Spectrally Adaptive Nanoscale Quantum Dot Sensors

The potential use of nanotechnology for hyperspectral (HS) and multispectral (MS) sensing and imaging is described in this article. It is noted how HS and MS sensors/imagers have great potential for a variety of applications important to the intelligence community; such applications range from monitoring chemical-agent production to identifying geographical terrain. It is also noted that, by sensing the spectrum of reflectance/transmittance of the agents in different wavelength bands—as can be done with HS/MS systems— image-analysis capability and detection probability can be greatly improved. However, traditional MS/HS systems are fairly bulky and expensive. The work described concerns advances in nanotechnology that offer potential solutions to these drawbacks. In particular, the creation of spectrally adaptive focal-plane arrays, based upon nanoscale quantum dots (QDs) in the mid-infrared regime (3-14 μm), hold promise of producing compact and relatively inexpensive systems. Such sensors use electro-optics and, thus, do not involve moving parts. The spectral adaptability is mainly attributed to the quantum-confined Stark effect that results from the QDs being placed in asymmetrical quantum potential wells. As a result, QD detectors can sense information over different spectrally overlapping bands as the electrical bias applied across the detector is judiciously varied. Signal-processing based algorithms are then developed and utilized to maximally exploit the biasdependent and diverse spectral response of the QD detectors for the purpose of target-spectrum reconstruction and target classification. Examples of these applications are also given in this article. Advances in hyperspectral (HS) and multispectral (MS) sensing and imaging in the infrared (IR) spectrum have enabled numerous remote-sensing applications. Technical article Wiley Handbook of Science and Technology for Homeland Security Article ID: IS24 Page 2 These include military surveillance (i.e., target recognition, identification and classification), medical imaging (i.e., medical diagnosis), and monitoring geographical terrain, only to name a few. Conventional HS/MS systems offer spectral information of a scene (target or an agent) in a spectral band by sensing a wide range of narrow segments of the IR spectrum in a spectral range of interest. This can be achieved by using a broadband IR detector in conjunction with dispersive optics (e.g., a bank of IR optical filters) that can be utilized to specify the spectral bands to be sensed. Alternatively, multiple sensors, each sensitive to a designated spectral range, can be employed to sense a wide spectral range. However, either one of these complex conventional methods is of relatively large physical size and high cost. Nanoscale and spectrally adaptable sensors are emerging as a highly desirable alternative to conventional MS/HS sensing strategies that feature simplicity through its single-detector nature (or array of identically fabricated detectors) without requiring dispersive elements. To this end, a new class of IR photodetectors based on nanoscale epitaxial quantum-dots(QDs) have recently been proposed and developed [1,2]. A key feature of this technology is that it exploits inter-subband transitions between quantum-confined energy levels in a self-assembled dots-in-a-well (DWELL) structure in an InAs/GaAs/AlXGa1-XAs semiconductor material system [3]. Potential advantages of this detector technology are low dark current, high operating temperature, and notably bias-controlled tunability [4]. The quantumconfined Stark effect applied to the system comprising dots in an asymmetric well, results in a bias-dependent spectral response and also introduces a red shift (spectral shift) with significant spectral overlap [5] as the bias is varied in nominal range. Hence, a single photodetector can be operated as multiple detectors simply by applying different bias: the bias-dependent photocurrents of a single detector can be regarded as the outputs, resulting from spectrally overlapping bands. Recently, DWELL-based focal plane array (FPA) grown and processed at the Center for High Technology Materials (CHTM) at the University of New Mexico had successfully demonstrated multicolor sensing capability [6] in both mid-wave infrared (MWIR) and long-wave infrared (LWIR) regions. Figure 1 shows representative imagery showing the DWELL-based FPA’s capability to sense MWIR and LWIR radiation. In order to maximally exploit the features of bias-dependent and spectrally overlapping spectra from the DWELL photodetector, two sets of signalprocessing algorithms were developed and tested to further bring about the following two extended enabling functionalities that are based on post processing of data. The first is the capability for continuous spectral tuning [7-9], which enables a so-called DWELL-based algorithmic spectrometer; and the second is the capability for application-specific, optimal hyperspectral feature selection, which in turn, enables target recognition [13]. The rationale behind either one of these algorithms is to judiciously fuse multiple bias-dependent photocurrents from a single DWELL detector based on precise mathematical rules. Technical article Wiley Handbook of Science and Technology for Homeland Security Article ID: IS24 Page 3 In this article, we report the principles, fabrication and operation of the spectrally agile and bias-tunable DWELL photodetector. Device growth and processing are briefly reviewed, followed by results on device characterization. Device optimization for improving the DWELL’s operating temperature is also described. In addition, two key post-processing strategies for maximal data exploitation are also reviewed and analyzed: the DWELL-based algorithmic spectrometer and hyperspectral feature-selection for target recognition. Principle of Operation for DWELL Photodetectors A DWELL detector is a smart hybrid of conventional quantum-well (QW) and QD infrared photodetectors. In a heterostructure, InAs QDs are embedded in InGaAs-GaAs multiple QWs, shown in Fig.2 [9]. Just as conventional QD detectors, a DWELL detector is inherently sensitive to normal-incidence radiation and photons. Lower dark-current levels are expected since the ground state is lowered with respect to GaAs band edge. Longer intersubband relaxation times in a DWELL structure can achieve a relatively high detectivity [11]. In addition, the reduced thermionic emission inherent in the DWELL technology leads to higher operating temperatures. Due to the quantum-confined Stark effect, a biasdependent spectral response is evident depending upon the asymmetric electronic potential of a geometrically asymmetric DWELL structure. Two main attributes of this geometry are the shape of the dot and the different thicknesses of the QW above and below the dot, which together lead to variation of the local potential as a function of the applied bias. A DWELL detector could provide better control over the operating wavelength and nature of the allowable energy transitions (bound-to-bound, bound-to-quasi-bound and bound-to-continuum (barrier)) in Fig. 3 [9]. All the DWELL devices considered in this article were fabricated and characterized at CHTM. Brief Descriptions of Device Growth and Processing The DWELL structures were grown by V-80 molecular-beam epitaxy (MBE) system, with an As2 cracker source. An average of 2.4 monolayers of InAs dots were deposited on the sample with a rate of 0.053MLs. Then the dots were Sidoped at a level of 1-5x10/cm. The DWELL consists of 30 stacks of InAs/GaAs/AlGaAs heterostructures between two n GaAs contact layers. DWELL detectors were then processed using standard contact-lithography, plasma-etching, and metallization techniques in a class 100 clean-room environment. Each 400μm square n-i-n mesas with top pixel apertures, ranging from 25 to 300μm in diameter, were lithographically defined in the top metal contact [4,9]. Device Characterization Spectral response measurements were performed on single pixel InGaAsDWELL detectors with a Nicolet 870 FTIR (Fourier Transform Infrared) Technical article Wiley Handbook of Science and Technology for Homeland Security Article ID: IS24 Page 4 spectrometer and a Keithley 428 current-amplifier, which controls the electrical bias to the detectors. Multiple measurements of experimental photocurrents and dark currents are taken at different biases using a HP parameter analyzer. The bias-dependent spectral measurements and the corresponding experimental photocurrents of the DWELL detector are shown in Fig. 4 [9]. Due to a red shift (spectral shift), there exists two different peaks at LWIR region, one around 9.5μm with negative bias and the other at 10.5μm with positive bias. However, the drawback of the current DWELL is the limited operating temperature because of the dominance of the dark-current at higher device temperatures [9]. In Fig. 5 [9], the spectral response of the DWELL starts degrading remarkably as the operating temperature of the device exceeds 60K. Also the working range of applied bias becomes narrower. Especially at 77K, it is to be noted that no spectral variation is observed for the applied bias range. More details of the characterization can be found in [9]. Higher Operating Temperature DWELL (“Double DWELL”) Higher temperature operation is most crucial to reduce device size and cost since the required cooling system is bulky and expensive. By achieving up to near-room temperature levels, this DWELL photodetector can be more effective due to its tunability as compared to the current state of art detectors. To further increase the operating temperature, the present InAs quantum dots/InGaAs/GaAs/AlGaAs DWELL growth structure is further modified by the increase in the shoulder of the GaAs well and the addition of shoulders on both sides of the InGaAs well, so it becomes the complete double DWELL (DDWELL) structure. The optimizations of the growth procedure and the processing technique can potentially lower the dark current level to obtain higher operating temperature of the device. In Fig.6 [11,12], the bias-dependent DDWELL spectral response is observed until the device temperature of 120K and the spectral shift are still present in LWIR. However, at 120K, the device photocurrent is dominated by noise mainly due to the high dark current level. Description of the DWELL-based Algorithmic Spectrometer The concept of the spectral-tuning algorithm in conjunction with the DWELLbased detector is thoroughly reviewed here to describe the role of algorithmic spectrometer, drawing freely from our earlier published works [7,8]. Assume that the object (target) of interest is illuminated by a black-body broadband radiation source and a DWELL detector probes the illuminated object applying different electrical biases, producing a group of bias-dependent photocurrents. The goal is to exploit these bias-dependent photocurrents to estimate (approximately reconstruct) the spectrum of the object of interest without the utilization of any physical dispersive optics or a spectrometer. The spectral estimation procedure is described as follows. First, a series of hypothetical, narrowband tuning filters, each with a prescribed center wavelength and transmittance, are defined by Technical article Wiley Handbook of Science and Technology for Homeland Security Article ID: IS24 Page 5 sweeping across the desired center wavelength of narrowband tuning filter in a spectral region of interest. Second, a set of superposition weights are calculated by estimating the spectrum of a DWELL detector with the choice of the tuning filters. Third, for each defined filter, the spectral reconstruction of object is performed by forming a weighted superposition of the DWELL spectral responses with pre-determined superposition weights. With this step completed, the so-called “synthesized photocurrent”, defined as the target reconstruction with weights, is shown to best approximate the ideal photocurrent obtained by sensing the same object of interest using an ideal broadband (with a spectrally flat response) detector looking at the object through the spectral tuning filter. This approximation results in the minimization of the mean-square-error (MSE) between the synthesized photocurrent and the ideal response. Finally, the third step is repeated for every tuning wavelength and the spectrum is reconstructed within the prescribed wavelength range. In Fig. 7 [9], the conventional spectrometer with an ideal broadband IR detector and a group of optical IR filters is schematically compared with the proposed algorithmic spectrometer for describing their functional equivalence. Mathematical Description Mathematically, the reconstructed target spectrum ∧ n I λ (Eq. (1) in [9]) at a desired tuning wavelength n λ , is formulated as