In recent years, nanostructures have demonstrated to offer new functionalities to traditional semiconductor devices. Quantum confinement changes the profile of the density of states notably, giving rise to new phenomena unobserved before. In low-dimensional structures, the surface effects become more important and their understanding is essential to model their optical and electrical characteristics. The nanostructure physics are severely affected by the interaction of the surface with the inner medium. Therefore, the impact of the surface is a major issue that needs to be taken into account in the analysis of the devices with nanostructures embedded. These devices are excellent platforms for finding experimental evidence of this interaction; their thorough study is therefore highly recommended to optimize device performance.

A very promising nanostructure to be incorporated in device optimization is the nanopillar, i.e. elongated structure with sidewalls defined by e-beam lithography and dry etching (top-down approach), by direct epitaxial growth with low lateral diffusivity, or by self-assembling techniques (bottom-up approaches). Each fabrication tool gives rise to a different type of surface whose properties depend on roughness, electronic states or relaxation degree. As in conventional devices, the use of external coatings and thermal/chemical treatments helps to control these properties, but light-matter interaction and carrier transport are unavoidably affected by them.

Light detectors and light-emitting diodes (LEDs) are examples of optoelectronic devices that can take advantage of the use of nanopillars. Ultra-high sensisitivities have been reported in nanowire avalanche photodiodes with detection limits of less than 100 photons. Moreover, quantum confinement effects have been previously reported in photodetectors with embedded nanopillars lower than 50 nm in diameter at low temperatures. On the other hand, nanopatterned samples have exhibited enhanced luminescence due to the extraction of guided modes from nanopillar arrays. III-nitride materials have demonstrated excellent characteristics for light detection and emission. Their wide bandgap tunability makes them attractive for many applications in the ultraviolet (UV) and visible ranges. Nitride-based LEDs are available in the market from the deep UV to the green, and light detectors with response cut-offs from 200 nm to the visible have been reported through the use of AlGaN and InGaN ternary alloys. Within detector technology, different device designs have presented photocurrent gain mechanisms, which help to enhance optical response, but none of them has been related with the use of embedded nanopillars.

In this work, we present the fabrication of submicron nanopillars by e-beam lithography and the planarization of the nanostructured surface for device processing. The devices are electrically characterized in order to determine the contribution of the nanopillars to the general optoelectronic performance. Our efforts focus on the study of surface effects and photocurrent gain mechanisms in structures without quantum confinement.

Samples were grown by metal-organic chemical-vapor deposition on double-side polished c-plane sapphire substrates. The structure of MSM devices consisted of an unintentionally doped 2 μm thick GaN layer with a residual electron concentration of ~10¹⁶ cm^-3. In contrast, PIN samples were grown on AlN templates. The structure comprises a p-i-n GaN junction with carrier concentrations of (5-10)×10¹⁷ cm^-3 for the p-type GaN:Mg layer and (3-5)×10¹⁸ cm^-3 for the n-type GaN:Si layer; the i-region consisted of an unintentionally doped GaN layer with a residual electron concentration in the 10¹⁶ cm⁻³ range. In order to minimize the leakage current, the thicknesses of the p-type, intrinsic, and n-type layers were adjusted to be 285, 200 and 200 nm, respectively.

Figure 1: (a) Top and (b) lateral views of the nanopillars fabricated from optimum e-beam settings. (c) Top and (d) lateral views of interconnected nanostructures fabricated from high defocus levels.

For nanopillar fabrication, samples were coated with polymethylmethacrylate (PMMA) and soft baked. Lithography was performed using a Leica LV-1 electron beam system. A 6 mm × 4 mm dot field was patterned in the PMMA, with a period of 1 μm. For an optimum dose of 20, the focus of the e-beam was varied to adjust the dot size. In order to ease the analysis of the surface effects, the target size was set to be 100 nm. This size allows us to build structures in the submicron range without quantum confinement effects. High defocus levels led to undesired proximity effects in the PMMA that distorted the dot field after resist development. The result of this phenomenon is illustrated in our pulications. The dot diameter in the latter is about 200 nm.

After e-beam lithography, a metal mask consisting of a 720-Å thick Ni layer was deposited, followed by a lift-off procedure. The pattern was then transferred into the GaN layer to form nanopillars via dry etching in a ECR-RIE system using a SiCl4:Ar plasma etch process. Scanning electron microscopy (SEM) revealed that the etch depth was 520 nm and the resulting nanopillar diameter was about 200 nm, resulting in a 1% fill factor (Figures 1(a) and 1(b)). The sidewall angle was 23o. The same recipe was applied to the samples with higher defocus values. The SEM images of the nanostructures obtained show 1) the elongation of the nanopillars in the writing direction and 2) the formation of transversal structures that connect the different rows of the dot field (Figures 2(c) and 2(d)). The result is a network of interconnected nanostructures, as identified in the top view of the sample surface (Figure 1(c)). Although these samples were not used for further device fabrication, they create an interesting scenario to study the lateral transport of charge among pillars.

Figure 2: Schematic diagrams of nanostructured (left) MSM detectors and (right) PIN photodiodes.

Insulating polyimide was spin-applied to the sample and cured, covering the nanostructured surface with a 1.9-μm thick layer. Oxygen plasma was then used to etch back the polyimide and reveal the top 100 nm of the nanopillars. SEM analysis of the etched GaN nanopillars after polyimide deposition and etch-back shows optimum filling of the space among nanopillars. Standard lithography was then used to define broad area (25-4225 μm²) circular pads formed by e-beam evaporation of a Ni/Au (400/1200 Å) dual layer and subsequent lift-off. These pads were distant each other more than 15 μm and serve as probing contacts in the MSM structures. For PIN photodiodes, one more step was needed to form the n-type contact: the polyimide in the areas that surround these contacts was completely removed by using oxygen plasma to allow for deposition of a Ti/Au (400/1200 Å) bottom contact using standard lithography, e-beam evaporation, and lift-off. Figure 2 illustrates the schematic diagrams of the devices after processing.

I-V measurements were performed on nanostructured PIN photodiodes with p-type contact areas of 625, 2025, and 3025 μm². Ten devices were measured for each of the different active areas and their characteristics were averaged. The resulting curves for the nanostructured diodes show strong rectifying behavior with about seven-orders-of-magnitude contrast between the current levels at 5V and -5V. The ideality factor obtained from the fitting of the I-V curves under low forward biases is 1.3-1.6, which represents a significant reduction from the values obtained in conventional p-i-n diodes, typically higher than two. A considerable reduction of the turn-on voltage in the nanostructured diodes is also noticeable. This improvement of the diode characteristics is related to the barrier found at the top surface of p-type layers under the contact layer. This barrier makes the diode depart from the ideality, as suggested by several authors. The formation of a nanostructured contact surface based on SiCl₄-based dry etching tends to increase the density of deep levels and the n-type character near the surface. Both effects contribute to enhance the tunneling mechanisms through the barrier, minimizing its effects.

At forward bias, the I-V characteristic under illumination shows an unexpected behavior. Despite the high carrier injection that takes place at those voltages, a clear photosignal of several microamperes is reproducibly found in contrast to conventional diodes. The photosignal must be the result of a gain mechanism capable to enhance the photogenerated currents of electrons and holes five orders of magnitude at a 2 V bias. As depicted in Figure 3:right, the gain curve saturates at higher voltages. A very high optical response has been also found in nitride low-dimensional quantum structures at forward bias and attributed to the modulation of the internal polarization fields. Moreover, this gain mechanism has been employed to study the detection of low photon fluxes, showing a detection limit of about 90 photons.

Figure 3:Green and blue curves respectively correspond to the measurement in the dark and under back-side illumination at 405 nm. Measurements were performed three times in a row to check for consistency; standard deviation is shown as ordinate bars on the data points. (Right) Gain curve calculated from the difference between the dark current and the photocurrent and divided by the photocurrent at zero bias.

MSM PDs presented dark currents < 1nA up to bias voltages of about 40 V. A strong increase of the optical response with bias was observed, leading to responsivities higher than 1 A/W. This behavior reveals the existence of a photoconductive gain mechanism. The relationship between this mechanism and the surface states at the sidewalls will be discussed. P-i-n PDs had peak responsivities (R_peak) of 35 mA/W at -4V and showed an abnormal increase of the response (R_peak>105 mA/W) under forward biases. A gain model is proposed to account for this behavior.