Merve
Balcı Leinen
,
Sebastian
Lindenthal
,
Daniel
Heimfarth
and
Jana
Zaumseil
*
Institute for Physical Chemistry, Universität Heidelberg, D-69120 Heidelberg, Germany. E-mail: zaumseil@uni-heidelberg.de
First published on 6th September 2022
Networks of semiconducting single-walled carbon nanotubes (SWNTs) can be used as the transducing layer for sensors based on water-gated transistors. To add specific sensing capabilities, SWNTs are often functionalized with additional moieties or selective membranes are applied, thus increasing the complexity of the fabrication process. Here we demonstrate that drop-cast networks of monochiral (6,5) SWNTs, which are commonly dispersed in organic solvents with the polyfluorene–bipyridine copolymer PFO-BPy, can be employed directly and without additional functionalization or ion-selective membranes to detect Cu2+ ions over a wide range of concentrations in aqueous solutions. The observed voltage shifts of water-gated transistors with these (6,5) SWNT networks directly correlate with the cupric ion concentration. They result from induced n-doping due to the complexation of positive copper ions to the bipyridine units of the wrapping polymer. Furthermore, the competitive binding of Cu2+ to the herbicide glyphosate as well as to biologically relevant pyrophosphates can be used for the direct detection and quantification of these molecules at nano- to micromolar concentrations.
In the above types of sensors the SWNT network itself is seen as largely neutral with respect to the analyte.13 However, the dispersion of SWNTs and formulation of inks already requires specific surfactants or polymers that may also interact with certain analytes. In particular, the selective dispersion of semiconducting carbon nanotubes in organic solvents relies on conjugated polymers often containing functional moieties along their backbone (e.g., pyridines, carbazoles, thiophenes)27–29 or sidechains,23,30 which may intentionally or unintentionally interact with ions or gas molecules. The most commonly used polymers for the selective dispersion of nanotubes are polyfluorenes31 and polycarbazoles.32 The highest selectivity for a single nanotube species – the (6,5) SWNTs – has been demonstrated with the polyfluorene–bipyridine copolymer PFO-BPy,33 which can be applied on a large scale34 and has also been used for the dispersion of large diameter semiconducting nanotubes for devices,35 including sensors.13
Here, we demonstrate that drop-cast networks of PFO-BPy wrapped (6,5) SWNTs can be used directly and without any additional functionalization in water-gated transistors to selectively detect Cu2+ ions over a wide range of concentrations. Furthermore, the competitive binding of Cu2+ to the herbicide glyphosate as well as pyrophosphate is used for their direct detection and quantification at low concentrations.
As illustrated in Fig. 1b, the transistors with polymer-wrapped SWNT networks were operated by water-gating. A custom-made PTFE reservoir was employed to enclose the interdigitated source/drain electrodes (channel length 20 µm, channel width of 10 mm) and a large side-gate (all Cr/Au). The reservoir was filled with de-ionized (DI) water or the aqueous solutions of the corresponding analytes as the electrolyte. Note that due to the autoprotolysis of water there is always a sufficient concentration of ions available for gating even without additional electrolytes.39
Prior to the acquisition of the presented transfer characteristics, all devices were conditioned by collecting ten consecutive cycles of transfer curves to ensure equilibration. The conditioning cycles of a transistor with a PFO-BPy wrapped SWNT network are shown in Fig. S2 (ESI†) including those for the corresponding analytes. They confirm that in all cases a stable state was reached already after the second cycle and no additional shifts occurred after that. All water-gated transistors showed only hole-transport within the investigated gate voltage range, as expected for large-bandgap SWNTs contacted by gold electrodes in water. The gate currents (leakage) were always at least one order of magnitude below the drain on-currents. Current hysteresis remained small. Only forward sweeps are shown and used in the following to extract voltage shifts.
After conditioning, transfer characteristics were acquired at low drain voltages (Vd = −0.1 V) by sweeping the gate voltage (Vg) between 0.8 V and −0.8 V as shown in Fig. 2a for a transistor with a PFO-BPy-wrapped SWNT network (PFO-BPy/CoMoCat). First, transfer curves were collected for pure water and recorded as the blank and reference for the cation analyte measurements. Next, aqueous solutions of Zn2+, Ni2+, Cr3+, Fe3+, Ag+ and finally Cu2+ (all with nitrate anions) at a metal ion concentration of 15 µM were introduced in the given order and the corresponding transfer curves were collected following the standardized conditioning steps (see Fig. S2, ESI†). After the removal of each analyte solution, another cycle of conditioning steps was performed with fresh DI water to ensure the consistency of the blank measurements before introducing the next analyte. A selection of transfer characteristics (only forward sweeps) is shown in Fig. 2a where a clear turn-on voltage shift to more negative values can be observed in the presence of Cu2+ ions accompanied by a slight increase in the overall on-current. In contrast to that, the presence of Ni2+ ions seems to have the opposite – albeit weaker – effect on the turn-on voltage with no significant impact on the on-current. The full range of tested cations is presented in Fig. 2b. Compared to the reference measurement with DI water, all investigated metal cations tend to shift the corresponding transfer curves toward slightly more positive gate voltages, except for Cu2+.
The response of the water-gated PFO-BPy/CoMoCat transistor towards various cation analytes was extracted as the gate voltage shift (ΔV) with respect to the blank measurement (PBC blank) at a constant drain current of −0.1 µA and an applied drain voltage of −0.1 V. Note, that we estimate a standard deviation of about 15–20 mV for the extracted voltage shifts based on the variation of the blank samples between measurements. These shifts are plotted in Fig. 2c. Evidently, all devices show a similar response (positive 50–70 mV shift) toward Zn2+, Ni2+, Cr3+, Fe3+ and Ag+ ions, whereas the response differs substantially in the presence of Cu2+ ions with a voltage shift of approximately −100 mV. This negative voltage shift indicates a clear molecular interaction of cupric ions with the PFO-BPy/CoMoCat network beyond simple variation of ionic strength of the electrolyte and hence Debye layer thickness. Based on previous reports on structurally similar molecules,40–42 as well as molecular bipyridine and 1,10-phenanthroline43 we presume that the Cu2+ ions interact strongly with the nitrogen of the bipyridine unit to form a stable complex. The binding configuration might be similar to that of the previously reported complex of Re(CO)5Cl with PFO-BPy.37,38 However, unwrapping of the polymer from the SWNTs is unlikely to occur due the insolubility of PFO-BPy in water.
To confirm the selective interaction of Cu2+ ions with the bipyridine units of the wrapping polymer PFO-BPy, a reference transistor was prepared with a drop-cast network of CoMoCat SWNTs dispersed with PFO as the wrapping polymer (see Fig. 1a), which does not contain any moieties that could form metal complexes. The same conditioning steps (see ESI, Fig. S3†) were applied prior to the acquisition of the transfer curves. These confirmed again that an equilibrium was reached after the first couple of cycles. However, compared to the transfer characteristics of PFO-BPy/CoMoCat network transistors, the hole currents were higher (probably because of slight variations of the network density) and the turn-on voltage was substantially more positive. Such p-doping of nanotubes in the presence of oxygen and water is expected44 but was not observed for PFO-BPy-wrapped nanotubes. This difference can be explained by the electron-donating lone-pair electrons of the nitrogen atoms in the bipyridine units of PFO-BPy, which slightly n-dope the semiconducting SWNTs as shown recently by Li et al.27 and thus counteract the usual p-doping of nanotubes in air/water.45
In contrast to the strong response toward Cu2+ by the water-gated transistors with PFO-BPy/CoMoCat networks, the nearly identical PFO/CoMoCat network transistors do not show any sensitivity toward Cu2+ or any other tested metal cation. Fig. 3a shows the transfer characteristics (only forward sweeps) of the latter devices measured under the same conditions using DI water, and aqueous solutions of 15 µM Ni2+ and 15 µM Cu2+. Regardless of the employed analyte, the transfer curves only show a small turn-on voltage shift to more positive values as also summarized in Fig. 3b and c for all other tested metal cations. The turn-on voltage shift with respect to the blank (PC blank) at constant drain voltage of −0.1 V and drain current of −0.1 µA stayed below 30 mV for all tested analytes and without any particular selectivity for any of them. The overall weaker shifts for all metal salt solutions compared to the PFO-BPy/CoMoCat devices may also be rationalized with the slightly smaller bandgap of (7,5) nanotubes (1.21 eV) compared to (6,5) SWNTs (1.27 eV).
Clearly, the presence of the bipyridine units is crucial for the observed negative gate voltage shift and selectivity towards Cu2+ ions by the PFO-BPy/CoMoCat network devices. Overall, we can rule out any generalized interaction of Cu2+ with semiconducting nanotubes, further corroborating the assumed complexation with the bipyridine unit. The presence of positively charged copper ions within the Debye length of the electric double layer around the nanotubes should result in additional n-doping (or compensation of p-doping) of the nanotubes and hence a shift of the turn-on voltage to more negative values. The opposite shift induced by the other cations might be explained by changes of the overall ionic strength of the electrolyte (including the nitrate anions) and thus the capacitance of the electric double layer.39,46 However, no systematic shift based on the expected total ion concentration was observed.
The impact of the Cu2+ concentration on the voltage shift of PFO-BPy/CoMoCat network transistors was investigated for a wide range of concentrations between 3 nM to 150 µM versus a blank measurement (PBC blank). Fig. 4a shows selected transfer characteristics (forward sweeps) corresponding to different concentrations of Cu2+ ions. As the concentration increases, the transfer curves shift substantially toward more negative voltages. The full range of concentrations is displayed in Fig. 4b, from which the voltage shifts at a fixed drain current (−0.1 µA) depending on the Cu2+ ion concentration were extracted and are summarized in Fig. 4c. The voltage shift of ΔV = −22 mV at a Cu2+ concentration of 0.1 µM increased to −116 mV for a concentration of 15 µM, reaching −200 mV for 150 µM. The logarithmic plot of ΔV versus the concentration does not give a clear linear dependence as might be expected. At very low concentrations the voltage shifts are still within the uncertainty limits and the plot remains relatively flat. However, for higher concentrations (>3 µM) a slope of about 100 mV per decade can be extracted, which is within the expected range for binding and detection of a doubly charged cation. Based on the uncertainty of the voltage shifts for these devices (see above) we may estimate the limit of detection for Cu2+ ions to be around 100 nM.
In summary, a simple dropcast network of PFO-BPy-wrapped CoMoCat SWNTs can be used to selectively and quantitatively detect micromolar levels of Cu2+ ions with a good detection range. Next, the response of the Cu2+/PFO-BPy/CoMoCat complex toward the still ubiquitous but harmful herbicide glyphosate47 in water was studied. Competitive binding of Cu2+ by glyphosate48 has been widely used as a detection scheme42,47,49,50 and should lead to a backshift of the transfer curves of PFO-BPy/CoMoCat transistors treated with Cu2+.
To test this sensing concept, a PFO-BPy/CoMoCat SWNT network was exposed to an aqueous solution of 300 µM Cu2+ and conditioned with 10 consecutive transfer curve cycles (sweep rate ∼50 mV s−1, total measurement time ∼13 min) to saturate all exposed bipyridine units with cupric ions and thus reach a reproducible starting point. After removing the Cu2+ solution and rinsing the nanotube network with DI water, a blank measurement (PBC-Cu blank) was performed with pure water. The blank measurement showed a large and stable negative turn-on voltage as expected due to the binding of the copper(II) ions. Subsequently, aqueous solutions with different concentrations of glyphosate were introduced. The corresponding transfer curves (forward sweeps) are shown in Fig. 5a. A substantial backshift of the turn-on voltages to more positive values with increasing glyphosate concentration is evident. The response to a wider range of glyphosate concentrations and a zoomed-in view of the forward sweeps are shown in Fig. 5b. The data clearly indicate the reversal of the initial shift caused by the cupric ions as they are bound by the herbicide. Fig. 5c summarizes the extracted voltage shifts (ΔV) at constant drain voltage (−0.1 V) and drain current (−0.1 µA) versus the glyphosate concentration.
The demonstrated Cu2+/PFO-BPy/CoMoCat SWNT network transistor exhibits excellent sensitivity toward glyphosate with large shifts. Even nanomolar concentrations of the herbicide can be detected with a possible limit of detection of 1 nM. The already substantial voltage shift of +117 mV at a glyphosate concentration of only 3 nM increases to up to +288 mV for 15 µM and can reach up to 300 mV at 150 µM of glyphosate. The detection range is thus within typical and useful limits of glyphosate contamination in water.47 The slope of the logarithmic plot is about 44 mV per decade (approximately half of that for Cu2+, see above), indicating that two glyphosate molecules are required to form a complex48 and remove or neutralize one Cu2+ ion from the PFO-BPy-wrapped nanotubes. Note that PFO-BPy/CoMoCat SWNT networks could be re-used for several glyphosate measurements but re-saturation of the exposed bipyridine units with cupric ions was necessary.
Similar to glyphosate, pyrophosphates (PPi) can also bind cupric ions. Abnormalities of pyrophosphate levels in the human body can be responsible for or indicators of various diseases.51,52 Hence, their quantitative detection (usually using fluorescence probes) has been investigated for some time.53–57 In analogy to glyphosate, we also employed the Cu2+-treated PFO-BPy/CoMoCat transistors to detect pyrophosphate anions in water as shown in Fig. S4 (ESI†). Nanomolar levels of pyrophosphate resulted in measurable voltage shifts. Overall a wide range of concentrations from a few tens of nM to 100 µM correlated directly with positive voltage shifts between +150 and +300 mV (slope ∼45 mV per decade).
So far, we have only shown transistor responses for analytes in DI water, which clearly is not a typical sample due to the low ionic strength. Hence, we also tested the responsivity of our devices to glyphosate in regular tap water. For this measurement, the Cu2+-treated PFO-BPy/CoMoCat SWNT network was formed as described above and a blank measurement was acquired with tap water (PBC-Cu blank). Five different concentrations of glyphosate were added to the tap water samples. The corresponding transfer characteristics are shown in Fig. 6a. The extracted voltage shifts of the forward sweeps (see Fig. 6b and c) were smaller compared to the DI water samples and the corresponding slope was lower (only ∼20 mV per decade). Nevertheless, the overall trend remained the same and should enable reliable glyphosate detection in regular water after suitable calibration.
Furthermore, the Cu2+–bipyridine complex on the PFO-BPy wrapped SWNTs can be used as a secondary probe for the detection and quantification of the herbicide glyphosate in water at nanomolar levels as well as biologically important pyrophosphates. Relevant concentrations of glyphosate could even be detected when using regular tap water as the electrolyte, demonstrating the applicability of this sensor concept for real-world samples in an environmental/agricultural context.
A Fluorolog 3 spectrometer (Horiba Jobin–Yvon GmbH) equipped with a xenon lamp (450 W) with a double monochromator for excitation and a cooled InGaAs diode array (800–1600 nm) for detection was employed to collect photoluminescence excitation emission (PLE) maps for the polymer/SWNT hybrid dispersions. The measurements were conducted under ambient conditions using a 4 × 10 mm quartz cuvette. The emission intensities in all maps were normalized to the power of the excitation light source at the corresponding wavelength and further corrected using the wavelength-dependent sensitivity of the detector.
To study the impact of different metal ions, a blank measurement was obtained with DI water before introducing any metal ion solution and used as the reference. Aqueous solutions of Zn2+, Ni2+, Cr3+, Fe3+, Ag+ and Cu2+ were used at a fixed concentration of 15 µM in the given order and repeated blank measurements were conducted using DI water between two metal ion solutions. Note that due to the valency of the cations the corresponding nitrate concentration was 15, 30 or 45 µM and the ionic strength varied from 15 µM (for AgNO3), 45 µM (for Cu(NO3)2, Zn(NO3)2, Ni(NO3)2) to 90 µM (for Fe(NO3)3, Cr(NO3)3).
For experiments with different Cu2+ concentrations, a blank measurement was recorded using DI water before introducing the Cu2+ solutions and used as the reference. Aqueous solutions of Cu2+ were prepared at concentrations of 0.003 µM, 0.03 µM, 0.1 µM, 0.3 µM, 3 µM, 15 µM, 30 µM, 75 µM and 150 µM. Starting from the lowest concentration, the measurements were conducted without blank measurements between the Cu2+ solutions.
For the detection of various glyphosate (Gly) concentrations, the nanotube networks were exposed to 300 µM Cu2+ solution and conditioned for 10 consecutive cycles to saturate the bipyridine units with Cu2+ ions. Next, a blank measurement was performed with DI water before introducing the glyphosate solutions and used as the reference. Aqueous solutions of glyphosate were prepared at concentrations of 0.003 µM, 0.03 µM, 0.1 µM, 0.3 µM, 3 µM, 15 µM, 30 µM, 75 µM and 150 µM. Starting from the lowest concentration, all measurements were recorded without blank measurements between the different glyphosate solutions.
The impact of pyrophosphate (PPi) was studied following the same procedure as for glyphosate, using a Cu2+ saturated nanotube network with the same concentrations as above.
For the detection of glyphosate in regular tap water, the devices were saturated with Cu2+ ions using a 300 µM Cu2+ solution and conditioned with 10 consecutive cycles. Next, the reservoir was filled with tap water (as obtained without further purification, Heidelberg-Neuenheim, Germany) and transfer curves were recorded as blank reference. Solutions of glyphosate were prepared at concentrations of 0.003 µM, 0.1 µM, 3 µM and 15 µM in the same tap water and used in the given order as the analyte. The transistor measurements were conducted without any blanks between the different glyphosate solutions.
Footnote |
† Electronic supplementary information (ESI) available: Spectroscopic characterisation of dispersions, conditioning of devices with different nanotube networks, pyrophosphate sensing. See DOI: https://doi.org/10.1039/d2nr02517e |
This journal is © The Royal Society of Chemistry 2022 |