Effect of KOH passivation for top-down fabricated InGaN nanowire light emitting diodes

Gallium nitride (GaN) is a wide-bandgap material typically used in light emitting diodes (LEDs) for state-of-the-art commercial lighting applications.¹ The advent of GaN LEDs enables efficient backlighting for liquid crystal displays (LCDs) as well as solid state lighting.² However, as conventional GaN LEDs are scaled down in size, the performance of these devices is lowered.³ Thus, nanowires (NWs) are emerging as a replacement solution for the planar configuration of LEDs.^4,5 Specifically, NW structures have received significant attention for optoelectronics due to their ability to achieve dislocation-free material and larger surface-to-volume ratios, which lead to enhanced light output power and reduced area consumption compared to planar structures.^6,7

However, the larger surface ratio of GaN NWs can lead to dangling bonds during NW formation. The induced surface defects can be mitigated through passivation, which will be explored in this study due to its importance in improving the performance of NW LEDs. The fabrication methods of NWs are generally categorized into two approaches: bottom-up growth and top-down etching.⁸ For the bottom-up approach, NWs are synthesized onto substrates, layer-by-layer, in the atomic scale through either molecular beam epitaxy (MBE) or metal organic chemical vapor deposition (MOCVD).⁹ Bottom-up growth can lead to both dense NW arrays if selective area growth is not employed and island growth incorporation.¹⁰ When using epitaxial growth, NWs may coalesce during growth instead of forming individual NWs. If a top-down approach is used instead, NWs can be precisely patterned and are formed through lithographic and etching steps.¹¹ In the latter, the surface states are formed by top-down etching that can create nonradiative recombination centers, which significantly degrade the quantum efficiency and output power of the NW LEDs.¹² The creation of nonradiative recombination centers in top-down NWs is largely due to the dry etching step, resulting in surface roughness and dangling bond damage on the NWs.¹¹ To prevent surface state formation, various surface passivation methods have been used previously. Atomic layer deposition (ALD) of various dielectric materials can be used to inhibit recombination by providing a surface passivation layer to the dangling bonds.¹³ ALD, however, does not fundamentally repair the surface damage, adds an additional step to the process, and creates an extra layer that may not be necessary to the structure. An alternative method is the use of hydroxyl-based solutions, such as potassium hydroxide (KOH), which has been shown to remove surface states through wet etching.^12,14,15 The KOH method not only gets rid of the surface damage by wet etching the outer layer of the GaN NWs but also vertically shapes and shrinks the diameter of the structure in a single batch process. However, there is no comprehensive analysis on the effect of KOH solution conditions on the PL of InGaN/GaN NWs. Unlike previous works, which have shown the ability of the KOH-based solution to achieve passivation, this study will focus on giving a more comprehensive report on the relationship between PL and surface topography with regard to different etching conditions such as temperature and KOH concentration.

The wet etching method with KOH has potential to improve the performance of top-down NW LEDs. The damage recovery aspect of KOH and other hydroxyl-based chemicals has been reported, but not previously quantified.¹² This section focuses on the physics of the etching mechanisms and the corresponding effects on the GaN NW-based LEDs.

As mentioned previously, during a dry etch process using a reactive ion etch (RIE) or inductively coupled plasma (ICP), damage can result in the form of dangling bonds. In other GaN-based electronic devices, leaving the etched surface unmodified, the absence of proper passivation has shown to increase the gate leakage current by up to 6 orders of magnitude and decrease the drive current by up to 15%.¹⁶ Here, these surface states emerge from the dangling bonds, which can lead to Shockley-Read-Hall (SRH) recombinations.¹⁷ Without passivation, surface states on LEDs will cause a loss of energy through phonon emission instead of photon generation, while for power electronics, the leakage current is increased.¹⁶ Since NW structures have a much larger surface area, they are more heavily impacted by dry etch damage compared to planar devices. NW LEDs with small diameters of under 80 nm are particularly sensitive, due to radiative recombination dominating on the surface of the wires, which is caused by Fermi level pinning.¹⁸ High power and low pressure dry etches, needed for anisotropy, typically result in a physical sputtering regime, which is more likely to damage the NW sidewalls.¹⁹ 

To study the progression of the KOH etch to NWs, we first prepared a batch of top-down GaN NW LED samples using RIE. The progression can be observed in Fig. 1. Here, the GaN LED NWs were prepared through nanosphere lithography using 3 μm SiO₂ spheres as the hard mask to form the initial wire shape. The coated spheres were then RIE etched with SF₆ and O₂ in order to radially shrink the sphere diameters to increase the spacing between NWs [Fig. 1(a)]. The sphere diameters were reduced from 3 μm to ∼2 μm. RIE etching with 30 standard cubic centimeters per minute (sccm) Cl₂, 20 sccm Ar, at 65 mTorr, and 225 W in a Lam 4600 is performed in order to form the NWs through a top-down etch. KOH wet etching is then performed to perfect the NW shape, and shrink the diameter, as shown in Figs. 1(b) and 1(c). The source for the KOH is the AZ400K developer that contains 2% by weight concentration of KOH according to the safety data sheet and is a widely available commercial source for KOH. Due to the small concentrations of KOH utilized, AZ400K is the preferred source of KOH.^11,12

In order to develop an overview of the process, the etching properties of KOH must be understood. KOH and other hydroxyl-based chemicals have the property of selectively etching certain crystal planes. In silicon, these solutions etch faster along the 〈100〉 direction, revealing the 〈111〉 planes that etch slower.²⁰ Unlike many semiconductors that have a diamond or zinc-blende lattice, GaN has a wurtzite structure.²¹ When vertical GaN NW structures are exposed to hydroxyl-based chemicals, the $⟨11¯01⟩$ plane etches rapidly forming a step pattern until the $⟨11¯00⟩$ plane is exposed. Once the $⟨11¯00⟩$ plane is revealed, etching continues, though at a slower rate. These crystal planes are identified in Fig. 2. The wet etching that occurs will leave the gallium (Ga) polar 〈0001〉 plane untouched as Ga repels the hydroxyl groups, meaning the top surface remains unetched. This is due to the polarity-selectivity of KOH that etches N-polar, semipolar, and nonpolar planes. The Ga-polar plane remains unetched due to differences in the surface bonding energies.¹⁵ The etching rate difference between the Ga-polar plane and the other planes is believed to be caused by the other N-polar, semipolar, and nonpolar planes having lower binding energies.^11,15 The etch rate differences result in N-polar, semipolar, and nonpolar planes being etched more easily.

RIE is used with chlorine-based chemistries to first etch the initial wire shape that often approximates the $⟨11¯01⟩$ plane [Fig. 2(a)]. Upon immersing the postetched samples into hydroxyl-based chemicals, the $⟨11¯01⟩$ plane is rapidly etched away until the $⟨11¯00⟩$ plane is exposed, all while leaving the surface 〈0001〉 plane unetched [Figs. 2(b) and 2(c)]. The $⟨11¯00⟩$ plane revealed from the etch leads to the formation of perfectly vertical NWs that are needed for the fabrication of vertical devices. The revealing of the $⟨11¯00⟩$ plane makes this etch an essential step in completing the NW fabrication, as it corrects the cone shape left by the dry etch in a controlled manner. Continued wet etching leads to the $⟨11¯00⟩$ plane being more slowly etched away compared to the $⟨11¯01⟩$ plane, allowing the diameter of the wire to be selectively tuned [Fig. 2(c)]. The $⟨11¯00⟩$ plane etches slowly due to an equal distribution of Ga and N on the surface.¹¹ When etching LED structures, the p-type region at the top etches slower compared with the unintentionally doped GaN (u-GaN) or n-type GaN, due to the depletion of holes at the electrolyte-semiconductor interface.²² For the LED NWs, extended KOH etching produces a slight “torch” type shape at the top of the NWs [Fig. 2(d)].¹² The controlled NW diameter tuning is a desirable effect, as it allows for device optimizations. The ability of the KOH solution to etch certain crystal planes faster has already been established, and similar chemical treatment has been shown to improve the performance.¹⁵ However, the passivation-induced improvement in NW LEDs etched with KOH has yet to be investigated in detail, which is the main purpose behind the following experiments.

Based on the fundamental understanding of the KOH etch on GaN materials, fully characterizing the wet etch step is vital as it can be used to achieve optimal passivation and improve device performance. As such, the goal of this work is to study temperature, time, and KOH concentration to uncover the ideal experimental conditions. To achieve this goal, extensive investigations on the effectiveness of KOH solutions with various concentrations and temperatures in suppressing surface states on InGaN/GaN NWs and achieving controllable NW diameters have simultaneously been performed.

The effects of time and temperature on the etching of the NWs were the first aspects studied. Samples on a sapphire substrate with a 6.2 μm thick u-GaN and 15.5 nm of In_0.13Ga_0.87N/GaN MQW were used for these experiments. NWs of 550 nm in diameter were formed through nanosphere lithography and RIE etched to a height of ∼1 μm. A concentration of 5% AZ400K was utilized for these tests. Temperatures of 23 °C, 45 °C, and 80 °C were investigated at various time intervals, recording both scanning electron microscopy (SEM) images and PL intensity at each step. Sample intervals of 30 min were chosen for the 23 °C samples due to the slower etching rate, while for 45 °C and 80 °C, a sample rate of 10 min was selected. Results of these etching tests are displayed in Fig. 3. Higher etch rates for the $⟨11¯01⟩$ and $⟨11¯00⟩$ planes occur as temperature increases. For the $⟨11¯00⟩$ plane, etch rates for 23 °C, 45 °C, and 80 °C were found to be 0.28, 0.97, and 2.43 nm/min, respectively. The etch proceeds, alternating between a rough and a smooth surface as hydroxyl groups peel away layers, continuing in a step pattern. The step pattern follows a “cavity” etch model.^23,24 In this model, the etching proceeds in a step pattern, which can be thought of as the inverse of step growth. Atoms near the step edges have smaller binding energy and are more readily removed, compared to atoms on the crystal surface.^23,24

PL data taken at each interval is shown in Fig. 4. Here, the results exhibit an oscillating effect as the etch proceeds, continuing downward from the initial starting PL. The loss of PL intensity as AZ400K wet etching time increases is due to the reduced active region volume from the NWs formed by the initial low-damage dry etch. A low-power dry etch leads to negligible surface damage. As a result, an increase in PL intensity is observed compared with a planar structure, which is due to the increased NW surface to volume ratio. The oscillations are caused by the transitions around the MQW region, going between a rough and a smooth surface, from the step etching, as illustrated in Fig. 4(a). The decay effect recorded is due to the loss of the active region as the wire diameter shrinks from the starting size. Temperature increases led to higher etch rates and more observable steps as the thermal energy available increases. From the morphologies observed in Fig. 3, an etch temperature of 45 °C appears to produce the best etch rate, morphology, and PL trade-offs.

Concentration was next varied to determine the interaction on etch progression. Samples with the same growth structure as used in Fig. 3 are utilized. These NWs similarly had a diameter of 550 nm, patterned with nanosphere lithography, and RIE etched to a height of ∼1 μm. A fixed time of 30 min was chosen, with a fixed temperature of 45 °C. The concentrations of AZ400K were varied from 5% to 80% in de-ionized (DI) water. SEM images were taken for each of the samples, with results shown in Fig. 5. During the initial RIE, the wires form a tapered base due to etch shadowing that reduces the total wire height. At low AZ400K concentrations, this tapered base etches inward leaving overhangs, which do not easily etch away [Fig. 5(a)]. As AZ400K concentration increases to 20% and above, the base is fully removed, increasing the effective wire height. With an increased concentration, the KOH solution facilitates a vertical etch along certain planes, which were not easily removed at lower concentrations. Moreover, the increased concentration results in the $⟨11¯00⟩$ etching morphology transitioning from a “step” pattern to a pillar shaped etch profile. Compared with low concentrations, high concentrations show this pillar type of etch as whole layers are removed from the sidewalls, which is opposite to whole layers being deposited through growth as more reactants become available.^11,23,24 The etch rate is also expected to increase for samples with higher concentrations as the amount of available hydroxyl groups for etching increases. The increased concentration effectively increases the removal of the more weakly bound atoms on the step edges. The concentration of 40% AZ400K provides the best trade-off between the etch rate and morphology, with complete tapered base removal.

As detailed in the Theory and Motivation section, the required dry etch step for top-down GaN NW LEDs results in defects that cause the nonradiative recombination to increase. In order to study the damage created by a low pressure and high power RIE etching step, and the potential for recovery through KOH passivation, PL data were recorded. The samples for these tests were a 6.2 μm thick u-GaN on a sapphire substrate. The NWs were patterned with nanosphere lithography, using spheres of 550 nm in diameter, and were dry etched to a height of ∼1 μm. The use of u-GaN NWs provides a better study of the effects of dry etching on the whole wire surface and base, as compared with just the MQW region. The first u-GaN NW sample was fabricated using a power of 125 W and a chamber pressure of 100 mTorr, considered to be low power and high pressure. Figure 6(a) shows a significant intensity increase, where the PL spectra almost double at the peak, compared to the planar surface, as minimal damage occurs during the NW formation. The NW structure is expected to lead to a boost in PL as the extraction efficiency increases. In contrast, the second u-GaN NW sample is subjected to a lower pressure, higher power, NW dry etch using 150 W and 80 mTorr. This sample is expected to be more impacted by the RIE step. Figure 6(b) demonstrates the surface damage on the formed NWs through a PL intensity drop by over 40%, compared to the planar surface. The reduction in PL is caused by the newly created surface states along the NW surfaces and base. NWs inherently produce more light due to higher extraction efficiencies due to the NW surface to volume ratio, though the PL boost is overcome by the damage created during a harsh etch. Dry etching parameters are critical to determine the PL intensities. Lower power and higher pressure lead to a more chemical etching regime instead of a more physical sputtering of the surface. Most applications desire high aspect ratio NW structures, which often require a low pressure, high power etch.

To study the damage recovery aspects, the same lower pressure, higher power etch was performed as shown in Fig. 6(b). After the NWs are formed, and correspondingly there is a fourfold PL reduction at the highest point, the sample is immersed in 5% AZ400K at 23 °C for 10 and 20 min. The low concentration and temperature are chosen due to a slow etch rate, keeping the original diameter with a minimal reduction. After just 10 min in the solution, 90% of the original PL intensity is recovered, as the surface states have been wet etched away [Fig. 6(c)]. Continuous etching for 20 min leads to no added improvement as the surface damage has already been passivated. Additional etching beyond 20 min will start to shrink the NW diameter and lead to an oscillating reduction in PL observed in previous samples [Fig. 4]. The recovery after 10 min indicates that the wet etch has etched the entire NW sidewall, which removes these surface states. The original PL intensity is not recovered past the planar measurements as the u-GaN 〈0001〉 surface remains damaged. The work performed here shows that the KOH etch step is an effective way to recover damage from a low pressure and high power etch. This research suggests that hydroxyl-based chemicals may be very beneficial for the fabrication of more efficient NW-based LEDs.

The work presented here explored various effects of temperature, concentration, and damage recovery on the wet etching of GaN NWs through the KOH solution. Studies of temperature highlight important variations, with a moderate temperature of 45 °C to have the best trade-offs between etch rate, morphology, and PL. These results at 45 °C provide a relatively high etch rate of 0.97 nm/min along with smaller etch step sizes as per SEM observations. Concentrations of 40% AZ400K create a collimated morphology, from the transition going from etch steps to etch pillars, with the complete removal of the NW tapered base. The impact of small changes in pressure and power on the dry etch induced surface damage to GaN NWs were highlighted. RIE induced damage was removed from GaN NWs by 5% AZ400K exposure, recovering up to 90% of the original PL intensity using a 10 min wet etch. The oscillating behavior of the PL intensity with regard to the etch time was demonstrated for the first time, allowing for a better understanding of the hydroxyl-based GaN etching process on the formation of NW LEDs.

This work is partially supported by the Office of Naval Research (ONR) under Award No. N00014-16-1-2524 and by the National Science Foundation (NSF) under Award No. ECCS 1751675.