Sidewall channel fabrication using membrane projection lithography and metal assisted chemical etching

Horizontal enclosed nanochannels have potential applications in nanofluidics, nanobiotechnology, DNA sensing, photonic crystals, and optical waveguides.^1–3 But, despite recent advances in top-down nanofabrication technologies to fabricate horizontal nanochannels in materials such as $SiO2$ (Ref. 4) and polymers,^3,5 less research has been focused on horizontal Si nanochannel fabrication.

In this report, we employ membrane projection lithography (MPL) as described by Burckel et al.⁶ to obtain submicrometer scale features on the sidewalls of $2μm$ deep silicon trenches and then apply metal assisted chemical etching (MacEtch)⁷ to obtain horizontal Si nanochannels.

MPL consists of standard planar lithography combined with a sequence of processing steps that together create out-of-plane features. The basic premise behind MPL is to create a patterned membrane positioned over a cavity and then use angle evaporation through the membrane to deposit metal instances of the membrane pattern on the sidewall of the cavity.⁶ Previous reports on sidewall patterning indicate that it is very difficult to obtain sidewall features and either the substrate has to be tilted or the incident light has to be deflected; the complexity rises if the sidewalls are narrower or the features are smaller.^8,9

The deposited metal instance acts as a catalyst for the MacEtch process. Due to the weak dependence of MacEtch on the crystal direction of semiconductor, features will etch in the direction perpendicular to the metal/Si interface.^7,10,11 MPL-formed patterns should, therefore, etch horizontally into sidewalls, creating the desired horizontal structures.

MacEtch is a relatively recently developed nanofabrication technique that can create very high aspect ratio semiconductor nanostructures (⁠$∼$200:1) with control of structural parameters such as orientation, length, and morphology.^7,10,12 It is a low cost wet etching process, which uses the catalytic activity of noble metals (here Au) deposited on top of the semiconductor (Si) to etch the semiconductor beneath it, in a solution of an oxidant (⁠$H2O2$⁠) and an acid (HF). Figure 1 is a schematic diagram of the MacEtch reaction. Au catalyzes the reduction of $H2O2$ to generate holes on the top of the Au layer, which are then injected into the Si substrate at the interface between Si and Au, thus oxidizing the Si layer beneath it. HF will dissolve the oxidized Si atom by forming silicon hexafluoride ions (⁠$SiF62−$⁠), thus etching away the Si underneath the Au layer. The metal layer descends into the surface as the semiconductor is being etched directly underneath the metal layer. As reported in the literature, MacEtch depends weakly on the crystal direction of the semiconductor and more strongly on the van der Waals forces between metal and semiconductor.^7,10,11,13 In this report, we explore the less studied horizontal etching capability of MacEtch in Si.

The fabrication steps used to create horizontal Si nanochannels are shown in Figs. 2(a)–2(h), and they can be divided into two major processing steps: (a) membrane projection lithography, which includes the steps to create a patterned inorganic membrane suspended on top of a hollow Si cavity, etched in a single crystal Si wafer using the Bosch process, and followed by metal evaporation through the membrane to deposit instances of the membrane pattern on the sidewall of the Si cavity and (b) metal assisted chemical etching, to transfer the pattern into the silicon sidewalls and obtain the desired horizontal nanostructures.

First, $2μm$ deep and 2–$8μm$ wide Si trenches were created in p-type Si (100) with a resistivity of 1–10 $Ω$ cm by the Bosch process using photolithography and an SPTS Rapier Si deep reactive ion etch (DRIE) system [Fig. 2(a)].¹⁴ The etch depth per cycle was 100 nm. Bosch etching was used instead of conventional reactive ion etching to ensure that the sidewalls were vertical.

The wafer was then cleaned using $O2$ plasma for 15 min (see the back-etching process below for details), followed by etching in a buffered oxide etch solution, and water rinse to remove the resist, the fluoropolymer sidewall coating, and any inorganic residue formed during the Bosch process. This cleaning process was essential to ensure that the planarizing layer adhered to the trench sidewalls.

The appropriate selection of the membrane material and the planarization material is a key step to this process. Here, LOR-20B (Microchem Inc.) is used to planarize the Si trenches. Since the planarization layer is temporary, the LOR-20B is later dissolved away in a standard photoresist developer and for the membrane material, ZEP-520A electron beam resist (Nippon Zeon Ltd.) was used; tests showed that it is unaffected by the LOR-20B developer and vice versa.

To planarize the Si trenches, LOR-20B was spin-coated at 1500 rpm for 45 s, soft baked at $180°C$ for 5 min, and then back-etched using $O2$ reactive ion etching (RIE) at 100 W and 60 mTorr for $∼15min$ using an Oxford Instruments PlasmaLab 80 Plus system to create uniformly planarized Si trenches as shown in Fig. 2(b). The back-etch time for LOR-20B is a crucial parameter and was found by repeated measurements of the remaining material using reflectometry, since underetching or overetching the planarization layer can result in the loss of the membrane layer when the LOR-20B is later removed. As the membrane material, ZEP-520A was spun-on at 1500 rpm for 45 s and soft baked at $180°C$ for 5 min to create a 100 nm thick membrane on top of the Si trenches as illustrated in Fig. 2(c). It was patterned using 50 keV Elionix ELS-7500EX equipped with a 20 MHz fixed clock at 100 pA with a $30μm$ final aperture and a 20 nm beam step size, using a base dose time of $2μs$ per dot, which resulted in an e-beam dose of $400μC/cm2$⁠. As the pattern dimensions are in the same range as membrane thickness, to obtain a circular pattern of diameter $∼200nm$⁠, an elliptical pattern of dimension 400 nm was chosen to compensate for the membrane thickness. Large openings near these small features were also exposed in order to facilitate later removal of the planarizing layer. The exposed ZEP-520A membrane was developed in o-xylene for 70 s, followed by rinsing in isopropanol for 30 s to get the desired patterned membrane as shown in Fig. 2(d). The underlying LOR-20B was then washed away in MF-319 developer (Microchem Inc.), which will not affect the ZEP-520A membrane. The samples were rinsed in water and four times in methanol before being dried using a Tousimis Inc. Series C $CO2$ critical point dryer to prevent the collapse of the membrane on drying [Fig. 2(e)].

The nanochannel noble metal patterns were formed by depositing 20 nm of Au by electron beam evaporation at a rate of 2 Å/s at $5×10−7Torr$ using a PVD-75 e-beam evaporator (Kurt J. Lesker). The samples were held at an angle of $45°$ to the deposition direction using a specially made fixture attached to the evaporator wafer platen [Fig. 3(a)]. Acetone was used to lift-off the metal coated ZEP-520A membrane and reveal the desired gold pattern on the Si sidewall as illustrated in Fig. 2(g).

Metal assisted chemical etching is a widely used wet etching process to create high aspect ratio features. Since the etch direction generally depends on the surface normal of the interface of the catalyst metal and Si, it can be etched horizontally if the metal is deposited on a sidewall. Two etchant concentrations were considered: a concentrated version with a 2:1:2 (v/v/v) solution of HF:$H2O2$⁠:$H2O$ and a diluted form, 50:1:50 (v/v/v) HF:$H2O2$⁠:$H2O$⁠, for 30 min at room temperature. Samples were held vertically in the etching solution using the fixture as shown in Fig. 3(b). They were then rinsed in water and dried with N$2$ to form the horizontal Si nanochannels as illustrated in Fig. 2(h).

Samples were imaged at critical points of the fabrication process with an FEI Strata DB235 focused ion beam (FIB) system and a JEOL-7500F scanning electron microscope (SEM) to study both top-down and cross-sectional morphologies. The FIB $Ga+$ beam was also used to create cross sections.

Figure 4(a) shows an SEM image of the cross section of a LOR-20B coated Si trench. It can be seen from the figure that a $4μm$ thick LOR-20B film fills the Si trenches and planarizes the top surface to within about 100 nm over the feature. Figure 4(b) shows the LOR in an isolated trench after back-etching the organic film with an $O2$ plasma. In this case, the planarizing layer is slightly overetched, since the bottom of the membrane layer lies slightly below the Si surface.

The membrane is formed by spin-coating ZEP 520A electron beam resist and then baking at a temperature that does not inhibit the subsequent dissolution of the LOR in the standard photoresist developer. No intermixing of the ZEP and LOR was observed. Figure 5(a) shows an SEM of cross section of an Si trench with the ZEP membrane on top, after the LOR support layer has been removed and the sample is dried using critical point drying, which prevented the collapse of the membrane. A scalloped, vertical Si sidewall, where Au will be deposited, is evident under the membrane.

Figure 5(b) shows a top-down SEM image of the patterned membrane over the Si substrate after electron beam lithography, development, and LOR removal. The elliptical holes in the resist are located over the Si trenches and positioned so that the subsequent angled evaporation will deposit Au on the sidewalls. A circular hole in a very thin membrane would project a circular spot on the sidewall when viewed at a $45°$ from the surface normal; however, the finite thickness of the ZEP 520A means that the circles should be elongated to form a circular shadow.

The images in Fig. 6 show Au disks deposited on an Si sidewall after resist removal. This feature is not circular and about $300×600nm2$ in size. The pattern design was apparently elongated too much to compensate for membrane thickness in this case. Stress in the membrane, after removal of the planarization layer, may also contribute to the stretching. Scalloping on the sidewall caused by the DRIE process is evident, which may create discontinuities in the Au disks, but the scallops did not appear to adversely affect nanochannel formation.

In Fig. 6(a), it can be seen that the gold disk is projected partly on the sidewall and partly on the trench floor. It is believed that there are two reasons for this (1) misalignment in the electron beam lithography step and (2) the fabricated trench widths are larger than their coded dimensions so that they shift the position of the wall away from the ideal membrane pattern coordinates.

Figure 7 shows nanochannels after MacEtch using the concentrated HF:$H2O2$ solution. The higher concentration generates more holes, which attacks the silicon around the disks and will result in enlarged channels and rough surfaces as can be seen in Fig. 7(a). This image also shows that the channel reaches from the one side of the wall to the other, indicated by the arrows. These samples help in the understanding the etching profile of the nanochannels. Figure 7(b) shows the cross section of one such channel, revealed by $Ga+$ focused ion beam that has cut through an Si structure. The darker region is Si, and the lighter region is the nanochannels covered with Ga and Si residues. The channel extends from one side of the Si wall to the other side but not in a straight line (see arrow).

Kong et al.¹³ showed that the concentrations of HF and $H2O2$ strongly affect how “straight” the etch path is; they found that an optimum 30:1:30 (⁠$HF:H2O2:H2O$⁠) solution will more reliably form vertical holes in Si when Au disks are deposited on the substrate surface. Figure 8 shows sidewall channels etched with a similarly diluted 50:1:50 solution. The arrows indicate the channels. These features more closely match the original Au dimensions and are about 300 nm in diameter. The surface of the Si structures is also much less rough than samples etched in the concentrated solution.

A higher magnification image of one of these channels, Fig. 8(b), shows the hole that was formed in more detail. The path of the nanochannel appears to be perpendicular to the Si wall, at least as far as this viewing angle can show. The small size of the channel makes it difficult to image a cross section using an focused ion beam. Not all of the channels etched completely to the far side in the etch times used for these samples. It was found to not be possible to reliably associate the hole on the one side of a wall with the equivalent hole on the opposite side using the existing patterns and SEM.

Future work includes creating patterns that will simplify identification of channels so that quantitative results may be collected and to develop the characterization techniques to follow the path of these channels through the Si. Further work is also required to reduce the diameter of the channels even more and to improve the smoothness of the Si sidewalls.

By combining the fabrication techniques of membrane projection lithography and metal assisted etching, 300 nm sized channels have been formed into the sidewalls of the Si structure, creating nanochannels that run under and parallel to the Si substrate surface. The MacEtch solution concentration influences the Si substrate surface morphology, and it was found out that a lower concentration of $H2O2$ in the etchant $HF:H2O2:H2O$ (50:1:50) yields more well formed channels than the higher concentration (10:1:10) solution.

This work was performed at the Singh Center for Nanotechnology at the University of Pennsylvania, a member of the National Nanotechnology Coordinated Infrastructure (NNCI) network, which is supported by the National Science Foundation (Grant No. NNCI-1542153).