Gallium Phosphide (GaP) reach-through avalanche photodiodes (APDs) are reported. The APDs exhibited dark current less than a pico-ampere at unity gain. A quantum efficiency of 70% was achieved with a recessed window structure; this is almost two times higher than previous work.
© 2011 OSA
Wide bandgap photodetectors have been investigated as possible replacements for photomultiplier tubes (PMTs) [1–3] in applications such as low-level ultraviolet detection in laser-induced fluorescence biological-agent warning systems. GaP APDs exhibit high responsivity in the wavelength range from 400 to 500 nm . They can potentially be applied to applications such as: detection under water at 400 nm, the wavelength at which water is transparent and detection of 440 nm wavelength emission from scintillation crystals (440 nm) that are used to detect gamma rays from missile and nuclear material . Currently PMTs are the most sensitive detectors available due to their high gain and low noise, but semiconductor photodetectors can achieve very high gains through the avalanche process with relatively excess low noise. Semiconductor photodetectors are also less expensive and more robust than PMTs.
Previously GaP avalanche photodiodes (APDs) were investigated for detection of fluorescence from NADH and flavin compounds (400-600 nm) . From that research, dark currents less than 1pA and a recessed-window structure with a peak quantum efficiency of 38% at 440 nm were reported. This paper presents a reach-through structure APD [7, 8]. It consists of undoped absorption and multiplication layers separated by a thin, highly-doped charge layer. The function of the charge layer is to tailor the relative field strength between the thin multiplication region and the thicker absorption layer. In this paper, the term “reach through” will refer to the state when the absorption, charge, and multiplication regions are fully depleted. A recessed-window approach is used to improve the quantum efficiency . The recessed-window reach-through GaP APDs exhibit dark currents less than 1 pA up to 20 V reverse-bias voltage, gain up to 10, and high peak quantum efficiency of 70% at 445 nm. Hamamatsu’s UV enhanced silicon photodiode, S1336-18BQ, has a maximum dark current density of 16.7 pA/mm2 at 10 mV reversed-bias voltage, a spectral range of 190 – 1100 nm with responsivites of 0.2 A/W and 0.5 A/W at 445 nm and 960 nm respectively. In comparison the GaP APDs presented in this paper have dark current densities of 6.4 pA/mm2 at 200 mV reversed-bias voltage, spectral range of 350 – 550 nm with equivalent responsivity at 445 nm for recessed-window devices.
2. Device structure and fabrication
The avalanche photodiodes were fabricated from MOCVD-grown GaP wafers. The following layers were grown sequentially on highly-doped n-type GaP substrate: 600 nm n-layer with doping concentration 1 x 1019 cm3, 200 nm unintentionally-doped “i”-layer (multiplication layer), 200 nm p-layer with doping concentration 1.3 x 1017 cm3 (charge layer), 800 nm unintentionally-doped “i”-layer (absorption layer) and a 300 nm p-layer (1 x 1019 cm3).
The mesas were defined by standard photolithography and etched by inductively coupled plasma to a mesa height of 2 μm. A 5 second etch in a solution of HNO3:HCl:H2O (1:1:1)  was employed to remove side-wall surface damage. The recessed-window devices required a second lithography step to define the window, which was dry etched to a depth of 250 nm followed by the 5 second chemical etch. A 220 nm SiO2 layer was deposited by plasma enhanced chemical-vapor deposition for sidewall passivation. This layer also served as an antireflective layer for peak quantum efficiency at 440 nm. The p and n contacts were defined by photolithography; a buffered oxide etch removed the SiO2 and AuGe-Ni-Au (40 nm, 10 nm, 110 nm) was deposited by electron-beam evaporation. Acetone with light ultra-sonic agitation was used for metal lift-off. The device structures of the recessed and non-recessed devices are shown in Fig. 1 .
3. Results and discussions
The current-voltage (I-V) characteristics for both devices were measured with an HP 4156B semiconductor parameter analyzer. For 100µm diameter devices dark currents < 1 pA were measured up to −20V (Fig. 2 ). At a gain of 10 the dark current was less than 20 pA for both recessed and non-recessed window devices. Gain up to 10 was achieved in both the recessed and non-recessed devices before breakdown. The reach-through voltage was determined to be 15 V from capacitance versus voltage measurements. Since 15 V is less than a third of the breakdown voltage and there is no significant increase of photocurrent relative to lower bias values, this point will be used as unity gain .
The external quantum efficiency was measured at unity gain. Light from a xenon lamp fed through a monochromator illuminated the device while responsivity measurements were made using a lock-in amplifier. The light-intensity was measured using a known calibrated silicon detector. The light was focused to a spot smaller than the size of the recessed window to ensure all the light was focused on the device. The recessed-window has been shown to increase the quantum efficiency of GaP  and SiC  APDs because it enables the incident light to be absorbed closer to the edge of the absorption layer . At 440nm (the peak responsivity) almost all the light is absorbed in the top p+ layer where recombination is strong. In this layer the electron diffusion length is short resulting in significant recombination before electrons can contribute to the photocurrent.
Figure 3 shows the spectral response of the non-recessed and recessed-window APDs; their external quantum efficiencies were 47% and 70% at 445 nm, respectively. The change in shape of the spectral response is consistent with the absorption coefficient in GaP. While data is not available for the diffusion length of electrons in our highly doped p-contact region, from our experimental data we can assume a diffusion length of ~15 nm. Using a simple model for absorption in each layer and assuming that all light absorbed in the depletion layer will be collected, we can calculate the expected quantum efficiencies for the two device structures. At 355 nm the absorption coefficient is 2.9 x 105 cm−1. There will be no contribution to the quantum efficiency in the non-recessed device since the light absorbed will recombine before reaching the depletion layer. For the recessed window device we can expect 18% quantum efficiency. At 445 nm the absorption coefficient is 3.3x104 cm−1 and we calculate quantum efficiency of 34% and 73% for non-recessed and recessed window devices, respectively. At longer wavelengths such as 540 nm the absorption coefficient is 3.6x102 cm−1, the light is absorbed far from the surface and both recessed and non-recessed window devices are expected to have a quantum efficiency of ~4%. Previously, we reported p-i-n structure GaP APDs with quantum efficiencies of 20% (non-recess) and 38% (recessed window). The significant improvement in quantum efficiency in the non-recessed structures can be attributed to the increase in the depletion region thickness from 300 nm in the previous structure compared to 1400 nm at reach-through in the presented structure. The introduction of a recessed window can cause a lateral decrease in the electric field strength in the active layer due to the high series resistance across the floor of the recessed window area , thereby resulting in a decrease in the speed response and also creating variation in the spatial uniformity. To ensure that there were no such problems with the current recessed window structure, the spatial uniformity of the recessed window device was measured at gains from 1 to 10 (see Fig. 4 ). For the spatial uniformity measurement, light from an argon laser at ~350 nm was scanned across the detector while readings were made using a lock-in amplifier. The light spot diameter was ~5 μm, which provides high spatial resolution. In Fig. 4 the center region represents the “active region” of the device. The valley directly outside the center region is the p-contact ring, which blocks the light signal. The outer ring is the edge of the mesa. The dip on the right outer ring is due to the probe tip. In the recessed window device high response to shorter wavelength light is expected in the window area. At a gain of 10 the standard deviation across the active area was < 30%.
The speed of the device was measured using a 265 nm Nd:YAG laser with a 400 ps pulse and 7.5 kHz repetition rate. The laser was focused on the active region of the device and the device probed with a low noise microwave probes and biased at unity gain. For 200 μm-diameter recessed and non-recessed window devices at unity gain, the 3dB bandwidth was 600 MHz; this is consistent with the calculated RC-limited bandwidth. The measured resistance was approximately 100 ohms and capacitance was 2.5 pF giving an RC-limited bandwidth of 640 MHz.
Another figure of merit for APDs is the excess noise factor, F(M). Lower values of excess noise are achieved when, k, the ratio of the ionization coefficients of holes (β) and electrons (α) is small. 200 μm-diameter recessed and non-recessed window devices were illuminated with an argon laser at 350 nm and 515 nm wavelengths, respectively. The measurements were made using an HP 8970B noise figure analyzer. Based on measured absorption coefficients for GaP , 350 nm light will be absorbed in the first 30 nm while for 515 nm light the absorption length is ~11 µm, which means that the light is absorbed much farther from the surface. It follows that 350 nm light results in pure electron injection into the multiplication layer, whereas 515 nm illumination creates mixed injection. Figure 5 shows a plot of the excess noise factor versus gain for both types of devices. The curves for k-values from 0.1 to 1.0 are included in the graph. A k of 0.9-1 has been observed at high gain for illumination at 350 nm. This is consistent with the reported k-value for GaP . However; for illumination at 515 nm (mixed injection) a k of 0.5-0.6 was observed. This lower k value associated with significant hole injection was observed by B. K. Ng et al. for SiC  and is proof that β > α for GaP.
Recessed-window, mesa-structure GaP reach-through APDs have been fabricated and characterized. Low dark currents (< 1 pA) and high quantum efficiency (70% at 445 nm), have been achieved.
This work has been supported by DARPA through the DUVAP program and U. S. Army Research Laboratory under cooperative agreement number W911NF-09-2-0019.
References and links
1. T. V. Blank, Y. A. Gol'dberg, E. V. Kalinina, O. V. Konstantinov, and E. A. Posse, “Ultraviolet radiation photodetectors based on structures consisting of a metal and a wide-bandgap semiconductor,” Semiconductors 37(8), 944–948 (2003). [CrossRef]
2. E. Monroy, F. Omnes, and F. Calle, “Wide-bandgap semiconductor ultraviolet photodetectors,” Semicond. Sci. Technol. 18(4), R33–R51 (2003). [CrossRef]
3. E. Muñoz, E. Monroy, J. L. Pau, F. Calle, F. Omnes, and P. Gibart, “III nitrides and UV detection,” J. Phys. Condens. Matter 13(32), 7115–7137 (2001). [CrossRef]
4. D. McIntosh, Q. G. Zhou, F. J. Lara, J. Landers, and J. C. Campbell, “Flip-Chip Bonded GaP Photodiodes for Detection of 400-to 480-nm Fluorescence,” IEEE Photon. Tech. Lett. 23(13), 878–880 (2011). [CrossRef]
5. R. C. Hughes, T. E. Zipperian, L. R. Dawson, R. M. Biefeld, R. J. Walko, and M. A. Dvorack, “Gallium phosphide junctions with low leakage for energy conversion and near ultraviolet detectors,” J. Appl. Phys. 69(9), 6500–6505 (1991). [CrossRef]
6. A. L. Beck, B. Yang, S. Wang, C. J. Collins, J. C. Campbell, A. Yulius, A. Chen, and J. M. Woodall, “Quasi-direct UV/blue GaP avalanche photodetectors,” IEEE J. Quantum. Electron. 40(12), 1695–1699 (2004). [CrossRef]
7. H. W. Ruegg, “An optimized avalanche photodiode,” IEEE Trans. Electron Dev. 14(5), 239–251 (1967). [CrossRef]
8. Q. Zhou, D. McIntosh, H.-D. Liu, and J. C. Campbell, “Proton-Implantation-Isolated Separate Absorption Charge and Multiplication 4H-SiC Avalanche Photodiodes,” IEEE Photon. Technol. Lett. 23(5), 299–301 (2011). [CrossRef]
9. T. Li, A. L. Beck, C. Collins, R. D. Dupuis, J. C. Campbell, J. C. Carrano, M. J. Schurman, and I. A. Ferguson, “Improved ultraviolet quantum efficiency using a semitransparent recessed window AlGaN/GaN heterojunction p-i-n photodiode,” Appl. Phys. Lett. 75(16), 2421–2423 (1999). [CrossRef]
10. A. R. Clawson, “Guide to references on III-V semiconductor chemical etching,” Mater. Sci. Eng. Rep. 31(1-6), 1–438 (2001). [CrossRef]
11. H. Nie, K. A. Anselm, C. Lenox, P. Yuan, C. Hu, G. Kinsey, B. G. Streetman, and J. C. Campbell, “Resonant-cavity separate absorption, charge and multiplication avalanche photodiodes with high-speed and high gain-bandwidth product,” IEEE Photon. Technol. Lett. 10(3), 409–411 (1998). [CrossRef]
12. H. D. Liu, D. Mcintosh, X. G. Bai, H. P. Pan, M. G. Liu, J. C. Campbell, and H. Y. Cha, “4H-SiC PIN Recessed-Window Avalanche Photodiode With High Quantum Efficiency,” IEEE Photon. Technol. Lett. 20(18), 1551–1553 (2008). [CrossRef]
13. D. E. Aspnes and A. A. Studna, “Dielectric Functions and Optical-Parameters of Si, Ge, Gap, Gaas, Gasb, Inp, Inas, and Insb from 1.5 to 6.0 Ev,” Phys. Rev. B 27(2), 985–1009 (1983). [CrossRef]
14. S. M. Sze and K. K. Ng, Physics of semiconductor devices, 3rd ed. (Wiley-Interscience, Hoboken, NJ, 2007), pp. x, 815 p.
15. B. K. Ng, F. Yan, J. P. R. David, R. C. Tozer, G. J. Rees, C. Qin, and J. H. Zhao, “Multiplication and excess noise characteristics of thin 4H-SiC UV avalanche photodiodes,” IEEE Photon. Technol. Lett. 14(9), 1342–1344 (2002). [CrossRef]