APPLIED PHYSICS LETTERS
VOLUME 78, NUMBER 15
9 APRIL 2001
High-performance InAsÕGaSb superlattice photodiodes for the very long wavelength infrared range H. Mohseni and M. Razeghia) Center for Quantum Devices, Department of Electrical and Computer Engineering, Northwestern University, Evanston, Illinois 60208
G. J. Brown Air Force Research Laboratory, Material Directorate, WPAFB, Ohio 45433-7707
Y. S. Park Office of Naval Research, 800 North Quincy Street, Arlington, Virginia 22217-5660
共Received 5 September 2000; accepted for publication 15 February 2001兲 We report on the demonstration of high-performance p-i-n photodiodes based on type-II InAs/ GaSb superlattices with 50% cut-off wavelength c ⫽16 m operating at 80 K. Material is grown by molecular beam epitaxy on GaSb substrates with excellent crystal quality as evidenced by x-ray diffraction and atomic force microscopy. The processed devices show a current responsivity of 3.5 A/W at 80 K leading to a detectivity of ⬃1.51⫻1010 cmHz1/2/W. The quantum efficiency of these devices is about 35% which is comparable to HgCdTe detectors with a similar active layer thickness. © 2001 American Institute of Physics. 关DOI: 10.1063/1.1362179兴
Recently, there has been considerable interest in type-II superlattices and heterostructures based on the InAs/Ga1⫺x Inx Sb material system, and successful lasers1 and detectors2,3 have been demonstrated in the mid- and longwavelength infrared ranges. Also, theoretical calculations4 predict high performance type-II infrared detectors in the very-long wavelength infrared 共VLWIR兲 range ( c ⬎14 m兲. Infrared detectors in this wavelength range are needed for space based applications such as pollution monitoring and space based astronomy. Currently, the only available detectors in the VLWIR range with high uniformity, high quantum efficiency, and high detectivity are extrinsic silicon detectors which operate below 10 K. Consequently, a threestage cryo-cooler must be used which, with its large volume, weight, and power consumption make its choice unappealing for space based applications since it significantly increases launch costs. Unlike extrinsic silicon detectors, type-II superlattice detectors are based on interband optical transitions and hence they can operate at much higher temperatures. Moreover, theoretical calculations5 and experimental results6 show that InAs/Ga1⫺x Inx Sb type-II superlattices have a similar absorption coefficient to HgCdTe, and therefore detectors with high quantum efficiencies are possible. We have previously reported type-II photoconductive devices grown on GaAs substrate in the c ⫽12 m to c ⫽22 m range operating at 80 K.7 Unlike HgCdTe,8 these detectors showed an excellent energy gap uniformity over a three-inch wafer area which is important for imaging applications. In this letter, we report on a series of high performance photovoltaic type-II superlattice detectors which show the same excellent uniformity in the VLWIR range. The main advantage of photovoltaic detectors is their suitability for staring, two-dimensional focal plane array 共FPA兲 a兲
Electronic mail:
[email protected]
applications, where low current bias circuitry significantly reduces the array power and heat dissipation requirements. The material is grown by an Intevac Modular Gem II molecular beam epitaxy equipped with As and Sb valved cracker sources on p-type GaSb substrates. The photodiode structures were grown at 395 °C according to a calibrated pyrometer. First, a 1 m GaSb buffer/contact layer doped with Be (p⬃1⫻1018 cm⫺3兲 was deposited. Then, a 0.5 m thick InAs/GaSb:Be (p⬃1⫻1018 to 3⫻1017 cm⫺3兲 superlattice was grown followed by a 2 m thick nominally undoped superlattice. Finally, a 0.5 m thick InAs:Si/GaSb (n⬃1 ⫻1018 cm⫺3兲 superlattice was grown and capped with a 100 Å thick InAs:Si (n⬃2⫻1018 cm⫺3兲 top contact layer. The growth rate was 0.5 monolayer/s for InAs layers and 0.8 monolayer/s for GaSb layers. The V/III beam-equivalent pressure ratio was about 4 for InAs layers and about 1.2 for GaSb layers. The cracker temperature for As and Sb cells was 800 °C. The selected thickness of InAs and GaSb layers were determined for specific cutoff wavelengths using a four-band superlattice k•p model. For devices with nearly a cutoff wavelength of 16 m, the thickness of InAs layers was 54 Å and the thickness of the GaSb layers was 40 Å. Although simple, this model has shown good agreement with our experimental results9,10 for a wide range of band gaps. Structural quality of the epitaxial layers was assessed using high resolution x-ray diffraction. Figure 1 shows the typical x-ray diffraction pattern of the photodiode structures. The mismatch between the average lattice constant of the superlattice and the GaSb substrate was below 0.06%, while the full width at half maximum 共FWHM兲 of the satellite peaks was below 60 arcsec for the grown devices. The surface morphology of the samples was also studied with an atomic force microscope 共AFM兲. Theoretical calculations show that surface and interface roughness lead to defect-like energy states inside the superlattice energy gap and broadening of the band edges.11 Also, experimental results show a
0003-6951/2001/78(15)/2107/3/$18.00 2107 © 2001 American Institute of Physics Downloaded 26 Aug 2005 to 129.105.5.195. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp
2108
Mohseni et al.
Appl. Phys. Lett., Vol. 78, No. 15, 9 April 2001
FIG. 1. High resolution x-ray diffraction of an InAs/GaSb 共54 Å/40 Å兲 superlattice grown on a GaSb substrate. FWHM of the satellite peaks is below 60 arcsec, and the mismatch to the substrate is below 0.06%.
strong correlation between the surface roughness and electrical performance of InAs/Ga1⫺x Inx Sb superlattice photodiodes.12 Figure 2 shows the gray-scale surface morphology of a sample. Wide atomic steps are visible which is an indication of excellent surface smoothness. We could routinely grow samples with a root-mean-square surface roughness below 4 Å over a 20 m ⫻20m area which is among the best reported values for this material system. The samples were then processed into 400 m ⫻400m mesas using standard lithography and wet etching. Ti/Au contacts were defined for top and bottom contacts with metal evaporation and liftoff techniques. No passivation or antireflection coating was used on the surfaces. The samples were then indium bonded to a copper heatsink and attached to the cold finger of a liquid nitrogen cryostat with KRS-5 windows. Absolute spectral responsivity was calculated from the measured spectral response of the device, using a Fourier transform infrared spectroscopy system, and it photoresponse to a calibrated blackbody setup. Figure 3 shows the typical spectral responsivity of the detectors with c ⫽16 m. The absorption features of CO2 and H2O are due to the small
FIG. 3. Absolute spectral responsivity of a device with c ⫽16 m. The 90% to 10% cut-off energy width is about 12 meV. The absorption features of CO2 and H2O are due to the small difference in the optical path length of the background measurement and the detector measurement.
difference in the optical path length of the background measurement and the detector measurement. The peak responsivity for the sample is about 3.5 A/W which leads to a quantum efficiency of ⬃35% at 12 m. The use of binary layers in the superlattice has significantly enhanced the uniformity and reproducibility of the energy gap. The 90% to 10% cut-off energy width of these devices is only about 2 kT which is at least four times smaller compared to the similar devices based on InAs/Ga1⫺x Inx Sb (0.15⬍x⬍0.26) superlattices.2,3,13,14 The major noise component at zero base is the Johnson noise, and hence the detectivity of the device with current responsivity R i at temperature T can be calculated from: D * ⫽R i
冑
R 0A , 4kT
共1兲
where R 0 is the zero bias differential resistance of the device, A is the device area, and k is the Boltzmann constant. The measured value for R 0 A product for the detectors with c ⫽16 m was about 0.072 ⍀ cm2 at T⫽80 K which leads to a Johnson noise limited detectivity of about 1.51⫻1010 cmHz1/2/W. In order to study the major components of the dark current at T⫽80 K, the current–voltage characteristic of the devices was modeled. Although the active layer of these devices consists of short period superlattices, bulkbased modeling of the dark current has been proven to give relatively accurate results.2,10,14 We used a similar formalism as reported in Ref. 10. Figure 4 shows the measured and modeled current densities versus the applied bias for devices with c ⫽16 m. The calculated current density, which consists of tunneling, generation recombination, and diffusion current densities, shows good agreement to the measured values for forward and reverse biases. We assumed an effective mass of m e ⫽0.03 m0 for electrons and m h ⫽0.4 m0 for holes based on previous theoretical calculations11 and experimental results.2,3 Based on the experimental measurements on similar devices,15 we also assumed an electron mobility parallel to the growth direction of e ⫽1000 cm2/Vs. The mobility of the holes is not significant in the diffusion current, since the device has a n ⫹ -p junction. The fitting parameters for the model were carrier lifetime e ⫽ h ⫽220 ns,
FIG. 2. AFM image of the surface of the device. Atomic steps with several micron width are clearly visible. Downloaded 26 Aug 2005 to 129.105.5.195. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp
Mohseni et al.
Appl. Phys. Lett., Vol. 78, No. 15, 9 April 2001
2109
In conclusion, we have demonstrated high performance InAs/GaSb photodiodes operating in the VLWIR range at T⫽80 K. Devices with c ⫽16 m showed a quantum efficiency of about 35% and Johnson noise limited detectivity of about 1.51⫻1010 cmHz1/2/W at 80 K. These values are comparable to the commercially available single-element HgCdTe detectors at similar wavelengths. Excellent band gap uniformity and reproducibility of these type-II superlattices over large areas make them an attractive choice for FPAs in this spectral range. This work is supported by Office of Naval Research under Contract No. N00014-99-1-0630 and Air Force under Contract No. F33615-00-C-5526. 1
FIG. 4. The measured and modeled dark current density of a device with c ⫽16 m at T⫽80 K. The model consists of the tunneling, generation recombination, and diffusion components of the dark current and indicates that generation recombination is the dominant source of the dark current around zero bias.
unintentional background doping level p⫽2.1⫻1015 cm⫺3, and generation-recombination lifetime in the depleted layer Gr⫽0.6 ns. In contrast to HgCdTe, tunneling is not significant even at high values of the reverse bias due to the higher effective mass of the electrons in type-II superlattices. However, generation-recombination current is the dominant source of dark current for these devices at T⫽80 K, and hence further improvement of the growth should increase R 0 A and detectivity. The value of R 0 A product versus temperature showed a diffusion limit behavior down to nearly 100 K, and then a generation-recombination limit behavior from 100 to 60 K. Below 60 K, the value of R 0 A increased even slower probably due to the defect related leakages. The ideality factor of the device was nearly 1.42 for small values of forward bias. The values for carrier lifetime and background concentration are among the best reported for type-II. The energy gap of the superlattices can be tailored for longer wavelengths and similar devices with cut-off wavelengths up to c ⫽25 m have been measured at 80 K.16
W. Bewley, H. Lee, I. Vurgaftman, R. Menna, C. Felix, R. Martinelli, D. Stokes, D. Garbuzov, J. Meyer, M. Maiorov, J. Connolly, A. Sugg, and G. Olsen, Appl. Phys. Lett. 76, 256 共2000兲. 2 J. L. Johnson, L. A. Samoska, A. C. Gossard, J. Merz, M. D. Jack, G. R. Chapman, B. A. Baumgratz, K. Kosai, and S. M. Johnson, J. Appl. Phys. 80, 1116 共1996兲. 3 F. Fuchs, U. Weimar, W. Pletschen, J. Schmitz, E. Ahlswede, M. Walther, J. Wagner, and P. Koidl, Appl. Phys. Lett. 71, 3251 共1997兲. 4 C. Grein, H. Curz, M. Flatte, and H. Ehrenreich, Appl. Phys. Lett. 65, 2530 共1994兲. 5 D. Chow, R. Miles, J. Schulman, D. Collins, and T. McGill, Semicond. Sci. Technol. 6, C46 共1991兲. 6 H. Mohseni, E. Michel, J. Sandven, M. Razeghi, W. Mitchel, and G. Brown, Appl. Phys. Lett. 71, 1403 共1997兲. 7 H. Mohseni, A. Tahraoui, J. Wojkowski, and M. Razeghi, Appl. Phys. Lett. 77, 1572 共2000兲. 8 A. Rogalski, Infrared Phys. Technol. 40, 279 共1999兲. 9 H. Mohseni, E. J. Michel, M. Razeghi, W. C. Mitchel, and G. J. Brown, Proc. SPIE 3287, 30 共1998兲. 10 H. Mohseni, J. Wojkowski, A. Tahraoui, M. Razeghi, G. Brown, and W. Mitchel, Proc. SPIE 3948, 153 共2000兲. 11 G. Bastard, Phys. Rev. B 25, 7584 共1982兲. 12 F. Fuchs, L. Burkle, W. Pletschen, J. Schmitz, M. Walther, H. Gullich, N. Herres, and S. Mueller, Proc. SPIE 3794, 41 共1999兲. 13 M. Young, D. Chow, A. Hunter, and R. Miles, Appl. Surf. Sci. 123, 395 共1998兲. 14 F. Fuchs, U. Weimar, E. Ahlswede, W. Pletschen, J. Schmitz, and M. Walther, Proc. SPIE 3287, 14 共1998兲. 15 L. Burkle, F. Fuchs, R. Kiefer, W. Pletschen, R. E. Sah, and J. Schmitz, Mater. Res. Soc. Symp. Proc. 607, 77 共2000兲. 16 M. Razeghi and H. Mohseni, Proceedings of the 45th Annual Meeting of SPIE, San Diego, California, July 30–August 4, 2000.
Downloaded 26 Aug 2005 to 129.105.5.195. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp
