Comprehensive characterization of hydride VPE grown GaN layers and templates

GaN community has recently recognized that it is imperative that the extended, and point defects in GaN and related materials, and the mechanisms for their formation are understood. This is a first and an important step, which must be followed by defect reduction before full implementation of this material and its allied binaries/ternaries in devices. This review is based on a recent concerted effort to establish benchmarks as far as defects are concerned, and identify the basic issues involved. Samples were analyzed for extended defects by TEM and chemical etches, for polarity by electric force microscopy and convergent beam electron diffraction (CBED), for point defects by DLTS, for optical quality and deep defects by photoluminescence (PL), for vacancies by positron annihilation, for donor and acceptor like states within the gap by ODMR and EPR, and for carrier transport targeted for defects and impurities by variable temperature and magnetic field-dependent Hall measurements.Hydride VPE samples grown at Lincoln Laboratories with 1.5, 5.5 and 55 μm thicknesses were investigated for defects by TEM, and their polarity was found to be Ga-polarity, as expected, by CBDE combined with simulations. The density of misfit dislocations at the substrate/EPI interface was determined to be on the order of 10¹³ cm⁻² based on high-resolution electron microscopy images. The threading dislocation density decreased gradually with distance from the interface, reaching a value of about 10⁸ cm⁻² at the surface of a 55 μm film. A 200 μm thick laser separated and free-standing HVPE grown GaN template grown at Samsung was also characterized similarly. The free surface and substrate sides were confirmed to be Ga- and N-polarity, respectively. The density of dislocations near the N-face was determined to be, in order, (3±1)×10⁷ and (4±1)×10⁷ by cross-sectional TEM and plan-view TEM, respectively. Identical observations on the Ga-face revealed the defect concentration to be less than 1×10⁷ cm⁻² by plan-view TEM and 5×10⁵ cm⁻² by cross-sectional TEM.Defects in a 10 μm thick GaN layer grown by HVPE at Lincoln Laboratory have been investigated by photo electrochemical (PEC) etching, and by wet etching in hot H₃PO₄ acid and molten KOH. Threading vertical wires (i.e. whiskers) and hexagonal-shaped etch pits are formed on the etched sample surfaces by PEC and wet etching, respectively. Using atomic force microscopy, one finds the density of “whisker-like” features to be 2×10⁹ cm⁻², the same value found for the etch-pit density on samples etched with both H₃PO₄ and molten KOH. Values agree well with TEM results.A free standing GaN template has been characterized for its structural and optical properties using X-ray diffraction, defect delineation etch followed by imaging with atomic force microscopy (AFM). The Ga-face and the N-face of the c-plane GaN exhibited a wide variation in terms of the defect density. The defect concentrations on Ga- and N-faces were about 5×10⁵ cm⁻² for the former and about 1×10⁷ cm⁻² for the latter, again in good agreement with TEM results mentioned above.High resolution X-ray rocking curves (omega scans) were measured. The [0 0 2] symmetric and [1 0 4] asymmetric peaks in 10 μm thick HVPE films had FWHM values between 5.8 and 7.9 arcmin, and 3.9 and 5.2 arcmin, respectively. The Samsung template investigated had wide diffraction peaks (20.6 and 24 arcmin for [0 0 2] and [1 0 4] diffractions, respectively) on the Ga-face, similar for the N-face, when a 2 mm slit size was used. When the slit size was reduced to 0.02 mm, the Ga- and N-face [0 0 2] peaks reduced to 69 and 160 arcsec. A bowing radius of 1.2 m was calculated to account for increased broadening with wider slits.In the HVPE layer studied, SIMS investigations indicate that both O and Si concentrations drop rapidly away from the surface into the sample — mainly due in part to the artifact of the technique and in part due to condensates on the surface of the sample, down to about 10¹⁷ cm⁻³ for Si and high 10¹⁶ cm⁻³ for O. The Ga-face profile in the Samsung template indicated levels below mid-10¹⁶ cm⁻³ for all three of the impurities. The picture is different for the N-side, however, as it was juxtaposed to the substrate during growth and was mechanically polished after laser separation. The impurity concentration on this face, depending on the species, is some 1–3 orders of magnitude higher than the case for the Ga-face.Transport properties as a function of the layer thickness, ranging from about 1 to 68 μm, while all the other parameters being the same, as they evolve from the sapphire/GaN interface and up were determined in epitaxial layers. A strong dependence on distance from the interface was observed with the averaged mobility figures increasing from 190 cm²/V s in the 5 μm film to 740 cm²/V s in the 68 μm film. The preliminary differential Hall measurements indicate that the mobility at the surface of the thick layer is about 1200 cm²/V s. Electron mobilities in free-standing template were 1425 cm²/V s at T=273 K and 7385 cm²/V s at T=48.2 K. By using the most recent unscreened acoustic deformation potential and allowing only the acceptor concentration to vary (2.4×10¹⁵ cm⁻³ for the best fit), one obtains an excellent fit to the measured mobility in the temperature range of 30–273 K by using an iterative BTE method. In addition, an excellent fit for the temperature-dependent electron concentration was also obtained utilizing the acceptor concentration determined from the fit to the Hall data, and the charge balance equation. This led to a donor concentration of 1.6×10¹⁶ cm⁻³, and activation energy of 26 meV, the latter being the highest reported in the literature for GaN.In the free-standing template, the Γ₅ and Γ₆ free excitons were identified from emission measurements by utilizing polarization geometries where the E field is perpendicular to the c-axis, favoring the Γ₅ exciton, and E field parallel to the c-axis (incident beam from the edge of the wafer) favoring the Γ₆ exciton. Focusing on the defect region of the PL spectrum, the N-face of the sample exhibited the usual yellow line. However, the Ga-face exhibited a broad band encompassing both yellow and green bands. The yellow luminescence in the free-standing template is weak and can be easily saturated. In contrast, the green luminescence is dominant and is attributed to the isolated defect involving gallium vacancy, whereas the yellow luminescence is related to the same defect bound to dislocation or surface-bound structural defect.Deep centers have been characterized by DLTS in HVPE-grown GaN epilayers of different thickness and dislocation densities, and templates. The main deep centers, such as A₁, B, and D, show higher concentrations in thinner samples, which suggests a correlation to the high dislocation densities. Based on the anti-correlation between A₁ and B, which is observed in thin HVPE-GaN layers, the defect B was tentatively attributed to N_Ga. Centers A₁ and E₁ found in thin HVPE-GaN are very similar to centers A₂ and E induced by electron-irradiation, indicating their point-defect nature. Centers A, C, and D are not affected by 1 MeV electron-irradiation, thus ruling out the possibility of these centers being identical to any EI-induced centers; however, their nature remains unknown. As the layer thickness decreases, an increase of deep centers, both in species and concentrations, was clearly observed, which is believed to be closely associated with the significant increase of threading dislocations as revealed by TEM. Based on a comparison with EI-induced centers and an observation of anti-correlation, A₁ is tentatively assigned to N_I, and B to N_Ga. The template exhibited a new trap B′, with parameters E_T=0.53 eV and σ_T=1.5×10⁻¹⁵ cm² on the Ga-face, in addition to the four traps commonly observed in various epitaxial GaN layers. For the N-face, an N vacancy-related trap E₁, with E_T=0.18 eV and σ_T=4×10⁻¹⁷ cm², was observed. On the other hand, the Ga-face sample contained trap C, with E_T=0.35 eV and σ_T=1.6×10⁻¹⁵ cm². This trap may be related to surface damage caused by the RIE process employed.Electron beam and optical depth-profiling of thick (5.5–68 μm) n-type HVPE-GaN samples have been carried out using electron beam-induced current (EBIC) and micro-PL to determine the minority carrier diffusion length, L, and minority carrier lifetime. The minority carrier diffusion length increased linearly from 0.25 μm, at a distance of about 5 μm from the GaN/sapphire interface, to 0.63 μm at the GaN surface for a 36 μm thick sample. The increase in L was accompanied by a corresponding increase in PL band-to-band radiative transition intensity as a function of distance from the GaN/sapphire interface. These observations in PL intensity and minority carrier diffusion length have been attributed to a reduced carrier mobility and lifetime at the interface and to scattering at threading dislocations.Positron annihilation experiments have been conducted in HVPE films with varying thicknesses from 1 to 68 μm. Mg-doped samples and bulk GaN platelets have also been investigated and the behavior of positron annihilation in Mg-doped samples established. Unlike the Mg-doped samples, the positron lifetime in the HVPE samples increased with decreasing lattice temperature. This was interpreted as acceptors in these n-type samples being due to Ga vacancies as opposed to relatively shallow acceptor impurities. The similarities in the behavior of these samples and those investigated previously where the III/V ratio was changed also lend support to the Ga vacancy argument. Previous investigations established that as the III/V is lowered by increasing the ammonia flow during the growth, the Ga vacancy concentration increased. Using Mg-doped samples as a standard, the vacancy concentration was determined to be about 10¹⁷ cm⁻³ near the surface for the layer with a thickness greater than 30 μm. Assuming that the growth parameters in the set of layers with varying thicknesses that were investigated are the same, the Ga vacancy concentration increases to mid-10¹⁹ cm⁻³ near the interface. Since the interfacial region is n-type and highly conductive, this region must also contain even larger concentrations of O and/or N vacancies which lead to n-type material. SIMS results already indicate mid-10¹⁹ cm⁻³ levels of O being present in this region. This has been attributed to O out-diffusion from sapphire as previously reported.FTIR, ODMR and EPR measurements have been performed in GaN layers and templates. In FTIR measurements, two absorption bands corresponding to binding energies of 30.9 (Si) and 33.9 meV were found. Splitting of the binding energies with magnetic field is consistent with an effective mass of 0.22m₀. Angular rotation studies were performed with the magnetic field oriented perpendicular and parallel to the c-axis to provide symmetry information. The ODMR on the 2.2 eV peak in a 5–10 μm thick GaN layer, the notorious yellow emission, showed signatures of shallow donor (effective mass like) and deep defect centers with g-values of 1.95 and 1.99, respectively. The 3.27 eV peak with resolved LO phonon replicas, which is the blue peak observed in many GaN films grown by a variety of methods, is attributed to transitions involving shallow acceptors with g2.1 and g2.0. ODMR on the 2.4 eV “green” PL band in the free-standing template also revealed evidence for shallow donors with a g-value of 1.95 and other deeper centers. The larger line width of the shallow donor signal from the template, relative to that found for the epitaxial layers, is indicative of a lower concentration of this center, which leads to an increased hyperfine interaction. EPR studies confirmed the notable difference between the epilayers and the template, particularly the larger line widths in the template due to the lower concentration of shallow donors. Specifically, the free-standing sample has about 6×10¹⁵ cm⁻³ uncompensated donors while the epilayers have a concentration about a factor of four higher.Calculations indicate that incorporation of Si has a negligible effect on the lattice constant, but O and Mg can lead to an observable expansion of the lattice. Since values of the GaN lattice constant have often been based on bulk crystals that are now known to contain large concentrations of oxygen, the “true” GaN lattice constant is actually smaller than what has been measured for such crystals. Boron is an unintentional impurity that can be introduced during MBE growth. There has been speculation about whether B might act as an acceptor in GaN; this would require it to be incorporated on the nitrogen site. Computations indicate that incorporating B on the N site is energetically unfavorable. Even if it did incorporate there, it would act as a deep, rather than a shallow acceptor. Formation energies of H in AlN and GaN have also been calculated. The behavior of H in AlN is very similar to GaN: H⁺ dominates in p-type, H⁻ in n-type. Surprisingly, H in InN behaves exclusively as a donor, i.e. it is not amphoteric as in GaN and AlN, but actually contributes to the n-type conductivity of the material.Scanning thermal microscopy (SThM) has been applied to measure the local thermal conductivity of epitaxial GaN as it is affected to a large extent by phonon scattering, and a closer to the true value of this parameter can be obtained by a local measurement in areas of lower defect concentration such as those found in the wing regions of lateral epitaxially grown GaN. The method relies on a thermo-resistive tip forming one quadrant of a Winston bridge. The bridge is balanced with the tip heated followed by bringing the tip in contact with the sample under test which cools down due to thermal dissipation. However, the feedback circuit attempts to keep the thermo-resistance and thus the tip temperature the same. The square of the feedback voltage necessary for this is proportional to the thermal conductivity. Accurate values can be obtained with calibration using known substrates such as GaSb, GaAs, InP, Si and Al metal. Using SThM, thermal conductivity, κ, values of 2.0–2.1 W/cm K in the wing regions of lateral epitaxially grown GaN, 1.70–1.75 W/cm K in HVPE grown GaN, and 3.0–3.3 W/cm K for free-standing AlN have been measured.