Detector with internal gain for short-

wave infrared ranging applications

Vala Fathipour
Hooman Mohseni

Vala Fathipour, Hooman Mohseni, “Detector with internal gain for short-wave infrared ranging applications,”
Opt. Eng. 56(9), 091608 (2017), doi: 10.1117/1.OE.56.9.091608.

Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 05/22/2017 Terms of Use: http://spiedigitallibrary.org/ss/termsofuse.aspx

 Optical Engineering 56(9), 091608 (September 2017)

Detector with internal gain for short-wave infrared ranging

applications
Vala Fathipour and Hooman Mohseni*
Northwestern University, Bio-Inspired Sensors and Optoelectronics Laboratory, Evanston, Illinois, United States

Abstarct. Highly sensitive photon detectors are regarded as the key enabling elements in many applications.
Due to the low photon energy at the short-wave infrared (SWIR), photon detection and imaging at this band are
very challenging. As such, many efforts in photon detector research are directed toward improving the perfor-
mance of the photon detectors operating in this wavelength range. To solve these problems, we have developed
an electron-injection (EI) technique. The significance of this detection mechanism is that it can provide both high
efficiency and high sensitivity at room temperature, a condition that is very difficult to achieve in conventional
SWIR detectors. An EI detector offers an overall system-level sensitivity enhancement due to a feedback
stabilized internal avalanche-free gain. Devices exhibit an excess noise of unity, operate in linear mode, require
bias voltage of a few volts, and have a cutoff wavelength of 1700 nm. We review the material system, operating
principle, and development of EI detectors. The shortcomings of the first-generation devices were addressed in
the second-generation detectors. Measurement on second-generation devices showed a high-speed response
of ∼6 ns rise time, low jitter of less than 20 ps, high amplification of more than 2000 (at optical power levels larger
than a few nW), unity excess noise factor, and low leakage current (amplified dark current ∼10 nA at a bias
voltage of −3 V and at room temperature. These characteristics make EI detectors a good candidate for
high-resolution flash light detection and ranging (LiDAR) applications with millimeter scale depth resolution
at longer ranges compared with conventional p-i-n diodes. Based on our experimentally measured device
characteristics, we compare the performance of the EI detector with commercially available linear mode InGaAs
avalanche photodiode (APD) as well as a p-i-n diode using a theoretical model. Flash LiDAR images obtained by
our model show that the EI detector array achieves better resolution with higher signal-to-noise compared with
both the InGaAs APD and the p-i-n array (of 100 × 100 elements). We have designed a laboratory setup with a
receiver optics aperture diameter of 3 mm that allows an EI detector (with 30-μm absorber diameter) to be used
for long-range LiDAR imaging with subcentimeter resolution. © 2017 Society of Photo-Optical Instrumentation Engineers (SPIE)
[DOI: 10.1117/1.OE.56.9.091608]

Keywords: photodetector; infrared imaging; infrared detector; electron-injection detector; phototransistor.

Paper 161631SSP received Oct. 19, 2016; accepted for publication Feb. 13, 2017; published online Mar. 22, 2017.

1 Introduction large area (∼m2 ) detectors are being built for the large hadron
A photon is defined as a fundamental excitation of a single collider (LHC)-collider.14–18 Imaging in ultraviolet or x-ray
mode of the quantized electromagnetic field.1 Planck was the region is easy to accomplish as the large photon energy is
first person to explain the blackbody radiation spectrum in converted into a number of charged pairs,12 which represent
1900. He explained this behavior by quantization of electro- the photon energy.19 Despite the fact that single-photon
magnetic radiation. Einstein utilized this concept to explain detection and imaging for longer wavelengths have also
the photoelectric effect2 in 1905, and Compton used it to existed for many years, they have until now developed very
explain the wavelength shift of scattered x-rays in 1923. slowly.20,21 This is because detection of longer wavelengths
Lewis introduced the term “photon” for the first time in requires utilization of semiconductors with a smaller energy
1926.2,3 The formal quantization of the electromagnetic bandgap (Eg) for the creation of the mobile charges, as
field was first introduced by Dirac in 1927.2 A single photon shown in Eq. (1):
in mode k has energy equal to hνk, where h is the Planck’s
constant and νk is its frequency.2 Eg ¼ hc∕λg;
EQ-TARGET;temp:intralink-;e001;326;221 (1)
The photon can be detected by a solid-state detector if it
can generate mobile charges (signal quanta) in the detector where λg is the cutoff wavelength, h is the Planck’s constant,
material. Single-photon imaging for high-energy photons and c is the speed of light. Reduction of Eg, however, results
(ultraviolet or x-ray) has existed for many years. It started in the exponential increase of the generation recombination
as early as ∼1912 using “cloud chamber,” which is an dark current density of the detector (J GR ); as such, semicon-
“imaging detector,”4–6 and switched to using solid-state ductor detectors have relatively poor performance at longer
electronic detectors in the 1970s (Fig. 1). Today, solid-state wavelengths. This is shown as
high-energy photon detectors excel technologically, and
J GR αT 3∕2 e−Eg∕2KT WðVÞ;
EQ-TARGET;temp:intralink-;e002;326;124 (2)

*Address all correspondence to: Hooman Mohseni, E-mail: hmohseni@

northwestern.edu 0091-3286/2017/$25.00 © 2017 SPIE

Optical Engineering 091608-1 September 2017 • Vol. 56(9)

Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 05/22/2017 Terms of Use: http://spiedigitallibrary.org/ss/termsofuse.aspx

 Fathipour and Mohseni: Detector with internal gain for short-wave infrared. . .

Fig. 1 Timeline of initial demonstrations of single-photon imager instrumentation. The cutoff wavelength
for each technology is shown by markers. The general trend that x-ray imagers with single-photon sen-
sitivity were demonstrated first, then visible, and then infrared shows the importance and the challenge of
imaging at the single-photon level at lower photon energy levels. Today, single-photon image sensors
operating at visible wavelength are increasingly available even to the consumer market. Data obtained
from Refs. 7–13.

where T is the temperature and W is the depletion width, applications as they rely on vacuum tube technology. This
which is a function of applied bias voltage, V. From Eq. (2), prevents their ability to be assembled in large arrays
it can be concluded that single-photon detection is severely with high pixel density. Furthermore, for infrared region,
hampered at the infrared regime, and room-temperature PMT performance is poor compared with the visible
single-photon electronic imaging is restricted to detector wavelength.
materials with λg below about 1.3 μm.19 Solid-state CCDs utilizing electron multiplication
Another chronic difficulty especially for visible and infra- (EMCCDs) became commercially available in 2001.23,24
red photon-starved imaging conditions, such as in low light Similar to CCDs, EMCCDs are inherently suited for imaging
surveillance imaging7 through high-speed, single-photon applications, while providing enhanced signal-to-noise ratio
imaging for the life sciences8,9 and in astronomical observa- (SNR) for signal levels below the CCD readout noise floor.
tions, is the readout noise. This is the additional noise added This is achieved in the EMCCD by increasing the signal
by the electronic charge detection circuit, where the detected through impact ionization. However, they exhibit an inherent
photoelectrons are converted into an output voltage or uncertainty in the multiplication process, which elevates their
current signal. The readout noise becomes more severe noise levels, and they present an excess noise factor
when high frame rate imaging is required. Detectors without F2 ∼ 2.2.25,26 Furthermore, due to their readout technique,
any internal amplification, such as the charged-couple they cannot be utilized in high frame rate imaging.7,27 The
devices (CCDs), which is the first widely commercialized highest data rates presented within this technology are oper-
electronic imaging technology,10 and the p-i-n diodes, are ations up to a clock frequency of 35 MHz.28 Extremely fast
inherently unable to provide accurate measurements of imaging can be performed with image intensified CCDs
fast low-intensity transients at high frame rates. In fact, (ICCDs).29 However, ICCDs are not a fully solid-state tech-
the ultimate sensitivity limitation of a CCD or a p-i-n detec- nology. As such, ICCDs are not a creditable alternative to
tor is set by the readout noise of the first-stage amplifier, EMCCDs in high frame rate imaging applications. Finally,
which becomes more severe at faster readout rates.11 To EMCCDs have a wavelength sensitivity of ∼0.3 to ∼1.1 μm
be able to respond to single photons, the photon detector and are not suitable for infrared imaging.
must exploit an internal multiplication process to achieve The APD, developed in the 1960s, uses a similar process
subelectron input referred readout noise.12 Although the to the PMT but in a semiconductor platform.30,31 APD detec-
noise contribution of such detectors is unavoidable and is tor arrays satisfy the significantly increased demands on high
always higher than a p–i–n diode, the contribution from pixel rates, and frame rates as high as 1 GHz have been
the amplifier can be lower than the detector in the presence reported in the literature.32 Unfortunately, APDs require
of gain in the detector.13,22 Detectors with an internal ampli- high bias voltages and are sensitive to material inhomogene-
fication of the charge carriers are suitable for high frame rate ity. As such, to benefit from their advantages in focal plane
single-photon imaging including photomultipliers (PMTs), arrays (FPAs), specifically designed readout circuits capable
avalanche photodiodes (APDs), and electron-multiplying of applying a high bias with very low noise are critical.
CCDs (EMCCDs). Furthermore, due to a superexponential gain characteristic,
While visible PMTs are impressive as a pioneering tech- the yield in achieving uniform arrays is low. For example,
nology, they cannot satisfy requirements of many modern a mere fraction of a percent variation in the epilayer

Optical Engineering 091608-2 September 2017 • Vol. 56(9)

Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 05/22/2017 Terms of Use: http://spiedigitallibrary.org/ss/termsofuse.aspx

 Fathipour and Mohseni: Detector with internal gain for short-wave infrared. . .

thickness or doping concentration results in sizeable shifts in Electron-injection (EI) detectors are an alternative
the APD output voltage that may render an array unusable.33 approach to above detection technologies and were intro-
The large electric field in the device leads to surface duced in 2007.52 They utilize the exact bandgap required for
breakdown mechanisms, and consequently guard rings SWIR detection and have a cutoff wavelength of 1700 nm.
are required to prevent this phenomenon. The guard rings Compared with the SWIR HgCdTe eAPDs, they require
increase the pixel pitch and reduce the fill factor. much less cooling. Similar to p-i-n diodes, they operate at
Furthermore, the APD pixels need to be spaced apart to pre- CMOS compatible bias voltages53 and have a low electric
vent cross talk due to carrier re-emission. Hence, realization field in the device (∼40 KV∕cm). This technology, together
of high-resolution imagers has remained a challenging with the stable detector characteristics, makes formation of
task for APD-based imagers. Recent advances in device large-format high-pixel-density FPAs less challenging for
design and epitaxial growth have made the formation of low photon flux applications. Similar to linear mode APDs
256 × 320 pixel arrays with 30-μm pitch possible.34,35 On and EMCCDs, they provide an internal amplification, which
the other hand, detector technologies with a lower electric suppresses the readout noise. On the other hand, unlike
field and with stable characteristics, such as p-i-n detectors, APDs and EMCCDs, due to an inherent negative feedback
are less sensitive to material inhomogeneities. As such, they inside the device, the amplification mechanism is avalanche-
can form high-density large-area imaging FPAs with array free and stable, and the devices show an excess noise of
sizes growing in proportion to the ability of readout circuit near-unity.54–57 Another benefit of this technology is that it
technology to read, process, and demonstrate the detector satisfies the significantly increased demands on pixel rates
signals. For example, currently, p–i–n-based FPAs with (micro/nanosecond level acquisition times is possible) as a
4096 × 4096 pixel arrays and 10-μm pixel pitch are avail- result of its large stable gain and high-speed response.58
able (Hawaii-4RG).36–38 Unfortunately, short-wave infrared Another advantage for EI detectors over HgCdTe eAPDs is
(SWIR) APDs based on III-V material exhibit typical noise that they are based on the mature InP material system and
factors of F ∼ 4 to 5,39 which is caused by the stochastic are realized by virtue of the widely available commercial
nature of the gain process.40,41 As such, a huge amount of III-V-based foundries.
work to reduce the multiplication noise in APDs has been
reported in the literature.42–44 The mercury cadmium telluride
(HgCdTe) material system has addressed this issue, and for 2 Layer Structure and Band Diagram
the midwave infrared (MWIR) and long-wave infrared spec- EI device schematic is shown in Fig. 2(a). The layer structure
tral ranges, detectors have an excess noise factor of near- consists of 1000 nm n− -doped In0.53 Ga0.47 As absorber,
unity.45–47 This is the result of a nearly exclusive impaction 50 nm pþ -doped GaAs0.52 Sb0.48 trapping layer, 50 nm-
ionization of the electrons.39 Unfortunately, in the SWIR undoped In0.52 Al0.48 As etch stop, 500 nm nþ -doped InP
region between 1 and 2.5 μm, which has vast applications injector, and 50 nm nþ -doped In0.53 Ga0.47 As capping layer.
including telecommunication, remote sensing, astronomical The epitaxial layers are grown with metal organic chemical
observation, spectro-radiometry, and spectro-photometry, vapor deposition on 2-in. InP substrates. The band diagram
the HgCdTe “electron-initiated APDs (eAPDs)” do not offer through the central axis of the device as a function of depth,
any gain.48 For the SWIR region, to obtain gain with a in darkness (blue) and under illumination (red), is shown in
near-unity excess noise, detectors with a bandgap in MWIR Fig. 2(b). When an electron–hole pair is generated in an
are typically cooled to ∼60 K.49,50 The extensive cooling is InGaAs absorber through optical excitation, the hole is
because HgCdTe eAPDs utilize a bandgap that is much trapped in a very small volume (GaAsSb trapping layer).
smaller than what is needed for SWIR detection. Further- This creates a high concentration that reduces the potential
more, low-pass filters are utilized to filter out the longer barrier and leads to the injection of many electrons to the
wavelengths.51 absorption layer.52,59

(a) (b)

Fig. 2 (a) The schematic diagram with a cross section showing the layer structure and (b) the band
diagram through the central axis as a function of depth: in darkness (blue) and under illumination
(red) for an EI detector.

Optical Engineering 091608-3 September 2017 • Vol. 56(9)

Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 05/22/2017 Terms of Use: http://spiedigitallibrary.org/ss/termsofuse.aspx

 Fathipour and Mohseni: Detector with internal gain for short-wave infrared. . .

3 Background on Electron-Injection Detector To obtain the range image, the return time of laser pulses
First EI imager with 320 × 256 pixel array and 30-μm pixel reflected from a target is measured. By steering the transmit-
pitch was demonstrated in 2010 using off-the-shelf CMOS ted light, many different points of an environment can be
readout circuit with 575 to 870 electrons rms noise.57,60,61 measured to create a full 3-D model. LiDAR has proven
As a result of the internal charge amplification mechanism to be a vital technology for a variety of applications, includ-
in the detector, the measured imager noise was reduced to ing autonomous vehicles, automated process control, target
28 electrons rms at a frame rate of 1950 frames∕s. These recognition, robots, collision avoidance, remote sensing,
devices were slow (a few KHz bandwidth) and had large aerial surveying, power grid facilities, architectural and
dark currents (∼6 μA at −1.5 V, which prevented long inte- structuring mapping, and oceanographic and archaeological
gration times in the camera. The shortcomings of the first- detection. Due to its numerous applications in urban areas,
generation devices were addressed in the second-generation SWIR is the wavelength of choice for a LiDAR system.
detectors. Compared with first-generation detectors, the Operating in SWIR (1.5 to 1.8 μm) would allow the maxi-
second-generation devices achieved 2 orders of magnitude mum eye-safe power to be at least 100 times higher than the
lower dark current and 4 orders of magnitude enhancement visible wavelength.
in bandwidth. Compared with the best-reported linear mode Despite the extreme usefulness and applications of
avalanche photo detector (SWIR HgCdTe), the second- LiDAR systems today, they are not on every site. This mainly
generation EI detector shows over 2 orders of magnitude stems from their bulkiness, high cost, and slow speed of
lower dark current density at all measured temperatures.54 operation. Furthermore, today’s systems lack millimeter
Second-generation devices achieved high gain (∼2000), scale resolution at longer ranges. Among the different
high bandwidth (fast rise time of ∼10 ns at 20 μW of optical components of a LiDAR receiver system, the optical detector
power), unity excess noise factor, and low leakage current directly affects the instrument sensitivity performance.
(internal dark current density of ∼1 μA∕cm2 at RT decreas- Current detector technologies for SWIR LiDAR systems are
ing to 1 nA∕cm2 at 210 K) at bias voltage of −3 V. A InGaAs p-i-n detectors,67 InGaAs APDs,68,69 or HgCdTe
performance comparison with other SWIR detector technol- eAPDs.70
ogies with internal amplification showed improved perfor- p-i-n detectors can offer extremely good range resolution
mance in terms of noise-equivalent sensitivity with existing as a result of their high-speed response and good timing
SWIR detectors, such as APDs.56 In the EI detectors, the resolution (transit time limited jitter). Jitter values ∼15 ps
optical gain dropped from ∼2000 to ∼30 across an input have been reported in small area devices (15-μm diameter).71
power of 10 to 1 nanowatt.62 The origin of the drop in optical However, these detectors are typically utilized in short-range
gain at low optical powers is related to the recombination LiDAR systems67,72 and require high-power lasers to be
current in the injector/trapping layer heterojunction and has able to obtain long-range imaging. The Army Research
been extensively discussed in our publications.58,62 Recently, Lab (ARL) reported the development of a compact short-
we have reduced defects at heterojunction and verified the range LiDAR system that utilized four 1-mm-diameter
above statement with our devices that have a gain of InGaAs p-i-n detectors on the receiver. The transmitter
∼1000 even at very low photon flux. used a tunable 200 to 400 kHz, 2.6-ns pulse width, and
These improvements have opened up applications for 1.5-μm laser with peak power of 1 KW. At 20 m, their
these detectors in the medical field [optical coherence system obtained 40-cm range resolution and had 256 ðhÞ ×
tomography (OCT)],63 remote sensing [light detection and 128 ðvÞ pixels. To increase the range, their future work
ranging (LiDAR)],64 and astronomy (exoplanet detection). included modifying their receiver using APD-based detec-
In Sec. 4, we elaborate on the application of EI devices tors instead of p-i-n detectors. Using the APD-based receiver
in LiDAR. design, they expected to be able to image targets with
a reflectivity of 0.1 at a 50-m range.
4 Utilization of Electron-Injection Detectors in Light LiDARs based on APDs operate at lower power levels or
Detection and Ranging longer ranges. Unfortunately, InGaAs APDs have an extra
Most of the modern imaging techniques can only image the noise mechanism compared with p-i-n photodiodes (the
two-dimensional projection of a scene and loose the depth so-called excess noise). The high excess noise factor and
information. However, our everyday lives involve three- the low gain prevent them from achieving their ideal shot-
dimensional (3-D) views, which is much more beneficial noise limited SNR performance when operated in linear
in apprehending the characteristics of the targets. 3-D imag- mode. As such, they are usually operated in the Geiger mode
ing has attracted increasing attentions especially in the field for LiDAR applications. The advantage to this mode of
of biology, medical settings, industrial, and consumer operation in a flash LiDAR is that variations in the gain
applications.65,66 Holography, which was demonstrated from pixel to pixel or as a function of operating parameters
after the invention of the laser, is one of the longest estab- or just due to statistical variations become irrelevant.
lished 3-D imaging techniques. OCT, discussed in Ref. 63, is Furthermore, to maintain an acceptable noise, InGaAs APDs
another emerging technology for noninvasive imaging; it is have to be gated. These nonlinear effects severely limit the
based on the principle of interferometry of light waves per- applications of InGaAs APDs in LiDAR, where the signal
forming high-resolution 3-D imaging deep into the tissue. dynamic range spans 2 to 3 orders of magnitude and the sig-
LiDAR is yet another 3-D imaging and spatial measurement nal arrival time is difficult to predict. Finally, APDs have a
technique that is an integral part of any autonomous vehicle non-Gaussian probability distribution of jitter. The InGaAs/
or robot and is becoming ubiquitous in many disciplines. InP APDs have jitter values of about 60 ps with single
In this approach, spatial coordinates associated with each photons,73,74 and researchers have reported achieving
pixel are recorded in a range image acquired by the detector. 30 ps with some trade-offs.75 The MIT Lincoln Laboratory

Optical Engineering 091608-4 September 2017 • Vol. 56(9)

Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 05/22/2017 Terms of Use: http://spiedigitallibrary.org/ss/termsofuse.aspx

 Fathipour and Mohseni: Detector with internal gain for short-wave infrared. . .

reported a LiDAR system that utilized 4 × 4 InGaAs Geiger- preliminary experimental results, which utilizes a small area
mode APD arrays, where each detector was 400 μm. Using a detector (diameter of 30 μm).
1-kHz, 30-μJ micro-Nd:YAG laser at 1 μm, they obtained a Detector gain, rise time, and jitter were measured at differ-
75-cm range resolution at a range of 58 m by averaging 200 ent optical power levels, and the results are shown in Fig. 3.
frames LiDAR images.76 As mentioned in Sec. 1, APDs Verification of LiDAR sensitivity enhancement: Based
based on HgCdTe have addressed this challenge and offer on the measured detector characteristics, a model was
unity excess noise factor.45,49,48 However, for SWIR region, developed in Python to validate the expected performance
detectors with a bandgap in MWIR are typically cooled to 60 improvement for EI detectors in a LiDAR system. Our
K.50,77 As well as the cooling requirement, unfortunately, the model assumes a 1550-nm source laser, with a spectral width
need for high bias voltages and the low yield in achieving of 0.01 nm, divergence of 1 mrad, beam diameter of 1 cm,
uniform arrays makes formation of large area high-resolution laser transmitter efficiency of 0.9, and 6 mJ of energy. The
flash LiDAR systems a challenging task for APD-based receiver diameter is 10 cm, with a field of view (FOV) of
imagers. LETI/SOFADIR used a 320 × 256 HgCdTe eAPD 40 mrad (full cone angle). The power returned to the device
array with a laser at 1.57 μm, pulse width of 8 ns, and pulse was calculated by assuming that the target was an ideal
energy of 8 mJ. They obtained a ranging resolution of about Lambertian surface with an albedo of 0.35 at 0.2 km.
15 cm at a range of 30 m. For this result, the detector had Figure 4 compares EI detector responses in this system for
avalanche gain of 23 and was cooled to 80 K.70 Table 1 com- two cases of single pulse and 100 shot averages. Flash
pares the obtained range, resolution, detector temperature, LiDAR images, assuming 100 × 100 pixels, are demon-
and peak laser power used in each of the above LiDAR strated in Fig. 4(a) with no averaging and in Fig. 4(b)
systems. with 100 times averaging. Cross-section profiles along the
Here, we utilize EI detector technology for high-resolu- middle row of single shot image and 100 shot images are
tion, long-range LiDAR systems. As mentioned in Sec. 1, provided in Figs. 4(c) and 4(d), respectively. For this
similar to p-i-n diodes, the EI detectors operate at CMOS calculation, a simple square target (of reflectivity 0.35)
compatible bias voltages78 and have low noise and low leak- was assumed with a perimeter background whose reflectivity
age current. The jitter in electron-injector detectors is also is half that of the target, just to show some contrast. The
transit time limited and detectors achieve (extremely good) power was assumed to be spread out uniformly over the
approximately tens of picoseconds of jitter performance.58 whole format.
The low electric field in the device together with a sublinear Figure 5 shows the pulse-response comparison of results
gain dependence on the bias voltage makes the formation of provided by the EI detector (a), Hamamatsu G8931-20 APD,
high yield, large-format high pixel density FPAs less chal- and the Hamamatsu 11193 p-i-n averaged over 100 shots.
lenging with this technology for low photon flux flash Figure 6 shows the images obtained with cameras with EI
LiDAR applications.38,59 Here, by providing images based detector array, APD array, and p-i-n array. Here, we assumed
on detailed theoretical modeling using experimentally 100 × 100 imaging arrays of each detector technology:
measured detector data, we show that EI detectors enhance (a) shows EI detector, (b) shows Hamamatsu G8931-20
LiDAR’s resolution and sensitivity. We have validated APD, and (c) shows Hamamatsu 11193 p-i-n array. For this
these results by comparing the LiDAR backscatter profiles plot, data are averaged over 100 shots.
obtained from the EI detector and commercial InGaAs Laboratory setup design: Here, we demonstrate an
p–i–n and APD detectors from Hamamatsu theoretically. We experimental measurement setup that would allow the
then demonstrate our initial laboratory setup for obtaining utilization of a second-generation EI detector in a LiDAR

Table 1 Performance comparison of EI detector array with the existing SWIR LiDAR detectors.

Range Pulsed laser λ T

Detector type (m) Resolution (cm) peak energy (μm) Receiver optics System size (K)

ARL72 Four InGaAs p-i-n 20 40 measured ∼2.6 μJ, 2.6 ns, 1.5 Tapered fiber bundle Fitted on iRobot 300
detectors (1-mm- 200 kHz, fiber laser PackBot
diameter detector)

MIT76 4 × 4 InGaAs Geiger- 58 75 measured 1 kHz, 30 μJ, micro- 1.0 10-cm parabolic/pair 13 × 21 in. box 300
mode APD (400-μm Nd:YAG laser telescope, reimaged
detector) into Maksutov
telescope
LETI/ 256 × 320 pixels and 30 15 measured 8 mJ, 8 ns 1.5 Telescope Optical table 80
SOFADIR70 30-μm pitch eAPD array mounted

NU EI detector 200 0.3 (calculated 40 ns, 6 mJ pulses 1.5 3-mm MEMS mirror Can fit on iRobot 300
(320 × 256 pixels and based on fiber laser PackBot
30 − μm pitch) measured
detector jitter)
Note: T denotes the temperature, λ denotes the wavelength, ARL denotes the Army Research Lab, MIT denotes the MIT Lincoln Laboratory, and
NU denotes Northwestern University.

Optical Engineering 091608-5 September 2017 • Vol. 56(9)

Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 05/22/2017 Terms of Use: http://spiedigitallibrary.org/ss/termsofuse.aspx

 Fathipour and Mohseni: Detector with internal gain for short-wave infrared. . .

(a) (b) (c)

Rise time (ns)

Jitter (ps)
Gain (AU)

Fig. 3 Second-generation EI detector characteristic at various optical power levels: (a) gain, (b) rise time,
and (c) jitter.

system, despite its extremely small (30 μm) active area. Such not only picks up a large background noise but also has to be
a system could then be transformed into the first flash large. Detectors larger than approximately a few millimeters
LiDAR based on EI detectors with a better sensitivity than are typically used to maintain the required FOV. The large
the reported performance. In a typical LiDAR system, while area of the detector increases the detector resistance capaci-
the transmitted beam is scanned over the scene, the receiver tance time constant and degrades jitter performance. An
should see the entire scene.72 In this approach, the detector InGaAs p-i-n detector with a 1-mm area has a transit time

Fig. 4 To assess the performance of the second-generation electron-injector detector in a LiDAR sys-
tem, a model was developed in Python based on measured experimental data. EI detector responses
are presented here: (a) flash LiDAR image single pulse and (b) flash LiDAR image 100 shot averages.
The cross-section profile along the middle row of image: (c) single shot and (d) 100 shot.

Optical Engineering 091608-6 September 2017 • Vol. 56(9)

Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 05/22/2017 Terms of Use: http://spiedigitallibrary.org/ss/termsofuse.aspx

 Fathipour and Mohseni: Detector with internal gain for short-wave infrared. . .

Fig. 5 Typical LiDAR 100 shot pulse average target at 0.2 km: (a) EI detector, (b) APD detector, and
(c) pin detector.

Fig. 6 Flash LiDAR images using 100 × 100 arrays with (a) EI detectors, (b) APD detectors, and (c) pin
detectors, using 100 shot averages.

limited jitter of approximately a few nanoseconds (assuming coupled into the instrument using a beam splitter and is
a saturation velocity of ∼6 × 107 cm∕s79). This results in scanned on the scene by the MEMS mirror. The same
∼40-cm depth resolution.70 To eliminate the need for a MEMS collects light from scene and sends it to detector.
large detector, one could use an extremely large numerical The system achieves 4-mm axial resolution, 3-mm depth res-
aperture (much larger than 1), which is not practical. olution, and 303 × 303 pixels (limited by the number of
In our approach, the detector has a small FOV. The FOV resolvable points on the mirror) at ∼15-m range with an eye
is, however, scanned actively together with the laser beam. safe laser. The laser source has a central wavelength of
The detector remains on axis and is not scanned. The system 1550 nm, with an average power of 8 W and 40-ns pulse
block diagram is shown in Fig. 7. In our setup, the beam is width. The pulse repetition frequency is variable between
100 kHz (80 μJ) and 1000 kHz (8 μJ), and the laser is lin-
early polarized with 15 dB PER and has 15% ASE. The laser
Fiber
HWP PBS QWP is fiber coupled with an NA < 0.08. As such, we utilize a
MEMs:3mm, +/-14 deg
NRPMEMS~303 beam expander (collimator) that can work with the given
Laser
D=3mm low NA fiber and produce the beam diameter of 3 mm,
Range~15 m
Objective
Axial Res~ 4 mm which is dictated by our MEMS mirror size and is the size
Detector
Depth Res~ 3 mm
FOV~1.5 m * 1.5 m
of our receiving aperture. All components in our setup are
Pixels ~303*303 rated for the average and peak power of laser. In specifying
Laser: ~
~ system parameters, given in Fig. 7, the laser transmitter effi-
Average power =8W
Pulse width=40ns
ciency is assumed to be 0.9. The power returned to the device
PRF=100 kHz was calculated by assuming that the target was an ideal
Peak power=20 kW
Energy= 80 uJ
Lambertian surface with an albedo of 0.35. The lateral res-
olution is dictated by the diffraction limited spot size on
Albedo=0.3 the scene.
The setup shown in Fig. 7 allows single element detectors
to be used in a LiDAR system with subcentimeter range res-
Fig. 7 LiDAR system. Not to scale. olution. In parallel, we have developed 320 × 256 pixels

Optical Engineering 091608-7 September 2017 • Vol. 56(9)

Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 05/22/2017 Terms of Use: http://spiedigitallibrary.org/ss/termsofuse.aspx

 Fathipour and Mohseni: Detector with internal gain for short-wave infrared. . .

with 30-μm pixel pitch EI detector arrays, which are being References
integrated with the readout integrated circuits. Once realized, 1. C. Cohen-Tannoudji, J. Dupont-Roc, and G. Grynberg, Photons and
the EI imager can replace the single detector in the setup Atoms: Introduction to Quantum Electrodynamics, Wiley-Interscience,
shown in Fig. 7 and improve performance specifications. New York (1997).
2. M. D. Eisaman et al., “Invited review article: single-photon sources and
Figure 7 gives a performance specification comparison based detectors,” Rev. Sci. Instrum. 82, 071101 (2011).
on our model between EI detector arrays with the detectors 3. G. Lewis, “The conservation of photons,” Nature 118, 874–875 (1926).
introduced in Sec. 5.64 4. “The cloud chamber,” 1912, Wilson, http://www-outreach.phy.cam.ac.
uk/camphy/cloudchamber/cloudchamber13_1.htm (27 February 2017).
5. N. Wermes, “Pixel detectors for particle physics and imaging applica-
5 Summary tions,” Nucl. Instrum. Methods Phys. Res. A 512, 277–288 (2003).
SWIR photon detection has become an essential technology 6. “The physics of particle detectors,” http://www.desy.de/~garutti/
LECTURES/ParticleDetectorSS12/L1_Introduction_HEPdetectors.pdf
in the modern world. Sensitive SWIR detector arrays with (27 February 2017).
high pixel density, low noise levels, and high SNRs are 7. G. B. Heim, J. Burkepile, and W. W. Frame, “Low-light-level EMCCD
highly desirable for a variety of applications including bio- color camera,” Proc. SPIE 6209, 62090F (2006).
8. E. Charbon, “Single-photon imaging in complementary metal oxide
photonics, LiDAR, optical tomography, and astronomical semiconductor processes,” Philos. Trans. Ser. A Math. Phys. Eng. Sci.
imaging. As such, many efforts in infrared detector research 2014, 372 (2012).
9. C. G. Coates et al., “Backilluminated electron multiplying technology:
are directed toward improving the performance of the photon the world’s most sensitive CCD for ultralow-light microscopy,” Proc.
detectors operating in this wavelength range. SPIE 4962, 319–328 (2003).
We review the history, principle of operation, present sta- 10. A. J. P. Theuwissen, Solid-State Imaging with Charge-Coupled Devices,
Kluwer Academic, Dordrecht (1996).
tus, and application of a sensitive SWIR detector technology, 11. V. Fathipour, S. J. Jang, and H. Mohseni, “New generation of isolated
which has demonstrated to be one of the most promising electron-injection imagers,” in 13th Workshop on Information Optics
paths to high pixel density FPAs for low flux applications. (WIO’14), Neuchatel, pp. 1–3 (2014).
12. H. Spieler, Semiconductor Detector Systems, Vol. 12, Oxford University
The so-called EI detector was demonstrated for the first time Press, England, U.K. (2005).
in 2007. It offers an overall system-level sensitivity enhance- 13. A. Pfenning et al., “Nanothermometer based on resonant tunneling
diodes: from cryogenic to room temperatures,” ACS Nano 9, 6271–
ment compared with the p-i-n diode due to a stable internal 6277 (2015).
avalanche-free gain. The amplification method is inherently 14. N. Wermes and G. Hallewel, “Technical design report of the ATLAS
low noise, and devices exhibit an excess noise of unity. The pixel detector,” CERN/LHCC/98-13 (1998).
15. W. Erdmann, “The CMS pixel detector,” Nucl. Instrum. Methods Phys.
detector operates in linear mode and requires only bias volt- Res. A 447, 178–183 (2000).
age of a few volts. The stable detector characteristics make 16. CMS, “The tracker system project,” ALICE Collaboration, Technical
formation of high yield, large-format and high pixel density Design Report, CERN/LHCC/98-006 (1998).
17. “ALICE technical design report,” CERN/LHCC 99–12 (1999).
FPAs less challenging compared with other detector technol- 18. T. Gys, “The pixel hybrid photon detectors for the LHCb-RICH
ogies, such as avalanche photodetectors. The detector is project,” Nucl. Instrum. Methods Phys. Res. A 465, 240–246 (2001).
19. P. Seitz and A. J. P. Theuwissen, Single-Photon Imaging, Springer
based on the mature InP material system (InP/InAlAs/ Series in Optical Sciences, USA, Vol. 160 (2011).
GaAsSb/InGaAs) and has a cutoff wavelength of 1700 nm. 20. S. S. Eikenberry, G. G. Fazio, and S. M. Ransom, “An SSPM-based
EI detectors have been designed, fabricated, and tested high-speed near-infrared photometer for astronomy,” Publ. Astron.
Soc. Pac. 108, 939 (1996).
during two generations of development and optimization 21. A. Rogalski, “History of infrared detectors,” Opto–Electron. Rev. 20(3),
cycles. The performance improvements in second-generation 279–308 (2012).
of devices have opened up many applications for these detec- 22. A. Pfenning et al., “Photocurrent-voltage relation of resonant tunneling
diode photodetectors,” Appl. Phys. Lett. 107, 081104 (2015).
tors in the medical field and remote sensing. We demon- 23. J. Hynecek, “Impactron—a new solid state image intensifier,” IEEE
strated the sensitivity advantage of using EI detectors in a Trans. Electron Devices 48, 2238–2241 (2001).
24. P. Jerram et al., “The LLLCCD: low light imaging without the need for
time-of-flight LiDAR system. This was carried out through an intensifier,” Proc. SPIE 4306, 178–186 (2001).
detailed theoretical models using experimentally measured 25. M. S. Robbins and B. J. Hadwen, “The noise performance of electron
device parameters. Using this model, we compared the per- multiplying charge-coupled devices,” IEEE Trans. Electron Devices
50, 1227–1232 (2003).
formance of the EI detector, the commercial p-i-n detector, 26. J. Hynecek and T. Nishiwaki, “Excess noise and other important char-
and the APD detector. The high internal gain, unity excess- acteristics of low light level imaging using charge multiplying CCDs,”
IEEE Trans. Electron Devices 50, 239–245 (2003).
noise, and excellent jitter of the EI detectors at room temper- 27. “Concepts in digital imaging technology electron multiplying charge-
ature allow the EI array to achieve good depth resolution and coupled devices (EMCCDs),” http://hamamatsu.magnet.fsu.edu/articles/
image quality. These also allow decreasing laser power and emccds.html (27 February 2017).
28. S. Ohta et al., “Characterization results of 1 k × 1 k charge-multiplying
increasing receiver sensitivity while shrinking package size CCD image sensor,” Proc. SPIE 5301, 99–108 (2004).
and power consumption. 29. M. S. Robbins, “The noise performance of electron multiplying charge-
coupled devices,” IEEE Trans. Electron Devices 50(5), 1227–1232
(2003).
Acknowledgments 30. A. Goetzberger et al., “Avalanche effects in silicon p-n junctions. II.
V.F. and H.M. would like to acknowledge partial support Structurally perfect junctions,” J. Appl. Phys. 34(6), 1591–1600 (1963).
31. K. Johnson, “High-speed photodiode signal enhancement at avalanche
from the National Science Foundation (NSF) Award No. breakdown voltage,” IEEE Trans. Electron Devices 12(2), 55–63 (1965).
IIP-1500314, Defense Advanced Research Projects Agency 32. M. A. Itzler et al., “Comparison of 32 × 128 and 32 × 32 Geiger-mode
(DARPA)/Army Research Office (ARO) Award Nos. APD FPAs for single photon 3D LADAR imaging,” Proc. SPIE 8033,
80330G (2011).
W911NF-11-1-0390, W911NF-12-1-0324, and W911NF- 33. W. R. Clark et al., “A 1 cm × 1 cm In 0.53 Ga 0.47 As-In 0.52 Al 0.48
13-1-0485, as well as computational resources and staff con- As avalanche photodiode array,” IEEE Photonics Technol. Lett. 18,
19–21 (2006).
tributions provided by the Quest high-performance computing 34. E. de Borniol et al., “HgCdTe-based APD focal plane array for 2D
facility at Northwestern University, which is jointly supported and 3D active imaging: first results on a 320 × 256 with 30 Aμm ^
by the Office of the Provost, the Office for Research, and pitch demonstrator,” Proc. SPIE 7660, 76603D (2010).
35. J. Rothman et al., “High operating temperature SWIR HgCdTe APDs
Northwestern University Information Technology. V. F. would for remote sensing,” Proc. SPIE 9254, 92540P (2014).
also like to acknowledge the Ryan Fellowship support from 36. G. Finger et al., “Evolution of IR detection and possible outcomes for
Northwestern University. astrophysics,” EAS Publ. Ser. 37, 343–353 (2009).

Optical Engineering 091608-8 September 2017 • Vol. 56(9)

Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 05/22/2017 Terms of Use: http://spiedigitallibrary.org/ss/termsofuse.aspx

 Fathipour and Mohseni: Detector with internal gain for short-wave infrared. . .

37. Y. Bai et al., “Teledyne imaging sensors: silicon CMOS imaging 63. V. Fathipour and H. Mohseni, “A new detector for high-speed swept
technologies for x-ray, UV, visible, and near infrared,” Proc. SPIE source optical coherence tomography,” Proc. SPIE 9609, 96090A
7021, 702102 (2008). (2015).
38. V. Fathipour, A. Bonakdar, and H. Mohseni, “Advances on sensitive 64. V. Fathipour et al., “Modeling of a sensitive time of flight flash Lidar,”
electron-injection based cameras for low-flux, short-wave infrared Proc. SPIE 9974, 99740N (2016).
applications,” Front. Mater. 3, 33 (2016). 65. T. Ahmed et al., “Optical quilt packaging: a new chip-to-chip optical
39. P. Feautrier, J. Gach, and P. Wizinowich, “State of the art IR cameras for coupling and alignment process for modular sensors,” in CLEO: 2014,
wavefront sensing using e-APD MCT arrays,” in First Light Imaging OSA Technical Digest (online), Optical Society of America, Paper
SAS, First Light Imaging SAS, France (2015). No. JTu4A.56 (2014).
40. R. J. McIntyre, “Multiplication noise in uniform avalanche diodes,” 66. T. Ahmed et al., “Label free detection of phospholipids by infrared
IEEE Trans. Electron Devices ED-13, 164–168 (1966). absorption spectroscopy,” Proc. SPIE 9166, 91660H (2014).
41. K. Lau et al., “Excess noise measurement in avalanche photodiodes 67. R. Moss et al., “Low-cost compact MEMS scanning ladar system for
using a transimpedance amplifier front-end,” Meas. Sci. Technol. 17, robotic applications,” Proc. SPIE 8379, 837903 (2012).
1941–1946 (2006). 68. S. D. Mayor and S. M. Spuler, “Raman-shifted eye-safe aerosol lidar,”
42. M. A. Saleh et al., “Impact-ionization and noise characteristics of thin Appl. Opt. 43, 3915–3924 (2004).
III-V avalanche photodiodes,” IEEE Trans. Electron Devices 48(12), 69. P. Yuan et al., “32 × 32 Geiger-mode ladar camera,” Proc. SPIE 7684,
2722–2731 (2000). 76840C (2010).
43. M. M. Hayat et al., “Boundary effects on multiplication noise in 70. E. De Borniol et al., “A 320 × 256 HgCdTe avalanche photodiode focal
thin heterostructure avalanche photodiodes: theory and experiment plane array for passive and active 2D and 3D imaging,” Proc. SPIE
[Al0.6Ga0.4As/GaAs],” IEEE Trans. Electron Devices 49(12), 2114– 8012, 801232 (2011).
2123 (2002). 71. A. Poloczek et al., “Integrated InGaAs pin-diode on exactly oriented
44. O.-H. Kwon et al., “Optimal excess noise reduction in thin heterojunc- silicon (001) substrate suitable for 10 Gbit∕s digital applications,”
tion Al0.6Ga0.4As-GaAs avalanche photodiodes,” IEEE J. Quantum in The 20th Annual Meeting of the IEEE Lasers and Electro-Optics
Electron. 39(10), 1287–1296 (2003). Society (LEOS), pp. 180–181 (2007).
45. G. Leveque et al., “Ionization energies in CdxHg1-xTe avalanche photo- 72. B. L. Stann et al., “Brassboard development of a MEMS-scanned ladar
diodes,” Semicond. Sci. Technol. 8, 1317–1323 (1993). sensor for small ground robots,” Proc. SPIE 8037, 80371G (2011).
46. V. F. Guelleca et al., “Linear photon-counting with HgCdTe APDs,” 73. M. Stipčević, H. Skenderović, and D. Gracin, “Characterization of
Proc. SPIE 8375, 83750Y (2012). a novel avalanche photodiode for single photon detection in VIS-NIR
47. http://www.first-light.fr/wp-content/uploads/2015/07/Datasheet-C-RED_ range,” Opt. Express 18, 17448–17459 (2010).
10.12.15.pdf 74. Y. Liang et al., “Low-timing-jitter single-photon detection using 1-GHz
48. A. Rogalski, “Recent progress in infrared detector technologies,” sinusoidally gated InGaAs/InP avalanche photodiode,” IEEE Photonics
Infrared Phys. Technol. 54, 136–154 (2011). Technol. Lett. 23(13), 887–889 (2011).
49. A. Rogalski, “Progress in focal plane array technologies,” Prog. 75. M. A. Itzler et al., “Single photon avalanche diodes (SPADs) for 1.5 μm
Quantum Electron. 36, 342–473 (2012). photon counting applications,” J. Mod. Opt. 54(2), 283–304 (2007).
50. J. Gach and P. Feautrier, “Advances in imaging: Electron-initiated APDs 76. R. Heinrichs et al., “Three-dimensional laser radar with APD arrays,”
improve high-speed SWIR imaging,” Laser Focus World 51(9), (2015). Proc. SPIE 4377, 106–117 (2001).
51. N. K. Dhar, A. K. Sood, and R. Dat, Advances in Infrared Detector 77. N. K. Dhar, R. Dat, and A. K. Sood, “Advances in infrared detector
Array Technology, INTECH Open Access Publisher, Croatia (2013). array technology,” in Optoelectronics–Advanced Materials and Devices,
52. O. G. Memis et al., “A photon detector with very high gain at low bias S. Pyshkin, Ed., InTech Open Access Publisher (2013).
and at room temperature,” Appl. Phys. Lett. 91, 171112 (2007). 78. V. Fathipour et al., “Approaching high temperature photon counting
53. V. Fathipour, S. J. Jang, and H. Mohseni, “New generation of isolated with electron-injection detectors,” Proc. SPIE 9220, 92200J (2014).
electron-injection imagers,” in IEEE 13th Workshop on Information 79. “Semiconductoor electronic properties,” http://www.ioffe.ru/SVA/NSM/
Optics (WIO), Neuchatel, Switzerland (2014). Semicond/GaInAs/highfield.html (27 February 2017).
54. V. Fathipour et al., “Isolated electron injection detectors with high gain
and record low dark current at telecom wavelength,” IEEE J. Sel. Top.
Quantum Electron. 20, 65–70 (2014). Vala Fathipour is a postdoctoral scholar at the University of
55. V. Fathipour and H. Mohseni, CLEO—Laser Science to Photonic California-Berkeley working on the development of state-of-the-art
Applications, OSA, San Jose (2015). LiDAR systems. She received her PhD in electrical engineering, solid-
56. V. Fathipour et al., “Highly sensitive and linear electron-injection state and photonics from Northwestern University. Her research inter-
detectors at the telecomm wavelength,” in Frontiers in Optics 2014, ests include design, fabrication, and characterization of single-photon
OSA Technical Digest (online), Optical Society of America, Paper infrared detectors, optical coherence tomography, and light detection
No. FW2A.4 (2014). and ranging system development. She has published 30 conference
57. O. G. Memis et al., “Sub-Poissonian shot noise of a high internal gain
injection photon detector,” Opt. Express 16(17), 12701 (2008). papers and 11 peer-reviewed articles.
58. Y. Movassaghi et al., “Analytical modeling and numerical simulation of
the short-wave infrared electron-injection detectors,” Appl. Phys. Lett. Hooman Mohseni is a professor of electrical engineering and
108, 121102 (2016). computer sciences at Northwestern University. He is the recipient of
59. V. Fathipour et al., “On the sensitivity of electron-injection detectors at several research and teaching awards, including the NSF CAREER
low light level,” IEEE Photonics J. 8(3), 6803207 (2016). Award, DARPA Young Faculty Award, and Northwestern Faculty
60. O. G. Memis et al., “Signal-to-noise performance of a short-wave infra- Honor Roll. He serves on the editorial boards of IEEE Photonics,
red nanoinjection imager,” Opt. Lett. 35(16), 2699 (2010). IEEE Selected Topics in Quantum Electronics, Optics Letter, and
61. O. G. Memis et al., “A short-wave infrared nano-injection imager with Frontiers in Material. He has published over 120 peer-reviewed
2,500 A/W responsivity and low excess noise,” IEEE Photonics J.
2, 858–864 (2010). articles in major journals, including Nature, Nano Letters, and ACS
62. V. Fathipour et al., “Impact of three-dimensional geometry on the per- Nano. He holds 14 issued US and international patents. He is a fellow
formance of isolated electron-injection infrared detectors,” Appl. Phys. of SPIE and OSA.
Lett. 106(2), 021116 (2015).

Optical Engineering 091608-9 September 2017 • Vol. 56(9)

Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 05/22/2017 Terms of Use: http://spiedigitallibrary.org/ss/termsofuse.aspx