Advanced Near-Infrared Cameras

Many imaging applications in the near-infrared (NIR) region of the spectrum require sensors that see beyond the traditional charge-coupled device (CCD) sensor’s spectral range. A standard uncoated CCD detector cuts off at a wavelength of around 1100 nm, where the silicon becomes transparent. Special wave-shifting coatings can push this cut-off wavelength out to 1600 nm and beyond, but the sensitivity of the coated sensor decreases dramatically. A better detector material is required for many applications that demand high sensitivity and a spectral response out beyond 1600 nm. Indium gallium arsenide (InGaAs) is that material.

Figures 1. Three images taken through the backside of a silicon wafer with an InGaAs camera. Various waveband filters at (a) 1000 nm, (b) 1050 nm, and (c) 1100 nm.

InGaAs detectors offer quantum efficiencies of ~85% in the 900-1680 nm waveband, which is higher than the competing technologies of lead-oxysulfide vidicons and coated CCD cameras. Commercially available NIR cameras built with InGaAs focal-plane arrays (FPAs) are rapidly eclipsing other sensors in this waveband. The superior performance of these cameras results from both the FPA design, which has ultra low noise on-chip amplification and the availability of high quality InGaAs material in bulk wafers. These sensors deliver excellent image quality over a wide dynamic range, including the very low light levels encountered in spectroscopy. Other applications include semiconductor wafer inspection, laser-beam characterization, and water and chemical detection.

Alternatives & Limitations

Silicon is transparent at wavelengths longer than 1100 nm, which makes silicon-based cameras ineffective at wave-lengths longer than that unless they are coated with a wave-shifting material. But the downside of the coating is that it drastically lowers quantum efficiency of the material to 1-2% in the 1100-1700 nm waveband. That limits its use primarily to laser beam profiling — an application in which intensities are high.

High sensitivity is important because the scenes typically viewed in NIR imaging cover a wide dynamic range, and the intensity of light signals can be very low. That is especially true in NIR imaging spectroscopy, where only a small portion of the InGaAs sensors’ passband is admitted.

Figures 2. Man in front of truck imaged by an InGaAs camera with (a) 900 - 1000 nm filter, and (b) 1400 - 1600 nm filter. (c) Shows the difference between the two images, false-colored in red and overlaid onto the 900 - 1000 nm image.

Indium antimonide (InSb) FPAs have high quantum efficiency in the NIR band, but they require cooling to cryogenic temperatures. Their spectral response is from 1.5 to 5.5 microns, requiring that the sensor be combined with a bandpass filter to make a camera that operates solely in the NIR band. These factors combine to boost the cost of this type of camera. In addition, mechanical cryocooler life is a limiting factor. The coolers cost about $10,000, making replacement costly in applications where the camera runs continuously.

In contrast, because the bandgap energy of InGaAs is much higher than for InSb, InGaAs FPAs can operate at temperatures around ambient (25 °C), with noise performance comparable to InSb sensors at liquid nitrogen temperatures. The thermoelectric coolers used with these FPAs cost about $20, making the overall camera much less expensive.

FPA Fabrication

InGaAs FPAs are fabricated by growing photodiodes on 3-inch-diameter wafers. The wafers are made by metal-oxide chemical vapor deposition (MOCVD) of InGaAs onto indium phosphide (InP) wafers. The FPA detectors are approximately 30 microns square, and are made by diffusing a p-type zinc through a diffusion mask into an n-type InGaAs substrate.

Figures 3a. Two containers of liquid: Methanol (left) and Water (right).

The next step is metal deposition and diffusion into the InGaAs layer to form ohmic contacts. A total of 81,920 cone-shaped indium bumps are deposited onto the contact metal pads, making a 320-by-256 pixel array. The same number of indium bumps is deposited onto a silicon mixed-signal IC die. Cold-forming then pressure-fuses the indium bumps on the InGaAs detector dies to the bumps on the read-out IC (ROIC) wafer.

Technicians bond wires between the FPA and a motherboard inside a dewar package. The motherboard traces are then wire-bonded to the package pins. The package includes a glass or quartz window, which has an anti-reflection (AR) coating that’s optimized for use in characterizing 1550 nm telecommunication lasers. Without the coating, interference fringes can be generated from reflections off the two window surfaces.

An AR coating is also deposited directly on the detectors. The combination of the AR window coating and the detector AR coating yields an extremely high transmission — in the 1300-1550 nm range. The vacuum package also incorporates thermoelectric cooling to stabilize the FPA at an operating temperature between 0 and 20 °C. The FPAs used in cameras have 99.5% operability and few cluster defects. Further, the few bad pixels are replaced using software algorithms that substitute neighboring good pixels for them. These semiconductor and software advancements result in image quality that competes favorably with that of silicon CCD detectors.

Wafer Inspection

In wafer inspection, the cameras deliver precise images of wafer surface features, even though the images are taken through the backside of the wafer. The underside of a wafer is opaque to visible light, but silicon becomes transparent at wavelengths longer than 1100 nm.

Figures 3b. NIR transmission spectra show the different signatures of the two liquids.

The transition from opacity to transparency is gradual for some CMOS wafers, occurring over a span of 100 nm. Figure 1a-c (wafer images) shows three images taken through the backside of a silicon wafer on which CMOS ROICs have been fabricated. The InGaAs camera is fitted with one of three bandpass filters, having center wavelengths of 1000 nm, 1050 nm and 1100 nm, respectively. The filters all have a width of 20 nm full-width half-maximum (FWHM).

The black line in the 1000 nm image is an ink mark on the wafer’s surface, and the circuitry is barely visible through the silicon. But at 1050 and 1100 nm, progressively deeper layers of circuitry are exposed, allowing voids and cracks within the wafer’s volume to be seen, and their positions in x, y and z coordinates to be determined.

Laser Beam Profiling

Internet demand has prompted a need for orders-of-magnitude increases in telecommunication bandwidth, which, in turn, has driven the development of higher-band-width optical fiber using dense-wavelength-division multiplexing, or DWDM. These developments have created a need for high-speed laser transmitters and receivers for long-haul data links in the 1300 nm and 1550 nm wavebands, where silicon optical fiber exhibits very low losses.

Fabricating the vertical-cavity surface-emitting laser (VCSEL) transmitters requires precise measurements of the laser energy’s spatial distribution to efficiently couple the energy into a fiber, waveguide or other optical device.

InGaAs detectors perform better than other materials in this beam-profiling application, which is best done using a two-dimensional detector array that is directly illuminated by the laser light. Uncoated silicon CCD detectors cannot be used for profiling 1300-1550 nm laser light. Lead-oxysulfide vidicon cameras — the relatively low-cost sensors historically used for imaging laser energy in these wavebands — suffer from various drawbacks, including high non-linearity, limited dynamic range, image retention, and a low damage threshold.

In contrast, InGaAs sensors offer high resistance to damage — typically 1 watt/cm2 — which is two orders of magnitude greater than is possible with lead-oxysulfide, and they don’t suffer from image retention.

Detecting Water

Another application of NIR imaging uses the wavelength-dependent reflectance of materials containing water to detect the presence of water in a scene. Water is strongly absorbing in the NIR band at wavelengths of 1350 nm or higher, but not nearly as absorbing at shorter wavelengths. Because of the molecular absorption by water, this results in certain objects that are imaged at wavelengths around 1100 nm to be quite reflective, while the same objects imaged at 1350 nm or higher can appear quite dark.

Figure 2a (man in front of truck) shows a person imaged using an InGaAs camera through a filter that passes 900-1000 nm light. The skin is quite reflective. Figure 2b shows the same scene imaged through a filter that passes 1400-1600 nm light. The exposed skin appears quite dark. Figure 2c shows the difference between the other two images, but false-colored in red and overlaid onto the 900-1000 nm image. Exposed skin is readily apparent because of its water content, as is the background vegetation, for the same reason.

Near-Infrared Spectroscopy

NIR spectroscopy is a quick and non-destructive method to analyze the chemical composition of materials. It measures the intensity of NIR light that is reflected or transmitted through a material as a function of wavelength. The resulting intensity-versus-wavelength curves are called NIR spectra, and the shapes of the curves provide information about the chemical species in the material. Imaging spectrometers are used in agricultural inspection, remote sensing, exhaust-gas analysis and pharmaceutical formulation analysis.

Figure 3a is an image of two containers of liquid. The one on the left is methanol, and the one on the right is water. The red boxes within each image were the points used to extract NIR transmission spectra, shown in Figure 3b. The two chemicals’ transmission curves have very different shapes, called signatures, which can be used to uniquely identify compounds. An application for this imaging might be the detection of spilled chemicals, such as gasoline, by an air-borne platform carrying an InGaAs camera.

As illustrated in this article, commercially available InGaAs cameras are performing many new applications in NIR imaging. They enable high-performance imaging in an important region of the electromagnetic spectrum at a cost that is lower than for cooled-sensor technology. The focal plane arrays the cameras incorporate are made using processes developed for the semiconductor industry that are inherently scalable to high-volume production. They also lend themselves to automation, which will enable further cost reduction, opening new markets for InGaAs imaging technology.

This article was written by Austin Richards, Ph. D., senior applications engineer at Indigo Systems Corporation, 50 Castilian Drive, Goleta, CA 93117. To contact Dr. Richards, please call Aileen Wrench of Indigo Systems at (805) 964-9797. Visit Indigo online at www.indigosystems.com .