Structural and optical characterization of infrared hot electron transistor

Hyeson Jung, Paul Pinsukanjana, Mitra Dutta, Kwong-Kit Choi, and Meimei Z. Tidrow Department of Electrical and Computer Engineering, University of Illinois at Chicago, Illinois 60607, USA Intelligent Epitaxy Technology, Inc., Richardson, Texas 75081, USA Department of Physics, University of Illinois at Chicago, Illinois 60607, USA U.S. Army Research Laboratory, Adelphi, Maryland 20783, USA U.S. Army Communications-Electronics Command, Night Vision and Electronic Sensors Directorate, Fort Belvoir, Virginia 22060, USA


I. INTRODUCTION
The research and development of long wave infrared ͑IR͒ detectors have included extrinsic doped germanium, Pb-SnTe, III-V semiconductor quantum well IR photodetectors ͑QWIPs͒, and HgCdTe. 1 One of the critical goals remains to be that of robust low cost detectors as well as those with room or near room temperature operation. 1,2Wide variety of QWIPs has been developed, and large format, high performance focal plane arrays have been made.Despite the well developed mature GaAs technology, relatively low operating temperature due to the high thermal generation rate of the QWIPs are under improvement.The performance of the QWIPs mainly depends on the dark current and responsivity.The thermal generation rate can be reduced by better detector design such as increasing a barrier height, 3 or placing a hot electron energy filter. 45][6][7] The key feature of an IHET is AlGaAs filter layer located between base contact and collector contact, 6 which acts as a high-pass filter for the photocurrent and blocks the tunneling dark current, with resulting satisfactory detectivity. 6,7The emitter and base supply an operating voltage to the QWIP.The filter is used to allow photoelectrons of certain energies into the collector and rejects the dark electrons from other lower energies.The rejected electrons will then drain through the base terminal.Many previous published experiments 2,[4][5][6][7] show that each individual QW is able to maintain a steady state of carrier distribution under properly adjusted bias over the full thickness of multiple wells which are composed of two distinct peaks generated by two independent excitation sources, thermal and optical.By accepting only electrons at particular energies, the filter reduces the dark current and increases the photocurrent to dark current ratio at the collector.Consequently, both the operating temperature and the responsivity of the detector can be increased.In this study, the dark current was calculated, and IHET structures were grown, and the structural, optical, and transport properties were investigated.

II. THEORY AND NUMERICAL CALCULATIONS
Before the describing the experimental results, it is useful to estimate the expected improvements.In QWIPs, the dark current, I d is dominated by the thermally assisted tunneling ͑TAT͒ current. 2 The TAT current originates from thermal excitation of the ground state electrons to yield a large in-plane energy, E in .Upon a scattering process, this large E in is transformed into a large E z perpendicular to the QWs, with which a larger tunneling probability ␥ results.Consequently, the TAT current is given by In Eq. ͑1͒, J͑V, T͒ is the measured dark current density, The current density per millielectron volt, ͓en TAT ͑E,V,T͒͑V͔͒ is the product of e, electron charge, n TAT ͑E,V,T͒ and ͑V͒ the electron drift velocity.The n TAT ͑E,V,T͒ is defined as the product of g 3D , the effective threedimensional density of states in the QW, f͑E, T͒ is the Fermi-Dirac energy distribution, and ␥͑E,V͒ the tunneling probability out of the QW.The value of ␥ is given by the usual Wentzel-Kramers-Brillouin ͑WKB͒ expression, and v͑V͒ is related to the low field mobility at low bias and the saturation velocity v sat at the high bias.Equation ͑1͒ satisfactorily explains the observed dark current, such as that of detectors A and B shown in Fig. 1͑a͒.There are several parameters in this equation such as the effective mass, m ‫ء‬ , the barrier height, the low field mobility, , and the saturated velocity.But since the barrier height in most cases is the same as that given by the Al mole fraction in the design of the structure and the m ‫ء‬ value and that of the saturated velocity are well known in GaAs and the latter is always taken to be equal to 1 ϫ 10 7 cm/ s in all the fittings, is the parameter that is used to fit the data as an adjustable parameter.
To obtain additional insights into the energies of the carriers, we plot the functions f͑E͒ and ␥͑E͒ at temperature T = 77 K and voltage V = 80 mV per QW period in Fig. 1͑b͒.Their product has a maximum just below the QW barrier height H͑V͒.The TAT current is, therefore, conducting just below the barriers, with only a small high energy tail above H͑V͒.Figure 1͑b͒ also shows the energy distribution of the photoelectrons, which is determined by the absorption spectra S͑E͒ of the detectors.At 80 mV, the photoelectron peak and the dark current peak are well separated in energy, especially in detector A. Since the dark current is conducting below H and the photocurrent is conducting above H, the two groups of electrons can be separated by an energy selective filter as depicted in Fig. 2. By attaching an additional barrier at one end of the QWIP ͑the base͒, one can regulate the passage of the electrons into a new collector terminal C according to their energies.With an appropriate collector bias V C , one can maximize the photocurrent to dark current ratio.
The IHET structure used is shown in Fig. 3.It is designed to have a 9.2 m wavelength cutoff.The calculated dark current transfer ratio, ␣ d , from the emitter to the collector is shown in Fig. 4͑a͒.The value of ␣ d is the fraction of dark electrons with energy above H͑V E ͒ that will not be filtered out by the filter barrier.This fraction of electrons is dependent on both V E and T. Similarly, the fraction of photoelectrons that has energy above H͑V E ͒ can be calculated from the absorption spectrum and the corresponding photocurrent transfer ratio, ␣ p , is shown in Fig. 4͑b͒.Its value depends on the emitter voltage only.After the preferential current filtering, the fraction of photocurrent I p versus the dark current I d in the total collector current I C is increased, which strengthens the signal to noise ratio.Numerically, the photocurrent to dark current ratio measured at the emitter, r E ϵ͑I p / I d ͒ E , and that measured at the collector r C , is related by r C = ͑␣ p / ␣ d ͒r E .Figure 4͑c͒ shows the value of ␣ p / ␣ d at different V and T. For example, at Ϫ8 V and 80 K, ␣ p / ␣ d is expected to be 5, and therefore, the S/N ratio will improve by the same factor by using the IHET structure.

III. EXPERIMENT
The epitaxial layers studied in the present work were grown on 4 in.GaAs ͑100͒ substrates by a molecular beam epitaxy which is equipped with an optical-based flux monitor and an absorption band-edge spectroscopy for monitoring the group III fluxes ͑Al, Ga, and In͒ and substrate temperature during the growth. 8,9Prior to the growth of the full In x Ga 1−x As/ Al y Ga 1−y As IHET structure which is almost 9 m thick, a In x Ga 1−x As/ GaAs QW superlattice ͑SL͒, a Al y Ga 1−y As bulk layer, and mini-IHET ͑ϳ2 m thick͒ were grown and characterized for the optimized growth condition calibration.The mini-IHET is a shortened version of the full structure IHET; 11 multi-QW ͑MQW͒ periods instead of 101.Silicon was the donor dopant with nominal concentration of 1 ϫ 10 18 cm −3 in the emitter layer, base layer, collector layer, and in the InGaAs wells, respectively.The MQW structure was sandwiched between a heavily doped GaAs layer ͑1 m thick͒ as the emitter contact and a heavily doped GaAs layer as the base layer ͑150 nm͒.On top of the base, there was an undoped AlGaAs filter layer ͑220 nm͒ followed by a heavily doped GaAs layer ͑200 nm͒ as the collector.The epitaxial layers were characterized by means of transmission electron microscopy ͑TEM͒, x-ray diffrac-FIG.2. The IHET structure with a thick barrier near the collector C as a high-pass filter.In this illustration, the higher energy photoelectrons created by optical transition is accepted into the collector while the lower energy TAT current is rejected into the base.tion ͑XRD͒, and photoluminescence ͑PL͒.The TEM experiments were performed with a JEOL JEM-3010, a 300 kV transmission electron microscope with a LaB 6 electron source and fitted with an ultrahigh resolution pole piece, which results in a lattice resolution of 0.14 nm.A sample was manually polished down to about 50 m, and then ionmilled for further thinning.The microscope is also fitted with a ThermoNoran Vista EDX system with a light element x-ray detector.XRD measurements were performed using Siemens Diffractometer D 5000 with Cu K ␣ − radiation ͑ = 1.5405Å͒.PL measurements were carried out using a single stage AC-TON SpectraPro 2500 spectrometer with macro-optics setup.HeNe laser was used for 632 nm excitation for the higher energy PL experiments.The maximum long wavelength range of the spectrometer is 1400 nm.The penetration depth of the 632 nm light is approximately 260 nm in GaAs, hence the top layer was etched to reveal the layers buried under top GaAs layer. 10Mixture of sulfuric acid, hydrogen peroxide and water ͑H 2 SO 4 ͒ : ͑H 2 O 2 ͒ : ͑H 2 O͒ in 1:1:25 proportion was chosen to etch the GaAs layer. 11A Horiba Jobin Yvon Fluo-roLog spectrofluorometer was used to measure fluorescence emission in the range between 800 nm to 3 m.

IV. RESULTS AND DISCUSSION
The cross-sectional TEM image of the QW mini-IHET structure of 3.5 nm/0.7 nm/50 nm In 0.1 Ga 0.9 As/ GaAs/ Al 0.21 Ga 0.79 As is shown clearly in Fig. 5, and each layer is marked in the figure.It was designed to have a target detection wavelength of ϳ9 m.The lattice images were observed for the electron beam incident along the ͓011͔ direction.The individual superlattice layers of eleven QWs, emitter layer, and the AlGaAs layer are well-defined.The dark area is the InGaAs, and bright area is the AlGaAs.As shown in the figure, the IHET structure consists of the absorptive QW superlattice, emitter, base, collector layers, and hot electron filter layer.The TEM images reveal the real lattice arrangement.A lattice image at higher magnification shows the perfect lattice match at In 0.1 Ga 0.9 As-GaAs-Al 0.21 Ga 0.79 As interfaces displayed in Fig. 6.The GaAs, InAs, and AlAs each with their zincblende structure, and their similar lattice constants as in Eq. ͑2͒, were epitaxial 12 a͑Al y Ga 1−y As͒ = 5.6533 + 0.0078y.͑2͒ Overall, TEM images shows that the IHET heterostructure is epitaxially well matched and no noticeable lattice defects were identified.
Figure 7͑a͒ shows XRD patterns of the IHET structure with the typical ͑200͒ and ͑400͒ reflections, associated with the substrate and MBE grown thin film, of IHET structure on GaAs͑100͒ substrate.For reference, XRD pattern of GaAs͑100͒ was obtained and plotted together; GaAs ͑400͒ reflection is observed for both Cu K␣ 1 and Cu K␣ 2 , and two unidentified peaks are found.The periodic structure of SLs give rise to satellite peaks in the vicinity of the ͑200͒ and ͑400͒ reflections. 13,14The peaks of IHET structure are shown in Fig. 7͑b͒ with the satellite marked with arrows.The main peak from the SL almost coincides with the reflection from the substrate and is the so-called order peak.Since Cu K ␣ − radiation was used, reflections from both Cu K␣ 1 and Cu K␣ 2 are observed.There are also more than sixteen satellite peaks including eleven on the lower angle side of the central Bragg peak, and five on higher angle side.On the higher angle side the peaks are rather distorted, due to the larger distance between Cu K␣ 1 and Cu K␣ 2 .The satellite peaks can be still be clearly observed.Using the formula below, the average layer thickness was extracted from the data, where d is the superlattice period ͑angstrom͒, l i,j is ith and jth diffraction order, i and j are Bragg diffraction angles, and is x-ray wavelength ͑1.5405 Å͒.Average thickness of period was found as 567.5 Å, while the target thickness was 549 Å including 500 Å of AlGaAs, 7 Å GaAs, 35 Å InGaAs, and 7 Å GaAs layer.These results indicate that the average width of QW is 1.47 Å thicker, less than one monolayer fluctuation than the nominal target width.Photoluminescence measurements were used to measure the energy level of the filter.The PL spectrum of the unetched IHET structure consists of a peak at 862 nm ͑1.439 eV͒ with full width at half maximum 76 meV, which agrees with the band gap of GaAs, is shown in Fig. 8͑a͒.Based on the penetration depth of the laser wavelength, the PL data is obtained only from the top GaAs collector layer ͑200 nm͒.The peak is shifted up in energy by 15 meV from the band gap ͑1.424 eV͒ of GaAs, due probably to the dopants and some amount of luminescence from the Al 0.19 Ga 0.81 As layer below.No other emission was observed.So the top GaAs layer was etched off to allow us to measure the composition of Al 0.19 Ga 0.81 As hot electron filter layer.Figure 8͑b͒ shows where y is Al composition. 10The calculated energy band gap of Al 0.19 Ga 0.81 As was 729 nm ͑1.699 eV͒ agrees well with the experimental data.The PL spectrum of one of the calibration layers, Al 0.21 Ga 0.79 As is the inset in Fig. 8͑b͒.measured energy band gap of Al 0.21 Ga 0.79 As is 719 nm ͑1.728 eV͒ agreeing well with the energy band gap of 718 nm ͑1.728 eV͒.In addition, a peak with emission at 1500 nm ͑0.83 eV͒ at the room temperature was observed as shown in Fig. 8͑c͒.Here the feature in Fig. 8͑c͒ was seen to be quite strong above the background, easily resolved as there are no other spectral features near it and is not instrumental in origin as it was observed on excitation with several wavelengths.
Engineering the AlGaAs hot electron filter is critical for the operation of the IHET. 4 The dark current filtered depends on the thickness of the filter layer and the energy level of the filter.The AlGaAs filter must be thick enough and high enough to block the dark current but too high of a filter height results in a reduction in photoinduced current.Emitter and collector spectral responses from a test array were measured at different V E and V C .The results are shown in Fig. 9.The emitter responsivity, which represents the response from the QWIP, matches the designed spectrum, confirming the QWIP material growth.The collector response R C is however about one order of magnitude less than the emitter response R E , showing that there is a large photocurrent reduction.Extrapolating the trend of R C versus V C , it would require a large V C of 2.5 V to capture ϳ80% of the photoelectrons.The apparent barrier height of the filter in the present material was higher than the designed value.The higher barrier height can be attributed to the finite p-type doping density in the material.For the level of background doping that gives rise to the apparent height of 1 V, the Poisson equation solution gives a very low doping level of 4 ϫ 10 16 / cm 3 , which is entirely possible according to the growth conditions.The presence of p-doing does not only raise the barrier height but also places the highest point of the barrier at the middle of the thick layer instead at the front.This increases the possibility of the photoelectrons to lose their energy before being selected.We can only say qualitatively that the higher sampling energy and the longer photoelectron traveling path make the large ␣ p reduction plausible.In Ref. 7, we had observed a photocurrent transfer ratio of 0.80.With the p-type background, we only see a ␣ p of 0.045 at 10 V in the present device.Therefore, there is a loss of efficiency of 18 times.But the dark current transfer ratio ␣ d is also lowered by the same p-doping.The value of ␣ d is 0.028 at 10 V. Therefore, we are still collecting more photoelectrons than dark electrons by 1.6 times.The unintentional p-doping lowered the electron carrier concentration substantially and reduced the intended n doping.It is this finite p-type dopant that is seen in Fig. 8͑c͒ in the near IR PL, showing a transition from a level ϳ40 meV below the barrier height to the energy level in the QWIP region confirming the existence of the finite p-type dopant.

V. CONCLUSION
We present here theoretical and experimental investigations of an IHET structure that was characterized by TEM, XRD, and PL.Measured photocurrent was seen to be significantly less than the theoretically calculated values for the specific design of the IHET device.This reduction in photocurrent was attributed to be finite p-type dopant in the filter barrier.However while there is a loss of efficiency of 18 times because of the p-doping background, we are still collecting more photoelectrons than dark electrons by 1.6 times.At 2 V, we are collecting six times more.As more photoelectrons are collected than dark electrons, the IHET will improve the signal to noise ratio and hence makes the detection more sensitive.A luminescence from an acceptor feature present in the band gap confirmed the presence of the p-type dopant.

FIG. 4 .
FIG.4.͑a͒ The calculated dark current transfer ratio and ͑b͒ the photocurrent transfer ratioand the photocurrent to dark current improvement factor ␣ p / ␣ d upon energy filtering.