Precise control of the size and noise of solid-state nanopores using high electric fields

The process for preparing a nanopore for a molecular recognition experiment involves the automated cyclic application of high electric field pulses produced by stepping the potential up to 6–10 V. As the conductance G of nanopores subjected to a potential bias greater than 4 V is no longer Ohmic and would yield unrealistic nanopore diameters (supplementary information figure S1 available at stacks.iop.org/Nano/23/405301/mmedia), a low voltage bias (400 mV) is applied after each high field pulse. This allows us to accurately measure the increase in nanopore conductance during growth. As nanopore conductance has been shown to directly depend on the nanopore size, a measure of the conductance can be used to infer nanopore diameter [16–19]. Here, we assume cylindrical pore geometry with an effective length of l_eff = 15 nm (half of the membrane thickness) using the relationship discussed by Kowalczyk et al (equation (1)) [16] describing the conductance of a nanopore of diameter d in a KCl solution of conductivity σ = 11.3 S m⁻¹. We note that this choice of l_eff is in good agreement with experimental data of DNA translocation, as discussed in section 2.3 of the text.

$$\begin{eqnarray} &&G=\sigma \left [\frac{4{l}_{\mathrm{eff}}}{\pi {d}^{2}}+\frac{1}{d}\right ]^{-1}. \end{eqnarray} \tag{ 1 }$$

Figure 1(b) shows that by applying 8 V pulses of 2 s duration, the ionic current through the nanopore increases in steps of approximately 1.3 nA after each pulse, corresponding to a increase in conductance of ΔG = 3.3 nS. The resultant increase in nanopore diameter per cycle under these conditions is ∼0.4 nm. While variability in membrane thickness, surface chemistry and initial nanopore geometry yields different growth rates from pore to pore, the magnitude and duration of each high field pulse can be adjusted to precisely control the amount of growth per cycle. Due to the relatively high capacitance of the nanopore device, measurement periods at 400 mV were extended to 7 s to allow for the capacitive current to dissipate and the faradaic current to stabilize after each voltage pulse. While an external power supply and current amplifier are used to provide the necessary high voltage and measure the resultant ionic current during enlarging, we can easily revert back to a low-noise experimental setup (see methods section 4.2). In the sub-1 V voltage range, enlarged pores act as ideal Ohmic resistors with linear I–V curves, as shown in figure 1(c). This trend is consistent over multiple enlarged nanopores (N = 20) and we find that the conductance of nanopores treated with high field pulses is stable up to several days after treatment (supplementary information figure S2 available at stacks.iop.org/Nano/23/405301/mmedia). We note that conductance values are constant even after unmounting a nanopore and flushing the system with fresh electrolyte solution, indicating that the observed increase in conductance is not a transient effect caused by local heating of the electrolyte in the nanopore.

In order to further understand the enlargement process, the dependence of nanopore growth on the electric field strength was investigated. From figure 2(a), varying the pulse strength from 6 to 10 V (0.20–0.33 V  nm⁻¹) produces quite different growth profiles. In each case, nanopores were immersed in 1 M KCl solution and the pulse duration was kept constant at 2 s. When subjected to a 6 V potential bias, nanopore growth is relatively slow with conductance increasing by ∼25 nS over 1000 s of exposure time (0.03 nS s⁻¹). In comparison, pores subjected to 8 V and 10 V increase in conductance at an average rate of ∼0.10 nS s⁻¹ and 0.63 nS s⁻¹, respectively. We note, however, that several different growth regimes are observed depending on the stage of enlargement and the applied bias. While nanopore growth at 6 V is approximately linear throughout most of the time window shown, the growth rate increases over time at 8 V and more drastically at 10 V. Additionally, nanopore growth appears to initially slow for the first 200 s of exposure to 6 V and 8 V, as highlighted in figure 2(b). For the 10 V case, this initial growth is characterized by a sharp increase in conductance from 10 to 14 nS after ∼85 s. From noise characteristics of the nanopore before and after enlarging, this jump in conductance is attributed to complete wetting of the nanopore, as discussed later in this paper.

Zoom In Zoom Out Reset image size

When subjected to pulses larger than 6 V, nanopore growth rate increases throughout the enlarging process; that is, larger pores grow faster than smaller ones. This is illustrated in figure 2(c), where the growth rate of the nanopores subjected to 8 and 10 V increases as a function of the conductance measured before each high electric field pulse. For the 10 V case, growth rate is plotted from data acquired after the initial 100 s of enlarging, following complete wetting of the nanopore. We find that in the conductance window shown, nanopore growth rate is approximately parabolic, ranging between 0.01–0.40 nS s⁻¹ and 0.01–11 nS s⁻¹ for 8 V and 10 V applied biases, respectively. Interestingly, this increase in conductance corresponds to a growth rate of less than 0.1 nm per second, assuming a cylindrical nanopore with an effective length of 15 nm. This method is thus very powerful for obtaining precise control over final nanopore size. It should also be noted that for nanopores of conductance greater than 150 nS, growth is less predictable for biases of 8–10 V, although pulse strength can be tuned to offer better control.

While it is clear from the different growth rates of the nanopores shown in figure 2 that the strength of the electric field plays a significant role in the enlarging of nanopores, a complete understanding of the growth process is not yet clear. In fact, several mechanisms are likely involved in the growth process. The initial slowdown in the increase of the growth rate (figure 2(b)), which seems common to many nanopores we investigated, could be related to the initial modification of the nanopore profile or to the removal of weakly bound residual carbonaceous contamination that may be left over from the TEM drilling and imaging process [20]. But as the nanopore is enlarged, the increasing growth rate for large conductance values in figure 2(c) seems to indicate a relationship between growth rate and the amount of current passing through the nanopore, as the applied bias and pulse duration are kept constant throughout the experiment. However, pore to pore variability in growth rate under identical experimental conditions (supplemental information figure S3 available at stacks.iop.org/Nano/23/405301/mmedia) as well as with varying ionic strength of electrolyte solution (supplemental information figure S4 available at stacks.iop.org/Nano/23/405301/mmedia) makes an elucidation of a complete mechanism difficult. Ascertaining the exact roles of power dissipation, electro-osmotic flow, resistive heating and electrochemistry occurring at the nanopore surface is beyond the scope of this article and is the subject of ongoing investigation. From a practical point of view, however, cyclic application of high field pulses provides a means of controlling nanopore size with exceptional precision, as the nanopore diameter can be increased in sub-nanometer steps in an automated fashion simply by varying the pulse strength and level of ionic current in situ under experimental conditions.

In addition to providing control over pore size, application of short high electric field pulses significantly reduces low frequency noise in nanopore current measurements. This so-called 1/f noise is thought to arise from fluctuations in the number of charge carriers occupying the nanopore in accordance with Hooge's phenomenological relation [12, 21, 25]. In solid-state nanopores studies, excessive 1/f noise is typically observed when nanopores fail to completely wet [22], or when, over the course of an experiment, translocating molecules adsorb to the nanopore wall [23, 24], often resulting in a partial clog. As these sources of noise can significantly complicate signal analysis, nanopores initially exhibiting high 1/f noise are typically discarded when standard procedures such as piranha cleaning and oxygen plasma treatment fail to mitigate these effects and produce clean signals [25]. While functionalization of the nanopore walls [26–28] is a potential option for preventing non-specific adsorption, the addition of organic molecules may not be desirable for some applications and involves additional preparation steps. Thus, due to the difficulty and expense associated with nanopore fabrication and preparation, a simple method for systematically and reproducibly reducing 1/f noise and rejuvenating clogged nanopores is of great practical importance for improving the yield of those usable for experiment.

By applying short high field pulses, we find that such pores are readily and rapidly cleaned in situ, without risking accidental damage to the membrane or requiring lengthy procedures associated with manipulating the nanopore chip. By maintaining a short pulse time ( < 500 ms), we are able to reduce low frequency noise while avoiding enlarging the nanopore. Figure 3(a) shows the power spectral density (PSD) plots of two nanopores for an applied bias of 200 mV both before and after the application 8–10 V pulses of 200 ms duration. As the dielectric properties of the SiN_x membranes and the amplifier headstage are the predominant sources of noise above 10 kHz, these portions of the PSD remain relatively unchanged from pore to pore. Below ∼1 kHz, however, low frequency noise varies as 1/f^α with 0.94 < α < 1.5 and can be significantly lowered upon treatment with high electric fields. For instance, the orange curve in figure 3(a) is a typical PSD of a pore that had become irreversibly clogged during a DNA translocation experiment and two separate washes in piranha solution at 75 °C for 30 min were unable to regenerate a low-noise current trace. The red curve, however, shows the PSD of an 11 nm (G = 47.0 nS) clogged pore that has been cleaned using 200 ms pulses of 10 V. In this example, 2 pulses were enough to remove the source of noise and leave a clean nanopore surface with a very stable conductance, illustrated by a three orders of magnitude decrease in low frequency noise in the below 500 Hz.

Zoom In Zoom Out Reset image size

To illustrate the effectiveness of this method for reducing 1/f noise in incompletely wetted nanopores, an unused ∼7 nm pore (as determined after TEM drilling) that failed to completely wet despite cleaning with piranha solution (blue curve in figure 3(a)) was treated using 8 V pulses of 200 ms duration. While a total of 7 pulses were applied, nanopore conductance did not significantly change after the first one, indicating that a single pulse was sufficient to completely wet the pore. The PSD of the nanopore upon treatment (green curve) exhibits nearly two orders of magnitude reduction in low frequency noise, which varies as 1/f^α with α = 0.94, consistent with previously characterized low-noise SiN_x nanopores [12]. Moreover, the nanopore exhibited a conductance of 19.6 nS, appropriate for a 15 nm long cylindrical nanopore of 7 nm diameter. Interestingly, we find that once a nanopore has completely wetted, low-noise characteristics are maintained through subsequent enlarging. figure 3(b) shows current traces of a 10 nm pore (as determined after TEM drilling) before and after complete wetting and enlargement to 27 nm (G = 126 nS) using 8 V pulses of 2 s duration as described above. Clearly, analysis of DNA translocation before treatment with high electric fields would be extremely challenging if not impossible, while the enlarged, treated pore exhibits substantial improvement in stability.

The complete wetting of a nanopore using a strong electric field has been previously observed and is understood as a lowering of the potential energy barrier for complete wetting of the pore via the alignment of the dipole moment of water vapor molecules [29, 30]. However, the nanopores investigated in those studies were hydrophobically modified for the application of nanopore gating and were either fabricated with conical geometry in polymer films or were too large to be suitable for single molecule detection. Moreover, the electric field strengths reached in these studies were over an order of magnitude less than those discussed here, a thorough investigation of the noise characteristics of treated nanopores was not performed and there was no indication of nanopore enlargement. While high electric fields have also been previously used to facilitate wetting of SiN_x nanopores capable of detecting and characterizing DNA molecules [31], to the best of our knowledge our study is the first to present data characterizing the process and experimental evidence revealing that low frequency noise can be effectively reduced. Furthermore, we find that treating nanopores with high electric field pulses produces low-noise current signals with a yield of over 90% (supplementary information figure S4 available at stacks.iop.org/Nano/23/405301/mmedia). Currently, it is common to reject a large portion of nanopores due to high noise, current rectification and incomplete wetting. This method shows great promise as a means for reproducibly preparing functional nanopores for molecular recognition experiments.

We further characterized these high electric field treated pores by performing DNA translocation experiments to evaluate their performances as single-molecule sensors. Pores enlarged to diameters ranging from 10 to 35 nm (40–490 nS in 2 M KCl) were used to detect λ DNA (48.5 kbp, New England BioLabs). Upon enlarging each pore, both electrolyte reservoirs were flushed with 2 M KCl (σ = 21.5 S m⁻¹) buffered at pH 8.0 (higher salt concentration was used to increase signal to noise of translocating molecules) and double-stranded (ds) λ DNA was introduced into the cis chamber at a working concentration of 1 μg ml⁻¹. Figure 4(a) shows that at a bias of 150 mV, conductance spikes become clearly visible upon the addition of DNA as molecules are electrophoretically driven through the 11 nm and 32 nm pores (87 nS and 432 nS in 2 M KCl, respectively). As expected for nanopores of these diameters, multiple quantized blockage depths are observed corresponding to the translocation of folded and unfolded DNA molecules [32]. In figure 4(b), the equidistant separation of peaks in histograms of the nanopore conductance during translocation events confirms the presence of multiple discrete blockage states during translocation. We note that the reduction of 1/f noise in the open pore current signal, a result of the enlarging process, produces histograms with easily resolvable peaks. Furthermore, while nanopores would eventually clog throughout the course of a translocation experiment, low-noise open pore conductance values were often regained with short (100–200 ms) pulses of 6–10 V. Nanopore rejuvenation for further experimentation through treatment with high electric fields is thus of great benefit for improving the efficiency and lifespan of nanopores.

Zoom In Zoom Out Reset image size

Interestingly, the change in conductance of ΔG = 1.7 nS produced by dsDNA translocating through the 32 nm pore is smaller than that of ΔG = 2.6 nS for the 11 nm pore. This is consistent with models described elsewhere [16] describing the increased role of access resistance in determining the change in nanopore conductance upon the translocation of dsDNA through large pores. This result confirms that the original nanopore has in fact been enlarged and that the rest of the membrane retained its integrity and did not conduct current, as otherwise DNA translocations would have exhibited a similar drop in conductance as the 11 nm pore. By fitting the ΔG values for enlarged pores with the model proposed by Kowalczyk et al [16] given by equation (2), we find that the effective pore length of treated nanopores can range from 11 to 16 nm, showing that the assumed pore length of 15 nm provides a good fit for dsDNA translocation through enlarged nanopores.

$$\begin{eqnarray} &&\Delta G=G(d)-G({d}_{\text{with~DNA}}). \end{eqnarray} \tag{ 2 }$$

Here, G is given by equation (1) and d_with DNA, given by (3) is the effective diameter of the pore containing dsDNA of diameter d_DNA = 2.2 nm.

$$\begin{eqnarray} &&{d}_{\text{with~DNA}}=\sqrt{({d}^{2}-{d}_{\mathrm{DNA }}^{2})}. \end{eqnarray} \tag{ 3 }$$

Nanopores used were fabricated in 50 μm × 50 μm, 30 nm thick free standing silicon nitride membranes supported on a 200 μm thick silicon frame purchased from Norcada (NT005X). In each membrane, we drilled a sub-10 nm pore using a tightly focused beam of electrons in a JEOL-2100 TEM operated at 200 kV accelerating voltage. Nanopores were drilled at 600k magnification using the condenser aperture of spot size=2 and α = 3.

Prior to mounting the nanopore chips in solution, nanopore chips were prepared for experiment by cleaning them for 30 min in piranha solution (3:1 H₂SO₄:H₂O₂) at 75 °C. The chips are then thoroughly rinsed in filtered, degassed deionized water and immediately mounted in the liquid cell containing filtered, degassed 1 M KCl 10 mM HEPES buffered at pH 8.0 at room temperature or the salt concentration indicated in the text. We used a polytetrafluoroethylene cell to mount the nanopore chip between two liquid reservoirs, as described in [12]. Custom-made silicone gaskets sandwich the chip to ensure a gigaohm seal. The cell is placed in a Faraday enclosure to reduce electrical noise. Ag/AgCl electrodes immersed on both sides of the pore were used to measure ionic current when connected to a current amplifier. A patch-clamp amplifier (Axopatch 200B, Molecular Devices) was used to record ionic current for power spectrum density calculations, I–V characterization and DNA translocations. Data acquisition was performed using custom-designed LabVIEW software controlling a National Instruments PCIe-6351 DAQ card. Data were low-pass filtered at 100 kHz using the Axopatch 200B 4-pole Bessel filter and sampled at 250 kHz. For applying voltage pulses up to 10 V between each electrode, the PCIe-6351 DAQ card was used, while a Keithley 428 current amplifier was used to measure the resultant current at a sampling frequency of 1 kHz. Alternating voltage pulses and conductance measurements were controlled using custom-written LabVIEW software.