**Curved Structure of Si by Improving Etching Direction Controllability in Magnetically Guided Metal-Assisted Chemical Etching**

#### **Tae Kyoung Kim 1, Jee-Hwan Bae 2, Juyoung Kim 2, Min Kyung Cho 2, Yu-Chan Kim 3, Sungho Jin <sup>1</sup> and Dongwon Chun 2,\***


Received: 1 July 2020; Accepted: 27 July 2020; Published: 30 July 2020

**Abstract:** Metal-assisted chemical etching (MACE) is widely used to fabricate micro-/nano-structured Si owing to its simplicity and cost-effectiveness. The technique of magnetically guided MACE, involving MACE with a tri-layer metal catalyst, was developed to improve etching speed as well as to adjust the etching direction using an external magnetic field. However, the controllability of the etching direction diminishes with an increase in the etching dimension, owing to the corrosion of Fe due to the etching solution; this impedes the wider application of this approach for the fabrication of complex micro Si structures. In this study, we modified a tri-layer metal catalyst (Au/Fe/Au), wherein the Fe layer was encapsulated to improve direction controllability; this improved controllability was achieved by protecting Fe against the corrosion caused by the etching solution. We demonstrated curved Si microgroove arrays via magnetically guided MACE with Fe encapsulated in the tri-layer catalyst. Furthermore, the curvature in the curved Si microarrays could be modulated via an external magnetic field, indicating that direction controllability could be maintained even for the magnetically guided MACE of bulk Si. The proposed fabrication method developed for producing curved Si microgroove arrays can be applied to electronic devices and micro-electromechanical systems.

**Keywords:** magnetically guided metal-assisted chemical etching; bulk Si etching; curved Si structure; catalyst encapsulation

#### **1. Introduction**

Fabrication techniques for micro/nano-structured Si, such as cantilever beams and bridge and buried microchannels fabricated using micro-electromechanical systems (MEMS) or nano-electromechanical systems (NEMS), have been intensively employed for a wide range of applications, including bio/chemical sensors [1,2], resonators [3], and power generators [4].

Dry-etching methods such as reactive ion etching (RIE) and deep reactive ion etching are important technologies for the micro/nano structuring of Si [5–7]. Recently, the Bosch process, which is a type of RIE, has been applied for the bulk etching of Si with a high aspect ratio, because it enables a high Si etching rate [8]. However, rough scallops on the surface and ion-induced defects impede the wider applications of this technique [9,10]. The deposition of C4F8, due to the passivation of the Si sidewall, is unfavorable when reactions are initiated at the Si surface, such as in battery anodes [10–13]. In addition, achieving control over the etching direction in dry-etching methods is difficult, thereby

hindering the application of these methods for the fabrication of complex Si structures so that a combined lithography process is required to fabricate MEMS/NEMS structures. An alternative approach is to employ wet Si etching methods using KOH or TMAH (tetramethylammonium hydroxide) etchants to fabricate micro/nano-structured Si; however, the etching direction in such approaches is significantly restricted by the crystalline orientation of Si substrates [14–16].

By contrast, metal-assisted chemical etching (MACE) of Si using noble metal catalysts has garnered substantial attention in the past decade for fabricating micro/nano structures because of its simplicity, low cost, and ability to fabricate high-aspect-ratio structures such as pillar and hole arrays without sidewall etching [17–21]. In general, noble metal catalyst such as Pt, Ag, and Au are used in MACE to significantly improve the etching rate. This is because the Si etching rate in an etchant solution without a catalyst is considerably low because of the low catalytic ability of the Si surface [22]; therefore, only the Si underneath the metal catalyst undergoes considerably etching. Hence, when noble metal catalysts are patterned on the Si surface and immersed in a solution of HF and H2O2 diluted in DI water, a faster Si etching rate can be obtained using MACE, as compared to the use of Si without a noble metal catalyst. This implies that the geometry of patterned Si can readily be controlled by the size of the noble metal catalyst, etching time, and etchant concentration in MACE. The MACE of Si has been to fabricate various Si structures such as nanowires and microstructures [23–29].

Ideally, MEMS structures such as Si cantilever and bridge arrays can be obtained via a simple MACE procedure, provided the etching direction can be easily manipulated during the etching process. However, similar to dry/wet etching methods, controlling the etching direction during MACE is difficult, as the etching direction can only be modulated via the crystalline orientation of the Si substrate [30–32] and the concentration of the etchant [33]. To fabricate complex Si structures such as cantilevers and bridges, a type of MEMS structure, the etching direction should be sequentially altered during the etching process. However, the limited control over the etching direction hinders the direct fabrication of the MEMS structure via the MACE of Si. Therefore, a versatile technique for achieving etching direction controllability during MACE is necessary to enable the fabrication of various Si structures that are in demand for industrial applications.

To improve the etching performance in terms of speed and direction control, we developed a magnetically guided MACE process utilizing a magnetic Fe layer encapsulated in a multi-layered metal catalyst [34–36]. Herein, we used a conventional Au/Fe/Au multi-layer metal catalyst instead of a single noble metal; this enabled the adjustment of the etching direction owing to the magnetic pulling force between Fe and an external hard magnet placed underneath the etching bath. The curved Si structures were fabricated by changing the external magnetic field direction during the MACE of Si, the pattern size of which was less than 1 μm [34,36]. Uniform curved Si structures were fabricated via this method; however, as the pattern size in this case exceeded 1 μm, non-uniform curved Si structures were fabricated owing to an inhomogeneous magnetic pulling force, which resulted from the corrosion of Fe due to HF [36]. In addition, the usage of Au-coated magnetic nanoparticles for the curved nanostructure during the magnetically guided MACE of Si is unfavorable if the pattern size needs to be manipulated on a larger scale [34]. Consequently, an appropriate method to suppress the corrosion of Fe due to HF is essential for the fabrication of curved and zig/zag Si microstructures with a massive production yield. This would also eventually facilitate the eventual direct fabrication of MEMS structures such as cantilevers and bridges via MACE.

In this study, we design a novel tri-layer metal catalyst, wherein Fe is encapsulated by Au to suppress its removal by HF, thereby producing a uniform magnetic pulling force. As a result, uniform curved Si structures could be obtained in a bulk scale via magnetically guided MACE. We demonstrate that the tri-layer metal catalyst with encapsulated Fe can suppress Fe corrosion, which, in turn, enables the production of uniformly curved Si structures, as compared to those achieved using a conventional tri-layer metal catalyst.

#### **2. Experimental**

Figure 1 summarizes the process of fabricating curved Si structures via magnetically guided MACE. A 500-μm-thick p-type Si (100) wafer (boron-doped; resistivity of 10–20 Ω·cm) was utilized in this study. The Si chip size used for the experiment was 2 <sup>×</sup> 2 cm2. The Si wafer was pre-cleaned in acetone and ethanol for 10 min each. To encapsulate the Fe in the tri-layer Au/Fe/Au catalyst, a photoresist (PR; S1827, MNX, Long Beach, CA, USA; approximately 2.5 μm) line patterning was produced on the Si wafer surface via photolithography (Karl Suss; MA6, SUSS MicroTec, Garching, Germany), as illustrated in Section 3. Subsequently, the Au catalyst layer (20 nm) was deposited on the PR grooves via e-beam evaporation (Temescal BJD 1800 E-beam evaporator, Ferrotec, Santa Clara, CA, USA)). After the PR was removed, the line patterning of Fe was performed to encapsulate Fe with Au, in which the patterning width was narrower than the lower Au layer. Thereafter, the Fe layer (10 nm) was deposited via e-beam evaporation, followed by the removal of the PR. Finally, the PR was patterned such that the pattern width was identical to the lower Au layer. This was also followed by Au deposition (20 nm) via e-beam evaporation to ensure that the Fe layer was protected by the etching solution during the magnetically guide MACE of Si.

**Figure 1.** Fabrication process for magnetically guided metal-assisted chemical etching (MACE) of Si: (**a**) photoresist (PR) patterning and bottom Au catalyst (20 nm thickness) deposition, (**b**) PR patterning and magnetic Fe layer (10 nm thickness deposition), (**c**) PR patterning and Au (20 nm thickness) deposition for Fe encapsulation, (**d**), (**e**) and (**f**) magnetically guided MACE of by changing permanent magnet position, (**g**) Si groove arrays after the removal of Au and PR by chemical etching.

After the fabrication of the Au/Fe/Au catalyst encapsulated in Fe via the deposition of an upper Au layer, a post annealing process was performed at 150 ◦C for 5 h to improve etching performance [37]. High-temperature annealing processes can enhance the ferromagnetic properties of the Fe layer, improving the etching speed and the controllability of the etching direction by increasing the magnetic pulling force. Furthermore, the porous surface structure of the catalyst can be produced via an annealing process that facilitates the diffusion of etchant through the metal catalyst.

A mixture of diluted hydrofluoric acid (2 M) and hydrogen peroxide (9 M) diluted in deionized water was used as an etchant. The magnetically guided MACE of Si was conducted in a Teflon bath; a neodymium iron boron (NdFeB) permanent magnet (4 cm × 4 cm × 1 cm height) with an energy product strength of 35 MGOe was placed outside and underneath this bath to adjust the etching direction during the process. Si chips with the patterned tri-layer Au/Fe/Au catalyst encapsulated in Fe were immersed in the etchant for the magnetically guided MACE. The Si covered by the catalyst was etched faster than that covered by the PR, which can be explained as follows [38].

The reduction of hydrogen peroxide (H2O2) on the noble metal catalyst surface generates holes (h+) as follows:

$$\text{H}\_2\text{O}\_2 + 2\text{H}^+ \rightarrow 2\text{H}\_2\text{O} + 2\text{h}^+ \tag{1}$$

Noble metal catalysts such as Au, Ag, and Pt facilitate the reduction of H2O2, thereby improving the etching speed by providing sufficient h<sup>+</sup> than the pure Si surface. The holes generated near the noble metal surface pass through the noble metal surface and are injected into the valence band of Si, because the valance band potential of Si is lower than the redox potential between H2O and H2O2 [22]. This causes the dissolution of Si with HF, as follows:

$$\rm{Si} + 4\rm{h}^+ + 6\rm{HF} \to \rm{SiF}\_6^{2-} + 6\rm{H}^+ \tag{2}$$

During the magnetically guided MACE of Si, the permanent magnet was shifted manually to alter the orientation of the magnetic field, which, in turn, changed the etching direction. The Si surface covered by the Au catalyst was etched along the direction of the external magnetic field because of both the enhanced etching rate through the catalyst reaction and the magnetic pulling force between the patterned ferromagnetic Fe and the hard magnet; the area covered by PR was protected because of its resistance to the etchant. The magnetically guided MACE of Si was performed at room temperature; after the etching process was completed, the sample was immersed in acetone and gold etchant to remove the remaining PR and metal catalyst. The etched Si chips were then washed with deionized water. Subsequently, the structure of the tri-layer metal catalyst of Au/Fe/Au and the etched Si structure were characterized using scanning electron microscopy (SEM; Regulus8230, Hitachi, Tokyo, Japan; Phillips XL30 ESEM, FEI, Hillsboro, OR, USA)). In addition, the surface morphology of the tri-layer metal catalyst was characterized via atomic force microscopy (AFM; Park XE7, Park Systems, Suwon, South Korea).

#### **3. Results and Discussion**

Figure 2 presents the top-view SEM and energy-dispersive spectroscopy (EDS) images of the tri-layer metal catalyst with encapsulated Fe after the annealing process. Figure 2a,b show the tri-layer metal catalyst with Au (50-μm wide)/Fe (40-μm wide)/Au (47-μm wide) line patterns with 100-μm spacing fabricated on the Si surface via photolithography. This was confirmed by the EDS images of Fe and Au, as shown in Figure 2c; a 45-μm wide Fe layer is patterned between the Au layers. The EDS spectrum at point #2 exhibits sharp peaks at 2.123 KeV and 1.740 KeV corresponding to Au (Mα1) and Si (Kα1), respectively, whereas only an Si peak was detected at the EDS spectrum of #1, indicating that the lower (50 μm) and upper (47 μm) Au patterns were fabricated successfully. The width of the upper Au layer is narrower than that of the lower Au layer; this discrepancy might be due to a fabrication error. For the EDS spectrum at point #3, Fe Lα<sup>1</sup> peaks at 0.705 KeV corresponding to Au and Si (Figure 2d) were observed. This indicates that the tri-layer metal catalyst with encapsulated Fe was fabricated successfully.

As the diffusion of the etchant depends on the surface morphology of the metal catalyst [21], porosity is an important factor influencing the etching speed. It has also been reported that enhanced porosity in the metal catalyst results in a higher etching speed [36]. Thus, to improve the porosity of the metal catalyst, we employed an annealing process after the formation of the tri-layer metal catalyst with Fe encapsulation. However, the changes in the surface morphology of the catalyst caused by annealing could not be clearly observed due to limitations of the image resolution of SEM. Hence, AFM was employed to analyze the surface morphologies before (Figure 3a) and after (Figure 3b) the annealing process. The root-mean-square roughness of the annealed metal catalyst was 1.038 nm (Figure 3b), whereas that of the as-deposited metal catalyst was 0.532 nm; this indicates that annealing provides a rough surface morphology that favors a higher etching during the magnetically guided MACE of Si, due to the shorter diffusion length of the etching solution [21].

**Figure 2.** Scanning electron microscopy (SEM) images and energy-dispersive spectroscopy (EDS) images and spectrums of the tri-layer Au/Fe/Au metal catalyst encapsulating Fe after the annealing process.

**Figure 3.** Atomic force microscopy (AFM) images of the tri-layer Au/Fe/Au metal catalyst with Fe encapsulation before (**a**) and after (**b**) the annealing process.

Figure 4 shows the cross-sectional SEM images of vertical groove arrays produced via the magnetically guided MACE of Si for 3 h (Figure 4a) and 5 h (Figure 4b). The tri-layer metal catalyst of Au (20-μm wide)/Fe (15-μm wide)/Au (20-μm wide) lines spaced at 100 μm intervals were patterned on the surface of Si encapsulated with Fe, and the MACE of Si was performed under the magnetic field. During the etching, the strong magnet was placed underneath the Teflon beaker, and its position remained fixed for vertical etching.

**Figure 4.** Cross-sectional SEM images of vertical Si groove arrays produced by magnetically guided MACE for etching durations of (**a**) 3 h and (**b**) 5 h.

Figure 4 confirms that the Au/Fe/Au tri-layer metal catalyst with encapsulated Fe is suitable for the magnetically guided MACE of Si in the vertical direction. The measured groove thickness for the etching of 3 h was 150 μm, while that for the etching of 5 h was 240 μm; these correspond to etching speeds of 0.83 μm/min (Figure 4a) and 0.8 μm/min (Figure 4b), respectively. Likewise, a reduction in the etching speed with etching time was observed in our previous results, which were obtained using a conventional Au/Fe/Au tri-layer metal catalyst [34,35]. A direct comparison between etching speed and catalyst structure is difficult because etching speed depends on not only the catalyst structure but etchant concentration and the size of the catalyst patterned on the Si surface during magnetically guided MACE. However, the reduction in etching speed with increasing time appears to be mitigated by the encapsulation of Fe, as compared to the etching speed for a conventional Au/Fe/Au catalyst [34,35]. This reduction in etching speed with the increase in etching time can be attributed to the reduction in the volume of Fe caused by the HF-induced corrosion, which decreases the magnetic pulling force between Fe and the permanent magnet. Unlike the conventional Au/Fe/Au tri-layer catalyst, the sidewalls of the Fe layer in our catalyst remain protected due to the encapsulation by Au; this effectively suppresses Fe corrosion. As a result, the reduction in etching speed with an increase in etching time is not appreciable during the magnetically guided MACE of Si. It should be noted that a slightly tapered structure was produced. The loss at both Au layers in the metal catalyst contribute toward the corrosion of Fe, resulting in a reduction of catalyst dimensions with increasing etching

time; this could potentially explain the tapered Si structure obtained. Moreover, this indicates that the corrosion of Fe could not be prevented completely, even when Fe is encapsulated by Au.

Figure 5 shows the Si groove arrays produced by magnetically guided MACE for 5 h using a conventional tri-layer catalyst (Figure 5a) and the tri-layer catalyst with encapsulated Fe (Figure 5b). The tri-layer metal catalyst of Au (50 μm wide)/Fe (40 μm wide)/Au (50 μm wide) line patterns were produced on the surface of Si encapsulated by Fe. To adjust the etching direction for fabricating a curved Si structure via magnetically guided MACE, the magnet underneath the beaker was moved gradually from a vertical position to a tilting angle of 30◦ during the etching.

**Figure 5.** Curved Si groove arrays produced by magnetically guided MACE for 5 h using (**a**) a Swiss-cheese-like Au/Fe/Au catalyst and (**b**) Au/Feencapsulated/Au catalyst (tilting angle: 30◦).

As shown in Figure 5a, 60-μm-thick and non-uniform Si grooves with rough surfaces were produced by etching with a conventional tri-layer catalyst, although rougher surface morphology was produced by annealing process, as shown in Figure 3, which reduces the diffusion length of the etchant. Changing the etching direction results in an unstable Si/catalyst interface, thereby facilitating Fe corrosion because the etchant can readily penetrate the interface and etch the Fe. Therefore, as the Fe is not sufficiently protected given the changing direction of the magnetic field, magnetically guided MACE of Si is infeasible due to the reduced and inhomogeneous magnetic pulling force. In contrast, uniform 150-μm-wide curved Si groove arrays with smooth surfaces were produced during the magnetically guided MACE of Si using the tri-layer metal catalyst with encapsulated Fe, as shown in Figure 6b; this indicates that the etching direction can be effectively controlled by changing the

direction of the magnetic field when the catalyst includes encapsulated Fe, thereby proving that the HF-induced corrosion of Fe can be effectively suppressed via encapsulation.

**Figure 6.** Curved Si groove arrays produced by magnetically guided MACE for 5 h using (**a**) a Swiss-cheese-like Au/Fe/Au catalyst and (**b**) Au / Feencapsulated /Au catalyst (tilting angle: 90◦).

The etching speeds calculated from the thickness of the measured grooves in Figures 5a and 4b are approximately 2 μm/min and 0.5 μm/min, respectively. Fe corrosion could result in partial Au ejection because of the strong adhesive force between Fe/Au, as compared with that for Au/Si. This would result in a lower etching speed and a rougher surface when using magnetically guided MACE of Si with a conventional tri-layer metal catalyst, as shown in Figure 6a. In contrast, a higher etching speed and a smooth surface morphology were obtained when the tri-layer metal catalyst with encapsulated Fe was used, indicating the partial Au ejection was successfully prevented by the suppression of Fe corrosion.

Figure 6 presents the curved Si groove arrays produced by magnetically guided MACE for 5 h using the tri-layer metal catalyst with encapsulated Fe. The tri-layer metal catalyst of Au (50-μm wide)/Fe (40-μm wide)/Au (50-μm wide) line patterns spaced at 100 μm intervals were fabricated on the surface of Si. First, the magnetically guided MACE of Si was performed by using a strong magnet placed vertically under a beaker. After this vertical etching for 1 h, the position of the hard magnet was gradually changed to a 90◦ tilting angle, and etching was conducted for 2 h under a lateral magnetic field.

As shown in Figure 6, a curved Si structure with a smooth surface was produced via magnetically guided MACE when the tri-layer metal catalyst with encapsulated Fe was used. As expected, the curvature of the Si structure was enhanced on increasing the tilting angle from 30◦ to 90◦, indicating that the etching direction could be adjusted effectively by using the external magnetic field.

Finally, the bulk micromachining of Si with a curved structure was successful conducted via the magnetically guided MACE of Si using the metal catalyst with Fe encapsulated in an Au (50 μm wide)/Fe (40 μm wide)/Au (50 μm wide) tri-layer. Therefore, it is evident that suppressing the corrosion of Fe, through Fe encapsulation, improves the controllability of etching direction during the magnetically guided MACE of Si.

#### **4. Conclusions**

Uniform curved Si groove arrays were produced using encapsulated Fe in a tri-layer metal catalyst (Au/Fe/Au); this catalyst improved the controllability of etching direction during magnetically guided MACE. A magnetic layer in a tri-layer metal catalyst was used to modulate the etching direction via a magnetic pulling force. To improve the direction controllability during magnetically guided MACE, the Fe layer was encapsulated with Au, which suppresses the HF-induced corrosion of Fe, thereby enabling the production of a uniform magnetic pulling force.

Uniform curved Si groove arrays were produced via magnetically guided MACE using this Fe encapsulation, whereby the etching direction was manipulated by moving the external strong magnet during this process. By contrast, non-uniform Si groove arrays with rougher surfaces were obtained when using a conventional tri-layer metal catalyst. This indicates that the encapsulation of Fe with Au can effectively protect Fe against corrosion due to HF, resulting in an enhanced direction controllability. Furthermore, it was observed that the curvature of Si groove arrays can be modulated by adjusting the tilting angle of the external strong magnet, when using encapsulated Fe in the tri-layer metal catalyst.

We also attempted to fabricate complex Si structures such as Si cantilever arrays and zig-zag microwire arrays; however, this could not be achieved using the proposed approach. On increasing the etching time, the controllability of etching direction decreases, which is likely caused by the corrosion of Fe due to HF. In addition, depletion of the etchant around etched areas between the metal catalyst and Si could result in a reduced etching rate as well as poor controllability over etching direction. To directly fabricate MEMS structures via the magnetically guided MACE of Si, magnetic alloy materials such as AuFe and CoFe need to be used as the metal catalyst in order to reduce the corrosion of the magnetic layer by HF. In addition, the etching process should be conducted under an etchant circulation system to ensure sufficient etchant reaches the reactive area which remain as future works. Although complex Si structures could not be achieved via magnetically guided MACE with Fe encapsulation, the results of this study prove that encapsulation helps prevent Fe corrosion, resulting in a uniform magnetic pulling force. Consequently, uniform curved Si structures can be fabricated in bulk by improving the controllability of etching direction during magnetically guided MACE.

**Author Contributions:** T.K.K., Y.-C.K., S.J., and D.C. designed the experiments; T.K.K., J.-H.B., and D.C. wrote the manuscript. T.K.K. and M.K.C. prepared Figures 2–6; J.-H.B. and J.K. prepared Figure 1, Figure 2, and Figure 6. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the National Research Foundation (NRF) of Korea under Grant 2018M3D1A1058793, funded by the government of Korea.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Black GaAs: Gold-Assisted Chemical Etching for Light Trapping and Photon Recycling**

#### **Paola Lova \*,**† **and Cesare Soci**

School of Physical and Mathematical Sciences, Division of Physics and Applied Physics,

Nanyang Technological University, 21 Nanyang Link, Singapore 637371, Singapore; csoci@ntu.edu.sg

**\*** Correspondence: lova.paola@unige.it; Tel.: +39-010-353-6192

† Current Address: Dipartimento di Chimica e Chimica Industriale, Università degli Studi di Genova, Via Dodecaneso 31, 16146 Genova, Italy.

Received: 17 April 2020; Accepted: 4 June 2020; Published: 5 June 2020

**Abstract:** Thanks to its excellent semiconductor properties, like high charge carrier mobility and absorption coefficient in the near infrared spectral region, GaAs is the material of choice for thin film photovoltaic devices. Because of its high reflectivity, surface microstructuring is a viable approach to further enhance photon absorption of GaAs and improve photovoltaic performance. To this end, metal-assisted chemical etching represents a simple, low-cost, and easy to scale-up microstructuring method, particularly when compared to dry etching methods. In this work, we show that the etched GaAs (*black GaAs*) has exceptional light trapping properties inducing a 120 times lower surface reflectance than that of polished GaAs and that the structured surface favors photon recycling. As a proof of principle, we investigate photon reabsorption in hybrid GaAs:poly (3-hexylthiophene) heterointerfaces.

**Keywords:** metal-assisted chemical etching; antireflection; *black GaAs*; photon recycling

#### **1. Introduction**

In thelast decades several antireflective coatings, scatteringlayers and photon recyclingstructures have been developed to enhance photovoltaic performances in devices based on inorganic semiconductors, [1–3] donor-acceptor organic [4] and hybrid [5,6] systems, perovskites [7], and for light transmitting panels based on luminescent solar concentrators [8]. Photon absorption can also be enhanced using scattering and light trapping surfaces [9–12] or photonic crystals [13–15] that aim to increase the light optical path, to confine light in the active layer, or to recycle emitted photons [16]. Most of these structures are obtained by lithographic methods that produce high quality and homogeneous patterns, but often do not provide the high throughput necessary for the fabrication of large area devices. To this end, wet chemical methods are a viable, low-cost, and easy to scale-up alternative to lithographic patterning. Among these processes, metal-assisted chemical etching (MacEtch) became a paradigm for silicon structuring to produce surfaces with near-zero reflectance [17–25]. Etched black silicon is indeed a commercial product, and entered mass production of solar cells and modules [26,27]. The high throughput demonstrated for silicon made wet etching processes interesting also for translation to other semiconductors [28].

MacEtch relies on the dissolution of a semiconductor surface catalyzed by a metal in a bath containing alkaline or acidic oxidizing agents [17,20,23,29–31]. The catalyst favors anisotropic substrate dissolution. Then, the substrate can be structured by patterned metal films [32,33] or etched randomly using dispersed metallic particles [24]. In recent years, several attempts aimed to extend MacEtch to III-V group semiconductors [32–43], which yield better device characteristics in light emitting diodes and solar cells compared to mainstream silicon and germanium [44,45]. GaAs structuring

via MacEtch has been reported in conjunction with catalyst vacuum depositions [46], or with metal patterning [32–41] by nanoimprint lithography [47], photolithography [20], and microsphere self-assembly [48]. While lithographic depositions allow highly controlled nanostructuring, they have limited room for scaling-up the fabrication. Recently, lithography-free MacEtch of GaAs was demonstrated by both electrode [49] and electrodeless deposition of gold nanoparticles [12]. In the first case, etching the (100) GaAs surface in a KMnO4:HF bath yields nanowire arrays similar to those reported for etched black silicon. This approach generates an effective medium with reflectance as low as 4% [49]. In the second case, etching in H2O2/HF baths induces crystal plane dependent dissolution rates and then light trapping surfaces with reflectance lower than 2%. [12] In this case, the crystal plane dependent etching rate is favored by the different reactivity of the Ga and As sp3 atoms [12]. A schematic of the etching reactions is illustrated in Figure 1. Specifically, a gold nanoparticle previously cast on the substrate is oxidized by the peroxide in the bath. The oxidized gold cations diffuse along the semiconductor surface and are selectively reduced at an arsenic site. Then, the precipitated catalyst can be further oxidized or form a electrochemical cell for gold deposition and surface etching [12,50]. Although catalyst ions diffusion and precipitation mechanism apparently contradict the etching mechanism reported in the seminal paper by Li and Bohn [17] and further confirmed by a number of researches [32], catalyst dissolution and redeposition have been demonstrated for two-step etching processes [50,51]. The selective auric reduction is favored by the higher reactivity of As with respect to Ga in the structure [52]. Indeed, the zinc-blende structure stabilizes trivalent Ga, and destabilizes As, which has valence five in electrophilic environment and an easy-to-oxidize pair of electrons. Arsenic is then oxidized to arsenic and arsenious acids, while Ga is complexed by fluorine anions allowing migration of the catalyst within the inner surface. Normally, monoatomic Ga planes, which only expose Ga atoms, are mildly reactive in the etchant solution. However, the catalytic effect of migrating gold ions is inhibited when a continuous catalyst film is cast on the semiconductor surface. In this case, the etching process does not depend on the crystal plane exposed to the bath [33,53–57].

**Figure 1.** Schematic of the metal-assisted chemical etching (MacEtch) process of a GaAs (111)B surface.

As mentioned above, while lithographic patterning limits the scalability and throughput of the etching process, lithography-free MacEtch is a start-to-end solution process that is easy to scale up to wafer size. The etching of different GaAs crystallographic orientations was previously reported and light trapping properties were reported for etched GaAs (111)B and (100) surfaces [12]. In this work, we investigate the possibility to employ the etched structures for photon recycling in hybrid polymer-inorganic heterointerfaces. Reabsorption of emitted photons is indeed particularly desirable in hybrid photovoltaic devices which suffer of low power conversion efficiency related to high charges recombination and low charge carrier mobility [58–60]. The possibility to reabsorb photons emitted upon charge recombination, which would otherwise increase losses, could improve the performance of these devices [58–60]. It was indeed demonstrated that charge generation at hybrid heterointerfaces occurs in both the polymer and in the inorganic material [6]. Photon reabsorption, therefore, could lead to larger generation yield in hybrid devices. To assess the possibility to employ the process for the fabrication of photon recycling surfaces, as a proof of principle, we investigate the photoluminescence reabsorption occurring at the heterointerface between structured GaAs and poly(3-hexiltiphene) (P3HT), which is a widely studied conjugated polymer in hybrid and tandem photovoltaic devices.

#### **2. Materials and Methods**

**Metal-assisted chemical etching**: the MacEtch was performed at room temperature on *n*-type (100), (110), (111)B, and (211) epi-ready GaAs wafers (Axt Inc, Fremont, CA, USA). In the process, gold nanoparticles were first deposited on the GaAs surfaces by immersion in a water solution containing 0.1 mM of AuCl3 (Sigma Aldrich, Saint Louis, MO, USA). The samples were successively blow-dried and etched in a bath of HF and H2O2 (4:1) in a Teflon® container (DuPont Wilmington, DE, United States) for about 10 min under vigorous stirring. The samples were then removed from the bath, rinsed in deionized water, and finally blow-dried. Au nanoparticles removal was performed by further etching of the (111)B microstructured surface in aqua regia (HNO3/HCl = 1:3) with different dilution in water (not diluted, 1:5, and 1:10).

**P3HT deposition:** P3HT was deposited on quartz substrates (Hellma Müllheim, Baden Württemberg, Germany) and on polished and etched GaAs by spin coating a 10% (*w*/*v*) solution in dichlorobenzene at a rotation speed of 800 rpm. All the samples were annealed for 15 min at 120 ◦C on a hotplate (VWR, Radnor, PA, United States). Deposition and annealing were performed in nitrogen environment to avoid polymer oxidation.

**Structural and optical characterization:** Gold nanoparticles and microstructure images of the etched GaAs samples were collected by scanning electron microscopy (SEM) with a field emission Jeol JSM-6700F (Jeol, Akishima, Japan) endowed with a secondary electron detector. Particle size distribution was retrieved manually. Representations of the GaAs crystal lattice were elaborated with the software VESTA (JP-Minerals, Ibaraki, Japan) [61] using data retrieved from the Crystallography Open Database [62]. Normal incidence reflectance was collected with a Bruker Vertex 80v Fourier transform spectrometer (Bruker, Billica, MA, USA) coupled with a Bruker HYPERION microscope (Bruker). Photoluminescence spectra were collected at room temperature using a Horiba Fluorolog spectrofluorometer (Horiba, Kyoto, Japan) with a CCD detector, a Xenon lamp as the exciting source, and a monochromator to select excitation wavelength. The measurements were collected exciting both P3HT and GaAs at 500 nm.

#### **3. Results and Discussion**

The effective electrodeless deposition of Au nanoparticles on the GaAs surface was confirmed by SEM measurements. Figure 2a reports a micrograph of the pristine particles, which are distributed densely and homogeneously on the GaAs surface. The diameter distribution shown in Figure 2c in black color displays that the catalyst has a broad size dispersion, whit diameters ranging from 3 nm to about 50 nm, while the larger fraction has a diameter approaching 7 nm. These data suggest the presence of particles aggregates on the semiconductor surface, which are not easily recognizable form single particles owing to the instrumental resolution. Figure 2b reports instead a SEM micrograph showing the gold particles on the surface after the etching process. In this case, they are not homogeneously distributed over the sample. Indeed after the etching, the catalysts are located within the etched features (see Figures 3 and 4). However, the particles dimension is more homogenous with respect to the previous case. The size dispersion shown in red in Figure 2c indicates a sharper distribution with diameters ranging from 6 to 35 nm, with the larger percentage having a diameter equal to 11 nm. The variations in the particles size and distribution can be attributed to the continuous

gold oxidation and reduction, which implies its dissolution and reprecipitation may decrease the diameter polydispersity.

**Figure 2.** SEM micrographs showing Au nanoparticles on the GaAs surface as casted (**a**) and after the etching process. (**c**) Particles size distribution before (black) and after (**b**) GaAs MacEtch.

Figure 3 shows the SEM micrographs collected after etching of the (100) (panels a, a'), (110) (panels b, b'), (111)B (panels c, c'), and (211) (panels d, d') surfaces. Similar results were discussed in a previous

work [12], but the brief analysis presented here provides a better interpretation of photon recycling mechanism of the etched surface. The top panels (a–d) of the figure display the cross-sectional micrographs collected for the different crystalline planes, while the bottom panels (a'–d') show a tilted view at lower magnification. Etching of the (100) GaAs crystalline plane provides facets with ca. 25◦ characteristic angle with the sample plane and size of about 30 μm (Figure 3a). While these facets are partly exposed on the sample surface, the micrograph also shows the formation of holes beneath the surface. Figure 3a' shows that these facets cover a large area, giving a homogeneous appearance to the surface. Figure 3b,b' displays the results for the (110) plane. It is clear from Figure 3b that the surface exposes facets with similar orientation and size than those observed for the (100) surface. Panel b' shows again a homogeneous distribution of etched planes on the entire sample surface. In the case of the (111)B surface, the etching generates randomly oriented hillocks (Figure 3c) and a homogenous surface (Figure 3c'). The facets on the etched (111)B orientation form larger angles than those observed in the (100) and (110) planes. Etching of the (211) plane digs conic-like holes into the surface and some spikes. These structures form again characteristic angles with respect to the sample plane (Figure 3d). Additionally, in this case, the resulting microstructures are homogeneously distributed and appear similar to those formed on the (111)B surface (Figure 3d'). Comparing the features size, it is possible to notice that the (111) surface provides less deep features than those arising from (211), (100), and (110) ones, which are instead similar in size. We suggest that the three surfaces showing deeper features are indeed etched at a faster rate, as monoatomic low-rate gallium planes are exposed to the etchant only at a later stage (i.e., the visible features in Figure 3a,b,d). Conversely, the etching of the highly reactive (111)B plane exposes the underlying (111)A plane (see Figure 1) where gallium atoms with mild reactivity can slow down the process.

The optical characteristics of the samples can be understood in terms of geometrical light trapping. The feature size, of the order of a few micrometers, is too large relative to visible and near infrared wavelengths to be treated within the effective medium theory. Light is therefore trapped within the etched surfaces upon multiple specular reflections. Figure 3e compares the reflectance spectra of the four etched samples with the one of a polished GaAs surface. The reflectance of polished GaAs approaches 0.7 in the range between 1100 and 925 nm. Moving toward shorter wavelengths, the values decreases to about 0.6 below 925 nm and remains rather constant until 700 nm where it increases and reaches ~0.85 at 550 nm. Conversely, the reflectance of etched GaAs approaches zero in the entire spectral range. The spectra are better visible in Figure 3f, where the intensity scale has been expanded. Within the GaAs transparency region, the minimum reflectance values are 0.013, 0.01, 0.008, and 0.007 for the (100), (110), (111)B, and (211) surfaces, respectively. Therefore, in the best case, MacEtch induces a 120-fold reduction of light reflection for the (211) surface at 550 nm (0.85/0.007).

**Figure 3.** *Cont*.

**Figure 3.** Cross-sectional (**a**–**d**) and tilted view (**a'**–**d'**) SEM micrographs of the (100) (**a**,**a'**), (110) (**b**,**b'**), (111)B (**c**,**c'**), and (211) (**d**,**d'**) GaAs crystalline planes. (**e**) Normal incidence reflectance spectra of polished (black) and etched GaAs surfaces: (100), red; (110), green; (111)B, blue; (211), cyan. (**f**) Reflectance spectra of the etched samples: A magnification of the reflectance scale allows to appreciate the spectral features.

Because the presence of residual gold would be detrimental for devices, we tested the possibility to dissolve the particles through a second etching step in aqua regia. Figure 4a reports the cross-sectional micrograph of the (111)B surface after the two MacEtch processes (see experimental section). There, gold is barely visible in the sample. However, several features on the surface appears broken. To reduce this detrimental effect, we also tested diluted etchants. Figure 4b shows the micrograph of a sample after etching with the acidic solution diluted 1:5 with water. Also in this case gold is barely visible, while the damages to the features are less evident than in the previous case. After further diluting the solution (1:10), the features are not damaged at all, but gold nanoparticles are still clearly visible in the micrograph as bright spots at the base of the etched features (Figure 4c). Then, we can infer that the dilution 1:5 can successfully remove the gold particles at a low expense for the microstructure quality.

**Figure 4.** Cross-sectional SEM micrographs of the (111)B etched surface after 30 min treatment in pure aqua regia (**a**), and in aqua regia diluted in water with proportion 1:5 (**b**) and 1:10 (**c**).

As mentioned earlier, light trapping is a promising mechanism for photon recycling in hybrid polymer-inorganic optoelectronic devices. To demonstrate that the etched GaAs favors reabsorption of emitted photons, as a proof of principle, we cast a thin film of P3HT on the etched (111)B surface and investigated the resulting optical properties of the hybrid system. Figure 5a compares the reflectance spectra of the bare *black GaAs* (black line) with the one collected for *black GaAs* covered with P3HT (red line). The presence of P3HT increases the overall reflectance. While the bare *black GaAs* reflectance approaches zero in the entire spectral range, addition of the P3HT film generates a structured spectrum. The data show a maximum at 1100 nm, where the reflectance approaches 0.15. Moving toward shorter wavelengths, the intensity decreases slowly until ca. 870 nm. At energies above the GaAs energy gap, the reflectance then decreases with a faster rate until 660 nm, where it approaches zero. On the short wavelength side, a broad peak is detected between 420 and 660 nm, with a maximum at 505 nm and structures at 470, 570, and 611 nm, which are assigned to P3HT [6,63–65]. Overall, these data confirm the effective coverage of the surface by the polymer. To assess the effectiveness of photon recycling, we excited both P3HT and GaAs at 500 nm and compared the emission spectra of flat and etched surfaces. Photon recycling is expected to yield reduction of the P3HT photoluminescence and enhancement of the GaAs photoluminescence signals upon photon reabsorption in GaAs, favored by the trapping geometry. Figure 5b compares the emission spectrum collected for bare P3HT (green line) with P3HT cast on polished (red line) and *black GaAs* (black line). The bare P3HT emission spectrum ranges between 600 nm and 900 nm with a maximum of intensity

at ~720 nm (green line) [66,67]. In the polished GaAs/P3HT spectrum no other features than those assigned to P3HT are evident and the emission intensity is lowered, suggesting electron transfer at the P3HT:GaAs interface (red line) [6]. The *black GaAs*/P3HT sample is characterized by stronger P3HT photoluminescence reduction compared to the flat sample, and by the presence of a peak at ~875 nm assigned to GaAs emission (black line) [12,33,68,69]. Quenching of the P3HT emission in the etched sample is consistent with geometrical light trapping discussed previously. Indeed, as sketched in Figure 5c, photons emitted from P3HT undergo several reflections between the surface features before they can leave the system, favoring reabsorption in GaAs. Because P3HT emission occurs below the HOMO-LUMO transition of the polymer, emitted photons cannot be reabsorbed by P3HT and therefore excite the underlying GaAs.

**Figure 5.** (**a**) Reflectance spectra collected for polished (red) and etched (black) GaAs (111)B surfaces covered with a P3HT thin layer. (**b**) Photoluminescence spectra collected for a P3HT film cast on a quartz substrate (green line), on polished GaAs (red line), and on *black GaAs* (black line). (**c**) Scheme of photon recycling at the *black GaAs*–P3HT interfaces.

These results are an important demonstration of a possible application of *black GaAs* as a photoactive material in hybrid solar cells. The superior semiconducting properties of GaAs, together with efficient light trapping and reabsorption of emitted photons shown here, could be employed to devise various hybrid structures for high performance photovoltaic devices.

#### **4. Conclusions**

We demonstrated that microstructured GaAs surfaces provide an efficient photon recycling platform to reduce losses associated with light emission occurring upon charge recombination in hybrid polymer-inorganic heterointerfaces. The microstructures were obtained by wet chemical etching of different GaAs crystalline planes in HF/H2O2, catalyzed by gold nanoparticles dispersed randomly on the semiconductor by electrodeless deposition. The process favors suppression of light reflection from the GaAs surface, with up to a 120-fold reduction. These are the best antireflective properties achieved for wet-etched semiconductors. Investigation of photoluminescence properties of GaAs:P3HT heterointerfaces confirmed that the microstructured GaAs reabsorb photons emitted by the P3HT cast on its surface. On the whole, the MacEtch process could be added to the fabrication workflow of photodetectors and solar cell devices to further improve their characteristics.

**Author Contributions:** Conceptualization, C.S.; methodology, P.L.; data analysis, P.L.; writing—original draft preparation, P.L.; writing—review and editing, P.L. and C.S. All authors have read and agreed to the published version of the manuscript.

**Funding:** C.S. would like to acknowledge the support of NTU Research Grant No. M4082409.

**Acknowledgments:** We thank Davide Comoretto for his unconditional support and for the useful discussions regarding this work. We also thank Valentina Robbiano and Franco Cacialli for their help in characterizing other properties of these samples.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Silicon Conical Structures by Metal Assisted Chemical Etching**

#### **Oscar Pérez-Díaz and Enrique Quiroga-González \***

Institute of Physics, Benemérita Universidad Autónoma de Puebla, Puebla 72570, Mexico; oscpedi@gmail.com **\*** Correspondence: equiroga@ieee.org; Tel.: +52-222-229-5610

Received: 27 March 2020; Accepted: 10 April 2020; Published: 11 April 2020

**Abstract:** A simple and inexpensive method to obtain Si conical structures is proposed. The method consists of a sequence of steps that include photolithography and metal assisted chemical etching (MACE) to create porous regions that are dissolved in a post-etching process. The proposed process takes advantage of the lateral etching obtained when using catalyst particles smaller than 40 nm for MACE. The final shape of the base of the structures is mainly given by the shape of the lithography mask used for the process. Conical structures ranging from units to hundreds of microns can be produced by this method. The advantage of the method is its simplicity, allowing the production of the structures in a basic chemical lab.

**Keywords:** silicon cones; metal assisted chemical etching; transversal pores

#### **1. Introduction**

Different techniques have been developed in order to produce Si structures to be used in applications such optoelectronics [1], energy storage [2], or sensors [3]. Among these techniques it is possible to find reactive ion etching (RIE), inductively-coupled plasma (ICP)-RIE, or chemical assisted ion beam etching (CAIBE); however, all of them require special equipment, like vacuum chambers or plasma generators. On the other hand, electrochemical etching of Si has proved to be a good option because it provides high control for Si dissolution. Nevertheless, etching complete wafers is complex, since high current densities in the range of amperes are required, which derive in undesired heating that makes necessary the use of high-quality cooling appliances.

On the other hand, the metal assisted chemical etching (MACE) technique does not require any special equipment or facilities, and makes possible the fabrication of complex structures [4–6]. MACE is performed by immersing a piece of semiconductor (commonly Si, but also other semiconductors like Ge [7] or III-V semiconductors [8,9] can be used), previously coated with a catalyst (usually the metals Au, Pt, or Ag [10]), in an HF based solution containing an oxidant agent (commonly H2O2, Na2S2O8 or KMnO4 [11,12]). The metal catalyzes the release of electronic holes from the oxidant and, depending on the potential energy difference between metal and semiconductor, it promotes their injection to the semiconductor. In the most common etching case, Si is oxidized beneath or around the metallized sections, and this oxide is dissolved by HF.

Despite it is possible to obtain different structures using MACE, the most of the reports on this technique indicate that the etching occurs most probably in crystallographic directions (there are fast etching planes and etch-stopper planes) or vertically [13]. However, it is also known that when the catalyst particles have diameters smaller than 40 nm, they produce pores either vertically or horizontally [14,15]. Furthermore, when MACE has been performed using particles with a dispersion of sizes from 10 to 400 nm, thin pores (with diameters below 100 nm) and wide pores (with diameters of hundreds of nanometers) were obtained vertically, while just thin pores were obtained horizontally. By eliminating the most of the particles of sizes below 100 nm, mainly vertical porosification was

obtained [16]. The vertical porosification of large particles is due to the much larger contact area below than at the sides of these particles, considering that they are spherical. For lightly-doped p-type Si wafers, for example, a larger contact area means a larger injection of electronic holes, which speeds the etching rate up. The contact area plays an important role in MACE [17]. When the particles are smaller, the probability is the same to etch either vertically or horizontally. With this equal probability, it is possible to think about a porous section growing upon the time in a direction with an angle close to 45◦ with respect to the vertical (the same amount of particles may move in the x than in the y direction, at similar velocity, producing that dy/dx = 1). To the knowledge of the authors, there are no reports to date taking advantage of this effect.

In the present work, a methodology to obtain Si conical structures by MACE is proposed. It consist of creating porous regions by MACE (previously defined by photolithography), which grow in angle with the time (due to lateral and vertical porosification obtained with catalyst particles smaller than 40 nm), and removing them afterwards. This methodology brings more flexibility to the MACE process. Additionally, it has the advantage of being simple, because the entire process could be performed in basic chemical labs without the need of complex facilities or equipment. Arrays of conical structures are important for different applications. They have been used as multielectrode sensing platforms for neuronal or cardiac tissue [18]. Additionally, arrays of complete or truncated cones have been used as antireflection layers [19] or to enhance the absorption of light [20], for different optical and optoelectronic applications such as solar cells. Moreover, such arrays have been used to modify the wetting properties of surfaces [21], achieving even super-repellency of hydrophobic surfaces [22].

#### **2. Materials and Methods**

p-type (100) Si wafers with resistivity of 15–25 Ω·cm were used as starting material. The fabrication procedure to obtain the conical structures of the present work consists of a sequence of steps: (a) Photolithography, (b) chemical deposition of Ag particles, (c) MACE etching, (d) dry oxidation of the porosified sections, and (e) dissolution of oxide. Alternatively, at the end of the process one can also dissolve the Ag particles in solutions of HNO3. The steps are schematized in Figure 1.

A quadratic pattern of circles was transferred by photolithography to a film of photoresist previously deposited on the Si wafers. The photoresist acts as masking layer for the metal deposition. The metallization can be done as sophisticated and controlled as in the case of thin film deposition by sputtering [23–25], or as simple as in the case of chemical deposition of metal particles using just a beaker [26]. For this report, it was used the simplest case. Ag particles were chemically deposited on the uncovered sections of Si by immersion in a solution 0.1 mM of AgNO3 in a mixture of HF (48%), H2O2 (30%) and H2O, in a proportion 2:3.4:94.6 *v*/*v*. High-density polypropylene beakers were used for this and all the subsequent processes with HF, since that material endures adequately this acid. The deposition time was 90 s, being performed in an ultrasonic bath in order to obtain a homogeneous distribution of particles.

The etching process was performed using an aqueous solution containing HF (48%), H2O2 (30%) as oxidant agent [27], and deionized water (DI), in a proportion 4:7:40 *v*/*v* at 30 ◦C. The etching time was 5 h. With this process, porous Si sections were obtained. In order to obtain the final structures, the porous sections of the Si samples need to be dissolved. To accomplish this, it is possible to use anisotropic [28,29] or isotropic chemical etching techniques [30]; however, in order to dissolve mainly the porous sections without important modification of the shape of the remaining Si, those techniques were avoided in this work. The samples were submitted to thermal oxidation at 850 ◦C under O2 flux (1 sccm) for 3 h. With this process, the porous Si sections were oxidized. To dissolve the oxide, the samples were immersed in a solution of HF (48%) and H2O in a proportion 1:9 *v*/*v* for 60 s. Silicon oxides are highly soluble in HF based solutions [31]. The final structures were analyzed with a JEOL JSM-7500F (Tokyo, Japan) field emission scanning electron microscope.

**Figure 1.** Schematic of the process to produce conical structures: (**a**) Photolithography, (**b**) deposition of Ag particles, (**c**) MACE, (**d**) dry oxidation, and (**e**) dissolution of oxide.

#### **3. Results and Discussion**

Figure 2a shows a SEM (scanning electron microscope) micrograph of a Si sample after Ag deposition. The Ag particles are the white sections in the micrograph. Their shape is semi-spherical, but sometimes the particles coalesce giving rise to ovoidal forms. Semi-spherical shapes are commonly obtained when depositing using low concentrated AgNO3 solutions [32]; the particles nucleate and start to grow until they coalesce and could form dendrites at longer deposition times [16]. It is also important to note that the particles are encrusted in Si. This happens because of the use of H2O2 during the deposition process: The Ag particles deposit on Si and catalyze the etching of Si at the same time, in the presence of the oxidant. However, the trenches are shallow because the deposition time is short (90 s). The deposited Ag particles have diameters in the range of 10 to 70 nm. A histogram of the particle size distribution (measured from SEM micrographs of the deposits) is presented in Figure 2b. As can be seen, the most of the particles have sizes below 40 nm. It was intended to have particles with sizes below 40 nm taking into account previous studies that suggest that with particles of those sizes the probability of etching vertically or horizontally is similar [16,18,19].

**Figure 2.** (**a**) SEM micrograph of the surface of a Si sample after deposition of Ag particles. (**b**) Size distribution of the Ag particles of the deposits.

Figure 3 shows a SEM micrograph of the structures obtained after the MACE process, dry oxidation and oxide dissolution. As can be observed, the structures are arrays of truncated cones. The bases and tops of the cones differ a bit from the circular shape. The diameters of the cones are 52 ± 5 μm for the top part and around 120 μm for the bottom part. They have a height of 60 μm.

Figure 4 shows a close-up to the structures. The walls of the cones is rough, with apparent porosity. This is an indication that the oxidation time was not enough to oxidize the whole porous Si sections. Because of this, the porous sections could not be completely dissolved during the last treatment in HF solutions (that dissolve SiO2). However, the porosity of the cone walls is a good indication of the transversal porosification. Taking a look at the surface of the cones (inset of Figure 4) helps confirming the existence of transversal pores. They grow in the <100> directions. In principle, one would not expect pores exactly at the surface; however, Ag particles may grow through the photoresist (the photoresist is partially permeable to Ag<sup>+</sup> ions during the deposition of Ag particles). It is important to mention that the photoresist used for the experiments of the present work is not HF resistant; nevertheless, it stands enough time for the Ag deposition, and it starts detaching during the etching process. It is not necessary that the photoresist stands the whole etching time, since no masking layer is necessary for this process (the etching rate in sections with Ag is hundreds of times faster than the etching rate in sections without catalyst). The few Ag particles grown beneath the photoresist could move in the X-Y plane during the etching process due to the availability of etchant in the surface (the photoresist does not stand HF, and the acid could diffuse through or below it); for this reason, it is possible to see transversal pores exactly at the surface.

**Figure 3.** SEM micrograph of the obtained structures after the whole process. An array of truncated cones is evident.

Taking a closer look to one entire truncated cone (see Figure 5), one can observe two different slopes of the cone walls. Going up to down, the first slope is of 2.8, while the second is of 1.3. The steep first slope is given by an excess of etchant; thus, the etching process is reaction-rate limited. As could be observed in the histogram of Figure 2, there is a good number of particles larger than 40 nm. Those particles have a higher probability of etching vertically. As they offer larger areas to catalyze the decomposition of H2O2, they inject a larger number of electronic holes to the semiconductor enabling a faster etching rate than with the smaller particles. It is known that the one dimensional (vertical) etching rate increases with the catalyst particle size (in particular with the coverage area of the catalyst) [10], but until certain limit of sizes, when the mass transfer beneath the catalyst particles is limited, and the etching rate starts to decrease [11]. After the first 24 μm of etching in depth, the process

is diffusion limited. It is common to observe diffusion limitation during a MACE process [33]. In this way, the etching process is mainly controlled by the availability of etchant, and the effect of the particle size is secondary. The difference between the vertical and the horizontal etching rate is about 30% in this depth range (27.5 μm of lateral etching vs. 36 μm of vertical etching, producing a slope of 1.3). The difference of etching rates in shallower depths is 180%. Following the tendency of the etching fronts, evoluting in angle, one can predict that if the MACE etching time is longer, complete conical structures (not truncated) could be obtained.

**Figure 4.** SEM micrograph closing up at the wall of the cones. Inset: Top view of the cones.

**Figure 5.** SEM micrograph of a truncated cone indicating its dimensions. The dashed lines indicate the two slopes of the cone walls.

Figure 6 shows a top view of the pattern of photoresist used during the etching process (photograph of the left), in contrast to the pattern of truncated cones obtained (SEM micrograph of the right). The photograph was captured with a portable optical microscope equipped with a CCD (charge-coupled

device) camera. The dots of photoresist deviate a bit from the circular shape due to the resolution of the photomask, which was fabricated with a conventional paper printer. The diameter of the dots of the original pattern is of about 120 μm, with a pitch of 230 μm. The final structures have an upper diameter of 52 ± 5 μm, with a lower diameter in the range of 120 μm (as the original pattern).

**Figure 6.** Left: Photograph of the pattern of photoresist used during the process. Right: SEM micrograph of a top view of the array of truncated cones.

The fact that the top of the cones does not have the same shape than the dots of photoresist could be explained with the fact that the etching occurs mainly in <100> directions (see Figure 4). The etching profile of lateral pores, saw from above, is schematized in Figure 7. The dots of the figure represent Ag particles, while the straight lines represent the pores. All the straight lines have the same length, considering equal etching rates in the [100] and [010] directions. It is clear that the region without lines (pores) is not exactly a circle. It is also possible to observe that there are sections with lines (pores) just in one direction, thus the density of pores in those sections is just the half. Sections with pores in two directions can be oxidized faster, due to the higher density of pores, which provide larger surface areas to be oxidized. These oxidized sections can be easily removed in HF solutions. One could still observe pores in the walls of the cones of this work. This should indicate that the oxidation of all porous sections was not complete (the oxidation of the sections with pores in just one direction takes longer). If the oxidation time would be reduced even more, the cones would tend to have a flower-like shape (the sum of the sections with no lines and the sections with lines in one direction of Figure 7).

**Figure 7.** Schematic of the lateral pores observed from above. The black dots represent Ag particles, while the straight lines represent pores.

To prove that the proposed process to produce conical structures works also in the micron range, an experiment was performed using a mask with a quadratic array of circular dots of 1.5 μm, with pitch of 3 μm. A micrograph of the resulting structures is shown in Figure 8a. Figure 8b shows a micrograph of the cross-section of one of the micro-cones. As can be observed, no pores cross the structure, supporting the theory of cone formation. Despite the surface of the cones looks porous, the bulk is solid. The lines observed in the cross section are a cleavage artifact.

**Figure 8.** SEM micrographs of micron sized truncated cones produced by the methodology of this work. (**a**) Overview; (**b**) cross-section of a cone.

#### **4. Conclusions**

Transversal porosification of Si by MACE using Ag particles of sizes smaller of 40 nm has been used as basis to produce conical structures. At etching depths smaller than 24 μm, porosification is controlled by the reaction rate, producing steeper cone walls. Deeper etching is limited by the diffusion of the etchant, producing a reduction of the slope of the cone walls. Transversal porosification occurs mainly in the <100> direction. Due to this, the final cross-sectional shapes of the cones do not follow exactly the shape of the patterns of the photolithography mask. It was proved that the methodology works to produce conical structures of sizes from units to hundreds of micrometers, and it could be developed in basic chemical labs without complex equipment.

**Author Contributions:** O.P.-D. made the experiments, performed SEM microscopy, and wrote the first manuscript draft. E.Q.-G. proposed and coordinated the project, performed the analysis of the results, and corrected the manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was funded by Projects CONACyT CB-2014-01-243407 and PROMEP BUAP-NPTC-377.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Fabrication of Ultra-High Aspect Ratio (**>**420:1) Al2O3 Nanotube Arraysby Sidewall TransferMetal Assistant Chemical Etching**

#### **Hailiang Li and Changqing Xie \***

Key Laboratory of Microelectronic Devices & Integrated Technology, Institute of Microelectronics of Chinese Academy of Sciences, Beijing 100029, China; lihailiang@ime.ac.cn

**\*** Correspondence: xiechangqing@ime.ac.cn; Tel.: +86-01-82995581

Received: 28 February 2020; Accepted: 30 March 2020; Published: 3 April 2020

**Abstract:** We report a robust, sidewall transfer metal assistant chemical etching scheme for fabricating Al2O3 nanotube arrays with an ultra-high aspect ratio. Electron beam lithography followed by low-temperature Au metal assisted chemical etching (MacEtch) is used to pattern high resolution, high aspect ratio, and vertical silicon nanostructures, used as a template. This template is subsequently transferred by an atomic layer deposition of the Al2O3 layer, followed by an annealing process, anisotropic dry etching of the Al2O3 layer, and a sacrificial silicon template. The process and characterization of the Al2O3 nanotube arrays are discussed in detail. Vertical Al2O3 nanotube arrays with line widths as small as 50 nm, heights of up to 21 μm, and aspect ratios up to 420:1 are fabricated on top of a silicon substrate. More importantly, such a sidewall transfer MacEtch approach is compatible with well-established silicon planar processes, and has the benefits of having a fully controllable linewidth and height, high reproducibility, and flexible design, making it attractive for a broad range of practical applications.

**Keywords:** Al2O3 nanotube; ultra-high aspect ratio; gold (Au) metal assisted chemical etching; atomic layer deposition; anisotropic dry etching

#### **1. Introduction**

In recent years, Al2O3 nanotube arrays possessing high surface-to-volume ratios have attracted much attention, owing to their potential applications in optoelectronics [1,2], biotechnology, and photocatalysis [3,4]. For example, several recent studies have indicated that Al2O3 nanotube arrays exhibit excellent dielectric properties [5] and good flexibility [6], as compared with other oxide nanotubes. Thus, Al2O3 nanotube arrays would be more advantageous for use as optical transportation media in optoelectronics. They can also be utilized as a new biomineralization nanoreactor in biotechnology, and so on. Various template-based strategies for the fabrication of Al2O3 nanotubes with a high aspect ratio have been proposed. These include hydrothermal reaction methods [7], coating of carbon nanotubes with aluminum isopropoxide [8], anodization of Si-based Al film [9,10], and etching of a porous anodic alumina (PAA) template in a NaOH solution and ultrasonic treatment of PAA membrane [11,12]. In addition to the Al2O3 nanotubes, atomic layer deposition (ALD) coatings of silicon oxide (SiO2) [13], zinc oxide (ZnO) [14,15], zirconium dioxide(ZrO2) [16], and titanium oxide(TiO2) [16,17] have also been successfully applied to the preparation of nanostructures.Using these methods, much progress has been made on achieving highly ordered nanotube arrays with a controllable size and short length. However, due to the poor consistency of the pillar size in the templates, it is not easy to confine and control the profile of the fabricated nanostructures. Moreover, separating hollow nanostructures from a template to obtain individual nanostructures remains a

challenge. Many kinds of nanotubes with high aspect ratios will lose their tubular structure after extraction from the template. With the continuous request for high performance nanotube-based devices, the aspect ratio of the fabricated nanotube arrays as obtained by tailoring electrochemical conditions is far from satisfactory.

Metal assisted chemical etching (MacEtch), first reported by Li and Bohn in 2001 [18], offers a promising wet etching solution for generating silicon nanostructures [19–22]. In addition, the MacEtch of germanium (Ge) [23] and III−V compound semiconductors, such as GaN [24,25], GaAs [26–28], GaP [29], InP [30], AlGaAs [31], InGaAs [32], and InGaP [27], have also been demonstrated. This method has the benefits of inherent simplicity, a low cost, high versatility, and high reproducibility, making it attractive for preparing silicon nanowire arrays. The conventional method for generating micrometer, submicrometer, and nanosized silicon structures with high aspect ratios [19–22] is by submerging a metal-coated Si sample into an etchant solution composed of hydrofluoric acid, hydrogen peroxide solution, and a diluting agent. Furthermore, large-area uniform micro-gratings with well controlled morphological features and depths as large as 80 μm have also been successfully produced by optimizing the MacEtch method [33]. Highly dense Si/TiO2 core/shell nanowire arrays have also been synthesized using a nanostructured Si template obtained by the MacEtch of the Si substrates and a layer of TiO2 deposited by ALD [34,35]. In the previous works, using Ti/Au nanostructures patterned with electron-beam lithography followed by ion beam etching, we fabricated vertical silicon nanopillar arrays with a period of 250 nm and an aspect ratio of 160:1 using MacEtch [36]. A 20 nm minimum feature size was also realized by MacEtch-based nanoimprinting [37]. Unfortunately, the MacEtch method cannot be directly adopted to generate oxide nanotubes.

Sidewall transfer lithography has been widely recognized as a promising technology that can fabricate deca-nanometer metal-oxide-semiconductor field-effect transistors (MOSFETs) and siliconnanostructures [38–40]. The key idea is to combine three well-established techniques (lithography, sidewall process, and dry etching) to create silicon nanostructures. In this work, we present a reliable means, called sidewall transfer metal assistant chemical etching, to fabricate Al2O3 nanotube arrays with an ultra-high aspect ratio. The key idea is to combine a low-temperature MacEtch and sidewall transfer process based on ALD and dry etching to simultaneously achieve their respective advantages. The former is used to generate vertically oriented silicon nanostructures that serve as a sacrificial layer, while the latteris used to transfer the silicon nanostructures to Al2O3 nanotube arrays with a higher aspect ratio and smaller feature size. The ultimate aspect ratio and feature size of the fabricated Al2O3 nanotube arrays can be controlled by modifying the thickness of the Al2O3 film deposited by ALD. The influence of the deposition temperature and annealing temperature on the structure and optical properties of Al2O3 thin films deposited by ALD are also examined.

#### **2. Materials and Methods**

The process flow of the proposed sidewall transfer MacEtchis shown in Figure 1. In our experiments, 4in.p-Type<100>CZsiliconwafers (Silicon Quest International, San Jose, CA, USA) with a resistivity of 2–10 Ω·cm and thickness of 500 μm were used. After the samples (Figure 1a) were cleaned in sulfuric acid and hydrogen peroxide to remove the surface native oxide layers, a 3-nmTi/20 nm Au thin film was deposited on the silicon substrates by a magnetron sputtering system (ACS-4000, ULVAC Company, Chigasaki, Japan). The working pressure was maintained at 4.5 <sup>×</sup> <sup>10</sup>−<sup>6</sup> Torr, and the temperature of the chamber was kept at 25 ◦C over the entire deposition process. Then, highly sensitive chlorinated electron beam resist ZEP520A was spin-coated on the Ti/Au layer to a thickness of about 400 nm, and was bakedon a hotplate at 180 ◦C for 2 min (Figure 1b). The resist was exposed with an electron beam lithography system (JBX-6300FS, JEOL Company, Tokyo, Japan) for patterning definition (Figure 1c), and ion beam etching (IBE) was performed to transfer ZEP520A resist pattern onto the Ti/Au layer, forming Ti/Au nanostructures that serve as a local cathode (Figure 1d). Then, the MacEtch processwas carried out to generate silicon nanostructures with a high aspect ratio (Figure 1e). The ALD process was used to deposit the Al2O3 film on the sidewalls and on top of the generated silicon

nanostructures (Figure 1g). Finally, the exposed tops of the Al2O3 film and its silicon nanostructures underneath were removed by dry etching, forming Al2O3 nanotube arrays with a higher aspect ratio and smaller feature size.

**Figure 1.** Schematic diagram of the sidewall transfer MacEtch process for the realization of the Al2O3 nanotube arrays with ultra-high aspect ratios: (**a**) silicon substrate; (**b**) deposition of the Ti/Au layer and ZEP520A resist; (**c**) electron beam lithography; (**d**) ion beam etching; (**e**) metal assisted chemical etching (MacEtch); (**f**) Si pillars template; (**g**) atomic layer deposition (ALD) Al2O3 film; (**h**) dry etching; (**i**) reactive ion etching of Si.

The morphology and structure of the fabricated Al2O3 nanotube arrays were characterized by a scanning electron microscope (SEM; SUPRA-55, Zeiss, Oberkochen, Germany). The information on the surface chemistry of the Al2O3 layer deposited by the atomic layer deposition (ALD) technique was investigated using X-ray photo emission spectroscopy (XPS) with a monochromatic Al Kα X-ray source. Both the Al2O3 film thickness and its refractive index profile were determined by a spectroscopic ellipsometry (Horiba Uvisel FUV, Kyoto, Japan) over the spectral range of 150 to 900 nm using the Cauchy model.

#### *2.1. Fabrication ofTi*/*Au Nanostructures with Low Aspect Ratio*

To obtain high-resolution resist patterns, electron beam lithography was performed at an accelerating voltage of 100 kV, with a beam current of 200 pA and an exposure dose of 400 μC/cm2. After electron beam exposure, the ZEP520A resist was developed using a standard developer N50D (ZEON) for 1 min at 18 ◦C, and rinsed with isopropanol (IPA) for 40 s to stop development. Then, the sample was dried with a steady stream of N2.

Next, a home-madeargon (Ar) IBE system was used to transfer the resist patterns into a Ti/Au thin film deposited onto the silicon wafer. The working pressure was maintained at 1.0 <sup>×</sup> 10−<sup>4</sup> Torr, and the substrate temperature was lower than 100 ◦C over the entire IBE process. The Ar<sup>+</sup> ion energy and the beam current density were 500 eV and 1 mA/cm2, respectively. The corresponding etching rates of Au and the resist were 120 nm/min and 20 nm/min, respectively, resulting in a selectivity ratio of 6:1. The etching time in our experiments was 12 s. After completion of the IBE process, the residual ZEP520A resist was removed by washing with a resist removal solution (ZDMAC, ZEON Company, Tokyo, Japan), followed by oxygen plasma ashing. The patterned Ti/Au nanostructures served as a catalyst in the subsequent low-temperature MacEtch process.

#### *2.2. Fabrication of Silicon Nanostructures with High Aspect Ratio*

A low-temperature Au MacEtch process was performed to generate silicon nanostructures with a high aspect ratio. It should be noted that samples must be kept clean and tidy before the Au MacEtch process. The Au MacEtch process was carried out in an etchant solution composed of hydrofluoric acid, hydrogen peroxide solution, and deionized water (4.1 M HF/0.15 M H2O2/45 M H2O) at 2 ◦C. The etching was conducted for 15 min. It is well known that the collapse of large aspect ratio nanostructures often occurs during the drying step. In our drying scheme, we used isopropanol with low surface tension, instead of deionized water (DI) water as a rinse solution. This reduced the capillary force acting on the silicon nanostructures. After the sample was rinsed with isopropanol, it was evaporated naturally and dried at room temperature.

#### *2.3. Al2O3 Film Deposition by ALD*

The ALD process allows one to precisely deposit highly uniform and conformal thin films onto complex three-dimensional topographies. Here, an Al2O3 film was deposited using Al(CH3)3 (TMA) and deionized water with a hot-wall atomic layer deposition system (Picosun R200, Espoo, Finland). The deposition temperature was 300 ◦C. An Al2O3 film was grown on the sidewalls and on top of the fabricated silicon nanostructures with an aspect ratio of 160:1, using TMA and deionized water. TMA (Sigma Aldrich, St. Louis, MO, USA) was used as the precursor and deionized water was used as an oxidant source during the ALD process. The TMA reactant exposure time, N2 purge time following TMA reactant exposure, water exposure time, and N2 purge time following the H2O reactant exposure were 0.5 s, 2 s, 0.5 s, and 2 s, respectively. The growth rate of the Al2O3 film was 0.089 nm/cycle, as inferred by spectroscopic ellipsometry. The sample was processed with 400 ALD cycles, and the corresponding Al2O3 film thickness was 50 nm.

#### *2.4. Annealing Processand Characterization of the Al2O3 Film*

The thermal expansion coefficient of Al2O3 film is 8.2 <sup>×</sup> 10−<sup>6</sup> ◦C−1, which is higher than that of silicon (2.6 <sup>×</sup> 10−<sup>6</sup> ◦C−1). To compensate the lattice mismatch between the Al2O3 ALD filmand silicon nanostructures, four samples were annealed at 700 ◦C, 800 ◦C, 900 ◦C, and 1000 ◦C under a vacuum for 90 min, seperately. The specifications of the Al2O3 film was investigated using X-ray photoelectron spectra (XPS). Characterization of the optical properties of our Al2O3 film from 150 to 900 nm wavelengths was also performed using spectroscopic ellipsometry.

#### *2.5. Formation of Al2O3 Nanotube Arrays with an Ultra-High Aspect Ratio*

First, the exposed tops of the Al2O3 film on the silicon nanostructures were removed by an inductive coupled plasma (ICP) etching system (ULVAC, Japan), with a mixture of BCl3 and Cl2 reactive gas. Secondly, the silicon nanostructures underneath the exposed tops of the Al2O3 film were removed by the same ICP etching system with SF6 reactive gas and an isotropic reactive ion etching mode. The etching parameters of the Al2O3 film and silicon nanostructuresa re summarized in Table 1.


**Table 1.** Recipes for Al2O3 and Si in etch system.

Process temperature is 25 ◦C for all of the processes.

The etch rates of the Al2O3 film and silicon nanostructures were 0.89 nm/s and 9 nm/s, respectively. The etching selectivity between the silicon nanostructures and Al2O3 film was as high as 66,000:1 for the Si etching recipe [41].

#### **3. Results and Discussion**

Figure 2 shows the dependence of the Au MacEtch process parameters (etchant solutions and etching temperature) on the quality of the fabricated silicon nanostructures. When the hydrogen peroxide concentration in the etchant solution was relatively high, the lateral etching rate increased and defects began to occur on the sidewalls, as shown in Figure 2a. This is because as the concentration of H2O2 increased, the number of the generated holes also increased, resulting in an increase in the silicon etching rate. In other words, when the generated holes between the Si and Ti/Au interfaces could not be completely consumed, the excess holes spread laterally, leading to lateral corrosion and the formation of defects in the sidewalls. Using an optimized H2O2 concentration, the transverse etching rate could be limited by the availability of the generated hole. Thus, vertical silicon nanostructures could be obtained, as shown in Figure 2b.

**Figure 2.** Cross-sectional scanning electron microscope (SEM) micrographs of silicon nanostructures using various etchant solutions composed of hydrofluoric acid, hydrogen peroxide solution, and deionised (DI) water with molar ratios of (**a**) 4.8:0.35:50, (**b**) 4.8:0.15:50, and (**c**) 4.8:0.15:50. The scale bars in (**a**–**c**) are 850 nm, 900 nm, and 1.9 μm, respectively. The etching temperatures in (**a**,**b**) are 22 ◦C, and the etching temperature in (**c**) is 2 ◦C.

Etching temperature also plays an essential role in the Au MacEtch process. When the Au MacEtch process was performed at 2 ◦C, a better morphology could be obtained than that at room temperature, at the expense of a much lower etching rate. In this work, a low-temperature Au MacEtch process could result in silicon nanostructures with an aspect ratio of up to 160:1, as shown in Figure 2c. Silicon nanostructure bottoms that are especially clean and flat can be obtained.

Figure 3 shows the binding energies of the Al2O3 film deposited at 300 ◦C. Two signatures of orbital, 74.4 eV for Al (2p) and 531.5 eV for O (1s), can be observed from an XPS wide scan. The O/Al ratio of the Al2O3 film is close to the expected value of 1.5, corresponding to the lattice oxygen of Al2O3 [42]. The difference between these two elemental peaksis close to the standard values in the literature for different forms of aluminum oxide [43,44]. A peak of binding energy of 1s carbon in the Al2O3 film could also be observed, indicating either an incomplete reaction or an insufficient N2 purge time.

Figure 4 shows the Al2O3 ALD film thickness and refractive index as a function of the annealing temperature. The refractive index of the Al2O3 ALD filminitially increased with an increase of annealing temperature from 700 ◦C to 800 ◦C, and later saturated with the increasing annealing temperature. When the annealing temperature reached 1150 ◦C, the refractive index approached a maximized value of 1.724, which is slightly smaller than that of crystalline sapphire (1.76). The measured data indicate that the amorphous Al2O3 ALD film in this experiment would densify further upon crystallization. This is consistent with the existing result [45]. Meanwhile, there was a 10% decrease in the Al2O3 ALD film thickness with increasing the annealing temperature from 700 ◦C to 800 ◦C, which was mainly

due to an increase inthe density and purity levels of the Al2O3 films deposited by ALD. The measured data also indicate the occurrence of Al2O3 crystallization after high-temperature annealing, which appears to be particularly advantageous for preventing the collapse of ultra-high aspect ratio Al2O3 nanotube arrays.

**Figure 3.** X-ray photoelectron spectra of the Al2O3 layer deposited at 300 ◦C. The two insets show the Al (2p) and O (1s) peaks, respectively.

**Figure 4.** Variation of film thickness and refractive index versus annealing temperature.

Figure 5a shows the SEM image of the top view of the generated Si nanostructures, followed by the Al2O3 film deposition by ALD and annealedat 1150 ◦C for 90 min. The results of the generated Al2O3 nanotube arrays without the annealing process are also given in Figure 5b. One can clearly see that the Al2O3 nanotube arrays bend and lean against each other.The bending is probably caused by a mismatch in internal stress between the Al2O3 layer and Si pillar. The thermal expansion coefficient of the Al2O3 film is 8.2 <sup>×</sup> <sup>10</sup>−<sup>6</sup> ◦C<sup>−</sup>1, while the thermal expansion coefficient of silicon is 2.6 <sup>×</sup> <sup>10</sup>−<sup>6</sup> ◦C<sup>−</sup>1. It should be noted that the Al2O3 ALD film was deposited by the temperature gradient method over a temperature range from room temperature to 300 ◦C, and the Si cores were removed afterwards using

a plasma reactive ion etching at 25 ◦C. By comparison, Figure 5c,d shows the SEM images of the top andcross section views, respectively, of the generated Al2O3 nanotube arrays after the whole process was finished. These uniform structures had a line width of 50 nm and height of 21 μm, corresponding to an aspect ratio as high as 420:1. The Al2O3 nanotube arrays were perfectly preserved from collapsing after the annealing process, confirming the successful pattern transfer of the sidewall transfer MacEtch process. The shape of the Al2O3 nanotube arrays exhibited very little deformation relative to that of the original Si nanostructures. Figure 5e shows a tilted SEM view of the Al2O3 nanotube arrays after the whole process. One can see that the silicon template was almost completely removed by plasma reactive ion etching at 25 ◦C.

**Figure 5.** A series of images demonstrating sidewall transfer MacEtch for the preparation of ultra-high aspect ratio Al2O3 nanotube arrays. (**a**) Top view SEM of the Si nanostructures after Al2O3 film deposition by ALD. The scale bar is 300 nm. (**b**) Top view SEM of the generated Al2O3 nanotube arrays without annealing process. The scale bar is 400 nm. (**c**,**d**) SEM images of the top and cross section views of the generated Al2O3 nanotube arrays resulting from the sidewall transfer MacEtch process. The scale barsin (**c**,**d**) are 250 nm and 1.6 μm, respectively. (**e**) Tilted SEM view of Al2O3 nano-tube structures after the whole process. The scale bar is 400 nm. The tilting angle is 35◦.

#### **4. Conclusions**

In summary, we have demonstrated a reliable sidewall transfer MacEtch process to fabricate ultra-high aspect ratio Al2O3 nanotube arrays with linewidthsas small as 50 nm, heights of up to 21 μm, and an aspect ratio of up to 420:1. This technique combines the advantages of the high aspect ratio nanostructure capabilities of the low-temperature Au MacEtch with thesidewall transfer process. The use of the sidewall transfer process has two advantages. First, it leads to higher-resolution and higher aspect ratio Al2O3 nanotube patterns than with MacEtch, and second, the line-width of the Al2O3 nanotubes can be precisely controlled by the cycle number of ALD for Al2O3. The sidewall transfer MacEtch provides a promising route to the scalable manufacturing of ultra-high aspect ratio

Al2O3 nanotube arrays, and is applicable to other kinds of oxide nanotubes, as long as the oxide can be deposited by ALD.

**Author Contributions:** H.L. and C.X. conceived the idea for the experiment; H.L. performed the experiments; H.L. and C.X. wrote the manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the National Key Research and DevelopmentProgram of China (grant no. 2017YFA0206002) and the National Natural Science Foundation of China (grant no. U1832217, no. 61804169, and no. 61821091).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

#### *Article*
