Development of Single Photon Counting Sensors Operating at Short Wavelength Infrared Wavelengths Joseph Boisvert*, Rengarajan Sudharsanan, Ping Yuan, Takahiro Isshiki, and Paul McDonald Spectrolab Inc., a Boeing Company, 12500 Gladstone Ave., Sylmar, CA, USA 91342 ABSTRACT This paper will review the development of single photon counting sensors at Boeing Spectrolab. Future development over the next five years will be discussed in the context of sensor requirements that have been established and will be established. Greater sensitivity through lower false event rates, higher bandwidths, lower after-pulsing rates, higher operating temperatures, and better uniformity are figures-of-merit that will be discussed in this presentation. We will present performance of large format InP/InGaAs Geiger mode avalanche photodiode arrays operating at 1.06 µm and 1.55 µm. 1. INTRODUCTION AND BACKGROUND Recent developments in three-dimension imaging have stimulated interest in single-photon counting avalanche photodiodes (APD) in the short wavelength infrared region. Geiger mode avalanche photodiode focal plane arrays (GM-FPAs) for short wavelength infrared applications have been reported by both MIT Lincoln Laboratories 1 and Boeing Spectrolab. 2 Most of the effort to date has been focused on photodiodes operating at 1.06 µm where high power emitters are available. For eye-safe Laser Detection And Ranging (LADAR) applications, there is also considerable interest in Geiger mode avalanche photodiodes (GM-APDs) operating at 1.55 µm. 2. SENSOR FLOWDOWN REQUIREMENTS Important figures-of-merit for GM-APDs are dark count rate (DCR) and photon detection efficiency (PDE) which together establish an upper limit on the signal-to-noise for the entire sensor system. For active 3-D imaging sensors FPA timing jitter is also a critical parameter as this sets an upper limit on range resolution. Other figures-of-merit of importance to the sensor system designer include afterpulsing and maximum frame rate. Afterpulsing, or temporal crosstalk between range gates, ultimately limits sensor efficiency. Frame rate limits sensor update time which in some applications is a critical performance parameter. Finally, spatial resolution is limited by pixel pitch and spatial crosstalk between pixels on the FPA. Together these parameters must be optimized to realize optimal sensor performance. In practice these parameters are not decoupled and often trade off against one another. A robust FPA read out integrated circuit (ROIC) can improve sensor performance by mitigating APD deficiencies. However, as is almost always the case, the FPA design is fundamentally limited by detector performance. Table 1 details the most important FPA figures of merit that impact single photon counting sensor systems.
Parameter Spec Test conditions Wavelength Format 1.06 µm 32x32 Pixel Pitch µm PDE 40% 240K, 4V overbias Fill Factor 70% DCR 20 khz 240K, 4V overbias Cross talk 2% Pixel operability 95% Optimized overbias for SCA RMS timing jitter 0.5 ns Table 1. Single photon counting focal plane array figures-of-merit for a 32x32 GM-APD array presently fabricated at Spectrolab 3. GM-APD FPA FUNDAMENTALS In Geiger mode operation the avalanche photodiodes are biased above the breakdown voltage, ~2-4 V, for a short period of time, typically a few nanoseconds gate time. Without any photon falling on the detector array, the noise in the detector array during the nanoseconds gate time operation is the dark count rate (DCR) and DCR increases with increasing overbias. Not all photons that fall onto the APD during this period of overbias are absorbed and converted into digital pulses that can be measured by the readout circuit. The fraction that is converted into signal is a figureof-merit called the photon detection efficiency. Figure 1 details the general concepts of Geiger mode operation. 1.E-03 Wafer 9002-10 Photon Input Geiger Mode Operation 1.E-04 Current (A) 1.E-05 1.E-06 1.E-07 Linear Mode Geiger Mode 10 Responsivity (A/W) Sensor Output t 1.E-08 1.E-09 40 50 60 70 Bias (V) 1 Single Photon Detection No Absorption or Insufficient Gain t Dark Count Figure 1. Geiger mode operation of an avalanche photodiode.
n-ingaas n-inp 500 Electric Field Profile n-ingaas n-inp i-ingaas InGaAsP Graded layer N-InP Charge layer E-Field (kv/cm) 400 300 200 i- InGaAsP i-ingaas Absorption layer InGaAsP Graded layer N-InP Charge layer i-inp Multiplier P-InP Buffer 2 P-InP Buffer 1 0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 Position (µm) i-inp Multiplier P-InP Buffer 2 P-InP Buffer 1 1.5 µm Gm-APD device design 1.06 µm Gm-APD device design Figure 2. Gm-APD device designs and the electric field profile of the designs Device Structure Figure 2 shows the 1.06 µm and 1.5 µm Gm-APD separate absorption and gain and multiplication (SAGM) device structures. For both structures InP layer is the multiplication layer and silicon doped InP layer is the charge layer. InGaAsP layer and InGaAs layers are used as the absorption layers for 1.06 µm and 1.5 µm device structures, respectively. Figure 2 also shows the electric field profile in these device structures. The device structure was designed in such a way that the electric field is maximum at the InP multiplication layer for high avalanche gain and it is minimum across the absorption layers so that the dark current will be low in these devices. In the case of 1.5 µm Gm-APD design, a thin InGaAsP layer was introduced between n-inp charge and InGaAs absorption layers to reduce the trapping of carriers due to the valence band discontinuity between InP and InGaAs. An N + InGaAs layer was used as a contact layer for the both designs. The nominal thickness of both InGaAs and InGaAsP absorption layers was 1.5 µm in order to have good quantum efficiency. The separate absorption and multiplication avalanche photodiode device growth and characterization details have been described elsewhere. 2 4. STATE OF THE ART AT SPECTROLAB Single Element Device Fabrication And Testing Results Single element mesa devices with different mesa diameters are typically fabricated using photolithography and the fabrication details were described earlier. 2 Devices with various diameters, from 20 µm to 200 µm, are fabricated to evaluate the quality of epitaxial wafers and also to optimize the device designs for improved device performance. Initial I-V testing of devices is done at room temperature to determine breakdown voltage, punch through voltage, and leakage current. Device optical gain and responsivity were measured by illuminating with 1.5 µm and 1.06 µm wavelength lasers. Figure 3 shows the I-V data of 1.5 µm and 1.06 µm devices with 25 µm and 50 µm diameter devices respectively. Both devices clearly exhibit low dark current and high gain near the breakdown voltages. The dark currents near breakdown voltage for the 1.5 µm and 1.06 µm devices are below 10 na and 50 pa, respectively. As expected for the small bandgap (1.5 µm) device the dark current is higher than the wider bandgap (1.06 µm) device.
Wafer 9002-10 Wafer 8702-11 Current (A) 1.E-03 1.E-04 1.E-05 1.E-06 1.E-07 Iph @ 6µW 1550nm Id w/ 25 µm Responsivity 10 Responsivity (A/W) Current (A) 1E-08 1E-09 1E-10 µm diameter 1.06 µm APD 0 10 Gain 1.E-08 1.E-09 40 45 50 55 60 Bias (V) 1 1E-11 1 0 20 40 60 80 Bias (V) a) 1.5 µm Gm-APD device b) 1.06 µm Gm-APD device Figure 3. Room temperature I-V data of; a) 1.5 µm and b) 1.06 µm devices Figure 4 shows the DCR measured at room and at low temperatures as a function of overbias, multiplication layer thicknesses, and pixel size for both 1.5 µm and 1.06 µm single element devices. As expected, DCR increases with increasing overbias for both 1.5 µm and 1.06 µm devices and decreases with decreasing pixel size, increasing multiplication layer thickness, and at reduced temperatures. The devices labeled as 8702 and 8704 had thin multiplication layer thickness (<1 µm) and they show DCR greater than 1 MHz at room temperature. Upon increasing the multiplication layer thickness to 1.5 µm DCR values decreased to less than 500 KHz for both 1.5 µm and 1.06 µm devices due to reduced electric field in the multiplication layer. A reduction in dark count rate with increasing multiplication layer thickness has been reported by Lincoln Laboratory. 3.00 DCR Data 8702-11C2 k 8702-11C3 8702-11C1 8702-11C5.00 8702-11C2 k 8702-11C3 8702-11C1 DCR Data- Room Temp 10.00 8704-13C1 8704-13C3 8704-13C5 8704-13C2 8704-13F9D4 8704-13L12D1 10.00 8702-11C5 8704-13C1 8704-13C3 8704-13C5 DCR (MHz) 1.00 0.10 8704-13L12D2 8704-13J11D2 8704-13J11D6 8704-13M12D2 9002-10A31B29 9002-10A32B27 9002-10A31B28 9002-10A31B25 9002-10A31F20 DCR (MHz) 1.00 8704-13C2 8704-13F9D4 8704-13L12D1 8704-13L12D2 8704-13J11D2 8704-13J11D6 0.01 9002-10A31F20 9002-10A23B23 9002-10 A2531 D45 9002-10 F25-1-273K 9002-10 F25-2-273K 9404-10 D30-1 0.10 8704-13M12D2 9569-9 D30-4 9590-11 D30-4 9590-11 CR30-2 9404-10 D30-3 9590-11CL30-2 0.00 0 1 2 3 4 5 6 7 Overbias (V) 9590-11 CL20-1 0.01 0 1 2 3 4 5 6 Overbias (V) Figure 4. The DCR at room temperature and 0 o C versus overbias characteristics taken on two types of GM-APDs. The 8702 and 8704 APDs were InGaAsP/InP APDs optimized for 1.06 µm, and their curves are shown with dotted lines and smaller markers. The 9002 and 9404 series APDs are sensitive at 1.55 µm and yet display a very low DCR even at room temperature. They are illustrated with solid lines. The DCRs of the 9002 APDs at 0 o C are shown with dash lines.
Figure 5. Photograph of a 32 x 32 Gm-APD detector array 32x32 GM-APD Array Fabrication And Testing Results A 32x32 element GM-APD array is shown in Figure 5. These detector arrays have been integrated with MIT-LL CMOS2002 ROICs and packaged into hermetically-sealed modules that include thermoelectric cooling. In order to improve the GM-APD performance temperatures, we have optimized the epitaxial growth, process, and device design for the Geiger-mode operation, and achieved an excellent DCR and greater than 40% PDE at operational temperatures. DCR data collected on recent APD arrays is shown in Figure 6. Optical crosstalk has been an issue with these GM-APD FPAs to date. Spectrolab has developed device designs to minimize crosstalk across these arrays. Crosstalk across the GM-APD arrays is shown in Figure 7 and has been demonstrated to be less than 2%. 1.E+07 2-30 1.E+07 2-30 3-30 7-30 4-30 5-30 8-30 6-30 9690-9 M10 9-30 1.E+06 7-30 9764-10 J12 8-30 1.E+06 on N-InP 11-30 9-30 on P-InP 14-30 10-30 11-30 16-30 DCR (Hz) 12-30 DCR (Hz) 17-30 1.E+05 13-30 1.E+05 14-30 18-30 16-30 19-30 1.E+04 18-30 21-30 20-30 22-30 1.E+04 25-30 24-30 26-30 28-30 29-30 27-30 30-30 28-30 32-30 1.E+03 36-30 1.E+03 30-30 37-30 33-30 1.E+02 38-30 34-30 39-30 40-30 35-30 1.E+02 36-30 200 220 240 260 280 300 320 200 220 240 260 280 300 37-30 320 Temp (K) Temp (K) 39-30 Figure 6. Dark count rate data collected on two recent Spectrolab 32x32 focal plane arrays. All data are taken with a 4 volt overbias on 30 µm diameter pixels.
9764-10 J9 Average Dark Count Rate (khz) 11.1 Crosstalk (%) 1.4 Figure 7. The measured cross talk data histogram and corresponding DCR on a recent Spectrolab 1.06 µm GM-APD focal plane array module. These data were collected at Boeing SVS. Timing jitter is a particularly important figure-of-merit for FPAs because this temporal noise establishes an upper limit on range resolution. Recent data collected at Boeing SVS on Spectrolab GM-APD arrays is shown in Figure 8. In these measurements a 1.06 µm laser was pulsed using an internal trigger from an MIT Lincoln Laboratory CM2002 ROIC. The laser pulse width was 660 ps and the system jitter was 600 ps. An FPA jitter of 500 ps would result in a measured jitter of greater than 1.5 ns. The data shown in Figure 8 show a FWHM of less than 2 bins indicating that the FPA has less than 500 ps of jitter. Not every parameter can be measured across all 1024 elements of a 32x32 FPA. However it is possible to measure breakdown characteristics across an array to insure that an optimized biasing condition can be applied to all elements to achieve optimal array performance. Figure 9 shows breakdown voltage characteristics across a typical Spectrolab 32x32 APD detector array. These data demonstrate excellent materials uniformity and control that result in uniform device characteristics. 5. SUMMARY AND PATH FORWARD Spectrolab, together with MIT Lincoln Laboratory, has been developing GM-APD sensor technology for SWIR applications. Continuous improvement in materials and device design has resulted in excellent FPA performance in a single photon counting technology that finds application in LADAR sensors. Dark count rate, photon detection efficiency, jitter, crosstalk and array uniformity continue to improve. While 32x32 arrays find applications in many sensor systems, development of 128x32 and 256x64 arrays have been reported by MIT Lincoln Laboratory. 3 In the near term Spectrolab will be developing these larger format arrays for emerging single photon counting sensor systems. These FPA modules will cover both 1.06 µm and 1.55 µm wavelengths and utilize Spectrolab s internal materials growth and characterization infrastructure. In addition, Spectrolab is developing a compact LADAR camera using these FPAs for many airborne applications.
Figure 8. Jitter data collected on a single pixel of a Spectrolab 1.06 µm GM-APD module. Each bin corresponds to 500 ps. After accounting for system jitter, a FWHM of less than 3 bins corresponds to a jitter less than 500 ps. Vbr Distribution 78.0 77.8 77.5 77.3 77.8-78.0 77.5-77.8 77.3-77.5 Vbr (V) 77.0 76.8 76.5 76.3 76.0 S29 S25 S21 S17 Column S13 S9 S5 S1 1 9 17 25 Row 77.0-77.3 76.8-77.0 76.5-76.8 76.3-76.5 76.0-76.3 Figure 9. The breakdown voltage characteristics across a Spectrolab 1.06 µm GM-APD array measured at 20 C.
ACKNOWLEDGEMENTS We would like to acknowledge the invaluable assistance provide by MIT Lincoln Laboratory over the last several years under various technology transfer programs. Measurement assistance provided by Boeing SVS under internal Boeing funding is also gratefully acknowledged. *jboisvert@spectrolab.com; phone 1 818 838 2610; fax 1 818 838 7474; www.spectrolab.com REFERENCES 1. K. A. McIntosh, et al., Arrays of III-V semiconductor Geiger-mode avalanche photodiodes Lasers and Electro- Optics Society, 2003. LEOS 2003. The 16th Annual Meeting of the IEEE. Volume 2, 686 (2003). 2. R. Sudharsanan et al., Single photon counting Geiger Mode InGaAs(P)/InP avalanche photodiode arrays for 3D imaging Proc. SPIE 6950, 69500N-1 (2008). 3. S. Verghese, et al., Arrays of InP-based Avalanche Photodiodes for Photon Counting IEEE J. Selected Topics in Quantum Electronics, 13, 870 (2007).