Wafer-scale integration of sacrificial nanofluidic chips for detecting and manipulating single DNA molecules

Wafer-scale fabrication of complex nanofluidic systems with integrated electronics is essential to realizing ubiquitous, compact, reliable, high-sensitivity and low-cost biomolecular sensors. Here we report a scalable fabrication strategy capable of producing nanofluidic chips with complex designs and down to single-digit nanometre dimensions over 200 mm wafer scale. Compatible with semiconductor industry standard complementary metal-oxide semiconductor logic circuit fabrication processes, this strategy extracts a patterned sacrificial silicon layer through hundreds of millions of nanoscale vent holes on each chip by gas-phase Xenon difluoride etching. Using single-molecule fluorescence imaging, we demonstrate these sacrificial nanofluidic chips can function to controllably and completely stretch lambda DNA in a two-dimensional nanofluidic network comprising channels and pillars. The flexible nanofluidic structure design, wafer-scale fabrication, single-digit nanometre channels, reliable fluidic sealing and low thermal budget make our strategy a potentially universal approach to integrating functional planar nanofluidic systems with logic circuits for lab-on-a-chip applications.


Supplementary Note 1: Conventional nanofluidic manufacturing methods
Unlike conventional nanopores where the fluid sealing is achieved by packaging, planar nanofluidic structures must incorporate a reliable sealing mechanism in the manufacturing process in order to control the fluidic flow. However, existing manufacturing technologies, which generally exploit selective sealing, wafer bonding, or sacrificial materials to create an enclosed nanofluidic system (Supplementary Figure 1), cannot meet all the requirements of patterning sub-5 nm features, producing complex planar fluidic structures, and also integrating metallic sensors over a wafer scale. Figure 1a) utilize dielectric deposition 1 or radiation induced melting 2 to form a continuous film covering the nanochannels surface but leave voids underneath the film for fluid transport. However, these methods demand special materials and nanostructure geometries (e.g., height, width, shape, etc.), and thus cannot be applied universally. In the case of the wafer bonding (Supplementary Figure 1b), nanopatterned substrates are sealed by a second substrate. Bonding to rigid materials 3,4 followed by a high-temperature (usually 300 to 1000°C) annealing can yield a strong bonding strength, suitable for applications requiring a precise control of structural dimensions. However, the stringent requirements of bonding surface cleanliness and high-temperature annealing process pose yield and metal integration challenges. Although bonding to soft materials (elastomers and polymers 5,6 ) can partially alleviate these difficulties, it has many drawbacks such as leakage, low bonding strength, clogging due to polymer deformation, and incompatibility with various chemicals.

Selective sealing methods (Supplementary
In comparison, sacrificial approaches (Supplementary Figure 1c) offers a promise of creating complex nanofluidic structures with embedded functional CMOS electronic components. Such approaches utilize a material "to be sacrificed", either organic polymers 7 or inorganic materials such as Si 8,9 and SiO2 10 , patterned into a reverse image of the desired nanofluidic structures (e.g. isolated pillars versus meshes), and selectively extract this sacrificial material at a later stage of processing to form the nanofluidic system by thermal decomposition 7 or wet chemical etching [8][9][10] . However, thermal decomposition has serious risks of structural damage at elevated temperatures. Wet etching processes are ineffective at nanometer scales and potentially destructive to the nanofluidic structures, because removing etched byproduct becomes exceedingly difficult and undesirable long processing time is needed (e.g. >40-80 hours for millimeter long channels at micron scales) 9 .

Supplementary Note 2: Substrate planarization
The key steps to inlay sacrificial Si microstructures include deposition of a sacrificial amorphous Si

Supplementary Note 3: Sacrificial Si nanostructure patterning
In this work, we utilize established procedures and recipes at IBM MRL lab during critical nanopatterning steps to maximize the feature uniformity and yield. For example, the plasma etch uniformity in our etch chamber has an etch rate uniformity of within 5% across a 200mm wafer and from wafer-to-wafer. The critical dimension (CD) in DUV lithography has a <15nm variation for a 200nm line/space standard design across a 200 mm wafer, and the yield is about 100% for the dimensions in this work (critical dimension ~200 nm). The high yield is achieved by printing in a controlled and fully automated environment of an ASML and TEL track without manual handling and by applying internal stepper diagnostics on a regular basis to control the focus and dose. To optimize the alignment accuracy between the EBL and DUV fluidic nanostructures, we patterned alignment marks on the substrate prior to nanopatterning, consisting of marks designed for EBL and DUV. We used the same alignment mark sets in both EBL and DUV nanopatterning, hence minimizing the alignment errors. Taking into account the alignment accuracies of different levels, i.e. <20 nm for EBL and DUV and about 1 µm for MUV, we carefully designed the corresponding fluidic structures with large enough tolerance.

Supplementary Note 4: Sacrificial Si extraction
The high optical contrast between Si and SiO2 allows us to conveniently monitor the XeF2 etching process of Si (Supplementary Figure 7). In our layout-design, the venting holes are separated by 60 µm (Figure 3). The sacrificial Si materials in the nanochannels did not have any venting holes patterned on the top; instead, they are extracted through the venting holes patterned on top of the micrometer-and nanometer-sized channels connecting the critical nanochannels. To investigate the effect of the feature size on the XeF2 etching process, we monitored the location of the Si etch-front by optical microscopy (Supplementary Figure 8 a). In the experiments, we tested the feature size from ~70 nm down to ~13 nm. Clearly, the XeF2 etching rate of amorphous Si in wider nanochannels was much faster than narrower channels (Supplementary Figure 8 b).
The size-dependent Si etching rate can be understood as a result of size-dependent vaporphase transport of the XeF2 precursor to Si surface and the volatile byproducts away from the Si surface. Obviously, the diffusion of XeF2 gas and by-product is slower within narrower channels. This can be attributed to higher probability of gas molecules to collide with the nanochannels sidewalls at vacuum (3 Torr XeF2, 15 Torr N2 in our experiment), in agreement with Knudsen diffusion model. Our experimental results also showed a linear dependence of etching rate versus channel dimensions, probably because the diffusivity is proportional to the critical dimensions of the nanochannels at the Knudsen diffusion regime. In spite of the slow etch-rate in the narrow (sub-20 nm) nanochannels, the successful etching can be completed by increasing XeF2 gaspurging time and cycles. In our experiments, 20 µm long, sub-20 nm wide, and 40 nm high nanochannels were successfully extracted.
The very narrow nanochannels were found to close under TEM observation. For example, an initially measured 14 nm wide channel were observed to close during TEM imaging (Supplementary Figure 9). Although not fully understood at this stage, this effect is probably attributed to the electron beam induced carbon deposition at the nanochannels and/or melting and coalescence of PECVD SiO2 under high-energy electron beam irradiation (possibly driven by surface energy).
Supplementary Figure 10 summarizes the SiO2 materials used in the sacrificial Si patterning, extraction, and sealing processes. SiO2 is deposited for three different purposes -substrate planarization, Si structure capping, and venting hole sealing. The PECVD SiO2 substrate and the planarization, capping, and sealing SiO2 films define the eventual microfluidic and nanofluidic device structures.

Supplementary Note 5: Nanofluidic structures and DNA hydrodynamic studies
Here we aim at demonstrating the capabilities of our sacrificial Si strategy of integrating complex and functional nanofluidic structures using a customized fluidic jig (Supplementary Figure 11). In our design, the nanopillar design is the key to achieving complex DNA hydrodynamic interactions, and the nanofluidic channel dimensions are not critical. The devices were fabricated following the strategy we detailed in previous section (Supplementary Figure 5), but here we chose DUV lithography rather than EBL to fabricate the nanofluidic structures ( Supplementary Figure 12), similar to our previous report. 4 Each fluidic chip was designed to have six isolated fluidic branches, which have identical pillar designs. Within each fluidic branch ( Supplementary Figure 12 a), the nanofluidic pillars and channels were patterned in an area of 700 µm × 400 µm and connected by microchannels on both sides. The nanofluidic structures included nanochannels in the middle surrounded by symmetrically arranged diamond-shaped nanopillars on each side (Supplementary Figure 12 b-c). The fluidic design featured diamondshaped nanopillars with abruptly designed interface to control DNA straddling interaction 11 and pillar gaps that are progressively reduced in dimensions from 1.4 µm to 240 nm, functioning effectively as cascaded two-dimensional fluidic network to pre-stretch the DNA. 3 With consecutively captured fluorescence images (exposure time 17.8 ms, frame cycle time 18.1 ms), we studied the single-molecule λ-DNA molecule translocation through the nanopillar and nanochannel regions ( Figure 5). The frame-by-frame speeds and extensions of these DNA molecules were derived by measuring the DNA head and tail locations. The extension is the fluorescently measured length corrected by DNA travel distance during exposure time using measured DNA frame speed through the relation . The average DNA speed in the imaged region is 140 / . Clearly from Figure 5, the extensions of DNA molecules are strongly correlated to the nanofluidic structure design. In this report, we do not focus on the detailed DNA hydrodynamic interactions with the diamond-shaped pillars, which have thoroughly analyzed in our previous report using similar structures. 4 In a different chip, the DNA molecules have similar hydrodynamic flow, straddling, and relaxation interactions with the nanopillars (Supplementary Figure 13). Clearly, the DNA molecule stretched much longer after entering the nanochannels (Supplementary Figure 13 a, frames 6-10) and also straddling nanopillars (Supplementary Figure 13 a, frames 14-18). This demonstration further illustrates the complexity of DNA hydrodynamic behaviors in nanofluidic structures, and also emphasizes the importance of our integration strategy in nanofluidics and single molecule studies.

Supplementary Note 6: Fluidic chip electrical test
The aim of this work is to demonstrate the feasibility of our strategy in wafer-scale integration of complex nanochannels and compatibility with single molecule fluorescence imaging. In this work, we also carried out ionic conductance measurement using our fluidic probe stations on two randomly selected chips at the edges of the 200 mm wafer. To evaluate the fluidic connection, we used a fluidic probe (Qmix, CETONI GmbH, Germany) to deliver a KCl buffer solution to the fluidic chip and simultaneously control the pressure and flow rate (Figure S14 a). An optical microscope with a CCD camera was mounted onto the probe station to visualize the fluidic channel regions during test (Supplementary Figure 14 b-c). The fluidic chips were first wetted by DI water under a typical pressure of 0.5 Bar. No leakage was observed using a pressure as high as 11 Bar, and no higher pressure was attempted. Then, the DI water was replaced by KCl solutions (pH 5.5) at different molarities, i.e. 1 mM to 100 mM. Using Ag/AgCl wires inserted in the fluidic probes in contact with the KCl buffer, we measured the electrical conductance of the fluidic chip across two fluidic access ports as 0.44, 2.3, 17.7 nS, respectively. The linear dependence of conductance on salt molarity indicated a complete wetting of the fluidic chips. In addition, the ionic current of the fluidic chips was stable over 11.1 hours (Supplementary Figure  14 f), indicating good device stability. From two wetted fluidic branches on each of the two randomly selected chips, we obtained very similar ionic current (variation <10%). The good agreement is attributed to a few reasons. First, the nanofluidic structures have uniform dimensions. Second, the nanofluidic channels are fully wet. Thirdly, the two-dimensional fluidic network in our design has many parallel channels connecting the inlet and outlet, and hence has a much more stable current compared to a single channel. The small variation is attributed to occasional air bubbles injected by the fluidic probes, interface resistance at the Ag/AgCl electrodes, etc.