Noise spectrum of the device structure. A low corner frequency below 1 kHz can be achieved under optimized conditions. 

Context in source publication

Context 1

... 1/f noise current is I n −1 / f = S I at f = 1 Hz and in 1 Hz bandwidth. I is the dark current, N is the number of carriers in the semiconductor and ␣ H is the Hooge’s factor. Expres- sion for S I ͑ f ͒ has been evaluated for HgCdTe diodes by 48 49 Kleinpenning and Van der Ziel et al. , and was related to diode diffusion and g-r current depending on temperature of operation and device length. Value of ␣ H varies from 1 ϫ 10 −3 to 5 ϫ 10 −5 . For the sake of calculations, taking maximum value of ␣ H = 1 ϫ 10 −3 and add I n−1 2 / f to total noise current and calculate total noise current as a function of fre- quency. Figure 5 shows total noise current as a function of frequency, frequency was varied from 1 Hz to 1 MHz. 1/f knee frequency evaluated from the intersection point of the tangents to 1/f noise portion and white noise portion of the curve is ϳ 875 Hz. For detector operation at frequency less than the knee frequency, 1/f noise power starts dominating its performance, although frequency variation in total noise current starts from ϳ 10 kHz. D ء was calculated to be ϳ 9.8 ϫ 10 9 cm Hz 1 / 2 W −1 for the គ + / ␲ / p គ + device for frequency Ͻ 1 / f knee. 1/f knee frequency for Auger suppressed long wave infrared ͑ LWIR ͒ and MWIR devices vary from 100 Hz to a several kilohertz. Low frequency 1/f noise is much 16 higher in LWIR devices as compared to MWIR devices. For f = 875 Hz, D ء was calculated to be 3.0 ϫ 10 10 cm Hz 1 / 2 W −1 . Thus, it is observed that although the 1/f noise current reduces detectivity, it is not a significant reduction. MWIR devices appear to be capable of achieving useful D ء values for imaging applications in high frame rate applications. In fact, 4 ␮ m cutoff HOTEYE thermal camera 50 operating at 210 K has been fabricated and demonstrated. For the extrinsic គ nonequilibrium device, thermal g-r processes and photogeneration are limited to the nearly intrinsic ␲ absorber layer. Heavily doped contact layers serve the purpose of excess carrier collection in addition to extraction and exclusion. Dependence of D * on doping of absorber and contact regions was analyzed and is shown in Fig. 6. Incident IR flux was same as mentioned in Sec. III B. Figure 6 ͑ a ͒ shows effect of changing doping concentration of ␲ absorber layer on detectivity of the គ + / ␲ / គ + double barrier device at T = 250 K. Donor and acceptor doping concentrations of គ + and គ + contact layers were 8 ϫ 10 16 cm −3 and 4.5 ϫ 10 17 cm −3 , respectively. D ء increases with acceptor doping starting at 1 ϫ 10 14 cm −3 and attains a maximum value near about 4 – 5 ϫ 10 15 cm −3 . Further increase in acceptor doping reduces D ء significantly because minority carrier extraction in ␲ layer gets reduced. Auger suppression becomes difficult for higher acceptor doping. Figures 6 ͑ b ͒ and 6 c , respectively, show the effect of varying the doping concentrations of គ + and p គ + contact layers. D ء is almost constant for គ + contact doping Ͼ 5 ϫ 10 16 cm −3 and falls of sharply for contact doping 1 cm . Heavily doped wider gap គ + contact is preferred for Auger suppression due 16 to Moss–Burstein band filling effect. With increased doping, the គ + contact becomes degenerate near ambient temperature and the Fermi level moves high into the conduction band. The threshold energy required for Auger-1 process increases and negligible optical absorption takes place in the គ + layer, which acts almost as a window layer. Hence with increased n-type doping, D ء nearly saturates. D ء increases with increase in គ + contact doping starting at 1 ϫ 10 14 cm −3 and attains a maximum value at concentrations near about 4 – 5 ϫ 10 17 cm −3 . The absorption increases in this case due to reduced band filling effect in conduction band, as the Fermi level moves further away from the conduction band. Fall in D ء is observed with further increased doping concentrations of គ + contact because of increased Auger-7 g-r process. Optically generated carriers in the narrow gap absorber layer diffuse to the two junctions and are collected by the wide gap heavily doped contacts. For efficient collection of photogenerated carriers, they must diffuse to the junction. Thus, the junction must be located at distance less than the diffusion length of excess carriers. Detectivity of the គ + / ␲ / p គ + device was evaluated for different absorber layer widths and is shown in Fig. 7 ͑ a ͒ . D ء increases with increase in ␲ width, attains a maximum for ϳ 4 ␮ m width and then decreases with further increase in ␲ width. The interplay between increasing quantum efficiency with increasing absorber width and decreasing collection efficiency of optically generated carriers with a large increase in absorber width is responsible for this observation. Hence an absorber region length of ϳ 3 – 5 ␮ m would suffice for efficient collection and absorption of incident radiation at 250 K. Plausible ex- planation is that at 0.1 V reverse bias device with absorber width of ϳ 3 ␮ m gets completely depleted of free carriers and the excess carriers generated within the absorber region diffuse rapidly to the contacts. Profiles of n ͑ x ͒ and p ͑ x ͒ for absorber width of 3 ␮ m are shown in Fig. 7 ͑ b ͒ for illustra- tion. For greater absorber layer width, extraction of carriers takes place, but at the same time photogenerated carriers have to diffuse longer in order to get collected. For LWIR Auger suppressed devices to operate at or near room temperature with BLIP performance, the requirements on doping levels, concentration of Shockley–Read centers and device design to prevent thermal generation at surfaces, interfaces, and contacts are more stringent than that for MWIR devices. 12 A comparison between MWIR ͓ គ + ͑ x = 0.35 ͒ / ␲ ͑ x = 0.3 ͒ / គ + ͑ x = 0.35 ͔͒ and LWIR ͓ គ + ͑ x = 0.3 ͒ / ␲ ͑ x = 0.22 ͒ / គ + ͑ x = 0.3 ͔͒ double barrier heterojunction device with similar acceptor doping concentrations is shown in Fig. 8. The variation in quantum efficiency, as a function of operating temperature between 150 and 300 K is shown in Fig. 8 ͑ a ͒ for both LWIR and MWIR devices. It is seen that the heterojunction photodetectors with similar design and under similar operating conditions show a decrease in quantum efficiency with temperature because of reduction in absorption coefficient. The quantum efficiency exhibited by MWIR detector varies between 60%–80% and that of LWIR detector varies between 35%–40%. A longer absorber region width and lower doping concentration would be required for LWIR detector to operate with similar performance at a particular temperature. Near ambient temperatures, HgCdTe becomes intrinsic due to high thermal generation of carriers, resulting in low minority carrier lifetimes due to Auger recombination processes. This low lifetime at high temperatures, results in high dark currents and high noise. Using the previously mentioned formulations for dark current density and the quantum efficiency, performance of the detectors operating in both regions was compared. Figure 8 ͑ b ͒ shows variation in dark current density and detectivity as function of temperature for the MWIR and LWIR detectors. The total dark current density increases with temperature, which is responsible for increased noise current and thus a decrease in detectivity. The dark current density and calculated values of detectivity are 5,38,37 in agreement with experimental results. MWIR photodetectors operating at near room temperature at frequencies above the knee frequency are capable of achieving BLIP performance with less severe demands on material parameters and device design. In practice, detectivities well de- signed of the uncooled LWIR detectors still remain below the g-r limit. A generalized model was used to assess performance of a 4.5 ␮ m cutoff; Auger suppressed double barrier heterojunction គ + / ␲ / គ + MWIR detector operating at 250 K in terms of device architecture, design, material properties, and operation temperature. Performance of was analyzed by numerical computations using SOR method. The study ad- dresses several issues: Profiles of carrier concentration, electrostatic potential, and g-r rates were simulated as a function of position, for doping concentrations and material parameters mentioned in Table I. Extraction and exclusion at the contact regions re- sulted in Auger suppression. Using the numerically simulated profiles of ␺ ͑ x ͒ , n ͑ x ͒ , p ͑ x ͒ , and G ͑ x ͒ -R ͑ x ͒ , detector dark current, responsivity, noise current, and D ء have been calculated. Detectivity values of the order of 3 ϫ 10 10 are obtained with optimized device parameters. Effect of 1/f noise current on performance was also modeled. For high frame rate applications, 1/f noise current did not seem to reduce the performance significantly. Performance parameters were also calculated as a function of doping concentration of the three layers, absorber layer width, and carrier lifetime. Absorber doping concentration Յ 5 ϫ 10 15 cm −3 and contact doping concentration of 1 ϫ 10 16 to 1 ϫ 10 18 gave the best results for absorber layer with of 3 – 4 m. MWIR photodetectors perform near BLIP limit with less stringent demands on material properties. It is concluded that the noise current arising at junctions and contacts can be further low- ered in a multijunction device. Authors would like to thank Director SSPL, Dr. R. Mu- ralidharan for his encouragement and kind permission to publish this work. Authors would also like to thank Dr. V. Dhar, Scientist ‘F’, SSPL, for his valuable suggestions and ...