyond around 1500 nm, they are impeded by a low quantum efficiency – and those based on germanium also suffer from a high dark current.