Effect of Dislocations on VLWIR HgCdTe Photodiodes

The effects of dislocations on very-long-wavelength infrared (VLWIR) HgCdTe photodiodes (cutoff wavelength >14 μm at 40 K) have been determined experimentally and analyzed. The photodiodes are in the back-illuminated configuration, fabricated from HgCdTe p-on-n double-layer heterostructure (DLHJ) films grown at BAE Systems by liquid phase epitaxy (LPE) onto lattice-matched (111) CdZnTe substrates. Arrays were hybridized to silicon ROICs to form focal plane arrays (FPAs). After characterization for dark current and response, the arrays were dehybridized and stripped of their metals and passivation layers. Dislocations were revealed using a Hähnert and Schenk (H&;S) etch. Pixel traceability was maintained throughout the analysis, permitting one-to-one correlation between photodiode performance and dislocation density measured within that photodiode. We found that response and dark current were correlated to etch pit density (EPD), which we assumed to be equal to dislocation density. Our results support earlier dislocation studies on larger-bandgap HgCdTe, which showed response was only weakly impacted by EPD, while dark current was strongly affected by EPD. Measured EPD values ranged from low 10⁵ to low 10⁷ cm^?2. Potential causes for this range in EPD are discussed.     

Dislocation profiles in HgCdTe(100) on GaAs(100) grown by metalorganic chemical vapor deposition

We studied dislocation etch pit density (EPD) profiles in HgCdTe(lOO) layers grown on GaAs(lOO) by metalorganic chemical vapor deposition. Dislocation profiles in HgCdTe(lll)B and HgCdTe(lOO) layers differ as follows: Misfit dislocations in HgCdTe(lll)B layers are concentrated near the HgCdTe/CdTe interfaces because of slip planes parallel to the interfaces. Away from the HgCdTe/CdTe interface, the HgCdTe(111)B dislocation density remains almost constant. In HgCdTe(lOO) layers, however, the dislocations propagate monotonically to the surface and the dislocation density decreases gradually as dislocations are incorporated with increasing HgCdTe(lOO) layer thicknesses. The dislocation reduction was small in HgCdTe(lOO) layers more than 10 μm from the HgCdTe/CdTe interface. The CdTe(lOO) buffer thickness and dislocation density were similarly related. Since dislocations glide to accommodate the lattice distortion and this movement increases the probability of dislocation incorporation, incorporation proceeds in limited regions from each interface where the lattice distortion and strain are sufficient. We obtained the minimum EPD in HgCdTe(100) of 1 to 3 x 10⁶ cm^-2 by growing both the epitaxial layers more than 8 μm thick.     

Photo-induced excess low frequency noise in HgCdTe photodiodes

We have investigated the properties of excess low frequency noise in illuminated mid wavelength infrared and long wavelength infrared HgCdTe photodiodes at zero bias. The current power spectrum (S_i) dependence is usually close to inverse frequency (f), but substantial variations have been observed. The magnitude of l/f spectra is voltage independent for small bias voltages, but is proportional to the square of the photocurrent (I). Consequently, the l/f knee increases, with photocurrent. Variable area device studies indicate that the noise sources are more closely associated with the device area (A_j) than perimeter, indicating bulk limitations. The power spectrum can be represented by an empirical relationship of the form S_i=α_phI ²/fA_j. This defines a figure of merit, α_ph which takes into the account the relationship between current dependence and device geometry. α_ph is device dependent, suggesting that randomly distributed defects play a role in the difference. This is also supported by noting that devices fabricated in material grown on lattice matched substrates have lower α_ph (10⁻¹⁶ cm²) than those fabricated in material grown on nonlattice matched substrates (10⁻¹⁴ cm²), which usually have two orders of magnitude larger dislocation density. We conclude that photo-induced l/f noise can be reduced via defect reduction and is not fundamental. Data on our best devices indicates that α_ph is somewhat lower for smaller band gap material. The temperature dependence of photo-induced excess low frequency noise is much weaker than that of bias induced excess low frequency noise, indicating unrelated generation mechanisms. In addition, photo-induced l/f adds in quadrature with bias induced l/f noise and is not well correlated in magnitude with either bias induced l/f noise or detector dark currents.     

1/f noise in HgCdTe photodiodes

1/f noise in HgCdTe photodiodes has been attributed to a variety of sources, most of which are associated with some form of excess current. At DRS, we have measured the 1/f noise in vertically integrated (VIP) and high-density vertically integrated photodiodes (HDVIP), over a wide range of compositions and temperature, for strictly well-behaved diffusion current limited operation. It is found that (1) the 1/f noise current is directly dependent on dark current density; (2) material composition and temperature are irrelevant, except in as much as they determine the magnitude of the current density; (3) in high-quality diodes, the 1/f noise is independent of background flux; and (4) surface passivation is relevant. These observations have been compared to the 1/f noise theory of Schiebel, which uses McWhorter’s fluctuation of the surface charge tunneling model to modulate diode diffusion current. Agreement is obtained with Schiebel’s theory for realistic surface trap densities in the 10¹²/cm² range, which will obviously be characteristic of the passivation used. The relevance of this work relative to high operating temperature phtodiodes is discussed.     

MWIR DLPH HgCdTe photodiode performance dependence on substrate material

Mid wavelength infrared p-on-n double layer planar heterostructure (DLPH) photodiodes have been fabricated in HgCdTe double layers grown in situ by liquid phase epitaxy (LPE), on CdZnTe and for the first time on CdTe/sapphire (PACE-1). Characterization of these devices shed light on the nature of the material limits on device performance for devices performing near theoretical limits. LPE double layers on CdZnTe and on PACE-1 substrates were grown in a horizontal slider furnace. All the photodiodes are p-on-n heterostructures with indium as the n-type dopant and arsenic the p-type dopant. Incorporation of arsenic is via implantation followed by an annealing step that was the same for all the devices fabricated. The devices are passivated with MBE CdTe. Photodiodes have been characterized as a function of temperature. R₀A_imp values obtained between 300 and 78K are comparable for the two substrates and are approximately a factor of five below theoretical values calculated from measured material parameters. The data, for the PACE-1 substrate, indicates diffusion limited performance down to 110K. Area dependence gives further indications as to the origin of diffusion currents. Comparable R₀A_imp for various diode sizes indicates a p-side origin. R₀A and optical characteristics for the photodiodes grown on lattice-matched CdZnTe substrates and lattice mismatched PACE-1 are comparable. Howover, differences were observed in the noise characteristics of the photodiodes. Noise was measured on 50 × 50 μm devices held under a 100 mV reverse bias. At 110K, noise spectrum for devices from the two substrates is in the low 10⁻¹⁵ A/Hz^1/2 range. This value reflects the Johnson noise of the room temperature 10¹⁰ Ω feedback resistor in the current amplifier that limits the minimum measurable noise. Noise at 1 Hz, −100 mV and 120K for the 4.95 μm PACE-1 devices is in the 1–2 × 10⁻¹⁴ A/Hz^1/2, a factor of 5–10 lower than previously grown typical PACE-1 n⁺-on-p layers. Noise at 120K for the 4.60 μm PACE-1 and LPE on CdZnTe was again below the measurement technique limit. Greatest distinction in the noise characteristics for the different substrates was observed at 163K. No excess low frequency noise was observed for devices fabricated on layers grown by LPE on lattice-matched CdZnTe substrates. Photodiode noise measured at 1Hz, −100 mV and 163K in the 4.60 μm PACE-1 layer is in the 1–2×10⁻¹³ A/Hz^1/2, again a factor of 5–10 lower than previously grown PACE-1 n⁺-on-p layers. More variation in noise (4×10⁻¹³−2×10⁻¹² A/Hz^1/2) was observed for devices in the 4.95 μm PACE-1 layer. DLPH devices fabricated in HgCdTe layers grown by LPE on lattice-matched CdZnTe and on lattice-mismatched PACE-1 have comparable R₀A and quantum efficiency values. The distinguishing feature is that the noise is greater for devices fabricated in the layer grown on lattice mismatched substrates, suggesting dislocations inherent in lattice mismatched material affects excess low frequency noise but not zero bias impedance.     

Planar p-on-n HgCdTe heterojunction mid-wavelength infrared photodiodes formed using plasma-induced junction isolation

Planar p-on-n HgCdTe heterojunction photodiodes have been fabricated using a plasma-induced type conversion process for device junction isolation. The technique is presented as a fully planar alternative technology to the commonly used mesa isolation structure. The starting material consisted of an indium-doped n-type mid-wavelength infrared (MWIR) HgCdTe absorbing layer that was capped by a 1-μm-thick wider bandgap arsenic-doped p-type layer. Junction-isolated p-on-n diodes were formed by selectively p-to-n type converting the p-type cap layer using a plasma process. Photodiode dark current-voltage measurements were performed as a function of temperature, along with noise and responsivity. The devices have cut-off wavelengths between 4.8 μm and 5.0 μm, exhibit diffusion-limited dark currents down to 145 K, give R₀A values greater than 1 × 10⁷Ω·cm² at 80 K and greater than 1 × 10⁵Ω·cm² at 120 K, and have negligible 1/f noise current at zero applied bias.     

Heteroepitaxy of HgCdTe(112) infrared detector structures on Si(112) substrates by molecular-beam epitaxy

High-quality, single-crystal epitaxial films of CdTe(112)B and HgCdTe(112)B have been grown directly on Si(112) substrates without the need for GaAs interfacial layers. The CdTe and HgCdTe films have been characterized with optical microscopy, x-ray diffraction, wet chemical defect etching, and secondary ion mass spectrometry. HgCdTe/Si infrared detectors have also been fabricated and tested. The CdTe(112)B films are highly specular, twin-free, and have x-ray rocking curves as narrow as 72 arc-sec and near-surface etch pit density (EPD) of 2 × 10⁶ cm⁻² for 8 μm thick films. HgCdTe(112)B films deposited on Si substrates have x-ray rocking curve FWHM as low as 76 arc-sec and EPD of 3-22 × 10⁶ cm⁻². These MBE-grown epitaxial structures have been used to fabricate the first high-performance HgCdTe IR detectors grown directly on Si without use of an intermediate GaAs buffer layer. HgCdTe/Si infrared detectors have been fabricated with 40% quantum efficiency and R₀A = 1.64 × 10⁴ Ωm₂ (0 FOV) for devices with 7.8 μm cutoff wavelength at 78Kto demonstrate the capability of MBE for growth of large-area HgCdTe arrays on Si.     

A comparison of gamma radiation effects on bromine- and hydrazine-treated HgCdTe photodiodes

In this study, we investigated the effects of gamma radiation on ZnS/CdTe-passivated HgCdTe photodiodes that were fabricated with one of two different surface treatments using bromine, Br₂, or hydrazine, N₂H₄. Unlike the ZnS-passivated HgCdTe photodiodes, the ZnS/CdTe-passivated HgCdTe photodiodes showed no degradation in resistance-area product at zero bias (R₀A) values after gamma irradiation of up to 1 Mrad. However, there is a significant difference between the bromine- and hydrazine-treated samples. Regardless of the dose of gamma radiation, there was little change in the forward current characteristics of the hydrazine-treated diode in comparison with the conventional bromine-treated diode. The hydrazine-treated diode showed fairly uniform R₀A values of >10⁷ Ω-cm² up to 1 Mrad of gamma irradiation, whereas the bromine-treated diode showed an abrupt change in R₀A values from ∼10⁶ Ω-cm² to ∼10⁷ Ω-cm² after gamma irradiation. Therefore, for use in a gamma radiation environment, the best combination for high-performance HgCdTe photodiodes is a ZnS/CdTe passivant that has been treated with hydrazine.     

On 1/f noise characteristics for 0.1 eV HgCdTe photoconductors

Sensitivities on 0.1 eV HgCdTe photoconductors with new electrode configuration and in different sizes were measured at 77K and under 10¹⁴phcm^?2s^?1 photon background conditions. After data for responsivity, generation-recombination noise (g-r noise) and minority carrier lifetime were reproduced by solving a one dimensional diffusion equation on excess minority carrier, discussions on 1/f noise were made and the following characteristics were concluded: (1) 1/f noise does not originate near electrodes for bias current. (2) 1/f noise hardly depends on sensor size and temperature in the 77–95K range, while g-r noise does. (3) 1/f noise is proportional to bias electric field, i.e. current density. (4) 1/f noise does not depend on photon background. From characteristics (2) and (4), it was concluded that 1/f noise has nothing to do with g-r noise. Finally, a new empirical formula was proposed for 1/f noise.     

MOVPE growth of HgCdTe for high performance 3–5 µm photodiodes operating at 100–180K

New results are reported on the growth of high performance medium wavelength infrared (3–5 μm) (MWIR) HgCdTe photodiodes in the three-layer P-n-N configuration. The detector structures were grown in situ by metalorganic vapor phase epitaxy (MOVPE) on (211)B oriented CdZnTe substrates. The mobilities of the single n-type layers with x-values of ∼0.30 are in the range of (3–4.5)×10⁴ cm²/V-s at 80K. The lifetimes on unpassivated films range from 1–5 and 4–10 μs at 80 and 180K, respectively, which are within a factor of two or less of the lifetimes calculated for Auger-1 and radiative recombination. The P-n-N films were processed into variable-area backside-illuminated diagnostic arrays and tested for quantum efficiency, spectral response, R_DA, I–V curves and 1/f noise in the 120–180K range. The internal one-dimensional quantum efficiencies are in the range of 85–100%. The optical collection lengths are typically ∼25 μm. I–V curves showed that diffusion current is the dominant junction current mechanism for temperatures ≥100K. R₀A values are at the one-dimensional limit for n-side diffusion currents over the 100–180K range. 1/f noise was measured to be very low at 120K and is the same as that measured in similarly processed arrays from recent LPE grown P-on-N heterojunctions. The results demonstrate that MOVPE growth can be used for large area, high performance MWIR HgCdTe detector arrays operating in the 120–180K temperature range.     

Dislocation Reduction of HgCdTe/Si Through <Emphasis Type="Italic">Ex Situ</Emphasis> Annealing

Current growth methods of HgCdTe/Cd(Se)Te/Si by molecular-beam epitaxy (MBE) result in a dislocation density of mid 10⁶ cm⁻² to low 10⁷ cm⁻². Although the exact mechanism is unknown, it is well accepted that this high level of dislocation density leads to poorer long-wavelength infrared (LWIR) focal-plane array (FPA) performance, especially in terms of operability. We have conducted a detailed study of ex situ cycle annealing of HgCdTe/Cd(Se)Te/Si material in order to reduce the total number of dislocations present in as-grown material. We have successfully and consistently shown a reduction of one half to one full order of magnitude in the number of dislocations as counted by etch pit density (EPD) methods. Additionally, we have observed a corresponding decrease in x-ray full-width at half-maximum (FWHM) of ex situ annealed HgCdTe/Si layers. Among all parameters studied, the total number of annealing cycles seems to have the greatest impact on dislocation reduction. Currently, we have obtained numerous HgCdTe/Si layers which have EPD values measuring ~1 × 10⁶ cm⁻² after completion of thermal cycle annealing. Preliminary Hall measurements indicate that electrical characteristics of the material can be maintained.     

High detectivity InGaN-GaN multiquantum well p-n junction photodiodes

InGaN-GaN multiquantum well (MQW) p-n junction photodiodes with semi-transparent Ni-Au electrodes were fabricated and characterized. It was found that the fabricated InGaN-GaN p-n junction photodiodes exhibit a 20-V breakdown voltage and a photocurrent to dark current contrast ratio of /spl sim/10/sup 5/ when a 0.4-V reverse bias was applied. The peak responsivity at 380 nm was 1.28 and 1.76 A/W with a 0.1- and 3-V applied reverse bias, respectively. Furthermore, an internal gain was found from our InGaN-GaN MQW p-n junction photodiodes possibly due to the long-lifetime of GaN based materials. Also, it was found that the low frequency noise of our photodiodes was dominated by the 1/f type noise. For a given bandwidth of 500 Hz, the corresponding noise equivalent power and normalized detectivity D/sup */ were found to be 6.34/spl times/10/sup -13/ W and 4.45/spl times/10/sup 11/ cm/spl middot/Hz/sup 0.5/ W/sup -1/, respectively.     

1/f noise studies in uncooled narrow gap Hg1−xCdxTe non-equilibrium diodes

We have studied the 1/f noise current in narrow gap semiconductor heterostructure diodes fabricated in mercury cadmium telluride (HgCdTe) and designed to operate in a non-equilibrium mode at room temperature. HgCdTe heterostructure diodes exhibit Auger suppression giving current-voltage characteristics with high peak-to-valley ratios (up to 35), and low extracted saturation current densities (e.g., 20 Acnr⁻² at 10 pm at 295K) but high 1/f knee frequencies (e.g., 100 MHz at 10 μm at 295K). A comparison is made with the noise levels found in room temperature non-equilibrium mode heterostructure InAlSb/InSb diodes. The devices are being used at high frequencies for CO₂ laser heterodyne detector demonstrators. For the devices to be useful in low frame-rate imaging arrays, the 1/f noise level must be sufficiently low that the signal is not swamped. Ideally, the knee frequency should be below the frame rate. The relationship between the noise current and reverse bias voltage, current density, and temperature will be examined in order to attempt to identify the principal 1/f generation mechanisms.     

Analysis of 1/f noise in LWIR HgCdTe photodiodes

We study the 1/f noise currents and dark currents in LWIR HgCdTe photodiodes. The measured dark currents of the diodes processed by post implantation annealing with different annealing times are analyzed using current model fitting methods. The different dark current components, such as diffusion current, generation-recombination current, band-to-band tunneling current, and trap assisted tunneling current, at various bias voltages can be separated from the measured dark currents. By the fitting analysis, some physical parameters are extracted and different annealing effects can be explained by the parameters. The improvements in diode characteristics by post implantation annealing can be explained by the changes of trap density, donor concentration, minority carrier lifetime, and generation lifetime. The 1/f noise currents are measured over a wide range of reverse bias voltages, and correlated with the extracted dark currents by superposition of the noise generated by the different dark current mechanisms. It turns out that the band-to-band tunneling has a smaller correlation with the 1/f noise than other current components, and the trap center seems to be responsible for the 1/f noise characteristics of the LWIR HgCdTe photodiodes.     

Effect of dislocations on performance of LWIR HgCdTe photodiodes

The epitaxial growth of HgCdTe on alternative substrates has emerged as an enabling technology for the fabrication of large-area infrared (IR) focal plane arrays (FPAs). One key technical issue is high dislocation densities in HgCdTe epilayers grown on alternative substrates. This is particularly important with regards to the growth of HgCdTe on heteroepitaxial Si-based substrates, which have a higher dislocation density than the bulk CdZnTe substrates typically used for epitaxial HgCdTe material growth. In the paper a simple model of dislocations as cylindrical regions confined by surfaces with definite surface recombination is proposed. Both radius of dislocations and its surface recombination velocity are determined by comparison of theoretical predictions with carrier lifetime experimental data described by other authors. It is observed that the carrier lifetime depends strongly on recombination velocity; whereas the dependence of the carrier lifetime on dislocation core radius is weaker. The minority carrier lifetime is approximately inversely proportional to the dislocation density for densities higher than 10⁵ cm⁻². Below this value, the minority carrier lifetime does not change with dislocation density. The influence of dislocation density on the R₀A product of long wavelength infrared (LWIR) HgCdTe photodiodes is also discussed. It is also shown that parameters of dislocations have a strong effect on the R₀A product at temperature around 77 K in the range of dislocation density above 10⁶ cm⁻². The quantum efficiency is not a strong function of dislocation density.     

Investigation of 1/<Emphasis Type="Italic">f</Emphasis> Noise Mechanisms in Midwave Infrared HgCdTe Gated Photodiodes

In this work, gated midwave infrared (MWIR) Hg_1–x Cd _x Te photodiodes are used to investigate the physical origin of 1/f noise generation. Gated photodiodes were fabricated on liquid-phase epitaxy p-type HgCdTe MWIR material with a vacancy-doped concentration of 1.6 × 10¹⁶ cm⁻³ and x = 0.31. CdTe was thermally deposited and used as both a passivant and a mask for the plasma-based type conversion, and ZnS was used as an insulator. Fabricated devices show a R ₀ A of 1–5 × 10⁴ Ωcm² with zero gate bias. Application of 2 V to the gate improves the R ₀ A by more than two orders of magnitude to 6.0 × 10⁶ Ωcm², which corresponds to the p-type surface being at transition between depletion and weak inversion. Trap-assisted tunneling (TAT) current was observed at negative gate biases and reverse junction biases. For gate biases greater than 3 V, a field-induced junction breakdown was observed. An I _n = α I ^β f ^−0.5 trend was observed above 200 pA reverse bias dark current, with α = 3.5 × 10⁻⁵ and β = 0.82, which corresponds to the TAT dominated region. Below 200 pA, junction generation-recombination (GR) current starts to dominate and this previously mentioned trend is no longer observed. Junction GR current was not seen to be correlated with 1/f noise in these photodiodes.     

Impact of critical processes on HgCdTe diode performance and yield

HgCdTe detector performance and yield are strongly dependant on CdZnTe substrate and HgCdTe epilayer properties, and on key device processes, especially for 8–12 μm application. Due to the correlation and optimization between these figures and diode performance, AIM has developed a mature HgCdTe technology for superior detector performance and high production rate. To meet high yield and performance for long wavelength (LW) HgCdTe diodes, dislocation densities of < 1 × 10^t cm⁻² both in substrate and epilayer have to be ensured. By a unique AIM substrate growth process, dislocation densities of 2 × 10⁴-9 × 10⁴ cm⁻² are achieved for all substrates and epilayers (100% yield). The etch pit density (EPD) on 〈111〉 epilayers is revealed by an AIM proprietary etching procedure. One critical effect is the dislocations in the diode area, which can originate from the substrate and epilayer growth and the subsequent device processes, respectively. Our studies have shown that device processes can cause additional dislocations in the diode area. Diode yield was clearly improved by a combination of wet and dry etching for diode contact etching.     

HgCdTe growth on (552) oriented CdZnTe by metalorganic vapor phase epitaxy

We report the growth of HgCdTe on (552)B CdZnTe by metalorganic vapor phase epitaxy (MOVPE). The (552) plane is obtained by 180 rotation of the (211) plane about the [111] twist axis. Both are 19.47 degrees from (111), but in opposite directions. HgCdTe grown on the (552)B-oriented CdZnTe has a growth rate similar to the (211)B, but the surface morphology is very different. The (552)B films exhibit no void defects, but do exhibit ∼40 μm size hillocks at densities of 10–50 cm⁻². The hillocks, however, are significantly flatter and shorter than those observed on (100) metalorganic vapor phase epitaxy (MOVPE) HgCdTe films. For a 12–14 μm thick film the height of the highest point on the hillock is less than 0.75 μm. No twinning was observed by back-reflection Laue x-ray diffraction for (552)B HgCdTe films and the x-ray double crystal rocking curve widths are comparable to those obtained on (211)B films grown side-by-side and with similar alloy composition. Etch pit density (EPD) measurements show EPD values in the range of (0.6–5)×10⁵ cm⁻², again very similar to those currently observed in (211)B MOVPE HgCdTe. The transport properties and ease of dopant incorporation and activation are all comparable to those obtained in (211)B HgCdTe. Mid-wave infrared (MWIR) photodiode detector arrays were fabricated on (552)B HgCdTe films grown in the P-n-N device configuration (upper case denotes layers with wider bandgaps). Radiometric characterization at T=120–160 K show that the detectors have classical spectral response with a cutoff wavelength of 5.22 μm at 120 K, quantum efficiency ∼78%, and diffusion current is the dominant dark current mechanism near zero bias voltage. Overall, the results suggest that (552)B may be the preferred orientation for MOVPE growth of HgCdTe on CdZnTe to achieve improved operability in focal plane arrays.     

Suppression of strain-induced cross-hatch on molecular beam epitaxy (211)B HgCdTe

Because the performance of HgCdTe-based photodiodes can be significantly degraded by the presence of dislocations, we have systematically investigated and suppressed lattice-mismatch-induced cross-hatch formation and the associated generation of dislocations in (211)B HgCdTe/CdZnTe. A series of HgCdTe epilayers were deposited simultaneously on pairs of substrates with differing ZnTe mole fractions. Epilayers’ CdTe mole fraction and substrates’ ZnTe mole fractions were measured using optical-transmission spectra. Lattice mismatch and residual strain were estimated from room-temperature, x-ray diffraction, and double-crystal rocking-curve measurements (DCRC). It was found that cross-hatch patterns were suppressed in epilayers deposited on nearly lattice-matched substrates (|Δa/a_sub|<0.02%). Such epilayers exhibited excellent crystalline quality as revealed by defect-decoration etching (etch-pit density (EPD)<10⁵ cm⁻²) and x-ray diffraction (full-width at half-maximum (FWHM) ∼10 arcsec). In addition to determining the upper limits of lattice mismatch needed to eliminate cross-hatch, we investigated the use of reticulated substrates as a means to suppress cross-hatch. We found that growth on reticulated mesa structures (<100 μm) with edges parallel to [01-1] resulted in epilayers with substantially reduced cross-hatch-line densities despite large lattice mismatch (Δa/a_sub <0.04%). The use of reticulated substrates could suppress cross-hatch because of lateral-alloy variation in large substrates and complex multistack epilayers (e.g., multicolor detectors).