Geiger-Mode Avalanche Photodiodes for Near-Infrared Photon Counting Mark A. Itzler, Rafael Ben-Michael, Xudong Jiang, Krystyna Slomkowski Princeton Lightwave Inc., 2555 US Route 130, Cranbury, New Jersey, 08512 E-mail: mitzler@princetonlightwave.com Abstract: We present the design and characterization of InP-based avalanche photodiodes optimized for single photon counting for wavelengths between 1.0 and 1.7 ȝm, and we discuss performance trade-offs and mechanisms responsible for present performance limits. ©2007 Optical Society of America OCIS codes: (040.5160) Photodetectors; (030.5260) Photon counting To serve emerging applications at near-infrared (NIR) wavelengths between 1 and 1.7 ȝm, there has been a recent surge of interest in NIR photon counting detectors[1]. Dominant applications are found in the fields of communications and imaging, and single photon sensitivity is proving to be critical for sub-fields such as quantum cryptography and 3-D imaging. Photon counting at 1.5 ȝm is optimal for optical fiber-based technologies and also supports more eye-safe wavelengths (>1.4 ȝm) for imaging and ranging. Photon counting at 1.06 ȝm complements mature laser sources in applications such as lidar, ranging, and imaging when greater sensitivity is required. Below its breakdown voltage V b , an avalanche photodiode (APD) operates in linear mode, for which output photocurrent is proportional to input optical power. If an APD is biased above V b , then a single photoexcited carrier can induce a run-away avalanche, giving rise to a detectable macroscopic current. Operating in this so-called Geiger mode, the detector is sensitive to a single photon and is referred to as a single photon avalanche diode (SPAD). In this paper, we describe our progress with SPAD design and performance and discuss mechanisms responsible for performance limitations through an analysis of activation energies found for dark count and afterpulse rates. In Fig. 1, we present a schematic cross-section of our SPAD design platform[2]. The absorption layer consists of either InGaAs or InGaAsP lattice-matched to InP. InGaAsP absorbers provide much lower dark count rates in return for shorter cutoff wavelengths (suitable in, e.g., 1.06 ȝm applications). Avalanche gain is achieved in an undoped InP multiplication region of ~1 ȝm thickness. To maintain high field in the multiplication region and low field in the absorption region, a moderately doped InP charge layer is placed between these regions. Since a heterointerface of InGaAs and InP tends to trap holes, a “grading” layer is employed to smooth this interface. Fig. 1. Schematic cross-section of InP-based single photon avalanche diode (SPAD) structure. Fig. 2. Dark count rate vs. detection efficiency for φ25 ȝm InGaAs/InP SPADs measured using 1 ns gating at 500 kHz, T=215 K, and Ȝ=1.55 ȝm. The most fundamental performance characteristics of a SPAD are the probability of detecting an input photon, or the detection efficiency (DE), and the probability of measuring a false count when no photon is input, or the dark count rate (DCR). Both quantities increase with overbias voltage V ov ≡ V bias - V b . For InGaAs/InP SPADs, V ov is generally applied during a short time (~1 ns to ~100 ns) in gated mode operation. As seen in Fig. 2, for 25 ȝm diameter SPADs measured with 1 ns gates at 215 K, typical devices exhibit DCR ~ 1x10 -5 ns -1 at 20% DE, while occasional outliers (data indicated by “○”) show reduction by as much as ~5X in DCR. Operation with 1 ns gates is desirable when photon arrival times are accurately known (e.g., in quantum communications). Higher DE can be achieved at the expense of higher DCR, and the optimal trade-off between DE and DCR is application-dependent. i-InGaAs or i-InGaAsP absorption n + -InP buffer n-InGaAsP grading n-InP charge i-InP cap p + -InP diffused region multiplication region SiN x passivation p-contact metallization n + -InP substrate anti-reflection coating n-contact metallization optical input 1E-7 1E-6 1E-5 1E-4 0% 5% 10% 15% 20% 25% Detection Efficiency Dark Count Rate (per ns) 215 K CMII1.pdf CLEO/QELS 2007 Baltimore, MD