The growth and performance of top-illuminated metamorphic In0.20Ga0.80As p-i-n photodetectors grown on GaAs substrates using a step-graded InxGa1-xAs buffer is reported. The p-i-n photodetectors display a low room-temperature reverse bias dark current density of ~1.4×10−7 A/cm2 at −2 V. Responsivity and specific detectivity values of 0.72 A/W, 2.3×1012 cm·Hz1/2/W and 0.69 A/W, 2.2×1012 cm·Hz1/2/W are achieved for Yb:YAG (1030 nm) and Nd:YAG (1064 nm) laser wavelengths at −2 V, respectively. A high theoretical bandwidth-responsivity product of 0.21 GHz·A/W was estimated at 1064 nm. Device performance metrics for these GaAs substrate-based detectors compare favorably with those based on InP technology due to the close tuning of the detector bandgap to the target wavelengths, despite the presence of a residual threading dislocation density. This work demonstrates the great potential for high performance metamorphic near-infrared InGaAs detectors with optimally tuned bandgaps, which can be grown on GaAs substrates, for a wide variety of applications.
© 2011 OSA
The achievement of maximum photodetector performance for many applications is dependent upon the proximity of the detector bandgap energy to the illumination energy. Nonetheless, p-i-n photodetectors, which are advantageous in their simplicity and lack of absorption geometry constraints (such as needed for QWIPs), are typically constrained in bandgap choice by lattice matching to conventional substrates, often meaning that photodetector bandgap energy is too low or not optimum for a desired illumination wavelength, and the energy difference usually contributes to excess noise. Examples of this abound. For instance, a vast array of applications, running the gamut from medicine to astronomy to manufacturing, employ high power neodymium- and ytterbium-doped yttrium aluminum garnet (Nd:YAG, Yb:YAG) lasers operated at 1064 nm and 1030 nm, respectively . However, there is a lack of high performance (low noise, high responsivity, high speed) photodetectors that can be operated at either of these wavelengths. Infrared-enhanced silicon photodetectors have a long wavelength cut-off at or below 1030 nm and the indirect bandgap of silicon necessitates very thick absorption layers to achieve even modest efficiencies, which limits the operating speed of these detectors, resulting in a low bandwidth-responsivity product. In0.53Ga0.47As photodetectors (lattice-matched to InP) can achieve very high efficiencies and cut-off frequencies but are hampered by the need to be grown on InP substrates, which are smaller, more expensive, more fragile, and for which device processing technology is less mature, than GaAs or Si substrate-based devices. Additionally, In0.53Ga0.47As photodetectors suffer from both higher dark currents in this wavelength range, due to the relatively small bandgap, and low laser threshold damage . Although Ge photodetectors can achieve very high speeds due to their high hole mobilities, they also experience high dark current and low shunt resistance due to a bandgap that is also too low, resulting in higher noise at the desired wavelengths. Dilute nitride semiconductors like InGaNAs [3,4] and InGaNAsSb  grown lattice-matched to GaAs have also been explored for operation in this wavelength range, since their bandgaps can be tuned by careful compositional control, but dilute nitride alloys, at this point in their development, are typically impaired by comparatively low material quality, reducing device performance. Low In composition (x < 0.25) InxGa1-xAs, which is approachable via metamorphic grading starting from a GaAs substrate, would ideally fill this need due to its direct bandgap (resulting in high absorption coefficients, thinner device layers, and higher speed) and tunable bandgap (lower noise), if high quality materials can be achieved.
There has indeed been much interest in achieving metamorphic (MM) InxGa1-xAs photodetectors on GaAs substrates, but the primary focus has been to reach a target composition of In0.53Ga0.47As, which enables high performances at the 1550 nm and 1300 nm wavelengths that are critical for telecommunication applications ; and the substitution of GaAs substrates for InP would reduce cost and enable larger area arrays. However, not all applications benefit from this composition, such as in the case of YAG laser detectors, which would perform optimally at a direct bandgap of ~1.1 eV, a value that can be achieved with a composition of In0.20Ga0.80As. The ideal target bandgap would also determine the lattice misfit that needs to be accommodated by the metamorphic buffer, which can then be designed to minimize the dislocation density in these materials in order to achieve the desired high performance. While the successful utilization of metamorphic InxGa1-xAs, with terminal lattice constants between that of GaAs and InP as the active material for device applications such as metamorphic solar cells has been reported , there is very little work on using low In composition InGaAs to achieve high performance photodetectors in the intermediate range of alloy compositions.
In this paper, we report the growth and fabrication of top-illuminated metamorphic In0.20Ga0.80As (Eg = 1.11 eV) p-i-n photodetectors lattice-matched to InxGa1-xAs step-graded buffers on GaAs substrates. By way of example, but motivated by the technological need described above, we chose to focus on detector development for the 1000 – 1100 nm range. High responsivity values of 0.72 A/W and 0.69 A/W were achieved at 1030 nm and 1064 nm wavelengths, respectively, without the use of any anti-reflection coatings. Excellent dark current densities, cut-off frequencies and noise equivalent powers are also reported. The quality of the metamorphic materials was studied using high-resolution X-ray diffractometry and transmission electron microscopy.
2. Experimental details
All growths reported here were performed in a modified Varian GEN II solid-source MBE system with valved P2 and As2 cracker sources. Idle background pressure in the UHV growth chamber was < 2×10−10 Torr. Substrate temperatures were measured via both in situ infrared pyrometry (calibrated to the GaAs native oxide desorption temperature) and with a substrate heater-mounted thermocouple. In situ monitoring of the growths was performed via reflection high-energy electron diffraction (RHEED).
All growths were performed on n-type (100)-oriented GaAs substrates with an intentional 6° off-cut toward the nearest (111)A. Three-step compositionally-graded InxGa1-xAs buffers with x = 0.10, 0.17 and 0.23 were grown to accommodate the ~1.5% misfit of the target photodetector materials with respect to the GaAs substrate. Each InGaAs step-layer was grown to a thickness of 0.5 µm at a growth rate of ~0.65 µm/hr with an As2:(Ga+In) beam equivalent pressure ratio of 24:1 and an n-type Si doping of 5×1017 cm−3. The In0.10Ga0.90As layer was grown at a growth temperature of 550°C, and all subsequent InxGa1-xAs layers were grown at 515°C to avoid In droplet formation. A growth stop was used after growth of each layer to change the In effusion cell temperature (and thus In beam flux), and/or to change the substrate temperature.
The In0.20Ga0.80As p-i-n photodetector structure consisted of a 0.05 µm p-type layer (5×1018 cm−3 Be-doped), a 1.0 µm undoped layer, and a 0.5 µm n-type layer (3×1018 cm−3 Si-doped). The device structures also incorporated a 0.1 µm In0.64Ga0.36P window layer to minimize the impact of surface recombination that could significantly reduce the responsivity at shorter wavelengths.
Material quality and properties were characterized by both structural and electronic measurements in order to correlate photodetector characteristics with physical materials properties and to guide the optimization of growth conditions. Triple-axis high-resolution x-ray diffractometry (HRXRD) measurements were carried out using a Bede Scientific Instruments D1 system with a CuKα1 x-ray source in order to quantify composition and strain state (i.e. relaxation) of each layer within the growth structure using a combination of ω-2θ rocking curves and reciprocal space maps (RSM). Cross-sectional and plan-view transmission electron microscopy (XTEM, PVTEM) samples were prepared using the focused ion beam technique in a FEI Helios NanoLab 600 dual-beam system. TEM was performed in a 200 kV FEI Tecnai F20 microscope using a two-beam bright field condition in the g(200) and g(220) reflections.
P-i-n photodetectors with square active area of 150 × 150 µm2 were fabricated. Photolithography and inductively coupled plasma etching using a mixture of BCl3 and SiCl4 gases were used to define the mesa. Photo-definable polyimide PI-2723 was used to perform several functions including planarization of the surface for front contact deposition, electrical isolation and device sidewall passivation. Ni/Ge/Au and Cr/Au metal stacks were deposited using electron-beam evaporation to form n-type and p-type Ohmic contacts, respectively. The Ni/Ge/Au contacts were annealed at 400°C for 4 minutes. No anti-reflection coating was applied. A cross-sectional schematic of the fabricated p-i-n In0.20Ga0.80As photodetectors are shown in Fig. 1 for illustrative purposes.
Fully fabricated photodetectors were characterized by current density vs. voltage (J-V), capacitance vs. voltage (C-V), and spectral response measurements, all at 300 K. The J-V measurements were performed with a programmable electrometer (Keithley, model 6514) with a baseline resolution of ~1 fA. C-V was done using a standard digital Boonton model 7200 capacitance meter operating at 1 MHz, and spectral response was obtained with a set-up that includes a 1000W quartz-halogen lamp with a Spex monochromator. The photocurrent was measured using a Keithley 617 electrometer and the incident light intensity was measured using a pyroelectric radiometer in conjunction with a calibrated Si photodetector.
3. Results and discussion
3.1 Structural materials characterization
HRXRD reciprocal space maps (RSM) taken at both the symmetric (004) and glancing-incidence asymmetric (224) reflections were used to determine epilayer compositions and in-plane and out-of-plane lattice constants, which in turn were used to calculate misfit and the residual strain state (i.e. level of relaxation) for the various layers. A (224) RSM for a typical In0.20Ga0.80As device structure is shown in Fig. 2 .
The RSMs show strong and distinct diffraction peaks for each layer with a degree of mosaic spread that is typical for high-quality compressively strain-relaxed metamorphic layers. Nearly full relaxation (> 95%) was obtained for each step, except for the final In0.23Ga0.77As buffer layer, which was only 80% relaxed. The presence of a small amount of residual strain (−0.25%) in the final In0.23Ga0.77As buffer layer was accounted for by growing an exactly lattice-matched In0.20Ga0.80As compositional step-back layer to ensure unstrained, lattice-matched conditions for subsequent growth of the device layers. The growth of the In0.20Ga0.80As step-back layer is critical since it has been shown that partially compressively strained InGaAs layers can have much higher deep level trap concentrations compared to unstrained InGaAs layers . Indeed, our own growth experiments confirm much poorer ultimate materials quality without this approach, based upon both structural and electronic property measurements. This “overshoot/step-back” method to achieve precise lattice-matching of device layer growth on metamorphic buffers has been documented elsewhere and in general is found to be critical for maximizing the performance of minority carrier devices, which are most sensitive to defect formation .
The resultant dislocation network was studied through TEM characterization of the graded buffer layers. It can be seen from Fig. 3(a) , which shows a cross-sectional view of the InGaAs step-graded metamorphic buffer, that the majority of misfit dislocations (MDs) were created at the interface between the GaAs substrate and the In0.10Ga0.90As layer. The goal of the InGaAs step-graded buffer is to “re-use” these dislocations at subsequent mismatched interfaces, thereby minimizing the nucleation of new dislocations . Figure 3(a) suggests that there is indeed no significant increase in the dislocation density in the subsequent InGaAs layers. Detailed dislocation counting via plan-view TEM reveals a threading dislocation density of 1.5×107 cm−2 in the terminal detector device layers, due to the somewhat aggressive grading rate (1% misfit/μm) used in this structure. The plan-view TEM image in Fig. 3(b) also reveals contrast lines that suggest a low level of phase segregation to be present in the tensile-strained In0.64Ga0.36P window layer. These contrast lines are not apparent in the XTEM image (Fig. 3(a)), which makes it unclear if the onset of phase segregation occurs in one of the InxGa1-xAs layers or if it is present only in the In0.64Ga0.36P layer. Further investigation on the exact origin and impact of the phase segregation is currently underway.
3.2 P-i-n photodetector characterization
A representative dark current density versus applied bias (J-V) plot for fully fabricated metamorphic In0.20Ga0.80As photodetectors is shown in Fig. 4 . The detectors demonstrate good rectifying diode behavior with sharp forward-bias turn-on and ideality factor of 1.2, indicating high material quality and a diffusion current limited diode characteristic.
The p-i-n devices also exhibit low room-temperature reverse bias dark current densities (at −2 V applied bias) of 1.4×10−7 A/cm2. This value is orders of magnitude lower than that previously reported for metamorphic In0.25Ga0.75As or InGaNAs p-i-n detectors that were designed for operation at 1064 nm wavelength . Low shot and thermal noise spectral densities of 5.55×10−13 A/Hz1/2 and 1.03×10−15 A/Hz1/2 were calculated  (without a matching circuit and not integrated over any bandwidth) at 1064 nm and −2 V, respectively. Despite the already outstanding performance demonstrated here, there is ongoing work to reduce the residual TDD (current value is 1.5×107 cm−2) further to evaluate the impact of TDD on detector performance.
Responsivity profiles at 300 K were measured for the In0.20Ga0.80As detectors as a function of applied bias voltage (VR), the results for which are presented in Fig. 5 . The responsivity shows only a small increase with VR and completely saturates at around −1 V. This is due to the intrinsic layer being almost completely depleted under zero bias, which is verified by the very small change in capacitance in the C-V profile and the corresponding carrier concentration versus depletion width plot shown in Fig. 6 . It is useful to also note that the achieved capacitance density of 2.5×10−8 F/cm2 under zero bias closely matches that achieved with lattice-matched In0.53Ga0.47As/InP detectors , displaying potential for high speed operation.
The long wavelength cut-off of 1115 nm for the In0.20Ga0.80As detector corresponds to its bandgap of 1.11 eV, as measured using room-temperature photoluminescence. A responsivity value of 0.72 A/W was achieved at 1030 nm for these GaAs substrate-based detectors. This compares favorably with the responsivities of commercial Si (0.65 A/W)  and In0.53Ga0.47As/InP (0.64 A/W)  p-i-n photodetectors at 1030 nm, whose bandgaps cannot be designed to maximize their responsivities at YAG wavelengths. At 1064 nm wavelength, the metamorphic detector responsivity is 0.69 A/W, which matches that achieved with In0.53Ga0.47As/InP detectors , and is still higher than state-of-the-art IR enhanced Si detectors (0.56 A/W) that have been optimized for this wavelength . While the high responsivities achieved at these wavelengths demonstrate the advantage of bandgap tuning via metamorphic epitaxy to achieve target-optimized detector materials, it must be noted that all of the detector comparisons presented in this paper are made in order to highlight performance at the specific target wavelengths (i.e. YAG laser emission) and not to compare the detectors (or materials) for their absolute performance, as their peak responsivities all occur at different material-dependent wavelengths.
An important figure of merit for a photodetector is the specific detectivity (D*). The shot noise dominated D* is defined as R·√A/In, the ratio of responsivity to the total noise current at a particular wavelength and bias voltage, normalized to detector area. High D* values of 2.3×1012 cm·Hz1/2/W and 2.2×1012 cm·Hz1/2/W were found for 1030 nm and 1064 nm wavelengths at −2 V, respectively, for the metamorphic In0.2Ga0.8As detectors. These values are comparable with the D* values achieved with InP substrate-based lattice-matched In0.53Ga0.47As/InP detectors at these wavelengths (calculated for these wavelengths using values from Ref. .). It should be noted that the D* values at −2 V as reported here are for comparison with other detectors for which D* values are available. The D* values are indeed even higher at zero bias due to only a very small decrease in responsivity with reduced bias voltage, but a large reduction in dark current density, and thus reduced noise. Compared to Si detectors at these wavelengths, however, the D* is still much lower and is likely a result of the exceedingly low defect densities in bulk Si devices, yielding very low noise currents. Nonetheless, this advantage for the Si detectors is offset by their lower responsivities and lower speeds at these wavelengths, which limits their usefulness for many applications.
The bandwidth of a p-i-n photodetector is determined by the RC time constant and the transit time for the photogenerated carriers. For applications that require high responsivity, high speed detectors in this wavelength range, Si is often not the material of choice, despite its excellent D* value, due to the need for thick device layers because of the low optical absorption coefficient (indirect bandgap). Although a very low RC value can be achieved for Si, the need for thicker material to maximize responsivity results in very long transit times, limiting the cut-off frequency of these devices to just a few MHz. In comparison, detectors with high speeds, on the order of a few GHz, can be achieved with both metamorphic In0.20Ga0.80As on GaAs and lattice-matched In0.53Ga0.47As on InP mainly due to much shorter transit times. The metamorphic In0.20Ga0.80As reported in this work was not optimized for high speed operation, and as such, wider bandgap p- and n-doped layers, which can mitigate the slow diffusion of carriers in the doped layers, were not used. Nonetheless, the capacitance and matching resistance measured at −2 V were 8.55 pF and 60 Ω, respectively. Assuming that the device is limited by its RC time constant (the transit time would be very small since the intrinsic layer is only 1 µm thick), this results in a theoretical 3-dB bandwidth of 3.1×108 Hz and a bandwidth-responsivity product of 0.21 GHz·A/W at 1064 nm. The low noise and high bandwidth-responsivity product achieved for the metamorphic In0.20Ga0.80As p-i-n detector demonstrates the high level of success achieved in accessing the target bandgap using the metamorphic buffer approach, while capturing the advantage of a GaAs substrate platform as compared with InP.
Low noise, high responsivity In0.20Ga0.80As p-i-n photodetectors fabricated on GaAs substrates using InxGa1-xAs metamorphic step-graded buffers have been demonstrated. These near-infrared photodetectors display low dark current densities of 1.4×10−7 A/cm2 at −2 V. Optical responsivities of 0.72 A/W and 0.69 A/W, and D* values of 2.3×1012 cm·Hz1/2/W and 2.2×1012 cm·Hz1/2/W were achieved for 1030 nm and 1064 nm wavelengths at −2 V, respectively, with an estimated 3-dB bandwidth of 0.31 GHz based on RC time delay. All of the metamorphic In0.20Ga0.80As/GaAs detector performance metrics at the target wavelengths (1030 nm and 1064 nm) compare favorably with those reported for In0.53Ga0.47As detectors that are grown on lattice-matched InP substrates.
The detector performance described here, coupled with both the use of larger area, more affordable, and robust GaAs substrates (compared with InP) as well as the natural bandgap tunability afforded by the metamorphic materials design to optimize for target wavelengths, clearly indicates advantages for metamorphic detectors for a variety of applications.
This work is supported by the Air Force Office of Scientific Research (K. Reinhardt) – grant # FA9550-06-1-0557 and grant # FA9550-10-1-0015, and also by the Army Research Office (J. Prater) – grant # DAAD-19-01-0588
References and links
2. S. Donati, Photodetectors: Devices, Circuits and Applications (Prentice Hall, 2000), pp. 109–214.
3. D. Jackrel, H. Yuen, S. Bank, M. Wistey, J. Fu, X. Yu, Z. Rao, and J. S. Harris, “Thick lattice-matched GaInNAs films in photodetector applications,” Proc. SPIE 5726, 27–34 (2005). [CrossRef]
4. W. K. Loke, S. F. Yoon, K. H. Tan, S. Wicaksono, and W. J. Fan, “Improvement of GaInNAs p-i-n photodetector responsivity by antimony incorporation,” J. Appl. Phys. 101(3), 033122 (2007). [CrossRef]
5. G. Lin, H. Kuo, C. Lin, and M. Feng, “Ultralow Leakage In0.53Ga0.47As p-i-n photodetector grown on linearly graded metamorphic InxGa1-xP buffered GaAs substrate,” IEEE J. Quantum Electron. 41(6), 749–752 (2005). [CrossRef]
6. S. Sinharoy, M. Patton, T. Valko, and V. Weizer, “Progress in the development of metamorphic multi-junction III–V space solar cells,” Prog. Photovolt. Res. Appl. 10(6), 427–432 (2002). [CrossRef]
7. D. Pal, E. Gombia, R. Mosca, A. Bosacchi, and S. Franchi, “Deep levels in virtually unstrained InGaAs layers deposited on GaAs,” J. Appl. Phys. 84(5), 2965–2967 (1998). [CrossRef]
8. S. P. Ahrenkiel, M. W. Wanlass, J. J. Carapella, R. K. Ahrenkiel, S. W. Johnston, and L. M. Gedvilas, “Optimization of buffer layers for lattice-mismatched epitaxy of GaxIn1−xAs/InAsyP1−y double-heterostructures on InP,” Sol. Energy Mater. Sol. Cells 91(10), 908–918 (2007). [CrossRef]
9. E. A. Fitzgerald, A. Y. Kim, M. T. Currie, T. A. Langdo, G. Taraschi, and M. T. Bulsara, “Dislocation dynamics in relaxed graded composition semiconductors,” Mat. Sci. Eng. B-Solid 67(1-2), 53–61 (1999). [CrossRef]
10. D. B. Jackrel, “InGaAs and GaInNAs(Sb) 1064 nm photodetectors and solar cells on GaAs substrates,” Doctoral dissertation (Dept. of Materials Science and Engineering, Stanford University, 2005), pp. 127–150.
11. S. M. Sze and K. K. Ng, Physics of Semiconductor Devices (John Wiley & Sons, 2007), Chap.13.
12. Hamamatsu Photonics, “InGaAs p-i-n photodiode G8941 series data sheet,” http://sales.hamamatsu.com/assets/pdf/parts_G/G8941_series.pdf.
13. Pacific Silicon sensor, “1064 nm enhanced silicon quadrant photodiode (Series Q) data sheet,” http://www.pacific-sensor.com/pages/sp_sq.html.