Modulated optical sensitivity with nanostructured gallium nitride

Gallium nitride (GaN) continues to provide many quality challenges resulting from numerous stacking faults and threading dislocations from lattice mismatches between growth substrates and subsequent epilayers. Growth techniques such as epitaxial lateral overgrowth,¹ intentional interface void generation,² and homoepitaxial epi-deposition³ have sought to reduce defect densities, currently on order of 5 × 10⁵–1 × 10⁶ cm⁻² but work continues in the drive towards higher quality GaN. Planar GaN is currently being used for light emitting diode stacks⁴ and high-electron mobility transistors⁵ which require high quality defect free substrates. In a push to improve device performance, low-dimensional GaN structures are being explored,^6,7 such as wires,⁷ rods,⁸ and pyramids.^9,10 In addition, 3D-nanostructured surface morphologies have been introduced in an effort to modulate the optical properties of GaN that are nominally limited in a purely planar format due to the critical angle of escape for photons.¹¹ 

Nanostructured surface fabrication primarily utilizes techniques such as reactive ion etching (RIE),¹² catalyst deposition,¹³ and chemical etching⁹ as methodologies of choice. Though RIE provides direct control universally throughout GaN regardless of orientation via top-down low throughput fabrication, some undesired complications such as physical damage, dangling bonds, and subsequent thermal zones are induced via RIE.^14,15 By utilizing chemically assisted etching, benefits such as high throughput and lack of thermal zones can be achieved by using chemicals such as molten KOH, phosphoric acid, sulfuric acid, or dilute aqueous KOH solutions.⁸ In addition to open cell etching, chemical etching can be assisted by UV sources and placed in a galvanic cell to improve etching characteristics.⁸ Owing to its natural chemical resistance, GaN offers unique etching rates based upon crystallographic orientation, which stems from the preferential removal of nitrogen versus gallium.¹⁶ This provides a way to develop unique nanostructures defined by specific crystallographic orientations.

In this study, chemical functionalization is used to modulate the optical response across planar and high aspect ratio GaN topographies. The GaN nanostructure utilized dislocations as site specific etching centers, termed defect-assisted etching, via dilute aqueous KOH solutions^8,9 to provide a repeatable and scalable fabrication technique. To compare the properties of nanostructured GaN, photoluminescence (PL) is used to identify changes to the critical photon escape angle as a result of faceting and formation of high aspect structures. In the case of nanorods or nanopyramids, the high aspect ratio provides unique opportunities for tailorable luminescence and binding sites from increased surface area,¹⁷ which is compared to both thin layer GaN and bulk free-standing GaN. The thin layer GaN was residing on a sapphire substrate,⁹ whereas the bulk free-standing GaN is a self-separated free of foreign substrate that has been subsequently chemically mechanically polished.² Chemical functionalization of GaN has shown unique properties such as modified optical,¹⁸ electrical,¹⁹ and unique surface hydrophobicity^17,20 by chemical termination. With recent interest in interfaces for applications such as organic light emitting diodes²¹ and sensors,²² chemical functionalization with desired adsorbates can provide tailored electrical, chemical, topological, and optical properties. Potential areas of application include biochemical sensors within microfluidic chambers and nanoscale optoelectronics.²³ In this investigation, the use of in-situ functionalization with 1H,1H,2H,2H-perfluorooctanephosphonic acid (PFOPA), which has been previously used on Zone,²⁴ ITO,²⁵ and GaN,²⁶ is used to modulate the optical responses of the three types of GaN structures in addition to identifying the resulting optical output following a lengthy environmental soak that potential devices might encounter. Recent studies on surface functionalization with PFOPA have shown reproducible binding²⁷ as well as unique modifications of the optical properties of free-standing bulk GaN,²⁶ indicating its potential for sensing applications.

The GaN samples are separated into three categories: planar layer GaN residing on sapphire, nanostructured GaN also residing on sapphire, and bulk free-standing GaN. The nanostructured GaN formations were produced from the planar thin layer GaN deposited on sapphire with n-type conductivity (Si doped) via metalorganic chemical vapor deposition (MOCVD). The samples were placed in an illuminated-galvanic cell utilizing a KSO-D (0.02M KOH + 0.02M K₂S₂O₈)⁹ aqueous solution and platinum counter electrode while being exposed to a 300-W xenon lamp for 60 min. The Ti contact protected about ¼ of the sample surface against etching and this area was used as a comparison versus the etched nanostructure.

The bulk free-standing GaN was grown via hydride vapor phase epitaxy (HVPE) on sapphire until reaching 1.2 mm thickness.² The thick GaN layers were then self-separated from the substrate during cooling down and were subsequently chemically mechanically polished and diced. The thin layer GaN material was characterized with high dislocation density in the range of 10⁹cm⁻², while the bulk GaN material possessed a low dislocation density in the range of 10⁵–10⁶cm⁻². The nanostructured material produced from the same thin layer GaN possessed similar high dislocation density to that of the initial layer due to the fact that the photo-etching process tends to take place around dislocations.²⁸ 

Prior to the secondary etching, all samples were placed in acetone, methanol, and dilute HCl baths. The samples (layered, bulk, and nanostructured) were separated into three groups: cleaned (no treatment) and etched with either a 50 vol. % H₃PO₄ or a 50/50 vol. % H₃PO₄/3 mM PFOPA solution. The samples were rinsed with deionized water, dried with nitrogen, and characterized. To replicate environmental conditions, samples were soaked in deionized water for 7 days and characterized again. We note that several samples from groups 1 and 2 were treated for statistical purposes and the results shown below are representative. The nanostructured samples were of limited numbers and in order to validate the results, the experiment was repeated.²⁹ 

Surface morphology of the three types of samples was evaluated by atomic force microscopy (AFM) using Digital Instrument Nanoscope IIIa in tapping mode and by scanning electron microscopy (SEM). The SEM images of the nanostructure were recorded with a Verios 460L SEM. Images were taken at a 4.4 mm working distance, 13 pA filament current, 2.0 kV accelerating voltage at a 30° tilt without a stage bias. PL at room temperature was done with a Horiba Jobin Yvon LabRam ARAMIS Raman/PL setup employing a 325 nm HeCd laser. Calibration was done using a Raman peak of 1295 cm⁻¹ from Teflon with a 40× UV objective at a 2400 grating/min resolution. The real-time display (RTD) exposure time, accumulated exposure time, and number of acquisitions for averaged signal were 0.5 s, 0.5 s, and 5, respectively, with a 90% signal filter. To ensure equal exposure, laser spot size was maintained at 10 μm (crossover) due to the different sample thicknesses. Each sample was evaluated at five random locations to account for surface variances with spectra capturing from 340 to 390 nm.

Figures 1(a) and 1(b) reveal smooth planar surface for both planar thin layer and bulk GaN materials, with root mean square (RMS) values were 0.7 and 0.2 nm, respectively. The lower value for the bulk GaN surface is consistent with lower dislocation density, reflected in lower number of surface depressions related to dislocations intersecting the surface. From Fig. 1(c), nanostructures with pyramidal shape, approximately 700 nm in length can be seen. Their uniform shape and cone-like tip structure (Fig. 1(c) inset) reveal gallium enriched facets frequently associated with semipolar planes.³⁰ The KSO-D formula is known to produce nanostructured surfaces of different shapes, e.g., nanorods, nanowires, and nanopyramids. The nanotextured surface can be controlled by the etching time,⁹ etching solution,⁸ and UV source power. In addition, the properties (dislocation density and carrier concentration/doping) of the initial GaN layer³¹ were found to enable different etching rates and nanostructures of various shapes, density, and aspect ratios. The moderate doping, i.e., carrier concentration of about 1 × 10¹⁷ cm⁻³ and high dislocation density in our GaN template, allowed pyramidal topography with semipolar facets, as opposed to the nanorod structure (with more uniform diameter and no semipolar surfaces) typically produced from highly doped GaN layer. It is important to note that although the RMS values were slightly increased, the surfaces of the planar samples remain smooth and unaffected by the treatment, which follows the previous works and associated supporting information.^26,32

The PL responses to the different treatments shown in Fig. 2 reveal that the near band-edge (NBE) for the layered GaN resides at 361 nm, while the NBE for nanostructured and bulk-freestanding GaN reside at 362 nm. The blue shift of the NBE emission from the thin GaN layer on sapphire is a typical feature related to the biaxial substrate induced strain. The similar spectral position of the PL spectra from the nanostructured and the bulk free-standing GaN can be explained by the reduced strain in both samples due to removed or reduced substrate effect, respectively. For the planar layered GaN and the bulk free-standing GaN, the NBE energy position does not change with different treatments both pre and post water soak. In contrast, the nanostructured samples allow for a greater response due to a better surface sensitivity of the semipolar surfaces to the chemical treatments as compared to the well-known chemical stability/resistivity of the Ga-face polar surface. By utilizing a dilute phosphoric acid, a noticeable NBE shift of 0.4 nm was seen with respect to the NBE position in the PL spectrum from a cleaned nanostructured sample, and an additional 0.2 nm shift was observed in the PL spectrum from the sample treated with PFOPA modified phosphoric acid. Concentrations below 43 vol. % were used to ensure that the nanostructured surface would not undergo additional etching as well as to prevent phosphonic acid agglomeration. Phosphonic acids, particularly those with unique electrical signatures, provide another avenue to modulate optical properties by utilizing native oxides.³³ The PFOPA chemical treatment was also found to facilitate oxide formation in our previous study²⁶ consistent with reports by other groups.³³ The PFOPA enables simplified characterization and is known to spontaneously form on oxides. The presence of oxide on planar GaN surfaces with different, polar and nonpolar, orientations treated with diluted phosphonic acids was experimentally proven by XPS analysis and described in detail elsewhere.^26,27 While the oxide formation on nanostructured sample was not possible to be directly measured the observed trends and the prior knowledge of higher affinity of the inclined crystallographic surfaces to oxygen³⁴ as compared to the polar basal plane of GaN, makes us speculate that oxide is most likely larger on the nanostructured samples surface. Taking into account the lattice parameter difference and prior reports on oxide formation on GaN surfaces,³⁰ the noticeable red shift of the NBE position in the treated nanostructured sample is explained by the compressive strain in the near interface region. In addition, a shoulder can be seen on the PL spectra from nanostructured and layered GaN samples at approximately 369 nm, which is associated with longitudinal-optical phonon replicas being expressed.³⁵ This shoulder is also present in the PL spectra of the bulk-free standing GaN but was not visualized due to size of scale used. As expected, the energy position of this feature shifts along the main peak upon each of the treatments.

Furthermore, Figs. 2(a) and 2(b) clearly reveal that the intensities of the NBE emission of all three types of samples were noticeably affected by the chemical treatments. For both the nanostructured and the thin layer GaN samples, the PL intensities were significantly improved by the chemical treatments. This is not a surprising effect considering the increased surface areas and improved extraction efficiency enabled by nanostructured surface and etching effect of the phosphonic treatment. In contrast, the chemical treatments have a decreasing effect on the PL intensity from the bulk GaN sample. We explain this observation by the highest smoothness of the bulk GaN sample as a result of the strongest chemical resistance of its Ga-face terminated surface and much lower density of defects³⁶ (as compared to the other two samples).

We have also studied the effect of sample exposure to water for 7 days. It was consistently observed that this treatment resulted in an overall decrease in PL intensity (Fig. 2(a) vs. Fig. 2(b)). Decreased PL signal, particularly of the NBE, is usually associated with increased nonradiative recombination events versus radiative recombination events.^36,37 In this case, increased hydroxide and oxide species from soaking increased the amount of nonradiative recombination sites at the surface, and thus lowering the overall emission intensity. This was particularly evident in the PL spectra of the thin GaN layer treated with phosphoric acid, which saw a 41% decrease pre vs. post soak. Unlike the PFOPA treated samples, phosphoric acid provides no passivation via oxide stabilization as phosphonic acids have already shown decreased leaching characteristics for GaN not available with unfunctionalized surfaces.³² The degree of emission intensity decrease between pre and post soaking provides relevant information for environmentally active optical devices.

For a more detailed look at the effect of treatments on the optical properties, we compared the relative changes in the maximum PL intensities versus individual treatments. In the pre-soaked samples (Figs. 3(a)–3(c)), the use of both treatments results in an increase in total photoluminescence for nanostructured and layered GaN, which was not the case for bulk free-standing GaN. As the nanostructured and layered GaN had already experienced both a polishing and etching process (with a protective wax layer for layered GaN), it is likely that the surface of bulk free-standing GaN retained a greater number of nonradiative recombination centers, resulting in diminished luminescence (Fig. 3(c)). It is not until after soaking that the nonradiative recombination centers are suppressed for bulk free-standing GaN (Fig. 3(f)). In the case of the nanostructured GaN, the PFOPA treatment demonstrated the ability to increase the PL signal versus an additive free treatment (Fig. 3(d)). Though the layered GaN displayed diminished luminescence for PFOPA treated presoak sample, the intensities of the spectra from post soaking samples were similar (Fig. 3(e)), indicating a stabilization of the surface optical properties not available with only phosphoric acid treatments. Based on these results, we conclude that the use of PFOPA provides a way for optical modulation of 3D GaN nanostructures with a stronger sensitivity as compared to the planar GaN surfaces. These results are of particular importance for sensing applications utilizing surface photoluminescence as a monitoring technique within aqueous environments that has been little explored to date.

In conclusion, additive assisted etching provides a unique way to modulate the optical properties of both planar and 3D GaN nanostructures. This was seen particularly in the increased PL intensity for nanostructured GaN treated with PFOPA as a result of the increased surface area, improved light extraction and possible higher semipolar sensitivity contribution versus polished polar GaN surface. All materials showed significant optical sensitivity to surface functionalization, even following lengthy soaking. It is beneficial to note that general trends, particularly for the defect rich layered GaN, remained consistent with treatments and subsequent soaking. Acid treatments provided NBE wavelength manipulation that was only seen with the high aspect ratio nanostructured GaN.

The authors acknowledge partial support for this work via NSF Grant No. DMR-1207075 and NCN Grant No. 2011/01/B/ST8/07569.