Characterization of HgCdTe MWIR Back-Illuminated Electron-Initiated Avalanche Photodiodes

This article reports new characterization data for large-area (250 μm ×  250 μm) back-illuminated planar n-on-p HgCdTe electron-initiated avalanche photodiodes (e-APDs). These e-APDs were fabricated in p-type HgCdTe films grown by liquid-phase epitaxy (LPE) on CdZnTe substrates. We previously reported that these arrays exhibit gain that increases exponentially with reverse bias voltage, with gain-versus-bias curves that are quite uniform from element to element, and with a maximum gain of 648 at −11.7 V at 160 K for a cutoff wavelength of 4.06 μm. Here we report new data on these planar e-APDs. Data from a third LPE film with a longer cutoff wavelength (4.29 μm at 160 K) supports the exponential dependence of gain on cutoff wavelength, for the same bias voltage, that we reported for the first two films (with cutoffs of 3.54 μm and 4.06 μm at 160 K), in agreement with Beck’s empirical model for gain versus voltage and cutoff wavelength in HgCdTe e-APDs. Our lowest gain-normalized current density at 80 K and zero field-of-view is 0.3 μA/cm² at −10.0 V for a cutoff of 4.23 μm at 80 K. We report data for the temperature dependence of gain over 80 K to 200 K. We report, for the first time, the dependence of measured gain on junction area for widely spaced circular diodes with radii of 20 μm to 175 μm. We interpret the variation of measured gain with junction area in terms of an edge-enhanced electric field, and fit the data with a two-gain model having a lower interior gain and a higher edge gain. We report data for the excess noise factor F(M) near unity for gains up to 150 at 196 K. We describe the abrupt breakdown phenomenon seen in most of our devices at high reverse bias.     

Shockley–Haynes Characterization of Minority-Carrier Drift Velocity,Diffusion Coefficient,and Lifetime in HgCdTe Avalanche Photodiodes

A combined study of the avalanche gain characteristics of HgCdTe electron-avalanche photodiodes (e-APDs) and of the minority electron properties in the p-type absorber using Shockley–Haynes (SH) measurements is presented for various Cd compositions x _Cd. Ideal gain performance associated with a low excess noise factor F = 1.2 have been measured at T = 80 K down to cutoff wavelengths of λ _c = 2.9 μm. The observation of both a record high, exponentially increasing gain of M = 600 in short-wave e-APDs and a low excess noise factor proved that the exclusive electron multiplication is stable down to x _Cd = 0.4. Zero-flux measurements at 80 K confirmed that the dark current tends to decrease at constant gain as x _Cd increases. Measurements using a readout integrated circuit allowed us to establish a new record in sensitivity for APDs: I _{eq_in} = 2 aA, corresponding to 12 e/s at gain of M = 24 in an e-APD with λ _c = 2.9 μm. SH measurements enabled direct estimation of the electron diffusion coefficient, drift velocity, and lifetime in the p-type absorber of the e-APDs as a function of electric field at temperatures between 80 K and 200 K. Measurements at 80 K yielded lifetimes consistent with the values expected for the nominal doping of the samples. The low-field electron drift mobility, estimated from the drift velocity, was found to be a factor of 0.4 to 0.5 lower than the mobility in n-type material with the same composition. In mid-wave (MW) infrared samples with λ _c = 5.3 μm, the mobility was observed to be μ _ep = 15 kcm²/Vs to 20 kcm²/Vs, being less than μ _en ≈ 40 kcm²/Vs to 50 kcm²/Vs. The reduction in mobility can, in part, be attributed to scattering by ionized acceptors and heavy holes. The diffusion mobility, estimated from the diffusion coefficient, was systematically higher than the drift mobility, indicating diffusion of hot electrons with a temperature higher than that of the lattice. The saturation velocity, v _{sat_ep} = 2 × 10⁶ cm/s to 6 × 10⁶ cm/s, did not correlate with the Cd composition in the samples. The measured saturation velocities made it possible to estimate the timing jitter in p-type absorbers with a built-in electric field. Jitter below 100 ps was estimated for SW and MW APDs with absorbing layer thicknesses up to 4 μm.     

HgCdTe MWIR Back-Illuminated Electron-Initiated Avalanche Photodiode Arrays

This paper reports data for back-illuminated planar n-on-p HgCdTe electron-initiated avalanche photodiode (e-APD) 4 × 4 arrays with large unit cells (250 × 250 μm²). The arrays were fabricated from p-type HgCdTe films grown by liquid phase epitaxy (LPE) on CdZnTe substrates. The arrays were bump-mounted to fanout boards and characterized in the back-illuminated mode. Gain increased exponentially with reverse bias voltage, and the gain versus bias curves were quite uniform from element to element. The maximum gain measured was 648 at −11.7 V for a cutoff wavelength of 4.06 μm at 160 K. For the same reverse-bias voltage, the gains measured at 160 K for elements with two different cutoff wavelengths (3.54 μm and 4.06 μm at 160 K) show an exponential increase with increasing cutoff wavelength, in agreement with Beck’s empirical model for gain versus voltage and cutoff wavelength in HgCdTe e-APDs. Spot scan data show that both the V = 0 response and the gain at V = −5.0 V are spatially uniform over the large junction area. To the best of our knowledge, these are the first spot scan data for avalanche gain ever reported for HgCdTe e-APDs. Capacitance versus voltage data are consistent with an ideal abrupt junction having a donor concentration equal to the indium concentration in the LPE film. U.S. Workshop on the Physics and Chemistry of II-VI Materials Newport Beach, California October 10–12, 2006.     

Experimental Performance and Monte Carlo Modeling of Long Wavelength Infrared Mercury Cadmium Telluride Avalanche Photodiodes

We evaluated the performance of long-wavelength infrared (LWIR, λ _c = 9.0 μm at 80 K) mercury cadmium telluride electron-injected avalanche photodiodes (e-APDs) in terms of gain, excess noise factor, and dark current, and also spectral and spatial response at zero bias. We found an exponential gain curve up to 23 at 100 K and a low excess noise factor close to unity (F = 1–1.25). These properties are indicative of a single carrier multiplication process, which is electron impact ionization. The dark current is prevailed by a diffusion current at low reverse bias. However, tunneling currents at higher reverse bias limited the usable gain. The measurements of the pixel spatial response showed that the collection width, and, especially, the amplitude of the response peak, increased with temperature. Furthermore, we developed a Monte Carlo model to understand the multiplication process in HgCdTe APDs. The first simulation results corroborated experimental measurements of gain and excess noise factor in mid-wavelength infrared (MWIR, x = 0.3) and LWIR (x = 0.235) e-APDs at 80 K. This model makes it possible for phenomenological studies to be performed to identify the main physical effects and technological parameters that influence the gain and excess noise. The study of the effect of the n ⁻-layer thickness on APD performance demonstrated the existence of an optimum value in terms of gain.     

Gated IR Imaging with 128 × 128 HgCdTe Electron Avalanche Photodiode FPA

The next generation of infrared (IR) sensor systems will include active imaging capabilities. One example of such a system is a gated active/passive system. The gated active/passive system promises target detection and identification at longer ranges compared to conventional passive-only imaging systems. A detector that is capable of both active and passive modes of operation opens up the possibility of a self-aligned system that uses a single focal plane. The mid-wave infrared (MWIR) HgCdTe electron injection avalanche photodiode (e-APD) provides state-of-the-art 3 μm to 5 μm performance for the passive mode and high, low-noise, gain in the active mode, and high quantum efficiency at 1.5 μm. Gains of greater than 1000 have been measured in MWIR e-APDs with a gain-independent excess noise factor of 1.3. This paper reports the application of the mid-wave HgCdTe e-APD for near-IR gated-active/passive imaging. Specifically a 128 × 128 focal-plane array (FPA) composed of 40-μm-pitch MWIR cutoff APD detectors and custom readout integrated circuit was designed, fabricated, and tested. Median gains as high as 946 at 11 V bias with noise equivalent photon inputs as low as 0.4 photon were measured at 80 K and 1 μs gate times. This subphoton sensitivity is consistent with the high gains, low excess noise factor, and low effective gain normalized dark-current densities, near or below 1 nA/cm², that were achieved in these FPAs. A gated imaging demonstration system was designed and built using commercially available parts. High resolution and precision gating was demonstrated in this system by imagery taken at ranges out to 9 km.     

0.8 V bulk-driven operational amplifier

A low-voltage bulk-driven CMOS operational amplifier is proposed in this paper. The inherent small transconductance of the bulk-driven devices is enlarged using a positive feedback, improving also the noise performance. The amplifier is designed using standard 0.18 μm n-well CMOS process. Although the amplifier is optimized for 0.8 V supply voltage, it is also capable to operate under supply voltage of 0.7 V. The amplifier consumes 130 μΑ, performing 56 dB open-loop gain, 154 nV/√Hz input-referred spot noise at 100 kHz, 80 dB CMRR at 100 kHz and IIP3 equal to −4.7 dBV.     

Latest Developments of HgCdTe e-APDs at CEA LETI-Minatec

In this communication, we report on the electro-optical characterization of planar back-side-illuminated HgCdTe electron initiated avalanche photodiode (e-APD) test arrays with cut-off wavelengths λ _c = 2.4 μm, λ _c = 4.8 μm, and λ _c = 9.2 μm. The e-APDs were manufactured at LETI using absorption layers grown by molecular beam epitaxy (MBE). We present measurements of the distributions in gain, noise, and equivalent input dark current. The mid-wave (MW) diodes yielded a very low dispersion (2%) and high operability (98%) for gains up to M = 200. The excess noise factor and equivalent input current (I _{eq_in}) operability were slightly lower, due to defects in the depletion region. The lowest measured value of I _{eq_in} = 1 fA corresponds to the lowest level measured so far in HgCdTe e-APDs and opens the way to new applications. The gain in the long-wave (LW) diodes was limited by tunnelling currents to a value of M = 2.4, associated with an average noise factor F = 1.2. A gain of M = 20 at a bias of −22.5 V was demonstrated in the short-wave (SW) e-APDs.     

X-Ray Diffraction and Photoluminescence Studies of InN Grown by Plasma-Assisted Molecular Beam Epitaxy with Low Free-Carrier Concentration

We report studies of InN grown by plasma-assisted molecular beam epitaxy. GaN templates were first grown on sapphire substrates followed by InN overgrown at 457°C to 487°C. Atomic force microscopy shows the best layers to exhibit step-flow growth mode of the InN, with a root-mean-square roughness of 0.7 nm for the 2 μm × 2 μm scan and 1.4 nm for the 5 μm × 5 μm scan.␣Measurements of the terrace edges indicate a step height of 0.28 nm. Hall measurements at room temperature give mobilities ranging from 1024 cm²/V s to 1904 cm²/V s and the electron concentrations are in the range of 5.9 × 10¹⁷ cm⁻³ to 4.2 × 10¹⁸ cm⁻³. Symmetric and asymmetric reflection x-ray diffraction measurements were performed to obtain lattice constants a␣and c. The corresponding hydrostatic and biaxial stresses are found to range from −0.08 GPa to −0.29 GPa, and −0.05 GPa to −0.32 GPa, respectively. Low-temperature photoluminescence peak energies range from 0.67 eV to 0.70 eV, depending on residual biaxial stress, hydrostatic pressure, and electron concentrations. The electron concentration dependence of the estimated Fermi level is analyzed using Kane’s two-band model and conduction-band renormalization effects.     

Impulse Response Time Measurements in Hg<Subscript>0.7</Subscript>Cd<Subscript>0.3</Subscript>Te MWIR Avalanche Photodiodes

The response time of front-sided illuminated n-on-p Hg_0.7Cd_0.3Te electron avalanche photodiodes (e-APDs) at T = 77 K was studied using impulse response measurements at λ = 1.55 μm. We measured typical rise and fall times of 50 ps and 800 ps, respectively, at gains of M ≈ 100, and a record gain-bandwidth (GBW) product of GBW = 1.1 THz at M = 2800. Experiments as a function of the collection width have shown that the fall time is strongly limited by diffusion. Variable-gain measurements showed that the impulse response is first-order sensitive to the level of the output amplitude. Only a slight increase in the rise time and the fall time was observed with the gain at constant output amplitude, which is consistent with a strongly dominant electron multiplication. Comparisons of the experimental results with Silvaco finite element simulations confirmed the diffusion limitation of the response time and allowed the illustration of the transit time and RC effects.     

Epitaxial Lead Chalcogenides on Si for Mid-IR Detectors and Emitters Including Cavities

Lead chalcogenide (IV–VI narrow-gap semiconductor) layers on Si or BaF₂(111) substrates are employed to realize two mid-infrared optoelectronic devices for the first time. A tunable resonant cavity enhanced detector is realized by employing a movable mirror. Tuning is across the 4 μm to 5.5 μm wavelength range, and linewidth is <0.1 μm. Due to the thin (0.3 μm) PbTe photodiode inside the cavity, a higher sensitivity at higher operating temperatures was achieved as compared to conventional thick photodiodes. The second device is an optically pumped vertical external-cavity surface-emitting laser with PbTe-based gain layers. It emits at ∼5 μm wavelength and with output power up to 50 mW pulsed, or 3 mW continuous wave at 100 K.     

HgCdTe Quantum Detection: from Long-Wave IR down to UV

A few years ago, visible detection was demonstrated using advanced substrate thinning processes on flip-chip hybridized HgCdTe focal-plane arrays, in both French and US laboratories. Constant quantum efficiency was demonstrated at LETI-Sofradir from the short-wave infrared (IR) (2.5 μm cut-off) down to the visible range in 2006, validating complete CdZnTe substrate removal. This paper presents and discusses HgCdTe photodiode spectral response characterization, focusing on the short-wavelength part of the spectrum. We confirm the extended sensitivity of middle- and long-wave diodes: constant quantum efficiency has been observed from 10 μm down to 230 nm in the ultraviolet (UV). Such a unique property may be useful for very large-bandwidth spectrometers requiring monolithic detectors. Avalanche gain of middle-wave avalanche photodiodes has also been investigated in both the visible and the UV range. We demonstrate here that the avalanche gain remains constant while keeping a very low excess noise factor. This result opens the way to low-flux applications in this wavelength range.     

Study of Spectroscopy and Microstructure in Nanocrystalline Cr-Doped Fibers Grown by the Drawing-Tower Technique

The spectroscopy and microstructure of Cr-doped fibers (CDFs) fabricated by the drawing-tower technique are investigated. The spontaneous emission spectrum exhibited broadband emission of 1.2 μm to 1.55 μm. High-resolution transmission electron microscope images showed that there were nanocrystalline structures in the core, surrounded by an amorphous matrix of SiO₂; they also revealed that the nanoparticle density was about 3.2 × 10³ μm⁻² near the core/clad interface, and 1.5 × 10⁴ μm⁻² near the core center. The results indicate preliminary success in fabricating nanocrystalline CDF with low transmission loss and crystal-like active properties.     

Study of the Au/In Reaction for Transient Liquid-Phase Bonding and 3D Chip Stacking

The latest three-dimensional (3D) chip-stacking technology requires the repeated stacking of additional layers without remelting the joints that have been formed at lower levels of the stack. This can be achieved by transient liquid-phase (TLP) bonding whereby intermetallic joints can be formed at a lower temperature and withstand subsequent higher-temperature processes. In order to develop a robust low-temperature Au/In TLP bonding process during which all solder is transformed into intermetallic compounds, we studied the Au/In reaction at different temperatures. It was shown that the formation kinetics of intermetallic compounds is diffusion controlled, and that the activation energy of Au/In reaction is temperature dependent, being 0.46 eV and 0.23 eV for temperatures above and below 150°C, respectively. Moreover, a thin Ti layer between Au and In was found to be an effective diffusion barrier at low temperature, while it did not inhibit joint formation at elevated temperatures during flip-chip bonding. This allowed us to control the intermetallic formation during the distinct stages of the TLP bonding process. In addition, a minimal indium thickness of 0.5 μm is required in order to enable TLP bonding. Finally, Au/In TLP joints of ∅40 μm to 60 μm were successfully fabricated at 180°C with very small solder volume (1 μm thickness).     

Interband Cascade Lasers with Wavelengths Spanning 2.9 μm to 5.2 μm

We report an experimental investigation of four interband cascade lasers with wavelengths spanning the mid-infrared spectral range, i.e., 2.9 μm to 5.2 μm, near room temperature in pulsed mode. One broad-area device had a pulsed threshold current density of only 3.8 A/cm² at 78 K (λ = 3.6 μm) and 590  A/cm² at 300 K (λ = 4.1 μm). The room-temperature threshold for the shortest-wavelength device (λ = 2.6 μm to 2.9 μm) was even lower, 450 A/cm². A␣cavity-length study of the lasers emitting at 3.6 μm to 4.1 μm yielded an internal loss varying from 7.8 cm⁻¹ at 78 K to 24 cm⁻¹ at 300 K, accompanied by a decrease of the internal efficiency from 77% to 45%.     

AlGaSb Buffer Layers for Sb-Based Transistors

InAs quantum wells can serve as the channel for high-electron-mobility transistors. Structures are typically grown on semi-insulating GaAs substrates with 1.5 μm to 3.0 μm buffer layers of AlSb and AlGaSb accommodating the lattice mismatch. We demonstrate that high electron mobility in the InAs (>20,000 cm²/V s at 300 K) and smooth surfaces can be achieved with Al_0.8Ga_0.2Sb buffer layers as thin as 600 nm, grown at rates of 1.5 monolayers/s to 2.0 monolayers/s. The use of thinner buffer layers reduces molecular beam epitaxial growth time and source consumption. The buffer layers also exhibit higher resistivity, which should reduce excess gate leakage current and improve device isolation.     

Interaction Between Ni and Cu Across 95Pb-5Sn High-Lead Layer

Ni/95Pb-5Sn/Cu ternary diffusion couples were used to investigate the cross-interaction between Ni and Cu across a layer of 95Pb-5Sn solder. High-lead solder layers with a thickness of 100 μm or 400 μm were electroplated over Cu foils. A pure Ni layer (20 μm) was then deposited over the as-deposited high-lead solder surface. The diffusion couples were then aged at 150°C to 250°C for different periods of time. With this technique, the diffusion couples were assembled without experiencing any high-temperature process such as reflow, which would have accelerated the interaction and caused difficulties in analysis. This study revealed that massive spalling also occurred during aging even though reflow was not used. The massive spalling began with the formation of microvoids. When the microvoids had congregated into large enough voids, intermetallic compounds (Cu₃Sn) started to spall from the interface. This spalling phenomenon occurred sooner with increasing temperature and decreasing solder volume.     

Study of the Transit-Time Limitations of the Impulse Response in Mid-Wave Infrared HgCdTe Avalanche Photodiodes

The impulse response in frontside-illuminated mid-wave infrared HgCdTe electron avalanche photodiodes (APDs) has been measured with localized photoexcitation at varying positions in the depletion layer. Gain measurements have shown an exponential gain, with a maximum value of M = 5000 for the diffusion current at a reverse bias of V _b = 12 V. When the light was injected in the depletion layer, the gain was reduced as the injection approached the N+ edge of the junction. The impulse response was limited by the diode series resistance–capacitance product, RC, due to the large capacitance of the diode metallization. Hence, the fall time is given by the RC constant, estimated as RC = 270 ps, and the rise time is due to the charging of the diode capacitance via the transit and multiplication of carriers in the depletion layer. The latter varies between t _10–90 = 20 ps (at intermediate gains M < 500) and t _10–90 = 70 ps (at M = 3500). The corresponding RC-limited bandwidth is BW = 600 MHz, which yields a new absolute record in gain–bandwidth product of GBW = 2.1 THz. The increase in rise time at high gains indicates the existence of a limit in the transit-time-limited gain–bandwidth product, GBW = 19 THz. The impulse response was modeled using a one-dimensional deterministic model, which allowed a quantitative analysis of the data in terms of the average velocity of electrons and holes. The fitting of the data yielded a saturation of the electron and hole velocity of v _e = 2.3 × 10⁷ cm/s and v _h = 1.0 × 10⁷ cm/s at electric fields E > 1.5 kV/cm. The increase in rise time at high bias is consistent with the results of Monte Carlo simulations and can be partly explained by a reduction of the electron saturation velocity due to frequent impact ionization. Finally, the model was used to predict the bandwidth in diodes with shorter RC = 5 ps, giving BW = 16 GHz and BW = 21 GHz for x _j = 4 μm and x _j = 2 μm, respectively, for a gain of M = 100.     

A 5.4-GHz high-<Emphasis Type="Italic">Q</Emphasis> tunable active-inductor bandpass filter in standard digital CMOS technology

A transistor-only CMOS active-inductor with an all-NMOS signal path is presented. By tuning the varactor-augmented parasitic capacitance at the only internal node the circuit losses from submicron MOSFETs can be partially or fully compensated to permit realizing unlimited values of Q, with little frequency and no power-consumption penalties. Transistor-only second-order bandpass filters using the active inductor were built in the TSMC 0.18-μm CMOS process, and high filter Q was obtained by tuning the varactor. The highest center frequency measured was f ₀ = 5.7 GHz for 0.2-μm gate lengths and the maximum repeatably measured Q was 665. Lower Qs can be obtained by reducing the capacitive compensation or by adjusting the circuit biasing. f ₀ and Q are tunable via separate varactors. IIP ₃ and input 1-dB compression point were simulated as 0.523 V_PP and 0.128 V_PP (−1.65 and −13.9 dBm from a 50-Ω source) at 5.7 GHz with Q = 100 and midband gain equal 4.7 dB. For the same conditions, the output noise and noise figure (R _S = 50 kΩ) were simulated to be 0.8 μV/Hz^1/2 and 25.6 dB, respectively. The filter core occupies an area of 26.6 μm × 30 μm and dissipates 4.4 mW at 5.4 GHz from a 1.8-V power supply. As the circuits use only MOSFETs they are fully compatible with standard digital CMOS processes. f ₀ statistics were obtained by measuring 40 chips at identical biasing condition.     

4-Bit Parallel-Input Exponential Digital-to-Analog Converter in CMOS 0.18 μm Technology

In this paper, a new digitally controlled linear-in-dB CMOS variable gain amplifier is proposed. The circuit employs the proposed novel approach in achieving a wide-range true-exponential transfer function e ^2X using a traditional pseudo-exponential amplifier followed by a variable gain stage, to expand the output dynamic range. A single digitally controlled variable resistor is used to tune the circuit accordingly by controlling X with a digital word. The result is a digitally controlled data conversion that yields a new type of non-linear digital-to-analog converter. Finally, a 4-bit converter is implemented in a TSMC 0.18 μm CMOS technology and displays a gain from about −21 dB to 36 dB in steps of 3.89 dB with an output linear error in [−0.66,0.45] dB and a static power consumption of 2.34 mW.