Comparative Benchmarking of Glass and Silicon Nitride Nanopores for Single-Molecule Detection

Abstract

In the rapidly evolving field of single-molecule sensing, solid-state nanopores have emerged as transformative tools for the label-free detection of biomolecules, ranging from DNA polymers to proteins. Yet, with two dominant platformsglass nanopores and silicon nitride (SiN _x ) nanoporesresearchers face a pivotal choice: which architecture best unlocks superior performance? Here, we deliver a head-to-head experimental comparison under comparable experimental conditions, benchmarking noise characteristics, signal-to-noise ratios (SNRs), and translocation dynamics for DNA and protein analytes across matched nanopore sizes. Our findings reveal compelling trade-offs: glass nanopores excel in DNA sensing, achieving record SNR values of >80 in 5 nm nanopores (4 M LiCl, 50 kHz filter cutoff) due to their conical geometry that focuses electric fields. In contrast, SiN _x nanopores dominate protein detection with SNR values of >120, leveraging thin membranes for enhanced current blockade from volume exclusion. Comprehensive performance metricsincluding unfolded DNA fraction, backward-to-forward translocation time ratio, translocation frequency, and perturbed eventsalso show distinct translocation behaviors of biomolecules in the two nanopore platforms. These insights, supported by finite-element simulations, establish a mechanistic framework for nanopore selection, favoring conical glass nanopores for polymeric analytes and SiN _x membrane nanopores for compact biomolecules. This work not only sets benchmarks for nanopore sensitivity but also enables the development of tailored sensors in diagnostics, sequencing, and beyond, advancing nanotechnology for high-resolution biomolecular analyses.

Introduction

Nanopore sensing has progressed from a promising idea with applications in single-molecule biophysics to a mainstream analytical platform in just over two decades. While biological nanopores were the first to demonstrate single-molecule detection, solid-state nanopores have also rapidly emerged as powerful tools. Solid-state nanopores offer distinct advantages, including enhanced robustness, tunability of pore dimensions, and compatibility with microfabrication techniques, which expand both their operating conditions and analytical applications. − Among the myriad solid-state implementations explored since, two families, “glass nanopores” ,, and “silicon-nitride (SiN _x ) nanopores”, ,− have matured into reliable sensors, outperforming their counterparts. Each embodies a distinct fabrication philosophy: glass nanopores exploit rapid, direct fabrication by pulling quartz capillaries, while SiN _x nanopores harness lithographic patterning to create ultrathin membranes with precisely located apertures. Both nanopore platforms enable label-free detection of single DNA/RNA polymers, − proteins, ,− nanoparticles, − or even entire viruses , by monitoring ionic current blockades as analytes translocate through a nanoscale orifice. Yet, despite dozens of studies reporting ever-increasing bandwidths, − signal-to-noise ratios (SNR), ,− and application niches, ,,− a systematic comparison of nanopore performance depending on the nanopore platform is still lacking. The charge and shape of the analyte molecule may influence when a researcher or an instrument designer should favor one nanopore architecture over the other. The absence of such a comparative map is increasingly conspicuous as nanopore technology migrates from proof-of-concept demonstrations in academic laboratories to deployed diagnostic and sequencing devices that must meet stringent performance, cost, and manufacturability requirements.

The popularity of glass nanopores stems from their straightforward production: a laser puller elongates a quartz capillary until its tip necks down to a few nanometers, creating an aperture that serves directly as the sensing pore. Laser pulling enables more rapid, lower-cost fabrication of nanopores with low electric noise compared to SiN _x nanopores, ,, making glass nanopores very attractive for sensing experiments. Beyond fabrication simplicity, the conical geometry inherently funnels ionic current, enabling asymmetric transport of ions and molecules within the nanopore tip, which results in ion current rectification (ICR) − that can be exploited for selective sensing, , analyte discrimination, − and ionic gating. , Moreover, the conical geometry and mechanical rigidity of glass nanopores make them ideal probes for scanning probe microscopy applications, enabling high-resolution control of the nanopore displacement. , Nevertheless, glass nanopores have trade-offs. A key challenge is the reproducible fabrication and subsequent filling of pores smaller than 5 nm. , For applications requiring multiple pores, multibarrel capillaries allow fabrication of double or quad nanopore arrays, , but scaling beyond this remains challenging. The elongated geometry of glass nanopores should also result in more extended electric fields and, hence, a larger sensing volume compared with SiN _x nanopores. As a result, the current blockades generated by analyte passage, as well as SNR, may be lower depending on the shape of the molecule and, hence, limit detection sensitivity.

SiN _x nanopores, in contrast, leverage the precision of semiconductor micro- and nanofabrication: silicon nitride films are deposited on silicon wafers, patterned lithographically, and back-etched to create freestanding membranes merely 10–30 nm thick, with individual pores subsequently drilled at predetermined locations using beam-based technologies ,, or formed by controlled dielectric breakdown. , Both techniques provide fine-tuned control over nanopore diameter and length, where the ultrathin membrane enhances sensing resolution by confining the detection volume. Equally important, the compatibility of SiN _x nanopores with semiconductor fabrication workflows opens pathways for sophisticated device integration, − from on-chip electronics to microfluidic systems. Furthermore, SiN _x nanopores are particularly well-suited for high-throughput applications requiring parallel sensing channels, − such as DNA sequencing, protein analysis, and viral particle characterization. Yet, this technological sophistication comes with barriers: the fabrication process requires expensive instrumentation and specialized expertise, ,, while production yields for sensing-grade nanopores remain suboptimal. Additionally, the thin SiN _x membranes are prone to damage during chemical cleaning or handling, and nanopore sizes may change when moving from vacuum to aqueous conditions.

Existing views of these two nanopore platforms emphasize fabrication workflow and cost. Yet, more fundamentally, their contrasting characteristics leave researchers with a critical question: which nanopore best serves their particular sensing application (Figure A)? Here, we address this decision-making challenge through direct experimental comparison under equivalent conditions.

1.

Comparison of glass and SiN _x nanopore platforms for biomolecular detection. (a) Conceptual illustration showing the two nanopore types: glass nanopores (left) and SiN _x nanopores (right), used for DNA and protein sensing. (b) Schematic illustrating the key differences between the two nanopore platforms. The nanopore size parameters d and h, are indicated. The cis side of the SiN _x nanopore corresponds to the locally thinned region, whereas the trans side corresponds to the trapezoidal structure (see Supplementary Figure S2). (c) Experimental workflow for nanopore fabrication, assembly, and measurement. Top panel: Glass nanopores were fabricated by laser-assisted pulling of quartz capillaries. Bottom panel: SiN _x nanopores were created by focused electron-beam drilling in a TEM on free-standing SiN _x membranes. Representative TEM images show nanopores with matched diameters (∼10 nm). Each nanopore was integrated into a custom microfluidic assembly for ionic current recording. Example ionic current traces of DNA translocation events are shown on the right.

Results/Discussion

Workflow of the Two Nanopore Platforms

Figure b summarizes the key differences between the two nanopore platforms in terms of their material and geometry. The glass nanopore is composed of silicon oxide (quartz), whereas the SiN _x nanopore is made of silicon nitride. The glass nanopore exhibits an asymmetrically conical shape from the cis to trans side, while the SiN _x nanopore is nearly symmetric, typically displaying a cylindrical or hourglass-like profile depending on the fabrication beam parameters.

We fabricated the nanopores of two types using the following materials and workflows. For the glass platform (Figure c, top), quartz capillaries (SiO₂; outer diameter 0.5 mm, inner diameter 0.2 mm) were processed by laser-assisted pulling to form nanopores with a diameter d (Figure b). Pulling parameters and the corresponding pore diameters are summarized in Supplementary Table S1; note that optimal parameters may vary slightly with laboratory conditions (temperature and humidity). For the SiN _x platform (Figure c, bottom), nanopores were created in free-standing SiN _x membranes by electron-beam drilling in a transmission electron microscope (TEM). The MEMS process flow used to fabricate the SiN _x membrane chips is provided in Supplementary Note 1 and Figure S1. Each chip contains a 20 × 20 μm² suspended SiN _x window with a locally thinned region (diameter ϕ = 2.5 μm; thickness h ∼ 15 nm) where the nanopore is positioned. A 1-μm SiO₂ layer between the SiN _x membrane and the Si substrate serves as an additional dielectric, reducing chip capacitance , (Supplementary Figure S2). The fabricated SiN _x nanopores have a diameter d and a length h (Figure b).

Nanopore diameters were matched across platforms (same diameter, d; Figure b). For SiN _x nanopores, this is the diameter of the constriction drilled into the membrane. For glass nanopores, this is the inner diameter at the tip of the quartz capillary. Representative TEM images of two 10 nm nanopores are shown in Figure c. Nanopores were then integrated into custom microfluidic assemblies (Figure c). A PDMS layer at the cis–trans interface in the glass nanopore device and two PDMS gaskets clamping the SiN _x chip between the cis and trans fluidic cells were employed to reduce the effective chip area exposed to the electrolyte and thereby decrease electrical noise in the recordings. ,

To benchmark performance, we measured DNA and protein analytes (Figure a). Example traces for a DNA construct carrying six barcoded markers are shown in Figure c. All measurements were performed in 4 M LiCl, 10 mM Tris-EDTA (pH 9, adjusted with LiOH). Data were sampled at 1 MHz and low-pass filtered at 50 kHz for the ionic current and 5 kHz for the voltage. Additional experimental details (fabrication of glass and SiN _x nanopores and nanopore measurements) are provided in Supplementary Notes 2–4.

Nanopore Noise and Electric Field Comparison

We first compared the noise performance of the two nanopore platforms by quantifying the root-mean-square (RMS) current noise. Figure a presents representative open-pore current traces at 0 mV for a glass and a SiN _x nanopore with matched diameters, as confirmed by TEM imaging (Figure b) and current–voltage (I–V) characterization (Figure b and c). The SiN _x nanopore exhibits a noise amplitude approximately four times higher than that of the glass nanopore (Figure a). We then evaluated RMS noise across batches of glass and SiN _x nanopores (N = 10), with each value calculated from a 3-s current trace under identical filtering conditions (Figure d). Because the low-pass filter cutoff strongly influences RMS estimates, we used a cutoff frequency of 50 kHz, higher than the commonly used 10 kHz, to enable a broader recording bandwidth. Across devices, the average RMS noise, < RMS _n >, was 5.8 pA for glass and 29.7 pA for SiN _x . The power spectral densities (PSDs) of the two platforms are shown in Figure i and j. In the frequency range of 10³–10⁵ Hz, the PSD of the SiN _x nanopore is ∼10^–1 pA²/Hz, compared with ∼10^–3 pA²/Hz for the glass nanopore. This frequency range is dominated by dielectric and capacitive noise, which scale with chip capacitance; the lower permittivity and thicker insulating layer of the glass nanopores reduce capacitance and thereby suppress high-frequency noise. We also examined the dependence of RMS noise on nanopore diameter (Figure k and l). For both platforms, RMS noise increases approximately linearly over the range up to 20 nm, consistent with thermal noise, which is inversely proportional to pore resistance. In addition, RMS noise exhibits a quadratic dependence on applied voltage (Supplementary Figure S3): for a 10 nm glass nanopore, it increases from 5.8 pA at 0 mV to 6.1 pA at ±600 mV, and for a SiN _x nanopore from 26.1 pA at 0 mV to 31–32 pA at ±130 mV, all for a cutoff frequency of 50 kHz (the voltage ranges for both nanopores are selected based on the “comparable experimental conditions” discussed later).

2.

Noise characterization and electric field distribution of glass and SiN _x nanopores. (a) Representative open-pore current traces recorded at 0 mV for glass and SiN _x nanopores of identical diameter, with a sampling rate of 1 MHz and a cutoff frequency of 50 kHz. (b,c) I–V characteristics of glass and SiN _x nanopores, respectively. (d) RMS current noise values of glass and SiN _x nanopores (N = 10 for each), calculated from 3-s open-pore traces filtered at 50 kHz, yielding average RMS noise values of ⟨RMS_n⟩ = 5.8 pA (glass) and 29.7 pA (SiN _x ). (e,f) PSD plots of the open-pore current noise for glass and SiN _x nanopores, respectively, measured at 0 mV with a 50 kHz cutoff frequency. (g,h) Dependence of RMS noise on nanopore diameter (0–20 nm) for glass and SiN _x nanopores. Each point is the average value measured from three independent nanopores. The buffer is 4 M LiCl, 10 mM Tris-EDTA (pH 9). (i,j) FEM simulations showing electric field distributions in the glass and SiN _x nanopores at the indicated applied voltages, with both devices modeled at a 10 nm diameter. (k,l) Axial electric field profiles along the nanopore axis for glass (200–600 mV) and SiN _x (50–150 mV) nanopores, respectively.

To ensure a meaningful comparison between the two nanopore platforms, we evaluated not only their pore diameters but also the electric field distributions near the pore entrance. Finite element method (FEM) simulations were performed to characterize the electric field distribution in glass and SiN _x nanopores under different applied voltages. The corresponding simulation models are shown in Supplementary Figure S4 and Note 7. Both nanopores were modeled with a diameter of 10 nm, while the geometric parameters of the glass nanopore, including the cone angle and axial length, were derived from previous experimental characterizations. , Figure e and f present the simulated electric field maps for three representative voltages in the glass and SiN _x nanopores, respectively. The results indicate that a substantially higher voltage is required for the glass nanopore to achieve the same peak electric field strength as that in the SiN _x nanopore of identical diameter. Specifically, applied potentials of 50, 100, and 150 mV in the SiN _x nanopore correspond approximately to 200, 400, and 600 mV in the glass nanopore, respectively. This relationship is further evidenced by the axial electric field profiles along the nanopore axis (Figures g and h). For instance, at 400 mV, a glass nanopore exhibits a peak field strength of 4.1 × 10⁶ V·m^–1, which is comparable to 3.5 × 10⁶ V·m^–1 observed at 100 mV in the SiN _x nanopore. While the SiN _x nanopore displays a relatively uniform electric field distribution, the glass nanopore features a gradually decaying field along its conical section, forming a long tail extending roughly 500 nm into the pore interior (enlarged views of the electric field strength profiles are provided in Supplementary Figure S5).

Interestingly, despite exhibiting similar peak field strengths, the two nanopore platforms yielded distinct biomolecular translocation behaviors. As shown in Supplementary Figure S6, double-stranded DNA molecules (dsDNA, 7.2 kbp) displayed a median translocation time of 2.16 ms at 400 mV in the glass nanopore, nearly twice that measured at 100 mV in the SiN _x nanopore (1.1 ms). This difference may stem from the gradually decreasing electric field along the conical geometry of the glass nanopores. Instead, the translocation speed at 500 mV in glass nanopores closely matched that at 100 mV in SiN _x nanopores, yielding a median dwell time of 1.06 ms (Supplementary Figure S6). Consequently, we define “comparable experimental conditions” in this study based on matched median dwell times as our driving-force proxy, utilizing applied voltages of 500 mV for glass nanopores and 100 mV for SiN _x nanopores as a consistent basis for all subsequent benchmarking. By standardizing the temporal scale of analyte transport, this approach allows us to evaluate the spatial detection capabilities of each platform, specifically current blockade and SNR.

Comparison of Sensing DNA

Next, we compared DNA sensing performance between glass and SiN _x nanopore platforms by measuring three DNA species: double-stranded M13mp18 DNA (7.2 kbp), topological variants of λ-DNA (knots and plectonemes), and a designed DNA construct. The construct comprises six evenly spaced groups of dumbbell motifs attached to the scaffold, with eight dumbbells per site (Figure c), and was assembled by hybridizing complementary oligonucleotides to a single-stranded M13mp18 scaffold (see Supplementary Note 5 for details). Representative translocation signals for all three species in glass and SiN _x nanopores are shown in Figure a–c; both platforms resolve fine DNA nanostructures and topological features (see Supplementary Figures S7–S10 for additional examples). Unless noted otherwise, these measurements were performed at 500 mV (glass) and 100 mV (SiN _x ). Compared with glass pores, SiN _x pores produce larger current blockades across these cases. All-points current histograms for linear dsDNA translocations (excluding folded events) in 10 nm pores (Figure d) yield mean blockade currents of 0.50 nA for SiN _x and 0.12 nA for glass (∼4× difference). The full width at half-maximum (fwhm) of the open-pore baseline is greater for SiN _x than for glass (Figure d), again indicating higher noise in SiN _x nanopores. Moreover, in glass nanopores, the blockade fwhm is comparable to the baseline fwhm, whereas in SiN _x nanopores, the blockade distribution is substantially broader than the baseline. Figure e summarizes the mean current blockade, Δ̅I, for a set of pores (means from the histogram fits), again showing larger Δ̅IΔ̅I for the three SiN _x nanopores than for glass. Using SNR = Δ̅I/RMS noise, the resulting SNRs for the 10 nm pores (Figure f) are similar for both platforms (∼20), indicating comparable effective sensitivity for DNA detection at this pore size.

3.

Comparison of DNA signal-to-noise performance in glass and SiN _x nanopores. (a–c) Representative translocation event signals of three DNA species: (a) double-stranded M13mp18 DNA (7.2 kbp), (b) topological λ-DNA variants (knot and plectoneme), and (c) a designed DNA construct containing six evenly spaced groups of eight dumbbell motifs. (d) All-points current histograms of linear dsDNA translocations in 10 nm nanopores, showing baseline and blockade current levels for glass (top) and SiN _x (bottom). See Supplementary Figure S22 for histograms of other DNA types. (e) Summary of mean current blockade values (Δ̅I) for three independent nanopores of each platform at 10 nm diameter. (f) Corresponding SNRs showing comparable sensing performance between platforms at 10 nm. Error bars represent the standard deviation (SD) obtained from the Gaussian fitting of the current blockade distribution. (g, h) All-points current histograms for DNA translocations through 5 nm glass (g) and SiN _x (h) nanopores. (i) Comparison of mean current blockades for both platforms at 5 nm, derived from the histograms. (j) Corresponding SNRs for 5 nm nanopores, highlighting that glass nanopores exhibit a larger increase in DNA sensitivity upon pore size reduction. Error bars represent SDs from measurements on three independent nanopores. Note that in some cases, the error bars are too small to be visible. Measurements were performed at 500 mV (glass) and 100 mV (SiN _x ). The sampling rate is 1 MHz, and the cutoff frequency is 50 kHz.

Given the dsDNA double-helix diameter of ∼2.2 nm, we evaluated both platforms using smaller pores (5 nm, characterized by TEM scanning; Supplementary Figure S11). Figure g and h show all-points current histograms for linear DNA translocations in a glass and a SiN _x nanopore, respectively, from which we extracted the mean current blockade Δ̅I (summarized in Figure i). In the glass nanopore, Δ̅I increases from 0.12 nA (10 nm) to 0.42 nA (5 nm; 3.5-fold). In SiN _x , Δ̅I increases from 0.50 nA to 0.72 nA (1.44-fold). Notably, the 1.44-fold increase in Δ̅I for SiN _x nanopores is much smaller than that for glass nanopores (3.5×). By contrast, when reducing the nanopore diameter from 10 to 5 nm, the SNR of SiN _x increases from ∼20 to ∼36, whereas that of glass rises from ∼20 to ∼81 (Figure j). In addition, as the diameter shrinks from 10 to 5 nm, the electric field strength in glass nanopores increases markedly (from ∼4.1 × 10⁶ to ∼8.8 × 10⁶ V·m^–1 at 400 mV), whereas the change is modest in SiN _x nanopores (from ∼3.5 × 10⁶ to ∼4.1× 10⁶ V·m^–1 at 100 mV; Supplementary Figure S14), due to the high aspect ratio of glass nanopores. See Supplementary Figure S23 for the SNR of a medium nanopore size (∼7 nm) for both platforms.

It is important to mention that if the glass nanopore were biased to match the SiN _x translocation speed (i.e., at higher voltage), the SNR would be even greater because the RMS noise increases only weakly with voltage (Supplementary Figure S3). We also characterized the glass nanopore at 300 and 400 mV (Supplementary Figures S12–S13), yielding mean blockades of 0.27 and 0.35 nA and SNRs of ∼54 and ∼70, respectively, again exceeding that for the SiN _x nanopore at 100 mV.

Beyond SNR, we evaluated additional factors that determine sensing performance. As summarized in Figure a, two primary attributes define a nanopore: geometry and material. These, in turn, set secondary properties such as surface charge density, hydrophilicity, roughness, and mechanical stability, all of which strongly impact device usability and performance. Having assessed the main performance metrics, including baseline RMS noise, current blockade, dwell time, and SNR, we further quantified four additional criteria: folded-event fraction, directional dwell-time asymmetry (τ_backward/τ_forward), translocation frequency, and the fraction of “perturbed” events. The durability of glass and SiN _x ⁵⁶ nanopores has been reported previously and is not discussed in this work.

4.

Comparison of DNA translocation dynamics in glass and SiN _x nanopores. (a) Summary of key structural and material properties that define a nanopore and the metrics used to evaluate its sensing performance. (b,c) Representative scatter plots of maximum current blockade (ΔI_max) versus dwell time (τ_d) for dsDNA translocation events in glass and SiN _x nanopores, respectively. (d) Schematic illustrating folded translocation through glass and SiN _x nanopores in the forward direction. (e) Folded-event fractions for three independent glass and SiN _x nanopores measured in 4 M LiCl. Error bars represent the SDs from three measurements per nanopore. (f) Schematic illustrating forward and backward translocation directions in a glass nanopore. (g) Comparison of folded-event fractions in conical glass nanopores under forward (tip-to-interior) and backward (interior-to-tip) translocation. Error bars represent the SDs from three measurements per nanopore. (h,i) Schematic illustrating forward and backward translocation directions in a SiN _x nanopore. Example forward and backward translocation events in SiN _x nanopores. (j) Example translocation events showing slower backward translocation (bottom) compared with forward translocation (top) in glass nanopores. (k) Dependence of τ_backward/τ_forward on pore diameter in glass nanopores. (l) 10-s excerpts of raw ionic-current traces for dsDNA in glass (top) and SiN _x (bottom) nanopores. (m) Comparison of translocation frequencies between glass and SiN _x nanopores. The sample concentration is 1.75 nM. Error bars represent SDs from measurements on three independent nanopores. (n) Representative unperturbed events in SiN _x nanopores. (o) Representative perturbed events in SiN _x nanopores. (p) Perturbed-event fractions for dsDNA (Mol A), the DNA construct (Mol B), and λ-DNA (Mol C), respectively. No perturbed events were observed in glass nanopores (0%).

Figure b and c present typical scatter plots of maximum current blockade versus dwell time of dsDNA translocation in glass and SiN _x nanopores, respectively. In both cases, two clusters are evident, corresponding to unfolded (linear) and folded translocation. As illustrated in Figure d, folded events arise when an internal DNA segment is captured at the pore mouth; these events complicate signal interpretation and are generally undesirable. We first compared the folded-event fraction under the operating directions used for each platform: from the tip toward the interior for the conical glass nanopore, and from the thinned membrane toward the Si support (trapezoidal base) for the SiN _x nanopore. In 4 M LiCl, both platforms yielded folded fractions of ∼75% (Figure e), indicating that, under these conditions, the prevalence of folded versus linear translocation is largely platform-independent. We then examined directionality in the conical glass nanopore by reversing the applied bias. In this context, forward translocation refers to motion from the tip toward the interior, while backward translocation denotes the opposite direction (Figure f). The folded-event fraction was approximately 75% in the forward direction but decreased to about 45% in the backward direction (Figure g). This reduction observed in the backward (interior-to-tip) configuration likely arises from the lower probability of internal loop formation within the confinement.

Direction also affected translocation speed, an asymmetric phenomenon we previously reported and leveraged for molecular “ping-pong”. , In SiN _x nanopores, forward and backward τ_d were nearly identical (∼1.1 ms; Figure h–i). In glass nanopores, however, backward events were markedly slower (∼4 ms; Figure j and Supplementary Figures S15–S16) than forward events (∼1.08 ms), giving a dwell-time ratio τ_backward/τ_forward ∼ 3.7. Across glass nanopores of different diameters, τ_backward/τ_forward increased as diameter decreased (Figure k), indicating enhanced directional rectification in narrower tips and demonstrating that the conical geometry can substantially slow biomolecule translocation.

We also compared capture behavior. Figure l shows 10-s excerpts of raw current traces for dsDNA in glass (top) and SiN _x (bottom) nanopores. Translocation frequencies extracted from 30-min recordings (Figure m) were 0.79 nM^–1·s^–1 for SiN _x and 0.47 nM^–1·s^–1 for glass, consistent with a lower effective free-energy barrier for the shorter SiN _x nanopore.

Finally, in SiN _x nanopores, we observed two distinct classes of translocation signals. Representative unperturbed eventsclean, well-defined blockades characteristic of complete DNA passageare shown in Figure n. In contrast, we frequently detected perturbed events, which appear as brief, shallow, and noisy blockades preceding the full translocation signal (Figure o). We propose that these perturbed events likely stem from transient interactions of the DNA with the exterior pore mouth. The perturbed-event fractions were 33.6%, 36.3%, and 29.8% for dsDNA, the DNA construct, and λ-DNA, respectively (Figure p). No perturbed events were observed in glass nanopores (0%), consistent with stronger electrostatic repulsion from the more negatively charged SiO₂ surface (∼−35 mC·m^–2) compared with SiN _x (∼−20 mC·m^–2) at pH 9. See Supplementary Figure S17 for example perturbed events of dsDNA and λ-DNA. These perturbed events introduce a practical disadvantage for quantitative sensing, as identifying and filtering them reduces data yield and complicates analysis. We utilized a pH 9 buffer to reduce DNA sticking without denaturing dsDNA at higher pH. While we used 4 M LiCl to improve the nanopore’s spatial and temporal resolution, further actions, such as decreasing salt concentrations to modulate electrostatic screening, may mitigate these transient interactions.

Comparison of Sensing Proteins

In addition to DNA molecules, proteins represent another major class of biological targets for nanopore sensing. Unlike DNA, which behaves as a long, uniformly charged polymer, proteins exhibit diverse shapes, sizes, and charge distributions. To evaluate protein sensing performance, we investigated four proteins with different molecular weights: Streptavidin (∼52 kDa), Bovine Serum Albumin (BSA, ∼66.5 kDa), Immunoglobulin G (IgG, ∼150 kDa), and an Alpha-fetoprotein dimer (AFP–hFc, ∼210 kDa), using both glass and SiN _x nanopores with 10 nm diameters. Figure a,c show representative ionic-current traces and scatter plots of translocation time (τ_d) versus current blockade (ΔI) for each protein. The mean current blockade values (Δ̅I), obtained from Gaussian fitting, are summarized in Figure e. In line with the DNA measurements, the Δ̅I values obtained in SiN _x nanopores are consistently larger than those in glass nanopores, yielding mean blockades of 50, 56, 142, and 205 pA for glass, and 265, 252, 707, and 946 pA for SiN _x , corresponding to Streptavidin, BSA, IgG, and AFP, respectively. The Δ̅I values exhibit a strong linear correlation with protein molecular weight for both nanopore types (Figure f). The corresponding SNRs for glass nanopores are 7.9, 8.7, 22.3, and 32.2, while those for SiN _x nanopores are 9.9, 9.5, 26.5, and 35.4, indicating consistently higher SNRs in SiN _x nanopores. The average SNR ratio of SiN _x to glass nanopores is ∼1.16 at a 10 nm scale.

5.

Comparative protein sensing performance of glass and SiN _x nanopores. (a–d) Representative ionic-current traces and scatter plots of translocation time (τ_d) versus current blockade (ΔI) for streptavidin, BSA, IgG, and AFP–hFc dimer measured using 10 nm glass (blue) and SiN _x (magenta) nanopores. (e) Mean current blockade values (Δ̅I, top) and corresponding SNRs (bottom) for the four proteins on both nanopore platforms. The error bars represent SDs obtained from the Gaussian fitting of the current blockade distribution. (f) Linear dependence of Δ̅I on protein molecular weight for both glass and SiN _x nanopores. (g) Representative current traces of BSA translocation through 5 nm glass (top) and SiN _x (bottom) nanopores. (h) Comparison of Δ̅I values between glass and SiN _x for BSA measured with 5 nm nanopores. (i) Corresponding SNR values for a 5 nm glass and SiN _x nanopore.

To further assess performance at smaller pore sizes, we employed 5 nm nanopores for comparative BSA sensing, as BSA’s effective diameter (∼4 nm) closely matches this pore size. This specific size-matching maximizes the volume exclusion effect, providing an ideal comparison for how each platform handles compact biomolecules approaching the pore’s physical limit. Figure g displays representative ionic current traces for BSA translocation events through glass (top) and SiN _x (bottom) nanopores. The Δ̅I values increase significantly as the nanopore diameter decreases, from 0.056 nA to 0.12 nA for glass and from 0.25 nA to 3.0 nA for SiN _x (Figure h and Supplementary Figure S18). Correspondingly, the SNR values increase to ∼22 for glass and dramatically to ∼121 for SiN _x , with the latter exhibiting approximately 5-fold higher SNR than glass nanopores of the same diameter. See Supplementary Figure S23 for the SNR of a medium nanopore size (∼7 nm) for both platforms.

We summarize the sensing performance of glass and SiN _x nanopores in Figure . At a 10 nm diameter, glass and SiN _x nanopores exhibit comparable performance for both DNA and proteins (Figure a–b). When the diameter is reduced to 5 nm, closer to the size of typical target biomolecules, the trends diverge: glass nanopores provide stronger DNA sensing (Supplementary Figure S19, SNR > 80), whereas SiN _x nanopores deliver higher protein sensitivity (Figure b, SNR > 120). These observations suggest a practical detection guideline: all else being equal, longer nanopores tend to favor long polymers, whereas shorter nanopores favor small molecules.

6.

Comparison of sensing performance between glass and SiN _x nanopores for DNA and protein detection. (a) Spider plots summarizing key performance metrics for DNA sensing, including RMS noise, current blockade (ΔI), SNR, translocation frequency, and unfolded event ratio. These performance metrics were measured in 10 nm nanopores. (b) Comparison for protein detection. Both ΔI and SNRs of 10 nm and 5 nm glass and SiN _x nanopores are listed. The comparison of DNA sensing performance between the two platforms in different nanopore sizes is provided in Supplementary Figure S19.

The mechanisms, however, are not explained by pore length alone. For DNA, the field strength is expected to increase when a DNA strand occupies the nanopore. In our simulation, the electric field at a 5 nm glass nanopore is approximately 2.1× that of a 10 nm pore (Supplementary Figure S14), compared with 1.17× in SiN _x nanopores. The greater increment of the electric field in the glass nanopore when decreasing nanopore size suggests that it will yield higher current density and, thus, higher current blockade in the signal.

For proteins, the higher SNR in SiN _x nanopores is consistent with a volume-exclusion mechanism. Proteins typically reside fully inside the pore, while only a fraction of a DNA molecule is confined at any instant. Because SiN _x pores are short and have a much smaller effective pore volume than tapered glass nanopores (whose effective length is on the order of hundreds of nanometers; Supporting Information Figures S5 and S14), the fractional volume exclusion, and thus the current blockade, is larger in SiN _x , yielding superior protein sensitivity.

Geometry also affects other performance metrics. The conical shape of glass nanopores enables a backward translocation configuration that extends DNA dwell times by ∼3–7× and substantially reduces the fraction of folded events. However, these benefits come at the cost of lower capture rates compared with short SiN _x nanopores. Material properties further differentiate the platforms. Owing to the lower dielectric permittivity of SiO₂, glass nanopores exhibit substantially reduced capacitive noise, leading to ∼4–5× lower RMS current noise under identical conditions.

Apart from sensing performance, the usability of a nanosensor also matters. , We compare that of glass and SiN _x nanopore platforms in Supporting Information Table S6. Glass nanopores are simpler to fabricate, whereas SiN _x nanopores require specialized instrumentation and expertise. Both platforms exhibit high repeatability for nanopores of >5 nm. Glass nanopores offer faster fabrication, though creating those of <5 nm presents challenges. Both types are reusable with proper storage. In terms of access to samples in the trans reservoir, glass nanopores are more restricted due to their narrow geometry, which complicates buffer exchange, sample handling, and the analysis of translocated biomolecules. By contrast, SiN _x nanopore chips typically provide straightforward access to both the cis and trans reservoirs, allowing more flexible experimental configurations. While both are mechanically stable, the capillary tips of glass nanopores and the SiN _x membrane are prone to breakage during handling.

We summarize previously reported SNR values for glass and SiN _x nanopores from the literature in Supporting Information Table S4 for comparison. It is important to note that SNR is influenced by multiple experimental parameters, including the filter cutoff frequency (which affects RMS noise), nanopore size, applied voltage, and analyte dimensions, which determine the magnitude of current blockade. Consequently, a fully parameter-normalized comparison with the prior literature is not straightforward. From another angle, this difficulty motivates our head-to-head evaluation of glass and SiN _x nanopores under the same experimental conditions.

Although our results suggest a general guideline that glass nanopores are better suited for DNA sensing, whereas SiN _x nanopores perform better for proteins, this criterion is not universal. For instance, if a protein is unfolded into an extended polymer chain, glass nanopores may offer higher sensitivity due to their geometry and field focusing. Conversely, for short oligonucleotides, the smaller sensing volume of SiN _x nanopores can provide superior detection performance.

Conclusions

In sum, this study provides a quantitative, mechanistic basis for choosing between glass and SiN _x nanopores, delivering a decision framework that links architecture to the analyte class, measurement priorities, and platform-level limitations. We anticipate that these insights will guide both sensor selection in current applications and the design of next-generation nanopore instruments.

Methods

Nanopore Fabrication

Silicon nitride nanopore chips were fabricated by using MEMS technologies. A SiN _x layer was deposited via chemical vapor deposition, followed by photolithographic patterning, localized membrane thinning, and silicon etching to release the thin, free-standing membranes. SiN _x nanopores were fabricated by TEM drilling of the membrane. Glass nanopores were fabricated by pulling quartz glass capillaries by using a laser puller to achieve conical geometries with nanoscale tip diameters. The pulled glass nanopores were subsequently integrated into custom microfluidic chips. Both nanopore platforms underwent plasma cleaning and hydrophilic surface treatments prior to use to ensure wettability and stable baseline current. Pore size characterization was performed by using TEM.

Nanopore Measurement

Single-molecule translocation experiments and electrical measurements were performed using a high-bandwidth patch-clamp amplifier. Ag/AgCl electrodes were immersed in the cis and trans reservoirs to apply a voltage across the nanopore and measure the resulting ionic current. Current–voltage characteristics were recorded to characterize pore size. Ionic current signals were sampled at 1 MHz and low-pass filtered at 50 kHz. Data acquisition and extraction of translocation dynamics, including current blockade, dwell time, and signal-to-noise ratio, were performed using custom-written Python scripts.

Finite Element Analysis

Finite-element analysis was performed using COMSOL Multiphysics. A 2D axisymmetric model was constructed to simulate the geometries of the SiN _x nanopores and the conical glass nanopore. By coupling the Poisson–Nernst–Planck (PNP) equations for ion transport with the Navier–Stokes equations for fluid dynamics, the spatial distribution of the electric field across the pores was simulated.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsnano.6c03089.

• MEMS fabrication process for SiN _x nanopore chip wafers; fabrication of glass nanopores and SiN _x nanopores; nanopore measurements; preparation of DNA samples; materials used in experiments; finite element method for analyzing the two nanopores; and supplementary figures (PDF)

F.Z. and Z.W. contributed equally to this work. F.Z. and U.F.K. conceived the concept of comparing glass and SiN _x nanopores. F.Z. performed DNA and protein measurements on both nanopore platforms. A.B. carried out TEM characterization of the nanopores. Z.W., R. H., and Q.Z. fabricated the nanopores using TEM drilling, and J.S. conducted MEMS fabrication of the SiN _x nanopore wafers. F.Z., K.C., and U.F.K. analyzed the data and cowrote the manuscript.

The authors declare no competing financial interest.