Removable Nanopore Sealing and In Situ Iron Oxide Nanoparticle Synthesis on Thin Silicon Nitride Membranes

Abstract

We report in situ synthesis of iron oxide particles inside silicon nitride nanopores via a chemical reaction, monitored by current readout. Nanopores were formed by electroporation on glass chips (diameters from 1.7 to 11.3 nm), transmission electron microscopy (TEM) drilling (diameters from 6.5 to 64.6 nm), or hydrofluoric acid etching (diameters from 12.6 to 36.2 nm) in 5–20 nm thick membranes. Nanopores seal on time scales from ∼1 ms to ∼3.6 s, across a range of sizes and concentrations. We show single and ∼5-pore arrays, as fabricated, after sealing, and after cleaning and pore recovery. These results are independent of the fabrication method. Energy-dispersive X-ray spectroscopy, aberration-corrected scanning TEM, and powder X-ray diffraction verify the synthesis of mixed magnetite and maghemite iron oxide. This work advances nanoparticle–nanopore chips for applications in biosensing, plasmonics, and photonics, where position and size control is required.

Introduction

Solid-state nanopores are increasingly explored for biomolecule detection, water desalination, , energy harvesting, and matter separation, among other uses. An interesting use of solid-state pores is as nanoscale reactors for particle synthesis. The so-called “template synthesis” of nanostructures involves using densely patterned nanoporous membranes to control the synthesis products and has been thoroughly demonstrated in track-etched polymer − and anodized aluminum oxide ,,− membranes (see Table S1). Venta et. al demonstrated electrically driven gold synthesis inside of single transmission electron microscopy (TEM) drilled silicon nitride (SiN _x ) pores. In contrast to nanoporous membranes typically used for template synthesis, single nanopores in SiN _x and 2D materials are fabricated in a deterministic fashion. Such devices and their few-pore counterparts are ideal for biosensing and single-molecule detection as well as in plasmonics − and photonics, , where pore number, size, and location must be controlled. For example, a single nanoparticle next to a nanopore can serve as an integrated optical antenna for plasmonics, and a single nanopore can be integrated within a field-effect transistor for high-bandwidth sensing. ,,

Here, we demonstrate positional control of particle synthesis as well as particle removal in single SiN _x pores and small arrays (4–5 pores) on 5 by 5 mm chips. Iron oxide particles inside nanopores are synthesized via a chemical reaction among ferric, ferrous, and hydroxide ions in electrolyte solution, monitored by current readout. We reproduce the main features of the nanopore “sealing” and pore “recovery” in 18 SiN _x membranes on glass and Si chips via in situ iron oxide synthesis (Figure ). From ionic current versus time traces, we characterize pore sealing times and dynamics. We also measure the resulting particle diameter for a range of precursor (Fe²⁺, Fe³⁺, and OH^–) concentrations, in low (50 mM KCl) and high (1 M KCl) electrolyte concentrations. Pore diameters are measured via TEM and/or estimated from the correlated ionic conductance. First, we discuss SiN _x pore fabrication using electroporation, TEM drilling, and hydrofluoric (HF) acid etching. We then show how iron oxide synthesis is monitored in real time via ionic current suppression across the pore. We use TEM imaging, energy dispersive X-ray spectroscopy (EDS), aberration-corrected scanning transmission electron microscopy (AC-STEM) imaging, and powder X-ray diffraction (XRD) to verify the synthesis of iron oxide particles within pores and characterize their structure. Finally, we recover open pores by rinsing with diluted sulfuric acid (H₂SO₄) or piranha cleaning to dissolve the iron oxide.

1.

Schematics of pore fabrication, pore sealing, pore recovery, and characterization methods. (a) Pores and arrays are formed by TEM drilling, electroporation, or hydrofluoric acid etching. (b) Iron oxide particle formation and pore sealing are initiated by adding NaOH to the trans side and FeCl₃ and FeCl₂ (molar ratio of 2:1) to the cis side. Trans and cis refer to the chambers in which the electrodes are placed. This process is monitored in real time by measuring the changes in ionic current through the pore. The iron oxide particles grow within the nanopore until partially or fully sealing it. Sealing occurs with zero and nonzero applied voltages. (c) Particles can be dissolved with sulfuric acid or piranha, recovering the nanopore again for ionic measurements. (d) The pores and iron oxide particles are characterized via direct TEM imaging and EDS. Parameters varied include nanopore size, electrolyte concentration, and precursor concentration (C ₀, 0.1 C ₀, and 0.01 C ₀).

The impact of this work spans several applications. First, it demonstrates a technique for controllable nanopore sealing and subsequent restoration of pore function. Nanopores can be sealed for storage and shipment and then later opened and cleaned. This is important for single nanopore sensors wherein the particles serve as “caps” that can be removed prior to measurement. A second benefit is in applications where the precise placement of a few to a single nanoparticles on a chip is beneficial in contrast to nanoporous membranes with high densities of pores. ,, Iron oxide nanoparticles are candidate materials to enhance electronic transport in 2D materials and to fabricate magnetic superlattices. Finally, electrical monitoring of nanoparticle growth can be exploited for the partial and sequential growth of composite particles beyond iron oxide and for the functionalization of nanopores with specific materials at their walls.

To demonstrate the versatility of this approach, we employed TEM drilling, HF acid etching, and electroporation to fabricate SiN _x pores (Figure a). TEM drilling is generally considered the gold standard for nanopore fabrication. , Electroporation is rapid and inexpensive. Here, we developed a simple passive-shunt resistor electroporation for low-noise glass chips featuring significantly smaller resistance–capacitance constants (Figures S1 and S2). Single pores were made in ∼20 nm thick membranes. Small nanopore arrays were fabricated in ultrathin (∼5 nm thick) localized circular regions patterned onto SiN _x membranes. In this case, SiN _x membranes were prepatterned by electron beam lithography and thinned by reactive ion etching (Oxford Instruments PlasmaLab 80+ RIE) to form locally thinned trenches (diameter ∼ 100 nm). Nanopores were opened in these trenches by HF etching. In all cases, pore formation was confirmed electrically (linear current–voltage (I–V), conductance-derived diameter estimate, and stable open-pore current), and iron oxide particle formation was not contingent upon the fabrication protocol.

Results and Discussion

Figure b shows SiN _x nanopore sealing by iron oxide. Sealing was verified by I–V measurements, TEM, or both. We demonstrate iron oxide synthesis inside 12 TEM-drilled nanopores (T1–T3, P2, P4, P6, P8 , P10–P13, and P 15), four electroporated nanopores (E1–E4), and two nanopore arrays (arrays A1 and A2). (Table S1 summarizes all nanopores included in this study. Figures S3–S10 show data obtained from both arrays and selected single pores.) Prior to particle formation, we measured the I–V curve of each device using a 50 mM or 1 M KCl solution and estimated the effective pore diameter from ionic conductance (d _{pore,before sealing}) using eq (S1) (Table S2). We show I–V curves of two representative devices (pore E2 and array A1) in Figure a,c and their corresponding real-time current traces during pore sealing in Figures b,d, respectively.

2.

Synthesis of iron oxide inside silicon nitride pores. (a–d) Electrical monitoring of particle growth. I–V curves of two representative devices (pore E2 and array A1) and their corresponding real-time current traces during iron oxide nanoparticle formation and pore sealing. (a) Current–voltage (I–V) traces for pore E2 (fabricated by electroporation) before particle formation (red squares), after particle formation and pore sealing (green circles), and after recovery and particle removal in sulfuric acid (blue squares). (b) Real-time current vs time trace and corresponding voltage during iron oxide nanoparticle formation. Steps include (1) 50 mM KCl was added to both sides, (2) 0.32 M NaOH was added to the trans side, (3) 0.08 M FeCl₃ and 0.04 M FeCl₂ (molar ratio of 2:1) were added to the cis side, and (4) voltage was then reversed to drive the reagents across the pore, and iron oxide nanoparticles form and seal the nanopore gradually. (c) Current–voltage (I–V) traces for array A 1 (nanopore array fabricated by HF etching) before (red squares) and after particle formation and pore sealing (green circles). (d) Real-time current vs time trace and corresponding voltage during iron oxide particle formation. Steps include (1) 50 mM KCl was added to both sides, (2) 0.32 M NaOH was added to the trans side, (3) 0.08 M FeCl₃ and 0.04 M FeCl₂ (molar ratio of 2:1) were added to the cis side, and (4) a voltage of +10 mV was then applied to check pore sealing. (e–j) TEM images of the pore P 10. The pore was sealed with no applied voltage using a precursor concentration of 0.8 mM FeCl₃, 0.4 mM FeCl₂, and 3.2 mM NaOH (0.01 C ₀) in 1 M KCl. (e–g) Bright-field TEM images of pore P10 before sealing (e, d _{pore,before sealing} ∼ 55 nm), after sealing (f, d _{pore,after sealing} ∼ 1.9 nm), and after recovery (g). (h) Annular dark-field (ADF) STEM images of the sealed nanopore. (i, j) Normalized EDS maps (Fe and O) after pore sealing. (k) Scatter plot of the TEM-measured particle diameter d _particle vs pore diameter d _pore from both single nanopores and individual pores in arrays performed at zero applied voltage. Pore diameters were estimated from either ionic measurements or TEM images. The diagonal line d _particle = d _pore is plotted, and points above (below) the line indicate particles with diameters larger (smaller) than the pore.

To initiate the synthesis reaction, NaOH was added to the trans side of the nanopore; FeCl₃ and FeCl₂ (molar ratio of 2:1) were added to the cis side. All reagents were dissolved in unbuffered 50 mM or 1 M KCl. Iron oxide (Fe₃O₄) particles likely formed via the reaction ,−

$${\mathrm{Fe}}^{2+}+2{\mathrm{Fe}}^{3+}+8{\mathrm{OH}}^{-}\to {\mathrm{Fe}}_3{\mathrm{O}}_4(\mathrm{s})+4{\mathrm{H}}_2\mathrm{O}$$ 1

When the pore was sealed, the reagents could no longer mix, and the ionic current fell to almost zero. We subsequently replaced the reagent solutions with KCl to confirm that there was little to no ion flow across the sealed pore (Figures a,c, green circles). Using this current trace, we also estimated d _{pore,after sealing} after pore sealing (Table S2) using eq (S1). (For nanopore arrays, this estimated diameter represents the effective pore diameter if the array was replaced with a single pore.) In a few devices, the ionic current fell substantially but did not reach 0. Any residual current was used to estimate an effective diameter d _{pore, after sealing}, corresponding to the pore area still accessible for ion transport. In all cases, d _{pore, after sealing} ≪ d _{pore, before sealing}.

Pore E2 in Figure b was fabricated by electroporation with a d _{pore, before sealing} of 5.6 nm. In the first ∼15 s, both the trans and cis sides of a flow cell were filled with KCl solutions, and I ≫ −7 nA at V = −300 mV (shown as 1. KCl in Figure b). At t = 40 s, while keeping V = −300 mV, KCl on the trans side was replaced with NaOH (in KCl), and I ≫ −5.5 nA (2. OH ^‑ added in Figure b). There was a nonzero current shift due to the chemical gradient created by the reagents. At t = 60 s, KCl on the cis side was replaced with FeCl₂ and FeCl₃ (in KCl) and the current dropped to −15 nA (3. Fe ²⁺ , Fe ³⁺ added in Figure b). The voltage was reversed to V = +500 mV at t = 70 s. This was accompanied by a sharp upward current spike due to the capacitive response of the system. After that, the ionic current continuously fell until reaching a plateau (4. Part icle formed and pore sealed in Figure b).

We also observed that iron oxide synthesis could proceed via reactant diffusional mixing without applied voltages. This is demonstrated in the nanopore array A1 (Figure d). NaOH, FeCl₂, and FeCl₃ were added to the flow cell. The current gradually approached zero, even with an applied bias (+10 mV). This suggested that the reagents approached the pore primarily by diffusional mixing, rather than by electroosmotic and/or electrophoretic forces. Therefore, we observe that iron oxide synthesis inside SiN _x pores does not necessarily require applied voltages.

To directly confirm particle formation, we include representative TEM images and EDS mapping of the pore P1 0. (A summary of EDS acquisition parameters is included in Table S2.) Figure e–g shows bright-field TEM images of the pore P10 before sealing, after sealing, and after recovery. Figure h is an ADF TEM image of the sealed pore corresponding to the Fe and O elemental maps in Figures i,j, respectively.

Figure k displays the TEM-measured particle diameter, d _particle, versus the pore diameter, d _pore, for single pores and for individual pores within two 3 by 3 arrays. Figure k also includes data points using single nanopore diameters estimated from the ionic conductance (open symbols in Figure k), and they lie close to the points using TEM-measured diameters. This good agreement from two correlated measurements validates the diameter values.

For single-pore chips, most data points lie on or slightly above the line d _particle = d _pore, indicating that d _pore ≤ d _particle. For the pores within the arrays (arrays A1 and A2), the particle diameters are typically ∼ 10 times larger than pore diameters. Data suggest that 50 mM KCl (used in array experiments) yields larger particles than 1 M KCl (used for single-pore chips), and this warrants a systematic ionic-strength investigation in the future.

We recovered the pores by dissolving the particles in 0.38 M H₂SO₄ for 10 min. (pore E2) or piranha cleaning for 10 min. (array A1) and rinsed the chip with water three times (Figure c).

The iron oxide nanoparticles can be removed via the reaction

$$\begin{array}{c}{\mathrm{Fe}}_3{\mathrm{O}}_4(\mathrm{s})+4{\mathrm{H}}_2{\mathrm{SO}}_4\to {\mathrm{Fe}}^{2+}+2{\mathrm{Fe}}^{3+}+4{\mathrm{SO}}_4^{2-}+4{\mathrm{H}}_2\mathrm{O}\end{array}$$ 2

We remeasured the I–V curve (Figure a, blue triangles) to estimate the d _{pore,recovered} after particle dissolution (Table S2). From our experiments, we see that the pore diameter before and after nanopore recovery remains constant or increases. Specifically, the nanopore diameter for samples E1, E2, E3, and E4 changed from 1.7 to 3.5 nm, 5.6 to 9.3 nm, 7.6 to 9.6 nm, and 11.3 to 11.4 nm, respectively. We observe an inverse relationship between the initial diameter and the relative change in diameter after pore recovery. That is to say, as the initial pore diameter increases, the absolute change in diameter decreases in pores E1, E2, E3, and E4 from ∼100 to ∼ 0%. These pores were exposed to KCl solution and repetitive voltage sweeps up to ± 700 mV. The observed pore expansion is highly consistent with previous measurements from Chou et al., who demonstrated that SiN _x nanopores expand due to etching in 1 M KCl (at a rate of about 1 nm per day) and that the application of a potential bias can accelerate this process. To eliminate pore expansion, future experiments can include diluting reagents used to remove the particles (e.g., piranha solution or sulfuric acid), diluting the electrolyte solution used to measure conductance and estimate pore diameter after particle removal, and/or coating the pores. Furthermore, we observed rectification in the recovered E2 (Figure a, blue triangles) evidenced by the deviation from a linear, symmetric I–V curve expected for an Ohmic pore. This is consistent with the residual iron oxide inside the pore after cleaning (Figure g), which modifies the pore surface charge and introduces asymmetry in the pore structure, thereby causing current rectification at a low KCl concentration (50 mM KCl for blue triangles).

We analyzed the sealed nanopores with TEM and EDS to verify the formation of particles and their composition. This was best demonstrated in array A1 with locally thinned regions for easier pore identification. The array consists of a 3 by 3 array of nine ∼100 nm-large, thinned (recessed) regions (∼ 5 nm thickness) in the SiN _x membrane, labeled from 1 to 9 (Figure (1a)). Nanopores were fabricated by HF etching in thinned regions labeled 2, 4, 6, 8, and 9 (Figure (1c)–(1g)). Regions labeled 1, 3, 5, and 7 remained intact as controls (Figure (1b)). Due to the irregular pore shapes, pore areas were obtained using ImageJ: ∼250, 236, 1030, 690, and 629 nm² for regions 2, 4, 6, 8, and 9, respectively (Figure (1c)–(1g) and Table S1). We also include a 3D numerical simulation demonstrating that a large separation ensures individual pores behave independently with negligible interaction , (Figure S11).

3.

Synthesis of iron oxide in the nanopore array A1 and subsequent recovery. (1) Pristine nanopore array before sealing. (1a) TEM image of a 3 × 3 array of ∼ 100 nm-large, locally thinned regions (∼5 nm thick), patterned by electron beam lithography and thinned by RIE, numbered 1–9. Locally thinned regions are circled. Colored circles correspond to locally thinned regions that contain pores. White circles correspond to locally thinned regions without pores. Note that the same colors (blue, yellow, pink, green, and purple) are used throughout the figure to identify the same locally thinned regions over time (2, 4, 6, 8, and 9). (1b–g) Higher magnification TEM images of the locally thinned regions 1, 2, 4, 6, 8, and 9, respectively. Nanopores inside of these thinned regions are outlined with colored dashed lines. (2) Nanopore array after sealing. (2a) TEM image of the 3 × 3 array after sealing with iron oxide particles. Iron oxide only fills the locally thinned regions with pores (regions 2, 4, 6, 8, and 9). (2b–g) Higher magnification TEM images of locally thinned regions 1, 2, 4, 6, 8, and 9 after sealing with iron oxide. These images confirm complete occlusion of the pores shown in 1b-g. (3) Nanopore array after recovery. (3a) TEM image of the 3 × 3 array after piranha cleaning to dissolve the iron oxide. (3b–g) Higher magnification TEM images of the locally thinned regions 1, 2, 4, 6, 8, and 9 after piranha cleaning. (4) ADF STEM images showing the sealed nanopore array and the elemental composition before and after recovery. (4a) ADF STEM image showing the sealed nanopore array. Regions 2, 4, 6, 8, and 9 are circled in different colors (blue, yellow, pink, green, and purple). (4b,c) Normalized EDS maps corresponding to the ADF image are shown in 4a. These maps confirm the presence of iron and oxygen. (4d) ADF STEM image showing the recovered nanopore array after piranha cleaning. (4e,f) EDS spectrum image maps corresponding to the ADF image shown in 4d. These maps confirm the removal of iron oxide particles. (5) Plot and linear least-squares fit of initial pore area versus resulting particle area. Pore areas were measured in ImageJ from images 1b–g. Particle areas were measured in ImageJ (2b–g). The error associated with these measurements was approximated to ±5% of the area. This plot shows that the particle size increases with pore size. The coefficient of determination (R ² = 0.75) indicates a sufficient fit between the experimental data and the linear regression model, y = 1.28 × 10^–4 x + 3.2 × 10^–2.

TEM images of the sealed arrays and thinned regions after pore sealing are shown in Figure (2a). TEM images confirmed iron oxide formation as the thinned regions were sealed by particles and particle aggregates, corroborating I–V measurements. The particles were formed only in the thinned regions with nanopores (regions 2, 4, 6, 8, and 9; Figure (2c)–(2g)), while the others (1, 3, 5, and 7) were unchanged from prior TEM imaging since they did not contain nanopores (Figure (2b)).

The growth of iron oxide can also extend past the locally thinned region (Figure (2c)–(2g)) if the reaction is long enough. We compared the pore area with the corresponding iron oxide particle area in Figure (5). The larger the nanopore is, the larger the particle that occupies it and its surrounding membrane. We repeated the experiment with 100 times diluted precursors than in Figure (3.2 mM versus 0.32 M for NaOH, 0.8 mM versus 0.08 M for FeCl₃, and 0.4 mM versus 0.04 M for FeCl₂) using nanopore array A2 (Figures S9 and S10) and found that the area of the iron oxide particles was smaller even for pores of greater sizes than in array A1. For example, in array A1, the pore area in region 2 was ∼250 nm² and the corresponding iron oxide particle area was ∼0.08 μm². In array A2, a larger pore (∼378 nm²) in region 1 resulted in a smaller particle of ∼0.016 μm² (Figure S9 and Table S2). Further experiments can include tuning the precursors, varying the pore sizes and spacings, and controlling the current to restrict particle formation. Partial nanopore filling can be of interest to create pores with desired properties, as well as modulate membrane porosity. In situ liquid cell TEM may be of interest to observe reaction dynamics within the pore in future studies. −

We used EDS to confirm the elemental composition of the synthesized particles. In Figure (4), we show ADF images of five nanopores (numbered 2, 4, 6, 8, and 9 in Figure (1a)) in A1 that were covered by particles (Figure (2a)). The corresponding EDS maps of iron (Figure (4b)) and oxygen (Figure (4c)) confirm the synthesis of iron oxide. Strong Fe and O signals exclusively occur across the pore and thinned regions, confirming the localized particle formation.

We further imaged array A1 to obtain EDS maps after the removal of the particles by piranha solution and the recovery of nanopores (Figure (3a)). Little to no particle residue in Figure (2) was observed in the recovered array and corresponding single nanopores in Figure (3). The ADF image of the same array in Figure (4d) and the weak EDS signals from Fe in Figure (4e) confirmed the removal of iron oxide. The widespread weak signal from oxygen across the array and the membrane (Figure (4f)) can be from a native silicon oxide, also observed in recovered nanopores in Figure (3c)–(3g).

To examine the crystal structure of the iron oxide, AC-STEM imaging was performed on a ∼20 × 20 nm² area in the center of particle 4 in array A1 (Figure a,b). Possible phases include crystalline magnetite (Fe₃O₄), maghemite (γ-Fe₂O₃), or a combination of both phases based on our synthesis involving the coprecipitation of Fe²⁺ and Fe³⁺ in the presence of OH^–. ,, AC-STEM images revealed the polycrystalline nature of the particles (Figure c–e). The inset fast Fourier transforms emphasize the nonuniform crystal orientations. A characteristic d-spacing was obtained by plotting an intensity profile from one of the atomic resolution images (Figure e,f) to determine the iron oxide phase. This intensity profile was Fourier transformed to produce the plot in Figure g. The first peak at ∼3.4 nm^–1 corresponds to a spacing of ∼ 2.9 Å between atomic planes. Magnetite and maghemite both exhibit d-spacing near ∼ 2.9 Å located at 2.967 and 2.95 Å, as determined by powder XRD, respectively. , Because magnetite can be oxidized to maghemite, a mixture of the two is likely present. ,,, Another means of differentiating magnetite from maghemite is by visual inspection. Magnetite appears black to the eye, whereas maghemite is reddish-brown. In our case, the precipitates appear blackish-brown, suggesting that both phases are present (Figure S12). Powder XRD data corroborated this hypothesis, confirming the presence of magnetite and maghemite (Figure h). Additional investigation into the phases of iron oxide formed could utilize Mössbauer spectroscopy, Raman spectroscopy, ,, and/or electron energy loss spectroscopy to characterize these nanoparticles.

4.

AC-STEM and XRD characterizations of iron oxide crystal structure in array A1. (a) ADF image of nanopore array A1 after sealing the pores with iron oxide. The red dashed box corresponds to the zoomed-in image shown to the right. (b) Higher magnification ADF image of the iron oxide particle in region 4 of array A 1. The blue box in the center of the image corresponds to the region shown in c. (c) Atomic resolution image of the center of the iron oxide formed in region 4, showing multiple crystal grain orientations around the black region in the center of the image with an inset FFT. (Note that the black region in the center simply corresponds to a thinner region of the iron oxide particle.) (d) Another region of the iron oxide, near the region shown in (c), with an inset FFT. This atomic resolution image emphasizes the differing crystal grain orientations present. (e) A third region of the iron oxide, near the regions shown in (c) and (d), with an inset FFT. This region is primarily covered by a single crystal grain orientation. An intensity profile was taken from the region boxed in white to produce (f). The intensity plot was Fourier transformed to produce (g). The first peak occurs at ∼3.4 nm^–1, corresponding to a d-spacing of ∼2.9 nm. (h) Powder XRD spectrum of iron oxide precipitates. All diffraction peaks correspond to the phases of magnetite (hkl planes labeled in blue) and maghemite (hkl planes labeled in green). The peaks were identified using the International Centre for Diffraction Data PDF card Nos. 01-082-3507 and 01-089-5894, respectively. The inset image shows the sample prior to measurement.

To illustrate the effect of precursor concentration, Figure compares results from two 3 by 3 arrays (A1 and A2) of thinned SiN _x regions (light-gray areas in Figure A) containing 5 and 4 pores, respectively. Nanopore regions are encircled. The precursor concentration for array A2 was 100 times smaller than for array A1, resulting in ∼2.5 times smaller particles on average, for similar-sized pores. Estimating from TEM, the average particle size in A1 was ∼ 370 nm versus ∼140 nm for A2 (Table S2).

5.

Iron nanoparticle growth: kinetics, concentration, and size dependence. (A) (a, b) Qualitative comparison of the effect of precursor concentration on particle formation in nanopore arrays A1 and A2. TEM micrographs of pristine, locally thinned arrays patterned into SiN _x with electron beam lithography and RIE. The colored circles indicate locally thinned regions with pores. (c, d) TEM micrographs after sealing arrays A1 and A2 with iron oxide. To seal array A1, 0.04 M FeCl₂, 0.08 M FeCl₃, and 0.32 M NaOH (C ₀) were used in 50 mM KCl. To seal array A2, 0.4 mM FeCl₂, 0.8 mM FeCl₃, and 3.2 mM NaOH (0.01 C ₀) were used in 50 mM KCl. The iron oxide particles formed in array A1 extend beyond the nanopore and locally thinned regions onto the bulk membrane. The iron oxide particles formed in array A2 are much smaller. (B) Representative current vs. time trace from Pore E2 shows that the current decays toward the baseline and is well fit by a single exponential $I(t)\sim \exp (-\raisebox{1ex}{$\mathrm{t}$}\!\left/ \!\raisebox{-1ex}{$\tau $}\right.)$ over 70–350 s from Figure b. τ is the characteristic time scale of sealing the pore. A larger τ means slower sealing (more gradual decay), while a smaller τ suggests faster sealing. (C) Scatter plot of τ vs. effective pore diameter estimated from ionic measurement d _pore for nanopore arrays and single-pore chips. The arrays exhibit longer sealing times than single pores at comparable “effective diameters” (a diameter of an equivalent single pore with the same conductance). (D) Scatter plot of τ vs. particle diameter estimated from TEM d _particle for nanopore arrays (averaged over individual pores in the array) and single-pore chips. The arrays exhibit sealing times ∼1–2 orders of magnitude larger than single pores at comparable d _particle.

We also investigated particle size and pore sealing time for single-pore and array chips. The growth is monitored as the current decays gradually to almost zero and is well-fitted by a single exponential $I(t)\sim \exp (-\raisebox{1ex}{$t$}\!\left/ \!\raisebox{-1ex}{$\tau $}\right.)$ (Figure B), where τ is the characteristic pore sealing time. Larger τ means slower sealing (more gradual current decay), while smaller τ suggests faster sealing. The τ values for single nanopores and arrays are summarized in Figures C,D as a function of the initial pore diameter and the final particle diameter; τ ranges from ∼0.001 to ∼3.6 s (Table S2). Overall, τ increases with the TEM-measured d _particle for both single-pore and pore-array chips, meaning that larger particles grow over a longer time. Together with Figure k, these data provide insights into characteristic sealing times as a function of nanopore and particle diameters from 8.5 to 64.5 nm and 7.1 to 441.1 nm, respectively. Collectively, these data reveal sample-to-sample differences as we observe aggregation/clustering, variable particle uniformity, and cases where the particle diameter exceeds the pore diameter. A range of parameters can be adjusted in the pore-sealing process to further control size and properties of particles. Future work can attempt to systematically vary the pore aspect ratios and electrolyte concentration. We presented a rapid current-monitored method for silicon nitride nanopore sealing allowing for easy pore recovery for biosensing experiments.

Conclusion

In conclusion, we demonstrated removable sealing of single nanopores and few-nanopore arrays in SiN _x using iron oxide particles. This work has relevance for both nanopore and nanoparticle-based technologies. Nanoparticle formation and pore sealing were monitored by ionic current suppression from the characteristic open-pore current to almost zero. Synthesis was independent of the nanopore fabrication methods used, and particles could be dissolved to recover the original pores. We characterized iron oxide particles, revealing the presence of magnetite and maghemite. Future work could further probe particle formation and properties as a function of pore diameter, thickness, and shape by sequential TEM imaging at different stages of pore sealing and by in situ liquid cell TEM imaging. Given our observations of gradual pore sealing, H₂SO₄/piranha treatment reversibility, and asymmetrical I–V curve rectification, the particle formation mechanism could be compatible with multiple mechanistic pathways. Iron oxide can nucleate in the pore interior, adsorb onto pore walls, and continue growing or nucleate in the pore interior and grow freely until the pore wall begins to constrict growth. , Future work can explore these details by using in situ imaging/EDS and surface coatings to localize precursors and develop models of particle growth across various pHs, ionic strengths, and concentrations. Our work advances particle synthesis to the single-pore regime and in pore arrays fabricated by top-down approaches, thereby opening the door to a range of advanced on-chip applications in biosensing, plasmonics, and photonics.

Experimental Methods

Materials

Unbuffered ionic solutions were produced by dissolving NaOH, FeCl₂, FeCl₃, and KCl (Alfa Aesar, MA) in Milli-Q (Millipore, Billerica, MA) water with a resistivity of 18.2 MΩ*cm to desired concentrations.

Ionic Current Measurements

Ionic measurements at bandwidths of 20 kHz were performed using the portable Nanopore Reader 200 kHz (Elements SRL, Italy). All measurements were conducted in unbuffered 50 mM or 1 M KCl. Data were analyzed using OriginLab (OriginLab, MA).

TEM, Energy-Dispersive X-ray Spectroscopy, and Aberration-Corrected Scanning Transmission Electron Microscopy Imaging

TEM characterization was performed in a JEOL F200 S/TEM, a JEOL NEOARM STEM, and a JEOL 2010F TEM, all operated at 200 kV. The JEOL F200 S/TEM is equipped with a Gatan OneView camera and JEOL dual SDD detectors for EDS data acquisition. The JEOL F200 was operated in TEM and STEM modes. The JEOL 2010F is equipped with a Gatan bright-field detector and was operated in the TEM mode. The JEOL NEOARM STEM is equipped with a Gatan bright-field and dark-field detector and was operated in the STEM mode (∼ 120 pA probe current). All TEM data were processed using Gatan Digital Micrograph and ImageJ.

Powder X-ray Diffraction Experiments

XRD measurements were performed using a Rigaku SmartlabSE diffractometer with Cu-Ka radiation (l = 1.541874 Å) equipped with a D/teX Ultra 250 detector. A 1D scan was collected with a step width of 0.01° and scan speed of 2° per minute at angles 2θ from 10 to 90°. To prepare our sample for powder XRD, we made solutions of 0.32 M NaOH, 0.08 M FeCl₃, and 0.04 M FeCl₂ in DI water. We mixed FeCl₂ and FeCl₃ solutions together until thoroughly dissolved and then mixed in the NaOH solution. We let the DI water evaporate out of the solution overnight and then scraped the precipitate into a powder for XRD.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsnanoscienceau.6c00032.

• Additional experimental details, materials, and methods (PDF)

#.

Z.X., P.B., and C.M. contributed equally to this work. B.D., D.N., and Z.X. developed electroporation. Z.X. and B.D. performed electroporation, ionic testing, and DNA translocation measurements. Z.X. and C.M. performed nanopore sealing. P.B. and C.M. performed TEM, AC-STEM, and EDS. Z.X., P.B., C.M., A.B-D., and M.D. performed ionic and image analyses. D.N. and P.B. performed chip fabrication. C.Y.L. performed COMSOL simulations. I.G. performed powder XRD. All authors discussed the manuscript and contributed to the writing.

The authors declare no competing financial interest.