High‐Density Plasmonic Nanopores for DNA Sensing at Ultra‐Low Concentrations by Plasmon‐Enhanced Raman Spectroscopy

Solid‐state nanopores are implemented in new and promising platforms that are capable of sensing fundamental biomolecular constituents at the single‐molecule level. However, several limitations and drawbacks remain. For example, the current strategies based on both electrical and optical sensing suffer from low analyte capture rates and challenging nanofabrication procedures. In addition, their limited discrimination power hinders their application in the detection of complex molecular constructs. In contrast, Raman spectroscopy has recently demonstrated the ability to discriminate both nucleotides and amino acids. Herein, a plasmonic nanoassembly is proposed supporting nanopores at high density, in the order of 100 pores per µm2. These findings demonstrate that the device has a high capture rate in the range of a few fm. The pore size is ≈10 nm in diameter and provides an amplification of the electromagnetic field exceeding 103 in intensity at 785 nm. Owing to these features, single‐molecule detection is achieved by means of surface‐enhanced Raman scattering from a solution containing 50 fm DNA molecules (≈4.4 kilobase pairs). Notably, the reported spectra show an average number of 2.5 Raman counts per nucleotide. From this perspective, this number is not far from what is necessary to discriminate the DNA sequence.


Introduction
Over the last two decades, novel biosensing devices based on various nanomaterials and different sensing schemes have been proposed for the detection of DNA and other molecules of interest DOI: 10.1002/adfm.202301934 in genetics and pathology.[3] In this regard, lateral flow nucleic acid assays are promising, rapid, and low-cost devices based on colorimetric detection. [4]Furthermore, electrochemical sensors based on magnetic nanoparticle assays or enzyme-labeled metallic nanoparticles have demonstrated low detection limits at the femtomolar level. [5]][9][10][11][12][13][14] Nanopores are generally based on a simple working mechanism, in which an electrical readout enables us to monitor the temporary blockage of the ionic current level during the passage of biomolecules.However, for complex molecules such as proteins, the electrical readout cannot provide the robust amount of information necessary for their molecular identification.In the context of solid-state nanopores, another important limitation is the minimum concentration required for analysis.As pointed out by Xue et al., solid-state nanopores require high analyte concentrations in the range 1-10 nm because of their low throughput. [15]Improvements in the capture rate have been achieved by concentrating the analyte close to the nanopore entrance using various strategies such as dielectrophoretic trapping, [16] the integration with a magneticbead-based sandwich immunoassay, [17] and the introduction of ionic concentration gradients. [18]Finally, among the bottlenecks affecting electrical sensing, limited bandwidth must be mentioned.
In summary, electrical sensing with nanopores faces various challenges including limited discrimination power, limited bandwidth (fast molecular translocation), and limited capture rate (low molar sensitivity).A possible solution involves switching to optical sensing that can manage temporal scales of microseconds or even shorter, with no significant limit in the bandwidth on those temporal scales.In fact, the typical translocation time of one nucleotide or amino acid is on the order of 1 μs, whereas commercial optical detectors can go down to the nanosecond range.Notably, optical sensing can benefit from the utilization of plasmonic nanostructures to exploit local enhancement of electromagnetic fields, referred to as plasmonic hotspots.[21][22][23][24][25] Single solid-state nanopores have been integrated with plasmonic nanostructures, such as metallic nanobowties, [25,26] nanogaps, [27][28][29][30] and nanotips. [31]It is worth to note that Raman scattering has been demonstrated to be able to discriminate the four DNA bases and all the twenty proteogenic amino acids. [20,32]Hence, plasmon-enhanced Raman sensing combined with solid-state plasmonic nanopores may achieve high discrimination power and bandwidth.Hence, in principle, all the aforementioned limitations affecting electrical sensing could be overcome.
However, there are challenges also associated with Raman readouts, such as i) the design and fabrication of plasmonic pores a few nanometers in diameter still capable of providing extremely high field enhancements (in analogy with plasmonic nanoparticles, small pores usually provide low polarizability and low enhancement), ii) the photon emission rate at moderate laser power necessary to prevent molecular damage, iii) optical noise from the metallic substrate, and iv) high analyte capture rate and molar sensitivity, which has not yet been addressed for plasmonic nanopores.
Here, we report on a free-standing high-density array of 3D plasmonic nanopores and related enhanced Raman measurements that address several of the mentioned limitations and potentially address all of them.First, the high pore density (100 pores per μm 2 ) enables a higher capture rate compared to single nanopore-based sensors and higher efficiency in both optical excitation and collection.Second, the 3D funnel shape of the pores, combined with their small diameter (10 nm), provides extreme amplification of the electromagnetic field, exceeding 10 3 , a value considerably higher than that achievable with conventional plasmonic pores.In addition, the funnel shape contributes to enhancement of the capture rate.Owing to these features, we achieved Raman detection of DNA molecules at the singlemolecule level from a solution with a concentration as low as 50 fm and a capture rate of ≈10 molecules per minute.At a concentration of 10 pm, the capture rate approaches 10 3 molecules permin.These values are approximately two orders of magnitude higher than the current state-of-the-art values.Other plasmonic nanopores reported in the literature, such as plasmonic nanoslits or bowtie nanoantennas integrated with solid-state nanopores, typically require concentrations of molecules in the nanomolar range or higher, similar to those required for electrical sensing. [33,34]Notably, under moderate laser power (2.5 mW) and with a low-energy wavelength (785 nm), we achieved an average optical count rate over the noise of more than two counts per μs (or per nucleotide) using a standard CCD camera.The signal-to-noise ratio ranged between 3 and 4, depending on the Raman peak.These values do not appear to be far from those required for single-nucleotide identification.Finally, the fabrication approach based on the assembly of nanoparticles is highly reproducible and potentially scalable for small-scale production without the use of expensive top-down nanofabrication tools.
In the following, after describing the fabrication approach, we first characterize the system performance using the electrophoretic delivery of DNA molecules and their electrical readout to investigate the transport properties, which are fundamental for flow-through-based devices.Subsequently, we discuss the optical characterization of the device and the related DNA detection by SERS.

Results and Discussion
[43][44] By taking inspiration from these approaches, we developed a robust protocol for realizing freestanding plasmonic nanopores.A PNA on a freestanding 2D nanoassembly was realized using nanosphere lithography and then transferred onto a supporting holey substrate, as schematically shown in Figure 1a.The details of the fabrication process are reported in the Experimental Section.Typically, the fabrication of freestanding nanoassemblies is based on four steps: the choice of building blocks (i.e., nanoparticles), a method to induce their assembly in a highly controllable way, [45][46][47] a strategy to connect neighboring building blocks, and finally, the transfer of the resulting nanoporous membrane onto a holey substrate.[50] In our study, neighboring polystyrene (PS) nanospheres of hexagonal close-packed (hcp) monolayers were connected by melting their reciprocal contact points with annealing above the glass transition temperature of PS. [51] The colloidal mask was then covered with 40 nm Au.Top-view and cross section SEM images of the resulting Aucoated colloidal masks are reported in Figure 1b and Figure S1a (Supporting Information), respectively.The Au layer provides the plasmonic properties of the PNA while also giving mechanical strength to the overall nanoassembly, which results in it being robust enough to be detached from the sacrificial substrate and transferred onto a thin insulating membrane (Figure S1b, Supporting Information), supporting a pattern of micron-sized holes.As obtained from the analysis of the SEM images of the PNA, this approach provides scalable fabrication of high-quality freestanding nanoassemblies supporting ordered nanopores ≈10 nm in diameter with a density of 100 pores per μm 2 (see Figure 1c).The shape and diameter of the nanopores were controlled by shrinking the triangular interstices between adjacent nanospheres to circular apertures using a controlled annealing process, as shown in Figure S2a-e (Supporting Information).To highlight the transfer process, a SEM image of the backside of the holey membrane after the transfer of the nanoassembly is shown in Figure 1d.Thus, the nanoassembly completely covered the patterned holes on the membrane.In this example, the membrane was patterned with several micron-sized holes to highlight the scalability and high quality of the fabrication approach across the sample.Furthermore, the deposition of the Au film has the additional benefit of completely plugging any void spaces and defects that occur during the packing of the PS nanospheres.Deposition was performed by direct current sputtering to fill the walls of the nanopores.A magnified SEM image of the backside of the PNA confirmed the presence of an Au crown on the backside of the nanopores, as shown in Figure 1e.The number of effective nanopores was tailored by changing the hole diameter in the supporting membrane.Although we focus on the device made by plasmonic nanoassembly transferred onto a holey membrane, the transfer process can be repeated on other types of substrates, such as micropipettes and gels, as shown in Figure S1c (Supporting Information).Before performing the flow-through measurements, the PNA was treated with oxygen plasma, as described in the Experimental Section, for two reasons: to completely etch away the PS nanospheres and hydrophilize the surface and to increase the wettability of the nanopores.
Then, we electrically characterized the pores and translocation dynamics to test the ionic transport properties and molecular capture rates.For this characterization, we used a PNA consisting of 20 nanopores 10 nm in diameter that were sealed with gaskets between two Teflon chambers, and both compartments were immersed in a potassium chloride (KCl) solution at different concentrations.The I-V characteristics, namely, current versus voltage plots, were acquired with a pair of Ag/AgCl electrodes, as described in the Experimental Section (see Figure 2a).As expected, the behavior of the ionic conductance as a function of the KCl concentration in the array of nanopores was similar to that reported in the literature for their single counterparts, [52] indicating that there are two transport regimes.At high KCl concentrations (> 10 mm), the current, and in turn the conductance, increases significantly as a function of the KCl concentration, and at low KCl concentrations (< 10 mm), the conductance is mainly constant and depends on the surface charge of the nanopore walls.Furthermore, at low KCl concentrations, rectification of the current occurred because of two effects: the funnel-like shape of the nanopores, which induced a different distribution of the surface charge on the two openings of each pore; and the small separation between neighboring pores, which may cause a different distribution in the ionic flux transported in the middle and at the edge of the array.The rectification ratio as a function of the applied voltage at low KCl concentrations is shown in Figure S3a,b (Supporting Information).The number of active pores affected the current measured at a fixed KCl concentration, as is evident from the IV characteristics collected from arrays of different sizes while in contact with the 10 mm KCl solution (see Figure 2b).In particular, the total ionic current did not increase linearly with the number of pores.In other words, it was not exactly equal to the sum of the currents from the isolated nanopores.Recent findings indicate that when the separation between neighboring nanopores is less than 50 times their diameter, the current flowing through the middle of a finite array of nanopores can be different from that flowing through the ones at the edge of the array. [53]sing DNA molecules of ≈4.4 kilobase pairs (kbps) diluted in 10 mm KCl at a final concentration of 500 pm in the cis chamber, DNA translocation through a PNA supporting 100 pores was observed under an applied bias of +200 mV, as shown in Figure 2c.The translocation events caused a temporary enhancement in the current, as observed from the current trace shown in Figure 2d.This effect has already been reported for DNA molecules diluted in low-electrolyte solutions and is a result of the net excess charge that the molecule brings inside the nanopores. [54]he current trace contains information on single, multiple, and consecutive translocations that can occur at any given time because of the presence of an array of nanopores.Translocations of significantly different amplitudes were observed and can be used to discriminate between single and multiple translocations at a given time, as is clearly visible in the orange and green insets in Figure 2d, respectively.The exponential fit performed on the corresponding count-normalized histogram of the dwell time showed that translocation events occurred with a dwell time of 2 ms at +200 mV, which is expected for DNA molecules of such lengths. [55,56]n the following paragraph, we show that high-density plasmonic nanopores can be used to capture and detect DNA molecules at ultra-low concentrations, thus showing an advantage over single-nanopore systems with electrical readouts, which fail to manage low concentrations of DNA.
Let us explain this important aspect of the difference between electrical and optical readouts.When electrical detection is used, the sensitivity is related to the ratio between the molecular volume and the sensing volume (i.e., the nanopore volume).This ratio determines the change in the blockade current flowing into the pore, that is, the readout.An array of closely spaced pores behaves as a single (large) nanopore with a total volume largely exceeding the molecular volume.Consequently, this collective configuration exhibits low sensitivity.In contrast, in the case of optical detection, this reasoning is inappropriate because the optical readout is unrelated to the pore volume.Hence, one may exploit an array of pores in close contact, which leads to an increased capture rate and molar sensitivity.Using an array of pores, it is possible to generate numerous plasmonic hot spots simultaneously.When pores are densely packed and confined in an area of 1-2 microns, they can be simultaneously excited with a high NA objective while the scattered light can be collected with the same objective in reflection mode or transmission mode by using with another objective.Thus, the probability of capturing and detecting a single molecule is significantly increased without compromising the ability to discriminate molecules that are associated with the Raman spectra.
To prove this capability, we investigated the dynamics of translocation with respect to the analyte concentration and measured the capture rate and translocation time.Figure 3a shows the current traces recorded for 10 s, which were obtained by progressively decreasing the DNA concentration in the -cis chamber from 500 pm to 5 fm, while an external bias of +200 mV drove the molecules toward the PNA.A histogram of the dwell time and current enhancement corresponding to the translocation events for the trace recorded with 500 pm DNA in the cis chamber are reported in Figure 3b.Histograms of the dwell time and current enhancement of the other current traces are shown in Figure S4 (Supporting Information).The capture rate as a function of the number of DNA molecules added to the cis chamber, calculated after 1 min of acquisition of the corresponding current traces, is shown in Figure 3c.The number of events per minute at 5 fm DNA is ≈10 ± 3 without turning to any pre-concentration method to locally increase the concentration of molecules nearby the PNA.These results are in agreement with the qualitative estimation of the translocation rate, which can be derived analytically, [57][58][59] as shown in Section S3.2 (Supporting Information).This proves that the array exhibits an enhanced capture rate compared to its single-nanopore counterpart, enabling us to perform high-throughput flow-through sensing.Additionally, at DNA concentrations below 1 pm, the probability of two translocation events coinciding is extremely low; therefore, the temporary enhancement of the ionic current can be interpreted as a single translocation event.Furthermore, this was confirmed by the distribution of the interarrival times, which followed Poisson statistics, as shown in Figure S7 (Supporting Information).Nevertheless, because the neighboring pores are in close proximity, a single molecule can affect the current of several pores when it does not translocate through a single pore in a linear head-to-tail state, and it experiences simultaneous capture, as observed in double-pore systems. [60]This is clear for certain translocations, which, apart from longer dwell times and higher current enhancements, show steps in the current event and, therefore, could be explained as capture of a single molecule in double-pore systems instead of simultaneous translocation of multiple DNA molecules.Further discrimination between the translocation events depending on the amplitude and shape of the current enhancement is shown in Figure S5 (Supporting Information) and discussed in the Supporting Information.Furthermore, it is worth mentioning that the enhancement of the capture rate arises not only from the high density of nanopores supported on the nanoassembly but also from their funnel-like shape.As discussed in the literature, with this type of 3D shape alone the translocation frequency of proteins and DNA through single nanopores can be enhanced up to ten times. [61]Further investigation of the influence of the number of nanopores and their funnel shape on the translocation event frequency is shown in Figure S6a,b (Supporting Information).
In the following section, we discuss the optical characterization and Raman measurements.The transmission spectra of the freestanding plasmonic nanoassembly are shown in Figure 4a for three different cases: i) large triangular pores 40 nm in size, ii) small circular apertures 10 nm in size, and iii) totally plugged circular pores.Here, the spectra were acquired after both the transfer of PNA onto the holey membrane and the removal of the PS beads with oxygen plasma treatment, as discussed in the Experimental Section.However, even before the detachment of PNA from the sacrificial layer, these three samples exhibited different colors, blue, purple, and goldish, respectively, as shown in Figure S8a (Supporting Information).This results from the fact that the interstices are slowly reshaped by increasing the duration of the annealing performed on the colloidal mask, eventually to the point that they are totally plugged.The PNA supports two modes: the first mode at shorter wavelengths, which is localized on the Au caps, and the second mode at longer wavelengths, which is confined inside the nanopores.This second mode depends on the size of the gap between the tips of the Au caps, that is, the diameter of the pores, and disappears if the nanopores are plugged.Hence, measuring the optical response of PNA is an alternative approach for estimating the geometry of nanopores without relying on SEM imaging.Furthermore, we performed optical simulations of the electric field enhancement at 785 nm by using COMSOL Multiphysics® software and considering the geometry of the PNA for the case of light impinging from the back side of the Au caps, as shown in Figure 4b.Because the electric fields are highly confined and enhanced inside the pores for light hitting the sample in this reversed configuration, it is clear that the nanoassembly supporting high-density plasmonic nanopores of 10 nm have the potential to perform surfaceenhanced Raman spectroscopy (SERS) with a 785 nm laser in a flow-through configuration (see the sketch in Figure 4c).The enhancement of the electromagnetic field inside the nanopores by up to a factor of 10 3 , namely in the gap between adjacent metallic nanospheres, arises from the coupling of their dipoles under resonant excitation. [62,63]Conversely, for light impinging in the opposite direction, namely in the forward configuration toward the Au caps, the electric field was enhanced around them and not inside the nanopores, as confirmed by the simulated elec-tric field distribution reported in Figure S8b (Supporting Information).The transmission, reflection, and absorption spectra of the PNA with 10 nm nanopores were simulated for both cases of light impinging the nanocaps in the forward and backward configurations, as shown in Figure S8c (Supporting Information).In particular, the simulated and experimental transmission spectra are in good agreement, except for the fact that the experimental transmission dip is broader due to some defects in the assembly, which are plugged from the Au deposition.
To test the optical sensing capabilities of the PNA, an in-housemade polydimethylsiloxane (PDMS) chamber was used to seal the chip supporting the PNA and form -trans and -cis compartments.Both chambers were filled with 10 mm KCl solution, and 50 pm DNA was introduced into the -cis chamber, as shown in Figure 4c.Upon laser excitation, the Raman signals of the DNA molecules translocated through the plasmonic nanopores were collected over time with an integration time of 10 ms, which was the minimum value allowed by our Raman setup (InVia Renishaw equipped with conventional CCD camera).Thus, the Raman time traces of the translocating DNA under the diffusion regime (without applied bias) and electrophoretic regime (with the bias set to 200 mV) were collected, as shown in Figure 4d.[66] Of course, the external bias increases the number of molecules translocating through the nanopore array because few rare diffusion events are observed without bias.Examples of SERS spectra acquired over time are shown in Figure 5a Figure 5. a) Examples of SERS spectra collected over time, when the DNA does not translocate through the nanopores (in gray), during translocation events for the diffusion regime (in blue) and under a bias of 200 mV (in red).Two SERS spectra labeled with "1" and "2" are reported as representative of two different translocation events for both the electrophoretic and diffusion regimes.The spectra are plotted with a Y-offset for better visualization.[66]  for three cases: i) no molecule translocating through the PNA (baseline signal), ii) during typical translocation events in the diffusion regime (no bias), and iii) during a typical translocation event under electrophoretic pull.The SERS spectra of the translocating DNA molecules detected in the hotspot sites of the nanopores exhibited a typical vibrational Raman signature of the DNA backbone and its constituents (the peak assignment is shown in Figure 5a).
By decreasing the DNA concentration in the cis chamber, the number of translocation events decreased, as inferred from electrical measurements shown in Figure 3c.Therefore, at lower DNA concentrations in the cis chamber, the time-averaged Raman signal decreases.This was observed from the Raman time traces shown in Figure 5b,c in which the DNA added to the cis chamber was 500 and 50 fm, respectively.In the latter cases, the intensities of the SERS signals collected under the electrophoretic pull were, on average, three times lower than those included in the Raman time trace in Figure 4d.Due to the negligible probability of observing multiple translocation events in the femtomolar regime, these results attest to the capabilities of the PNA to perform SERS in the flow-through even in the singlemolecule regime.For example, the intensity of specific bands of DNA at 1080 cm −1 showed visible spikes over time, indicating the passage of a single molecule into the pore array, as shown in Figure 5d.We noticed that the capability to detect translocating DNA molecules at fm concentrations is a remarkable result compared to the current state-of-the-art, which is in the nanomolar range.This high molar sensitivity is due to the large number of pores combined with their high density (100 μm 2 ).This combination enables the illumination and the collection of the scattered light within the area delimited by the laser spot, which has a diameter of ≈1 μm.In addition, during single-molecule translocation, the total number of Raman counts integrated along the full spectrum varies from one translocation to another.Quantitatively, after noise subtraction, the number of counts per DNA molecule range from 9000 to 12 000.Such a variation is reasonably due to the fact that translocation times may vary from one molecule to another, as determined by electrical measurements (Figure 3b), and is well known in the literature.In addition, a molecule passing close to the pore wall is subjected to a plasmonic field higher than that of a molecule passing through the pore center; hence, it emits a higher signal.This demonstrates the importance of reducing the pore size to ≈5 nm, at which the pores are uniformly filled with the plasmonic field.
By analyzing the data in Figure 5b, one can notice that the number of detected translocations is on the order of 10 events per minute, while the electrical readout estimated a value of 70, which is approximately seven times less.This is due to the fact that in our Raman setup, there is a "dead" time (the detector is off) between two consecutive acquisitions that are equal to 200 ms.Since the integration time is equal to 10 ms, we are currently able to get only 5% of the total events (the system acquires for 10 ms, and then it is off for 200 ms before the next acquisition).This clearly explains the reduction in the capture rate in the optical experiments owing to the technical limitations of our system.This shows that the plasmonic array can capture all events, although the camera records only 5% of them.This limitation can be overcome using state-of-the-art optical systems, such as SPAD or PMT cameras, that currently reach bandwidths of tens of MHz.
To understand whether this device can be used to discriminate between biomolecular constituents, it is important to evaluate the number of photons emitted by a single nucleotide in DNA/RNA molecules or by an amino acid in a protein, and whether this number is large enough to discriminate between individual components, or at least some of them.For example, it has been estimated that 90% of human proteins can be identified by recording sequences of only two amino acids. [67]As aforementioned, we recorded an average of ≈10 4 counts per DNA molecule that is composed of ≈4.4 kbps, hence an average of 2.5 Raman counts per nucleotide.From this perspective, it is reasonable to think that by reaching an average of at least 50-100 counts per nucleotide (or amino acid), it would be possible to reconstruct their relative sequences to identify the molecule or, in principle, even sequence it.In this regard, many improvements are feasible in the medium term, such as reduction of the pore size and improvements of the plasmonic response, tailored and better performing optical systems, fast cameras with high bandwidth and quantum efficiency in NIR, and data processing.In addition, as in the case of electrical sequencing, signal quality can be improved by summing the signals of many consecutive molecules and taking advantage of artificial intelligence and advanced data analyses. [68,69]

Conclusion
We proposed a method for the realization of a plasmonic freestanding array of nanopores.The fabrication method is robust and reproducible.Based on colloidal lithography, it is potentially scalable for small-scale production.The nanopore assembly can be moved onto various holey substrates for optical sensing in a flow-through configuration.Before optical measurements, we investigated the electrical currents flowing into the pores during DNA electrophoretic translocation.In this way, we characterized the DNA translocation dynamics and confirmed single-molecule translocation.The nanopore density reached 100 pores μm −2 .The large number of pores enabled a significant capture rate of the DNA molecules, even at concentrations well below 1 pm.We showed that an assembly of 100 pores could be optically excited and collected simultaneously using a single high-NA objective in reflection mode.By exciting the pores at a wavelength of 785 nm, we achieved a local field amplification of up to 10 3 in intensity.Such an enhancement, combined with the high capture rate, enables the detection of DNA molecules at the single-molecule level down to a concentration of 50 fm, which is considerably better than the current state-of-the-art methods.Notably, the reported spectra show an average number of 2.5 Raman counts per nucleotide.From this perspective, this value is not far from that necessary to discriminate the sequences of individual nucleotides in DNA molecules.However, a significantly faster optical camera is required to achieve this goal.Commercial SPAD cameras are currently being developed rapidly; hence, this goal is not unrealistic.Furthermore, by adding thin coatings to the Au surface, it is possible to decrease the electrostatic interactions between the nanopore walls and molecules to better control translocation events. [70]With further optimization and development, this approach can be applied not only to DNA but also to RNA and proteins.

Experimental Section
Preparation of PNAs and Transferring on a Holey Supporting Substrate: First, PS nanospheres were assembled at the water-air interface of a water bath using a previously developed interfacial self-assembly method to form hexagonally close-packed (hcp) arrays of nanoparticles. [71,72]Briefly, nanospheres were first spread to form hexagonal close-packed (hcp) grains on a hydrophilized silicon wafer at an angle before being transferred to a water bath.Negatively charged PS spheres with a diameter of 140 nm were acquired from microParticles GmbH.These spheres were dispersed in 5 wt.% aqueous solution.This process of PS nanosphere assembly could be optimized by increasing the pH of water to 9 and adding a solution of sodium hydroxide to the water bath.This promoted proper packing of the charged nanospheres and reduced the presence of defects in the grains.The hcp monolayer of PS nanospheres was transferred onto a Si substrate, which was previously coated with a sacrificial layer of aluminum oxide (Al 2 O 3 ) deposited using an atomic layer deposition (ALD) system (thickness: 40 nm, temperature: 80 °C, Oxford Instruments FlexAL).Following the assembly process, adjacent PS nanospheres were cross-linked with a thermal annealing performed with a hot plate at 117 °C, above the glass transition temperature of polystyrene (≈105 °C).Consequently, the triangular interstices between neighboring spheres were reshaped into circular apertures, and their average size was reduced from an initial value of 50 to 10 nm by adjusting the heating time.The sample was then coated with 40 nm of gold by sputtering (rate:22 nm per min, Quorum Sputter Coater Q150T ES).To transfer the PNA, an aqueous potassium hydroxide (KOH pellet from Sigma-Aldrich, 6% w/v in DI water) solution was prepared to soak the sample and chemically etch the sacrificial layer.The PNA was then transferred to the air-water interface of DI water with gentle immersion of the sample.Finally, the floating PNA was carefully fixed on a holey silicon nitride (SiN x ) membrane, which was previously patterned with micron-sized holes, using a focusing ion beam scanning electron microscope (FEI Helios NanoLab 650, voltage: 30 kV, current: 25 nA).
Morphological and Optical Characterization: A focused ion beam scanning electron microscope (FEI Helios NanoLab 650) was used to capture images of the samples.Transmission spectra were recorded in the spectral range of 400-1000 nm using a Woollam V-VASE ellipsometer.
Electrical Measurements: The chip was sealed with two gaskets between two Teflon chambers filled with a KCl electrolyte solution of potassium chloride.A pair of Ag/AgCl electrodes were then placed to apply an external bias and monitor the ionic current through the nanopore arrays.The electrodes were connected to an amplifier (Axopatch 200 B, Axon Instruments) with a digitizer operating at 200 kHz and a low-pass filter with a cutoff frequency of 10 kHz.Before filling the cis and trans chambers with the electrolyte solutions, the hydrophilicity of the nanopores was increased by oxygen plasma treatment performed with an oxygen plasma cleaner (power:100 W; oxygen flow rate:20 sccm) on both sides of the PNA.Subsequently, both chambers were filled with ethanol, which was replaced with a degassed electrolyte solution.This procedure enhances the wettability of nanopores and prevents the formation of bubbles, which were typical problems during electrical measurements of nanopore-based devices.
PDMS Encapsulation: A microfluidic chamber made of polydimethylsiloxane (PDMS, Dow Corning SYLGARD 184 silicone elastomer) and cured at 65 °C for 2 h was used to encapsulate the sample and form two compartments in a vertical configuration suitable for SERS measurements using the Raman setup.
SERS Measurements in Flow-Through: A Renishaw inVia Raman spectrometer equipped with a Nikon 60× water immersion objective and a 785-nm laser were used to perform the Raman measurements.The laser beam was adjusted to have a power of 10% and a spot size of 1 μm.The integration time was set to 10 ms.For translocation experiments, double-stranded plasmid DNA molecules of ≈4.4 kbps were purchased from Thermofisher (pBR322 Plasmid DNA, purified by using chromatography) and diluted in buffer (10 mm Tris-HCl, 1 mm EDTA) to a concentration of 5 μg mL −1 and stored at −20 °C until use.Furthermore, according to the analysis provided by the supplier, 90% of the molecules were in the coiled state.They were then diluted in a potassium chloride solution to the desired concentration and introduced into the -cis chamber.By applying a voltage of 200 mV across the two compartments during Raman measurements, the translocation of DNA molecules was promoted by electrophoresis.
Optical Simulations of the PNA: The COMSOL Multiphysics® software was used to perform optical simulations. [73]The refractive index and extinction coefficient of gold were taken from the work of Rakić et al., [74] and the refractive index of the environment (water) was set to 1.33.The incident light from 0.4 to 1 μm was illuminated along the Z-axis.Periodic boundary conditions and perfectly matched layers were applied perpendicular and parallel to the Z-axis, respectively.The PNA structure was built based on the SEM images.
Statistical Analysis: The count distributions of the dwell times and current enhancements were obtained by analyzing the current traces using a half-amplitude threshold method integrated in the Clampfit software.Two input values were set for baseline and trigger, and the software identified the start of an event by crossing the trigger level.The next crossing of the trigger level determined the end of the event.Therefore, the duration of each event determined the dwell time, whereas the maximum current value determined the current enhancement during the translocation event.Events faster than 0.1 ms were discarded because they were not related to real translocation events, but to scattering processes near the entrance of the nanopores.The capture rate as a function of DNA concentration shown in Figure 3c was obtained by averaging the total number of translocation events detected for three current traces, each with a duration of 1 min.The standard deviation of the average value is also displayed in Figure 3c.

Figure 1 .
Figure 1.a) Scheme of the fabrication of the free-standing plasmonic nanoassembly and its' transfer on a holey supporting substrate.b) Top-view scanning electron microscopy (SEM) image of the nanoassembly after reshaping the interstices of the array and Au coating (scale bar: 300 nm).c) Histogram of the nanopores diameter supported on the plasmonic nanopore array (PNA).d) Back-view SEM image of the PNA transferred on a membrane patterned with 1.5 micron-sized holes (scale bar: 4 μm).e) High magnification SEM image of the PNA (scale bar: 100 nm).

Figure 2 .
Figure 2. a) IV characteristics as a function of KCl concentration measured from a PNA supporting 20 pores of 10 nm.b) IV characteristics measured from PNAs supporting different number of active pores in a solution of 10 mm KCl.c) Scheme of the PNA sealed between the -cis and -trans compartments.DNA molecules are added in -cis compartments and driven toward the pores by the applied bias while the translocation events are tracked over time through instantaneous changes of the ionic current.d) Current trace obtained by adding DNA (4kbps; 500 pm) in the -cis chamber and applying a bias of +200 mV toward a PNA supporting 100 pores.The insets in orange and green show single and multiple events of DNA translocating through the nanopores, respectively.e) Time-normalized histogram of the dwell time obtained from the analysis of the current trace.

Figure 3 .
Figure 3. a) Current traces with a duration of 10 s collected by adding various concentrations of -DNA in the -cis chamber under an external bias of +200 mV.b) Counts distribution of (i) the dwell time and (ii) the current enhancement, obtained from the translocations of DNA diluted at a concentration of 500 pm in the -cis chamber.c) Capture rate as a function of DNA concentration.

Figure 4 .
Figure 4. a) Transmission spectra of the PNA with different shapes and sizes.b) Simulated electric field distribution at 785 nm for the case of light impinging the Au caps from the backside.c) Scheme of the optical setup to perform SERS measurements of translocating DNA under an external bias.d) Example of Raman time trace of DNA translocating through the PNA collected with 10 ms integration time under the diffusion regime (bias off) and the electrophoretic regime (bias on).The concentration of DNA introduced in the -cis chamber is 50 pm.
Figure5.a) Examples of SERS spectra collected over time, when the DNA does not translocate through the nanopores (in gray), during translocation events for the diffusion regime (in blue) and under a bias of 200 mV (in red).Two SERS spectra labeled with "1" and "2" are reported as representative of two different translocation events for both the electrophoretic and diffusion regimes.The spectra are plotted with a Y-offset for better visualization.SERS bands of DNA are labeled according to literature.[64][65][66]Abbreviations: A: adenine, T: thymine, C: cytosine, G: guanine, pb: phosphate backbone.b,c) Raman time traces of translocating DNA under the electrophoretic pull.The concentration of DNA introduced in the -cis chamber is 500 and 50 fm, respectively.d) Intensity versus time of a specific DNA bands at 1080 cm −1 .