Normal-Incidence InAs Self-Assembled Quantum-Dot Infrared Photodetectors With a High Detectivity

An InAs/AlGaAs quantum-dot infrared photodetector based on bound-to-bound intraband transitions in undoped InAs quantum dots is reported. AlGaAs blocking layers were employed to achieve low dark current. The photoresponse peaked at 6.2 m. At 77 K and 0.7 V bias, the responsivity was 14 mA/W and the detectivtiy, , was 10 cm Hz /W. M IDand far-infrared (3–20 m) detection is a key technology for numerous commercial, military and space applications, e.g., night vision, thermal imaging, chemical analysis, nondestructive detection, remote sensing, and missile guidance and defense. Due to the long carrier capture and relaxation times, quantum-dot infrared photodetectors (QDIPs) have the potential for lower dark current and higher photoresponse than quantum-well infrared photodetectors (QWIPs). Most importantly, the three-dimensional (3-D) confinement of electrons in the quantum dots permits QDIPs to operate in the normal incidence mode, unlike QWIPs which are not sensitive to radiation that is incident perpendicular to the quantum wells [1]. To date, there have been several papers on InAs/GaAs, InGaAs/GaAs, and InGaAs/InGaP QDIPs [2]–[11]. Most of the devices employed a doped active region, which resulted in high dark current. In this paper, we report an InAs/GaAs QDIP with unintentionally doped active region and AlGaAs barrier layers. Our previous study on QDIPs with doped active region ( 2 electrons per quantum dot) show that the dark current is higher than QDIPs with an unintentionally doped active region. The AlGaAs layers act as blocking layers [6]–[11] for dark current, as first demonstrated in [6]. The devices reported here have demonstrated low dark current, low noise, and high detectivity. The InAs QDIP studied in this work belongs to the class of n-i-n structure QDIPs (Fig. 1) [6]–[8]. The samples were grown on semi-insulating GaAs (001) substrates by solid-source molecular beam epitaxy. Five layers of 3-monolayer (ML) InAs quantum dots were inserted between highly Si-doped top and bottom GaAs contact layers. The punctuated island growth technique was used to grow the quantum dots [12]. The GaAs spacer layers between the contact layers and the nearest quantum-dot layer had a thickness of 219–239 ML. 30-ML GaAs regions Manuscript received November 20, 2001; revised May 31, 2002. This work was supported by AFOSR under the MURI program. Z. Ye and J. C. Campbell are with Microelectronics Research Center, University of Texas at Austin, Austin, TX 78712 USA (e-mail: [email protected]). Z. Chen, E.-T. Kim, and A. Madhukar are with the Department of Materials Science and Physics, University of Southern California, Los Angeles, CA 90089 USA. Publisher Item Identifier 10.1109/JQE.2002.802159. Fig. 1. Schematic of InAs/GaAs QDIP structure. were used as the quantum-dot cap layers. In order to reduce the dark current, four pairs of AlAs/GaAs (1 ML/4 ML) were introduced below the quantum-dot layers and on the top of the GaAs cap layers. The size of the pyramidal-shaped quantum dots was estimated with atomic force microscopy (AFM) and a cross-sectional transmission electron microscope (XTEM): the height was 59 17 Å and the base width was 210 Å. The dot density was 625 40 / m . Device fabrication followed standard procedure: photolithography, wet chemical etching, metal deposition and lift-off, and rapid thermal annealing. Mesas having a diameter of 250 m and a height of 1.4 m were defined with an etch of H PO : H O L H O (8 : 1 : 61). A 50m-diameter top contact and the bottom contact were formed by evaporation and liftoff of Au/Ni/AuGe. The contacts were then annealed at 430 C for 20 s. In the following discussion, “positive” bias means that a positive voltage was applied to the top contact. The normal-incidence spectral response was measured with a Nicolet Magna-IR 570 Fourier transform infrared (FTIR) spectrometer and an SRS 570 low-noise current preamplifier. Fig. 2 shows the spectral response at 0.8-V bias and at temperatures of 63 K, 77 K, and 100 K. The intraband photoresponse peaks occurred at 6.2 m for all three spectra. The full-width at halfmaximum (FWHM) of the spectrum, , was 0.4 m, from which it follows that %. The narrow spectral width is consistent with our previous results [8], [13]. These results indicate that the electron transitions are intraband transitions from 0018-9197/02$17.00 © 2002 IEEE YE et al.: NORMAL-INCIDENCE InAs SELF-ASSEMBLED QUANTUM-DOT INFRARED PHOTODETECTORS 1235 Fig. 2. Normal incident photoresponse of the QDIP sample at the bias of 0.8 V and temperatures of 63 K, 77 K, and 100 K. Fig. 3. Peak responsivity at 77 K, 100 K, and 120 K. a lower bound state to a higher bound state [13]. The observed spectral width reflects the uniformity of the size of the quantum dots. The QDIP exhibits the highest photoresponse at 77 K. This can be explained as follows. As the temperature increases, more electrons occupy the lower states of the quantum dots. As long as there are unoccupied excited states available, the electrons in the lower states can participate in photon induced intraband transitions. However, a further increase in the number of electrons in the quantum dots, which results from the increase in dark current at higher temperature, will cause a decrease in the number of unoccupied excited states and, consequently, a decrease in the photoresponse. Additionally, a decrease in photo-excited electron lifetime at higher temperature can also result in a decrease in the photoresponse. The absolute spectral responsivity was calibrated with a blackbody source ( 995 K). Since the blackbody spectrum included near infrared radiation, which could result in interband transitions, in addition to midand long-wavelength photons, optical filters were placed next to the aperture of the blackbody to block radiation with wavelengths less than 3.5 m. Fig. 3 shows the peak spectral responsivity versus bias at temperatures of 77 K, 100 K, and 120 K. With an increase in positive bias, the responsivity increased from 0.33 mA/W at 0.1 V to 280 mA/W at 1.7 V. For negative bias, the responsivity increased near four orders of magnitude from 5.2 10 mA/W at zero bias to 418 mA/W at 1.6 V. Negative differential responsivity [8] was not observed within the bias range from 1.6 to 1.7 V. The different responsivity curves for the positive and negative bias are attributed to the asymmetric shape of the quantum dots along the growth direction and the wetting layers beneath the Fig. 4. Dark current density at temperature ranging from 60 K to 296 K. Fig. 5. Measured noise current (dots) at 77 K and 100 K, and calculated thermal noise current at 77 K. quantum dots. Consequently, electrons in the quantum dots experience different barrier heights, depending on whether transport is toward the top or bottom contacts. Dark current density versus voltage characteristics are shown in Fig. 4 for temperature in the range from 60 K to 296 K. The structural asymmetry of the quantum dots also results in asymmetrical dark current density for positive and negative bias. At low bias, the increase in dark current density is due to the fact that as the bias increases, more electrons occupy the quantum dots, which results in an increase in the average sheet electron density. When a large fraction of the quantum-dots states are occupied, further increase in bias does not significantly alter the sheet electron density. This causes a lowering of the energy barrier for injected electrons at the contact layers, which results in the nearly exponential increase of the dark current. At 0.7-V bias, the dark current density was 2.5 10 A/cm at 60 K. With increasing temperature, it increased over seven orders of magnitude to 11.1 A/cm at room temperature. Similarly, at 0.7-V bias, there was an increase of over eight orders of magnitude from 1.6 10 A/cm at 60 K to 14.4 A/cm at 296 K. Compared to a similar structure without the Al Ga As blocking layers, the dark current has been suppressed by over three orders of magnitude [8]. For bias 0.7 V and 100 K, the dark current increased exponentially with temperature, which suggests that in this temperature range, the dark current originates from thermionic emission. The calculated activation energy was 196 meV at zero bias, which was close to the energy corresponding to the cutoff wavelength (193 meV) of the sample. For temperatures lower than 100 K, sequential resonant tunneling and phonon-assisted tunneling are probably the dominant components of the dark current. 1236 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 38, NO. 9, SEPTEMBER 2002 Fig. 6. Peak detectivity at 77 K and 100 K. The noise current was characterized with low noise current preamplifiers and a SRS 760 fast Fourier transform spectrum analyzer. For V, the noise current was measured with a SRS current preamplifier. However, below 0.6 V, the photodetector noise current was below the noise floor of the instrument. Near zero bias, a low noise current preamplifier with high gain was used. However, restricted by the input power limitation of this current preamplifier, in the bias range from 0.1 to 0.5 V and 0.5 to 0.1 V, the noise current was interpolated. Fig. 5 shows the noise current of a 250m-diameter device at 77 K and 100 K. The calculated thermal noise current at 77 K is also shown. The thermal noise current can be expressed as , where is Boltzmann’s constant, is the absolute temperature, and is the differential resistance of the device, which was extracted from the dark current. At V, the calculated thermal noise current (3.2 10 A/Hz ) was very close to the measured noise current (2.9 10 A/Hz ), which indicates that thermal noise is significant in the low bias region. As the bias was increased, the noise current increased much faster than thermal noise. The detectivity is given by , where is the device area, is the responsivity, is the noise current, and is the bandwidth. Fig. 6 shows the peak detectivity versus bias at 77 K and 100 K. The best performance was achieved at 77 K and 0.7 V where the peak detectivity was 10 cm Hz /W. The corresponding responsivity was 14 mA/W. With increase in temperature to 100 K, the peak detectivity dropped to 1.1 10 cm Hz /W at 0.5 V, due to the decrease in responsivity and increase in noise current. In conclusion, we have demonstrated QDIPs based on bound-to-bound intraband transitions. These QDIPs were sensitive to normal-incident infrared radiation and exhibited a low dark current with cm Hz /W and mA/W at 0.7-V bias and 77 K. In contrast, the QDIPs with the same structure, except with a GaAs barrier layer, exhibited cm Hz /W at 77 K.