In
Figure 2b, we plot the transfer curves (conductance
G versus liquid gate potential
) for a gold-coated nanoribbon functionalized with a SAM of
ligands (
) measured in buffered solutions with varying NaF concentration from
to
. The curves shift to the right indicating the adsorption of negatively charged
ions. To quantify the shift, we extract the threshold voltage
for each transfer curve using our well-established method [
6,
7,
12,
30,
31] by reading out the value of
at a constant conductance value of
in the subthreshold regime of the transistor (black arrow in
Figure 2b). In the following, we use
to quantify the response of the different nanoribbons to changes in electrolyte concentration. In total, the response of a subset consisting of 14 out of 48 nanoribbons (4
, 4
and 6
) was measured in order to minimize the measurement time. For the sake of clarity, we discuss here the results for a specific NR triplet consisting of one
, one
and one
as depicted in
Figure 3a. More information on the reproducibility and distribution of responses is given in the
Supplementary Materials. The ribbons were chosen as such to represent functioning devices, showing a similar behavior in the control measurements in KCl and pH solutions as observed in previous measurements [
6]. In the following, we compare the response of these three devices measured for increasing salt concentration (
mM to
M) of NaF, NaCl and KCl and changing pH from pH 3 to pH 9 (
Figure 3b–e). In particular, we investigate whether we can discriminate between sodium and fluoride ions by comparing the response of
and
with the control
.
Figure 3b shows the threshold voltages for the selected NR triplet in NaF solution. For the sake of readability, the experimental points from each NR were shifted along the vertical axis, leading to
. The original data is shown in the
Supplementary Materials. Green squares correspond to
for
shown in
Figure 2b. The threshold voltage
increases with salt concentration. We define the total change of the threshold voltage as
. For
,
as indicated in
Figure 3b. Additionally, the threshold voltage of
(black triangles) and
(red circles) are shown. Note that
exhibits a response to changes in NaF concentration with
mV. We attribute this response to the non-specific adsorption of fluoride ions at the bare gold surface, similarly to what we observed in our previous work for chloride ions [
6,
12]. Interestingly,
shows even a smaller
over the investigated concentration range. The observed behavior of the three different surfaces agrees well with the following picture: the largest response is observed for
due to the adsorption of fluoride ions at the SAM. The smaller response of
corresponds to the non-specific adsorption of fluoride ions. Therefore, we conclude that the response measured for
partially includes contributions from non-specific adsorption of fluoride ions at the gold surface. The smallest response is observed for
due to the additional adsorption of
ions in the crown ether, partially compensating the effect of non-specific fluoride adsorption. We repeated the measurement for the same set of NRs for increasing NaCl (
Figure 3c) and KCl (
Figure 3d) concentration. For both salts,
shows a response to changes in concentration due to the non-specific adsorption of chloride ions, in agreement with our previous work [
6,
7]. Furthermore, all three NRs exhibit a similar response to pH, as shown in
Figure 3e, which is attributed to the presence of a low density of oxidized gold surface atoms [
6].
To account for the non-specific anion adsorption at the gold surface, we follow the differential approach as introduced in our previous work [
6,
7]. Thereby, we subtract the threshold voltage of
(
) from the two active NRs (
and
) leading to the differential signal
for
and
for
as shown in
Figure 4. It reveals the response of the two ligands (
Figure 4a:
ligand,
Figure 4b:
ligand) and allows a quantitative comparison of the different surfaces. Negligible or weak responses to pH and changes in KCl concentration are observed for both ligands. This indicates that the functionalization does not influence the pH response and that neither potassium nor chloride ions bind to the two ligands. When changing NaCl and NaF concentration, however, a clearer differential response of
mV per decade (mV/dec) in salt concentration is observed for
, which is due to the sensitivity of the
ligand to sodium. Note that the sign of the differential response indicates the adsorption of positively charged sodium ions. While
shows only a differential response when sodium ions are present, a similar behavior is expected from
for fluoride ions. However, due to the negatively charged fluoride ions, a positive differential response is predicted in this case. Indeed, we find for
a differential response of
mV/dec in NaF due to the adsorption of
at the SAM. Therefore, the simultaneous detection of sodium and fluoride ions in NaF is achieved. Finally, we also observe a differential response for
of
mV/dec in NaCl which points towards some non-specific adsorption of sodium ions at the SAM. However, cation adsorption is not expected from the structure of the
ligand and further measurements are needed to verify this finding.
We observe that the obtained responses are smaller than the maximum Nernst limit of
mV/dec in ion concentration. This is a disadvantage compared to ISEs where thick membranes allow a Nernstian response over a large concentration range [
32]. The sub-Nernstian response might be due to the relatively low ligand density at the sensor surface achieved. Using a very simplified site-binding model, we estimate the lower value of the density of
ligands on
to be
and the lower value of the density of
ligands on
to be
. Note, these values are the lower estimates of the ligand density as discussed in the
Supplementary Material. However, the studied ion receptors have not been optimized to achieve a high density on the surface, e.g., by minimizing their size. Comparing different NRs of the same surface reveals large variations in response (see
Supplementary Materials). This indicates that our method of functionalization is prone to variations in final ligand density, which has a pronounced influence on the response, as described in our previous work [
7]. The quality and the reproducibility of the SAM are therefore key elements for the further success of the presented approach. Although not demonstrated in this work, our approach could allow for the detection of mixed analyte solutions, where several types of anions and cations are present, given the response of individual NRs to the specific analytes is known. However, cross sensitivity limits the universality of this system and has to be taken into account. Nonetheless, our functionalization method results in an integrated sensing platform, and the simultaneous detection of sodium and fluoride ions is demonstrated.