1. Introduction

Rapid etching of semiconductor substrates has received growing demand with the advent of hybrid platforms such as III-V on silicon. These applications require thinning of substrates to a few microns thickness [1]. A combination of mechanical grinding and wet etching is a common technique [2], but is undesirable because of the imperfections associated with lapping and the dust produced in processing toxic materials such as GaAs. Fast wet etching is a simple and reliable method; with the use of light, the etching rates can be enhanced by several times [3]. Light-assisted or photochemical (PC) etching can also be used for etching patterns without masks and for producing deep highly anisotropic structures [4,5], microlens arrays [6], grayscale structures [7], laser mirrors [8], diffractive optical elements [9], microfluidic devices [10], integrated lenses on light-emitting diodes [11], or through-wafer vias [12]. However, it is challenging to maintain smooth surface morphology during PC etching due to the fast etch rate. One feature that has been observed in wet etching and electrochemical etching of GaAs is the formation of micron-sized crystalline features [13,14]. In this paper, we discuss how to spatially localize the crystal formation in a PC etching experiment, and we also analyze the main factors that affect this process. With these results, it becomes possible to control the PC etch process to obtain fast etches with or without these crystal structures.

2. Experimental details

For the digital projection PC etching process, a pattern was first drawn using Microsoft PowerPoint and then was projected from a commercial digital light processing (DLP) projector (ViewSonic PJD7820HD, 3000 lumens, 1920 x 1080 pixels) onto the semiconductor sample after 8x demagnification by a lens system. The sample was placed in a chemical solution inside a Teflon container. Two types of chemical solutions were used: etchant A – a dilute hydrochloric acid and hydrogen peroxide mixture in the ratio of 1:1:50 HCl: H₂O₂: H₂O and etchant B – a dilute sulfuric acid and hydrogen peroxide mixture in the ratio of 4:1:80 H₂SO₄: H₂O₂: H₂O. These are two commonly used etchants for GaAs. We will compare them to determine the effect of etchant composition on microcrystal formation. Similar to most PC etching processes in the literature, no stirring was done during the etching because the moving liquid can distort the focus of the image. The samples used in all the experiments were n+ GaAs wafers, with Si-doping at 10¹⁸ cm⁻³. Both chemical solutions etched GaAs even in the absence of light, but in the regions with illumination, the etch rate was much faster because of the PC mechanism of etching. The etch rate in the absence of light using either etchant is in the range of 60-80 nm/min, whereas with white light it increases to 2-5 µm/min. Further, to study the effect of the substrate quality on the microcrystal formation, we used a mechanical grade 3” wafer from University Wafer for one experiment; everywhere else we used prime grade 3” wafers by AXT. No pre-treatment was done to the samples except general cleaning with acetone, isopropyl alcohol, and de-ionized water. The experiments were done in the dark with only the projected image shining onto the sample. The ambient environment was air. The projected patterns were all white in color, but their shapes and sizes were varied across experiments; this changed the average power of the light incident on the sample, but the local intensity at each pixel was constant.

The surface morphology was observed with a Hitachi S4800 scanning electron microscope (SEM) after various stages of etching. Chemical analysis of the microcrystals was done using an Energy Dispersive Spectroscope (EDS) built into a Hitachi S4700 SEM tool.

3. Results

The SEM images of the surfaces of three different samples etched with only etchant A, with A and then B, and with only etchant B, respectively, are shown for comparison in Fig. 1. The goal of this study was to determine the effect of etchant composition on microcrystal formation. The samples were PC etched using a pattern of a white circle of nearly 600 µm diameter, for 40, 15 + 30, and 35 minutes, respectively. The illumination pattern restricted the formation of micro-features to selective areas of the wafer.

 

Fig. 1 Sample surface of prime grade wafers after PC etching for (a) 40 minutes in etchant A, (b) 15 minutes in etchant A and then 30 minutes in etchant B, and (c) 35 minutes in etchant B. The slight differences in total etch time is not significant given the long total time and the slow microcrystal growth rate.

Download Full Size | PDF

Etching with etchant A forms etch pits and hillocks. In contrast, etching with etchant A and then B or with only etchant B shows a different morphology consisting of microcrystals, several tens of microns in size. It has been studied previously that wet etching GaAs can leave the surface rich in arsenic [13]. It has also been confirmed in the literature that these microcrystals are mainly composed of arsenic oxide (As₂O₃) [15]. Further, similar octahedral crystals are seen even as bulk arsenic oxide, known as arsenolite. Like most by-products produced during the processing of GaAs based semiconductor wafers, arsenic oxide is a toxic carcinogen. The materials safety data sheet (MSDS) describes health hazards and safety precautions and also indicates that arsenic oxide is soluble in dilute HCl [16]. Thus, we believe that the etch pits and hillocks seen in Fig. 1(a) are microcrystals that have formed but that have been partially dissolved by the dilute HCl of etchant A. The presence of oriented linear pits in Fig. 1(a) suggests that the dissolution reactions strongly attack specific planes and/or corners of the microcrystals.

An EDS analysis of the sample from Fig. 1(b) was performed to determine the surface chemical composition in different regions of the wafer. The results are shown in Fig. 2. They confirm that the microcrystals have distinct peaks for As and O and only small amounts of Ga (red curve), whereas non-microcrystal regions showed both Ga and As peaks (blue and black curves). The material composition in regions outside the microcrystals but within the PC etched zone (black curve) shows a larger As and O content than the non-illuminated “dark-etch” zone (blue curve), implying that PC etching with sulfuric acid leaves the surface rich in arsenic oxide. These two EDS results suggest that these microcrystal structures are rich in arsenic oxide, which is an intermediate in the etching process. Arsenic oxide may have a lower solubility than gallium oxide, thereby leading to arsenic oxide terminated surfaces.

 

Fig. 2 EDS spectra of different regions (red: microcrystal, black: PC etched, blue: dark etched) on a prime grade sample that produced microcrystals by being PC etched for 15 minutes in etchant A and 30 minutes in etchant B.

Download Full Size | PDF

To study the time-dependence of the growth of these crystals both in size and in number, we PC etched a single sample in etchant B for 5 minute intervals after one initial 15-minute etch in etchant A. The sample was taken out and cleaned each time before SEM imaging. It was then returned into a fresh solution of etchant B. Figure 3 shows the sequence of these images.

 

Fig. 3 SEM of prime grade sample surface etched for 15 minutes using etchant A followed by etching with etchant B for (a) 0, (b) 5, (c) 10, and (d) 15 minutes. Figure 3(a) was taken at a slight rotation compared to the other images.

Download Full Size | PDF

As before, after the first etch with etchant A, there are a few partially dissolved microcrystals. See Fig. 3(a). After 5 minutes in etchant B, an isolated microcrystal appears near the bottom of Fig. 3(b). Etchant B also creates significant surface roughness. Additionally, by tracking some of the microcrystals, we can see that their size grows in time. This has been observed previously [17] but not intensively quantified. The observed non-uniform growth suggests several more factors at play. It has been previously hypothesized that the microcrystals form because of the migration of surface arsenic oxide and its preferential precipitation onto existing crystallites [13]. Our results are consistent with this explanation. A recorded video of the growing microcrystals that supports the conclusion that they are formed by the migration of surface arsenic-oxide is presented in the Supplementary section. See Visualization 1. That sample was PC etched in etchant B only, without initially being in etchant A. The same white circle was shined onto the sample; it appears to shrink in size because as the etching proceeds the edges of the circle do not remain flat and, thus, light does not get reflected vertically. From the video, we observe that the microcrystals form and grow in size while the etching is taking place, even without removing the sample from the solution and exposing it to air.

From the time-dependent growth study in Fig. 3, we can further posit that the etch area is also a key factor. Between 0 – 5 minutes, only 1 microcrystal is formed and we attribute this to a low surface arsenic-oxide composition. Between 5 – 10 minutes, a number of medium-sized microcrystals appear, possibly because the surface arsenic-oxide content has increased. And finally between 10 – 15 minutes, the already-present microcrystals grow in size but a larger number of smaller microcrystals form, possibly because of the smaller migration area available due to occupation by the older microcrystals. The extremely tiny microcrystals in Fig. 3(d) (15 minutes in etchant A plus three 5 minute sessions in fresh etchant B) were not observed in Fig. 1(b) (15 minutes in etchant A plus 30 minutes continuously in etchant B). We hypothesize that this difference is because the etch process was interrupted for the Fig. 3 sample. In addition to fresh etchant being introduced, the migration of surface arsenic-oxide is modified when the sample is taken out and cleaned for SEM imaging.

The size of the etching area also affects the size of microcrystals formed. To confirm this, we performed another experiment where we illuminated patterns of the same intensity but different sizes in an array configuration. Both samples were first etched for 15 minutes using etchant A followed by 30 minutes of etching using etchant B. The SEM images are shown in Fig. 4. By projecting a 4 x 5 array of squares of side length 80 µm, only partially dissolved microcrystals are observed. For the 2 x 2 array of circles of diameter 230 µm, typically four microcrystals were found within each circle. This suggests that microcrystal formation and dissolution are affected by the etch area, possibly because a minimum number of arsenic-oxide molecules may be required for the migration to lead to crystallization.

 

Fig. 4 SEM of prime grade sample surface after PC etching with etchant A for 15 min and etchant B for 30 min for a projection of (a) a 4x5 array of white squares of 80 µm each, and (b) a 2x2 array of white circles of 230 µm each.

Download Full Size | PDF

Substrate-inherent defects also affect the quality and morphology of wet etched surfaces [18]. Several types of wet etchants have been used to reveal defects in crystal structures because the etching rate is different for the normal and defect regions. Because PC etching has a strong dependence on the band structure of the crystal, it has been used to qualify substrates by delineating many types of defects [19]. Therefore, we expect to see a difference in the microcrystal formation for substrates of different quality. In this study, we only used etchant B and did not pre-treat the surface with etchant A because we wanted to isolate the effect of wafer quality. As shown in Fig. 5, a network of microcrystalline features was observed for mechanical grade substrates rather than the individual microcrystals seen in Fig. 1(c) for prime grade substrates. The mechanical grade substrates have a higher etch pit density (EPD). Our observations of the network of microcrystalline features suggest that the crystal defects may promote nucleation sites for the crystallization of the arsenic oxide octahedrons.

 

Fig. 5 SEM of mechanical grade sample surface after PC etching with etchant B for 30 minutes.

Download Full Size | PDF

4. Discussion

The EDS spectrum distinguishes between the surface compositions of the three types of surfaces: a microcrystal surface from PC etching, an illuminated surface that did not form microcrystals during PC etching, and a non-illuminated surface that underwent wet chemical etching. PC etching is an electrochemical process due to the transfer of charge carriers involved in addition to chemical reactions. The absence of the As-O peaks in the non-illuminated wet chemically etched surface unlike on the PC etched surfaces suggests that surface residues are affected by the etch rate (much faster for PC etching). It has been observed in the past that the microcrystals form during exposure to air after wet etching or electrochemical etching with concentrated HF solution [13, 15]. It is clear from our results that it is not only the electrochemical or wet etching mechanism but also the type of solution, the etch rate, the area of the illuminated region, and the substrate quality that are key factors that determine the surface morphology with respect to these microcrystals. Moreover, the microcrystals can form in solution even without exposing the wafer surface to air.

5. Conclusions

In summary, the selective area formation of arsenic-oxide microcrystals on GaAs during PC etching was studied. Etch rate, solution type, area of illumination, and substrate quality were identified as major factors that affect crystal formation. A better understanding of the mechanism of the formation of these microcrystals was developed by using EDS to determine the arsenic-oxide content in different regions of the sample and by performing SEM imaging at different times during the etching process. For rapid etching of selective areas of a substrate or entire substrates, PC etching is a promising candidate. To obtain smooth featureless etches, appropriate measures must be taken to suppress microcrystal growth. These may include limiting the etch rate or the illumination area, using an appropriate solvent for arsenic oxide (e.g. dilute HCl), and/or controlling the relative solubility of arsenic oxide and gallium oxide.