Early-stage detection prevents disease progression and complications in treatment procedures, especially for infectious diseases. This requires rapid and accurate sensing technologies and techniques that remove the need for expensive and time-consuming sample preparation and transfer to the labs and the running of multiple experiments. To that end, point-of-care (POC) testing has been introduced for quick disease diagnostics that enables caregivers to start early treatment, leading to improved health outcomes. Here, we introduce a tunneling current bio-sensing technology based on a metal–insulator–electrolyte junction, which is highly sensitive to charge variations at the insulator–electrolyte interface. The charge variations occur as a response of hybridization of complementary DNA sequences to DNA primers immobilized on the insulator surface. This leads to the label-free detection of as little as tens of DNA molecules or, equivalently, samples with 0.01 fM concentrations. Since the sensing is based on a single terminal measurement of current with respect to a reference electrode, our technology can enable the creation of novel compact medical and portable POC devices for real-time disease detection.
INTRODUCTION
Detection and real-time amplification of nucleic acids have profoundly transformed biomedical research and molecular diagnostics, which have a wide range of applications in areas such as quantification of gene expression,1–5 clinical microbiology,6,7 cancer diagnostics,8,9 phytopathology,10,11 and genotyping.12–14
The majority of the existing technologies for detection and rapid amplification of DNA molecules require optical sensing in addition to fluorescently labeled sequence-specific probes or fluorescent dyes for labeling DNA molecules,15–19 which makes the devices bulky and expensive, preventing them to be widely accessible by physicians and patients.
Over the past decade, highly sensitive (attomolar) nanowire sensors have evolved for detecting molecular (DNA or protein) hybridization.20–23 However, these sensors require complicated fabrication techniques and the connection of two electrodes to the nanowire to measure the conductance change across the wire. Moreover, there are several obstacles limiting their development into widely used portable point-of-care (POC) diagnostic devices, such as high costs and nonuniformities in fabrication.20 In this paper, we introduce a novel compact sensing technology to overcome the above challenges—a tunneling-based charge sensor of hybridized molecules using a single terminal measurement of the tunneling current through a Metal–Insulator–Electrolyte (MIE) junction. We show that we can use our tunneling sensor to detect DNA molecules in solutions with concentrations as small as one hundredth of a femtomolar concentration. Moreover, this technology has the potential to detect attachment of the other charged biomolecules, such as proteins, on the surface.
The organization of this paper is as follows. First, we introduce fabrication process, device preparation, and surface modification steps. Then, we discuss the working principles of the device. Additionally, device architecture for DNA sensing is presented. Then, equivalent circuit (EC) parameters of the sensor are extracted from an Electrochemical Impedance Spectroscopy (EIS) instrument and analyzed. Finally, data from multiple devices tested with a precise source meter are presented and studied. We end with some brief concluding remarks.
MATERIALS AND METHODS
Device fabrication process
In this research, tunneling current through a very thin layer of aluminum oxide is studied. A commercial 4-in. silicon wafer is used as the substrate. To prevent electric shorting, a layer of 1000 Å oxide is grown on the silicon wafer in a dry oxidation tube. Then, photolithography is performed using a Shipley 1827 positive photoresist. Next, aluminum electrodes are deposited using an e-beam evaporator tool. The width of each electrode is 500 µm. Finally, a 40 Å aluminum oxide layer is deposited by using the Atomic Layer Deposition (ALD) process. The size of the device is 9 × 11 mm2. The E-beam evaporation process is performed at UC Irvine Materials Research Institute (IMRI), and all other steps are done at the Integrated Nanosystems Research Facility (INRF) at UC Irvine.
After the fabrication process, multiple devices were tested and a curve fitting technique was used in I–V plots to evaluate the tunneling layer. Thickness of the insulator film, effective mass of an electron in Al2O3 ALD, and barrier height of Al/Al2O3 were utilized as the parameters in the curve fitting. Several identical devices were studied, and the curve fitting results were almost identical. By applying 1 V across the insulator, the electric field is calculated as 2.5 MV/cm. The strength of the field indicates that direct tunneling dominates at the junction;24 thus, the corresponding equation is used for the fitting. The effective mass of an electron is calculated as 0.1993me, which is consistent with the literature, ∼0.2me.24–27 This number also indicates that aluminum oxide is more susceptible to tunneling than silicon dioxide or silicon nitride with the effective mass of ∼0.5 me.24 The thickness of the Al2O3 ALD was computed as 4.1 nm. Since the native oxide grows on aluminum films at room temperature, the oxide layer was expected to be slightly thicker than 4 nm. Moreover, the thickness of the aluminum oxide layer was measured by using an ellipsometer [UVISEL Spectroscopic Phase Modulated Ellipsometer (SPME), Horiba] at UC Riverside to confirm the fitting results. Finally, the barrier height was calculated as 2.89 eV. The error function for the fitting is ∼1 × 10−16.
Device preparation and surface modification
The process for covalently attaching DNA primers to the 4 nm Al2O3 atomic layer deposited on the electrode surfaces is described in what follows. First, the device is washed clean with a solvent, followed by a de-ionized water (DI) rinse. Then, the oxide surface is activated by either 60 min UV–ozone or 1 min O2 plasma treatment. Immediately after surface activation, the device is incubated in a 5% v/v APTES in 100% ethanol for at least 3 h at room temperature, followed by backing at 120 °C for 15 min on a hotplate. For APTES, 99% (3-Aminopropyl) triethoxysilane from Sigma-Aldrich is used in this research. Next, the device is incubated in a 10 mM biotin linker in 1× PBS for 45 min, followed by incubation in 100 μg/ml streptavidin in 1× PBS for 75 min at room temperature. In between each step, the device is washed in 1× PBS for 2 min, three times. EZ-Link NHS-PEG4-Biotin and streptavidin from ThermoFisher Scientific are used in this work. Finally, the device is incubated in 5 µM biotinylated DNA primer solution in 1× PBS overnight at 4 °C or 1 h at room temperature. To verify surface modification, a sample device is chosen for the attachment of dye labeled DNA primers and is checked with a fluorescence setup when the process is completed.
RESULTS AND DISCUSSION
Working principles of device
The device architecture is a Metal–Insulator–Electrolyte (MIE) design with a metallic counter electrode (CE). DNA primers are immobilized on the insulator surface for the detection of target DNA molecules. The potential distribution across a MIE device with a positive potential applied to the working electrode is modeled as a linear drop across the insulator, a linear drop across the Stern layer of adsorbed ions on the surface of the insulator in the electrolyte, and then a diffuse layer potential drop embodied by the Gouy–Chapman–Stern model. When DNA is attached to the insulator surface, the potential distribution of the MIE junction changes due to the negative charge of DNA’s phosphate backbone. Landheer et al. modeled this layer as an ion-permeable membrane and used the Poisson–Boltzmann equation to calculate the potential drop behavior.28 The potential across the MIE device with an attached DNA “membrane” is shown in Figs. 1(a) and 1(b) with positive and negative biases, respectively. Immobilized charges on the insulator surface and ion concentration of the electrolyte are major factors affecting potential distribution throughout the double layer area.28 Figure 1 shows the origin at the left-hand side of the membrane, which lies between x = 0 and x = w. ψ0 is the potential at the insulator–electrolyte interface, ψm is the potential at the membrane, and ψe represents the potential at the extremum, xe, if an extremum happens. The potential at the bulk of the electrolyte is considered zero. The potential distribution ψ(x) and the parameters mentioned in Fig. 1 can be calculated by using the Poisson–Boltzmann equation for the electrolyte and membrane. In the case of positive bias, when negatively charged DNA is attached to the surface of the insulator, the local potential drop across the insulator increases. The more the DNA is added, the greater the potential drop and, thus, the electric field. There might be an extremum in the potential distribution if there is a high concentration of negative charge attached to the surface and when the electrolyte ionic concentration is low. There will be an extremum in Fig. 1(b) if enough positive charges are immobilized on the insulator surface.
Device architecture for DNA analysis
Figure 2(a) shows a three-electrode device used in this research. Each of the electrodes is used as working electrodes independently. The counter electrode is an aluminum wire with the diameter of 250 µm inserted inside a fluidic channel placed on top of the working electrodes while running the experiments.
PDMS fluidic channels were mounted on top of the DNA-functionalized aluminum electrodes using UV curable glue. As shown in Fig. 2(c), each device has either three or six independent active areas, three per channel. The width of the channels is 1.75 mm. Each channel contains 12 µl solution. Aluminum electrodes fabricated on the chip are the working electrodes, while the aluminum wires inserted in the solution and stabilized by UV-glue are counter electrodes. Figure 2(d) illustrates the 3D printed holder used for measurement. A Keithley source meter, model 6517A, in the voltage mode was used to measure the tunneling current. To validate the results, several devices were measured using an EIS instrument as well. More details are presented in the section titled Equivalent circuit parameters.
Equivalent circuit parameters
Figure 3 shows the EIS data from a device before and after DNA hybridization. Experiments were performed using a precise biologic instrument in the frequency range of 1 Hz to 1 MHz at VAC = 30 mV and VDC = 1 V. The two-electrode setup was used in all the experiments because the counter electrode potential does not drift significantly over the course of the experiment. This is generally true in systems that exhibit very low currents.30 The device was functionalized with single-stranded DNA forward primers and was first measured in the mastermix solution and then after adding 78-base DNA target solution and performing the Polymerase Chain Reaction (PCR), in order to completely saturate the surface with an extended forward primer. Key DNA sequences used in this study are introduced in the supplementary material. Mastermix, also known as supermix or ready mix, is a premixed solution that has all the components for a polymerase chain reaction that are not sample-specific. Solid markers represent the measured data, while the dashed lines represent the fitted curves using an equivalent circuit model.
The data are presented in a Nyquist plot in Fig. 3(a) where the real component of the impedance (Z′) is the abscissa and the imaginary part (Z″) is the ordinate. Both devices have the same general features in the Nyquist plots, first a high frequency semicircle with low real impedance (see the supplementary material, Fig. 1) associated with the counter electrode (CE), followed by a broad semicircle feature at lower frequencies with large real impedance arising from contributions from both the oxide layer and the double layer. This broad feature can be deconvoluted into two distinct semicircles for each process. When DNA is attached to the surface of the oxide layer, the broad semicircle is depressed. Furthermore, with attachment of DNA, both the real and imaginary impedance decrease. The Bode plots are shown in Fig. 3(b), where the AC frequency is the abscissa and log |Z′| is one ordinate and phase angle is the other. When DNA is attached to the surface of the oxide layer, the frequency of the maximum phase angle is shifted to higher frequencies.
We next performed equivalent circuit (EC) modeling of the EIS data. The EC used in this study is shown in the inset of Fig. 3(a) and contains a series resistor (Rs) to account for bulk resistance of the electrolyte and contact resistance, a parallel resistor (RCE) and a constant phase element (CPECE) for the impedance at the counter electrode, a parallel resistor (Rox) and a constant phase element (CPEox) to account for the resistance and the capacitance of the oxide layer, a parallel resistor (RDL) and a constant phase element (CPEDL) to account for the charge transfer resistance and the double layer capacitance, and a serial Warburg element W to account for diffusional impedance. The biologic EC-Lab software was used for EC modeling; the factor was smaller than 0.02 in all the fittings. The results of the EC fitting are shown in the supplementary material, Fig. 2.
. | Before DNA . | After PCR DNA . |
---|---|---|
. | hybridization . | hybridization . |
Cox (nF) | 25.39 | 24.52 |
Rox (MΩ) | 37.54 | 28.21 |
CDL (nF) | 125.6 | 119 |
RDL (MΩ) | 1.977 | 2.18 |
Rs (Ω) | 405.5 | 581.5 |
. | Before DNA . | After PCR DNA . |
---|---|---|
. | hybridization . | hybridization . |
Cox (nF) | 25.39 | 24.52 |
Rox (MΩ) | 37.54 | 28.21 |
CDL (nF) | 125.6 | 119 |
RDL (MΩ) | 1.977 | 2.18 |
Rs (Ω) | 405.5 | 581.5 |
The difference between calculation and measurement may result from a slightly larger PDMS fluidic channel (active area), a thinner oxide layer—or a combination thereof.
The EC model results show a decrease in Rox from 37.54 to 28.21 MΩ after DNA hybridization. When a DC positive potential is applied to the insulator side of the junction, the electric field increases within the insulator. After accumulation of negatively charged DNA on the oxide surface, the electric field will increase. Thus, the resistance to transferring charge through the oxide layer should decrease compared to the device without DNA, and it did. This decrease in Rox with DNA attachment to the oxide surface enables the sensing application of the device. The trend is consistent with the data measured by using the Keithley source meter.
Double layer capacitance (CDL) slightly decreased after DNA hybridization on the surface. Moreover, charge transfer resistance (RDL) increased by almost the same factor as the double layer capacitance decreased. These results are similar to those observed in other studies of direct detection of DNA hybridization.31–34 No significant change was observed in the time constant of the measured current by using the source meter. The dominant time constants before and after DNA hybridization are 0.298 and 0.31 s, respectively, based on the data presented in Table I. It is an approximation, calculated from the product of charge transfer resistor and effective capacitance of the oxide and double layer capacitance. There is a second time constant that is the product of electrolyte resistor and a series combination of the oxide and double layer capacitance, but it is too small to be detected.
Rs accounts for bulk resistance of the electrolyte and contact resistance. Since the inserted aluminum wire in the solution, counter electrode, was renewed in between tests, a slight change was predictable in contact resistance and, therefore, in the value of Rs.
Since the active area is 0.875 mm2, Stern capacitance per unit area is 190 nF/mm2 or 19 µF/cm2. This value falls within the range of values typically measured for aqueous electrolytes of 5–20 µF/cm2.35
Exponential dependency of the oxide tunneling current to the potential across it enabled us to detect very small alterations in the potential caused by variations in the charge accumulated on the oxide surface at the insulator–electrolyte interface. It is worth mentioning that most of the tested devices showed a change of a factor of two in resistance after hybridization. However, in the devices measured by EIS, the change was a factor of 1.33 due to experimental limitations since the measurements had to be done in two laboratories.
On-chip real-time charge sensitive detection
The real-time charge detector can operate under three main conditions: (1) by measuring the peak instantaneous current on application of a bias voltage, (2) by monitoring the steady state current after a few seconds of application of the bias voltage, or (3) by measuring the change in AC impedance of the device at the bias voltage resulting from DNA binding. In this study, we used the data measured after 3 s of application of the bias voltage.
After validation of the results with EIS, we focused on studying the change in the tunneling current measured by using the Keithley source meter. Figures 4(a) and 4(b) illustrate tunneling current vs voltage and time for different target solutions, respectively. The data were taken from a device, while the solutions were pipetted into the PDMS fluidic chamber from the lowest to the highest concentration and washed with the 1× PBS buffer between each test. As shown in the graphs, the change is detectable for solutions with concentration as small as 0.01 fM containing ∼72 target molecules. A 16% change is seen for this solution compared to a mastermix solution. The device saturates with tens of femtomolar concentrations, containing millions of DNA molecules, and shows the maximum change of a factor of ∼2.5 for this device. Positive voltage was connected to the insulator side of the junction, and 1 V is used for all the current vs time measurements. This procedure is reversible, meaning that by removing the attached DNA target molecules from immobilized primers at the insulator surface and washing the chamber with buffer, the current vs voltage graph will be back to the original value with no target molecules. Moreover, additional experiments were done with non-complementary DNA target molecules and no significant change was observed in the measured current, as expected.
In addition, a comparison between time constants calculated based on EIS and Keithley measurements is presented. Figure 5 shows current vs time for the same chip tested for EIS measurement. MATLAB fitting application was used, and a time constant of 0.297 s with the largest R-squared of 0.96 was extracted. The time constant calculated based on the values of the EIS data in Table I is 0.298 s before DNA hybridization.
In this study, multiple devices were tested with complementary and non-complementary target DNA strands, which are presented in the supplementary material. With the non-complementary DNA molecules, there was almost no change measured in the current tunneling through the junction. In this work, with the use of complementary target DNA molecules, a change was detectable in solutions with 0.01 fM or ∼72 target molecules. The limit of detection of the device is the change in the measured current that translates into an amount of charge added to the surface, which is dependent, but not limited to the length of the target DNA molecule. By utilizing solutions containing more target molecules, the change increases until the surface saturates, at which point the current is ultimately 2.5 times larger than the one measured with an original mastermix solution and zero added charges; see Fig. 4(c). Thus, it is expected that by improving the saturation capacity of the surface, the sensitivity of the device increases. The tests were repeated with multiple devices and different DNA primers and target molecules, mentioned in the supplementary material, and the results were consistent. Moreover, additional tests were done under the negative bias and the current showed a decrease as expected and is explained in Fig. 1(b). Attomolar and ultrasensitive DNA biosensors have been emerging in recent years and perform mainly based on graphene field-effect transistors.36–39 Additionally, several novel detection methods, such as single-stranded DNA-coated carbon nanotube sensors, single walled nanotubes, and traveling-wave dielectrophoretic force on microparticles, have shown promising results for DNA and biomolecule detection.40–42 However, this study has shown comparable ultra-high sensitivity while utilizing simple fabrication steps. However, our study has shown comparable sensitivity while utilizing simpler fabrication steps. Furthermore, coupled with solid-phase PCR,43 this device can lead to the development of a quick and precise portable POC diagnostic device in the future.
CONCLUSION
In this work, we developed a precise charge sensor capable of detecting complementary DNA molecules in solutions with one hundredth of a femtomolar concentration. The base of the sensor was a metal–insulator–electrolyte–metal junction. Since the oxide layer was extremely thin, it performed in the tunneling regime. Due to the log-linear relationship between current and voltage across the tunneling oxide, the device is highly sensitive to any electric field variations within the insulator. The oxide surface was functionalized with single-stranded DNA primers; hence, the hybridization of the primers with complementary target DNA molecules in the solution generated charge variations on the surface. As a result, the electric field strength within the oxide either decreased or increased depending on the sense of the voltage applied to the counter electrode. This technology can detect any charge attachment on the surface; it can thus be applied to detect other biomolecules, such as proteins employing antibody–protein pairs.
SUPPLEMENTARY MATERIAL
See the supplementary material for DNA sequences used in this study and raw data captured using the biologic EC-Lab software and fitting curves.
ACKNOWLEDGMENTS
All the fabrication was performed at the UCI Integrated Nanosystems Research Facility (INRF), except for e-beam evaporation, which was done at UC Irvine Materials Research Institute (IMRI). This work was partially supported by W. M. Keck Foundation Laboratory for live cell genomics (Grant No. KF-52356).
The authors declare no competing financial interest.
DATA AVAILABILITY
The data that support the findings of this study are available from the corresponding author upon reasonable request.