Analyst View Article Online

Published on 30 June 2014. Downloaded by University of Chicago on 27/10/2014 17:08:23.

PAPER

View Journal | View Issue

Cite this: Analyst, 2014, 139, 4947

Selective ion-sensing with membranefunctionalized electrolyte-gated carbon nanotube field-effect transistors† a b b b a K. Melzer,*a A. M. Mu ¨ nzer, E. Jaworska, K. Maksymiuk, A. Michalska and G. Scarpa

In this work the ion-selective response of an electrolyte-gated carbon-nanotube field-effect transistor (CNT-FET) towards K+, Ca2+ and Cl
in the biologically relevant concentration range from 10
1 M to 10
6 M is demonstrated. The ion-selective response is achieved by modifying the gate-electrode of an electrolyte-gated CNT-FET with ion-selective membranes, which are selective towards the respective target analyte ions. The selectivity, assured by the ion-selective poly(vinyl chloride) based membrane, allows the successful application of the herein proposed K+-selective CNT-FET to detect changes in the K+ activity in the mM range even in solutions containing different ionic backgrounds. The sensing mechanism relies on a superposition of both an ion-sensitive response of the CNT-network as well as a change of the effective gate potential present at the semiconducting channel due to a selective and ion activity-dependent response of the membrane towards different types of ions. Moreover, the Received 22nd April 2014 Accepted 30th June 2014

combination of a CNT-FET as a transducing element gated with an ion-selective coated-wire electrode

DOI: 10.1039/c4an00714j

offers the possibility to miniaturize the already well-established conventional ion-selective electrode setup. This approach represents a valuable strategy for the realization of portable, multi-purpose and

www.rsc.org/analyst

low-cost biosensing devices.

1. Introduction Since the introduction of the ion-sensitive eld-effect transistor (ISFET) in the 1970s by Bergveld et al.,1 there has been growing interest in applying the concept of eld-effect transistors (FETs) to convert chemical signals into electrical ones. In the last few years this concept has also been applied to devices on the nanoscale, e.g. nanowire and carbon nanotube FETs. In this context much effort was put into various sensing applications such as pH sensing or the label-free detection of biomolecules.2–9 The sensing principle is based on the adsorption of charged molecules on the active surface of the sensor contributing to electrostatic gating of the semiconductor, which further modies the current owing in the semiconducting channel of the FET. Moreover, the possibility of down-scaling the device and the solution processability makes FETs based on carbon nanotube networks (CNT-FETs) a promising platform to meet the demand of cost-effective, scalable and exible biosensors.10–12 Unfunctionalized CNT-FETs have already proven to be very sensitive towards changes in the hydroxonium (H3O+) concentration or the ionic strength of an electrolyte.10–12 However, especially the

a Institute for Nanoelectronics, Technische Universit¨ at M¨ unchen, Arcisstraße 21, 80333 M¨ unchen, Germany. E-mail: [email protected] b

Faculty of Chemistry, University of Warsaw, Pasteura 1, 02-093 Warsaw, Poland

† Electronic supplementary 10.1039/c4an00714j

information

(ESI)

This journal is © The Royal Society of Chemistry 2014

available.

See

DOI:

emerging eld of point-of-care diagnostics and environmental monitoring sets high standards for the development of biosensors, namely the selective sensing of pH and physiological electrolytes (K+, Na+ and Ca2+) or other biological species and ions in complex samples. In the eld of medical and environmental applications the concept of ion-selective electrodes (ISEs) has already been wellestablished, since these sensors offer high selectivity and sensitivity coupled with an easy preparation of the receptor layer.13 Apart from the traditional internal solution arrangement of ISEs, this type of sensor can also be prepared by applying ion-selective membranes onto conducting polymers or carbon nanotube layers,14–16 yielding the so-called all-solid-state potentiometric sensors. However, the usage of conventional ISEs is limited in applications where a small size, low power consumption, integration in a miniaturized ow system, low sample consumption and a rapid response for in situ monitoring are further requirements for the selective detection of certain ions or the pH value.17,18 In this context ISFETs, which are modied to achieve not only an ion-sensitive but also an ionselective response, become competitive. A big advantage of ISFETs over conventional ion-selective electrodes is that the high impedance signal from the high resistive ISE can be circumvented in ISFETs by in situ signal amplication, generating a low-impedance output signal and additionally providing low power consumption.19 One can already nd several examples in the literature for inorganic

Analyst, 2014, 139, 4947–4954 | 4947

View Article Online

Published on 30 June 2014. Downloaded by University of Chicago on 27/10/2014 17:08:23.

Analyst

ISFETs,20–23 where the gate insulator was functionalized with polymeric ion-selective membranes to selectively detect certain ions. But these sensors most oen suffer from the bad adhesion of the polymeric membrane to the surface of the gate materials.21,24 One more recent example of an ion-selective eld-effect transistor was published by the group of C. Sch¨ onenberger.25 They modied silicon nanowires with a thin gold layer, which was further functionalized with a self-assembled monolayer of thiol-modied crown ethers to selectively detect sodium ions in a concentration range between 10
3 M and 1 M. By using carbon nanotubes or semiconducting polymers, instead of inorganic semiconductors, as transducing layers with the accompanied good biocompatibility26,27 of both, the transducing layer as well as the polymeric ion-selective membrane, the use of these sensors for biological and medical measurements (e.g. monitoring of cell–cell communication events by detecting the two prominent second messengers for cell–cell communication K+ and Ca2+(ref. 28)) is possible and it opens up the way towards low-cost disposable systems. Cid et al.29 previously applied a valinomycin-based K+-selective membrane on top of a CVD-grown network of single-walled CNTs, which was used as an active semiconducting layer in a eld-effect transistor. With online measurements of the source-drain current (ISD) vs. time in the conventional back-gated conguration they could prove a selective response of their device towards potassium in the relatively small concentration regime of 10
8 M to 10
6 M. Very recently the rst ion-selective organic eld-effect transistors have been reported.30–32 The group of Z. Bao32 integrated Hg2+-selective binding sites into an organic eld-effect transistor by incorporating DNA-functionalized gold nanoparticles on the surface of the semiconducting polymer PII2T-Si (a polyisoindigo-based polymer with siloxane-containing solubilizing chains). They could prove a selective response of the OFET towards Hg2+ in seawater samples against other important seawater contaminants (Zn2+ and Pb2+). A polymeric membranemodied ion-selective P3HT (poly(3-hexylthiophene)) FET was reported by Schmoltner et al.,30 where a Na+-selective membrane separates the P3HT FET and a reference electrolyte with a given ion concentration from the sample solution and a conventional Ag/AgCl-reference electrode. Since the proposed setup30 resembles a lot an organic eld-effect transistor gated with a conventional ISE, miniaturization of the whole sensor, which is a requirement for further integration of the sensor in a microuidic system, is still restricted by the relative huge size of the reference electrode. This limitation can be circumvented by combining the concept of an ion-selective coated-wire electrode33 as a gate-electrode together with a eld-effect transistor as a signal converter, which allows the realization of a miniaturized ion-selective sensor. Recently we could demonstrate an ionselective response towards the biologically relevant cations K+ and Ca2+ by applying this concept and functionalizing the conventional metal gate-electrode of an intrinsically ion-sensitive P3HT FET with different polymeric ion-selective membranes.31 These sensors could be successfully used to detect concentrations of the respective primary ion in the mM range even in solutions with a xed ionic background.

4948 | Analyst, 2014, 139, 4947–4954

Paper

However, in all cases either the drain current (ISD) response29,31,32 over time was analyzed or the ISD vs. time response was translated into a potentiometric signal.30 Since an electrolyte-gated FET in contrast to a conventional ISE exhibits a multi-parametric response towards changes of the electrochemical potential at the semiconductor/electrolyte and the gate-electrode/electrolyte interface, additional information about the sensing mechanism, which get neglected during online measurements, can be obtained by analyzing the transistor characteristics under various conditions. In this work, we utilize carbon nanotube eld-effect transistors in the electrolyte-gated conguration, where the semiconducting layer is a randomly distributed CNT network deposited by spin-coating an aqueous nanotube ink solution on top of a substrate with already predened source and drain electrodes (see also the schematic shown in Fig. 1, further details on the fabrication process are reported in ref. 11 and in the Experimental section). By functionalizing the gate-electrode of the electrolyte-gated CNT-FET, which is one possibility to achieve a selective response of an electrolyte-gated FET,34,35 with poly(vinyl chloride) ion-selective membranes of different compositions, we take advantage of a special kind of ISEs previously developed33 which allow the construction of very small and durable solidstate electrodes. In this coated-wire electrode (CWE) conguration the polymeric membrane is directly coated onto a conductive surface, such as a metal wire.33 Analyzing the transistor characteristics and extracting the two main parameters, the threshold voltage Vth and the transconductance, allows us to distinguish between primary and interfering ions. We demonstrate a selective response of the CNT-FET towards K+, Ca2+ or Cl
in the biologically relevant concentration range of 10
1 M to 10
6 M. Moreover the response of the model K+-selective CNTFET towards potassium is not affected in solutions with different ionic background contents. The sensing mechanism relies on the superposition of both, an ion-sensitive response of the CNT-network as well as a change of the effective gate potential present at the semiconducting channel due to a selective and ion activity-dependent response of the membrane towards different types of ions. This means that the overall response of the device depends on changes of the

Device structure and measurement setup. The schematic representation of an ion-selective electrolyte-gated CNT-FET. The conventional electrolyte-gate (platinum wire) is functionalized with an ion-selective polymeric membrane.

Fig. 1

This journal is © The Royal Society of Chemistry 2014

View Article Online

Paper

Analyst

electrochemical potential at both active interfaces, the interface of the membrane-functionalized gate-electrode with the electrolyte as well as the interface between the CNT-network and the electrolyte. This combination of an ion-selective coated-wire electrode as gate-electrode together with an intrinsically ionsensitive CNT-FET as a signal converter allows the construction of a micro-scale, portable, low-cost ion-selective sensing device.

Published on 30 June 2014. Downloaded by University of Chicago on 27/10/2014 17:08:23.

2.

Experimental section

2.1. Device fabrication The water-based carbon nanotube ink was prepared by dispersing 0.05 wt% of a commercially available CNT powder, purchased from SWeNT (SG65, chirality 6.5 (>50%) and >90% semiconducting tubes) in deionized water with 1 wt% carboxymethyl cellulose (CMC)36 as the dispersant. Sonication with a horn sonicator (Branson Digital Sonier 450, 30 min) further enhanced the dispersion of the carbon nanotubes. The dispersion was then centrifuged to separate the sediment from the well dispersed CNT ink. Directly prior to the deposition of the nanotube lm, the substrates with the already predened IDE structures (5 nm chromium, 40 nm gold, channel length 50 mm, W/L ¼ 900) were exposed to an oxygen plasma (30 s) to guarantee homogeneous lm formation on the hydrophobic silicon dioxide underneath layer. Finally the ink was spin-coated (3000 rpm, 50 s) onto the substrate. Aer the CNT deposition, the substrates were immersed overnight (>12 h) in diluted nitric acid (68% HNO3 : H2O ¼ 1 : 4) to remove all CMC residuals, followed by thorough rinsing with deionized water and drying on a hot plate. Subsequently the carbon nanotube network was removed everywhere on the chip except for the active sensing layer by another oxygen plasma treatment step to prevent the occurrence of parasitic currents. To avoid direct contact of the source and drain electrodes with the electrolyte they were passivated with a photoresist layer (AZ 4562, Clariant GmbH).

2.2. Sensor functionalization The ion-selective gate-electrodes were fabricated by cleaning a Pt-wire with acetone and isopropyl alcohol before coating the Pt-surface with the membrane cocktail (the exact compositions of the membranes are listed below) until a homogeneous layer with a thickness of 100–200 mm was achieved. Aerwards the electrodes were dried overnight under ambient conditions to guarantee a complete removal of all solvent residues. To obtain an optimal response, the freshly prepared membranes require a conditioning procedure to achieve an adequate saturation with water and an uptake of primary ions, which further improves the detection limit.37 Therefore the functionalized electrodes were placed for about 12 h in conditioning solutions containing 10
3 M of primary ions diluted in deionized water before the rst measurement was performed. In between different measurements the electrodes were stored dry aer rinsing them with deionized water and drying them with nitrogen. Prior to

This journal is © The Royal Society of Chemistry 2014

each new measurement they were conditioned again for 2 h in the conditioning solution. 2.3. Membrane preparation Here the fabrication of the PVC-based ion-selective membranes is described. All membrane components were purchased at Sigma Aldrich (selectophore purity grade). K+-selective membrane cocktails contained (by weight) 1.26% of sodium tetrakis [3,5-bis(triuoromethyl)phenyl]borate salt (NaTFPB), 3.09% valinomycin (selectophore purity grade), 64.99% of the plasticizer dioctyl sebacate (DOS), and 30.66% poly(vinyl chloride) (PVC). In total 206.8 mg of membrane components were dissolved in 6 ml of tetrahydrofuran (THF). Ca2+-selective membrane cocktails contained (by weight) 0.7% of sodium tetrakis [3,5-bis(triuoromethyl)phenyl]borate salt (NaTFPB), 1% N,N-dicyclohexyl-N0 ,N0 -dioctadecyl-3-oxapentanediamide (ETH-5234, selectophore purity grade), 67.7% of the plasticizer 2-nitrophenyl octyl ether (oNPOE), and 30.6% poly(vinyl chloride) (PVC). In total 280 mg of membrane components were dissolved in 6 ml of tetrahydrofuran (THF). Anion-selective membrane cocktails contained (by weight) 4.9% of tridodecylmethylammonium chloride (TDMAC), 64.3% of plasticizer 2-nitrophenyl octyl ether (oNPOE), and 30.8% poly(vinyl chloride) (PVC). In total 280 mg of membrane components were dissolved in 6 ml of tetrahydrofuran (THF). 2.4. Electrical characterization in liquid For all electrical measurements a PDMS ring was mounted around the active area of the CNT-FET to serve as a compartment for 100 ml of the electrolyte, which was exchanged manually. Transfer curves (ISD vs. VEG) were recorded by sweeping the gate voltage, applied to the gate-electrode immersed in the electrolyte, between 0.6 V and
0.6 V with a sweep rate of 0.008 V s
1, using a Keithley source meter (System 2636A). During the whole measurement the drain voltage was xed at a constant voltage of
0.1 V. In between two succeeding transfer measurements the system could equilibrate for 5 minutes aer exchanging the electrolyte. KCl ($99.0%, Sigma Aldrich), NaCl ($99.5%, Sigma Aldrich), CaCl2 (CaCl2$2H2O, $99%, ACS reagent) and MgCl2 ($98%, Merck) were dissolved in deionized water (resistivity 18.2 MU cm) and were further diluted. Measurements in liquid with an ionic background were either performed with a constant background of 150 mM NaCl or in a mixed ionic background containing 112 mM NaCl, 0.41 mM CaCl2, 0.28 mM MgCl2 and optionally 3.7 mM KCl. The mixed ionic background equals the maximum concentration range of these cations (Na+, K+, Ca2+ and Mg2+) in blood.38

3.

Results and discussion

Fig. 2 shows the source-drain current ISD vs. the electrolyte-gate potential VEG of a CNT-FET gated with either an unfunctionalized Pt-wire as the gate-electrode or with a membrane-coated Ptelectrode recorded in a 10
2 M solution of KCl as the electrolyte.

Analyst, 2014, 139, 4947–4954 | 4949

View Article Online

Published on 30 June 2014. Downloaded by University of Chicago on 27/10/2014 17:08:23.

Analyst

Fig. 2 Transfer characteristic ISD vs. VEG of a pristine Pt-gated CNTFET (black) and an ion-selective CNT-FET gated with a Pt-wire coated with a K+-selective PVC-membrane (blue). The red dashed lines show the fit of the linear region of the drain current ISD, which was further used for the determination of the threshold voltage Vth. The arrows indicate the sweep direction.

The transfer curves exhibit typical eld-effect characteristics with a slight hysteresis: for the Pt-gated CNT-FET we obtain an anti-clockwise hysteresis, whereas the membrane-functionalized CNT-FET shows a clockwise hysteresis. The transfer curve of the ion-sensitive Pt-gated CNT-FET (Vth ¼
0.26 V) is shied by about 100 mV towards more positive gate voltages (Vth ¼
0.16 V) by functionalizing the gate-electrode with a polymeric ion-selective membrane, in this case a K+-selective membrane. These results clearly demonstrate that the functionalization of the gate-electrode does not impair the device performance of the electrolyte-gated CNT-FET: on-current values of several tens of mA and an on/off-ratio of about 150 are achieved. Fig. 3a shows transfer curves ISD vs. VEG recorded with a CNTFET gated with a Pt-wire, which was modied with a K+-selective polymeric membrane. With a decreasing potassium (K+) activity in the solution, the transfer curve shis towards more negative gate voltages (indicated by the arrow). To quantify the shi, we extract the threshold voltage Vth from the transfer curve by tting the linear regime (red dashed lines) and extracting the

Paper

intercept with the x-axis. In Fig. 3e the values of Vth are plotted against the logarithm of the ion activity present in the electrolyte (Vth vs. log a). We obtain a linear response of +39.3 mV per dec between 10
1 M and 10
7 M. For lower K+ activities the threshold voltage shi saturates. A change in the ion activity results in a change in the open circuit potential at the membrane/solution interface. This potential either lowers or increases the effective gate potential, which is present at the semiconducting channel. The incorporation of positively charged primary ions (i.e. the ion to which the membrane is primarily responsive to) results in a membrane potential, which gets more positive by an increasing primary ion activity in the sample.39 The maximum possible shi in the membrane potential is given by the Nernst equation, which predicts a shi of +59.5 mV per dec at 300 K for a monovalent cation.40 The lower detection limit of 10
7 M of the primary ion is not an intrinsic limitation of the transducing ionsensitive CNT-FET, but rather it is pre-determined by the properties of the used type of membrane (see ESI, Fig. S1†).41 The achieved detection limit is well comparable with other sensors using this type of receptor layer.13–15,37 For the control measurement with an unfunctionalized Ptgated CNT-FET we obtain a shi of the transfer curve towards more positive gate voltages with decreasing K+-activities (Fig. 3b and 2e). This intrinsic ion-sensitivity of the CNT-network can be attributed to an electrical double-layer dependent, and therefore ion concentration dependent, screening of negative charges on the surface of the CNTs, which probably arise from the acidic treatment required during the device fabrication (see also Experimental section).10–12 A lower ion concentration leads to a thicker electrical double layer at the interface of the CNTnetwork and the electrolyte, which reduces the amount of these surface charges getting screened by salt ions. As a consequence the effective gate potential at the CNT surface gets increased and the threshold voltage shis towards more positive gate voltages.10,42–44 One has to keep in mind that the threshold voltage of an ionselective electrolyte-gated transistor is not only related to the potential drop formed on the surface of the ion-selective

Fig. 3 Selective potassium sensing with membrane modified CNT-FET. (a–d) Transfer curves recorded in decreasing activities of different ions: (a) K+ -selective CNT-FET in KCl, (b) unfunctionalized CNT-FET in KCl, (c) K+-selective CNT-FET in NaCl and (d) K+-selective CNT-FET in CaCl2. For the K+ -selective CNT-FET in KCl a shift of the transfer curves towards more negative gate-voltages (indicated by the arrow) occurs, whereas the control measurements show a negligible signal or a shift of the transfer curves towards more positive gate-voltages (unfunctionalized CNTFET in KCl). (e) Threshold voltage Vth vs. ion activity. Vth was extracted from the linear fits (red dashed lines) of the different transfer curves (a–d).

4950 | Analyst, 2014, 139, 4947–4954

This journal is © The Royal Society of Chemistry 2014

View Article Online

Published on 30 June 2014. Downloaded by University of Chicago on 27/10/2014 17:08:23.

Paper

Analyst

membrane given by the Nernst equation but also the ionconcentration dependent change in the gate capacitance at the CNT-network/electrolyte interface plays an important role. Moreover, other factors like surface states and the doping level in the channel have an impact on Vth,45 but they are predetermined by the properties of the semiconducting material and can be assumed to stay constant during the experiments. To prove the K+-selective response of the functionalized CNTFET we also performed control measurements in electrolytes containing different activities of either the interfering monovalent cation Na+ (Fig. 3c) or the interfering divalent cation Ca2+ (Fig. 3d). Potassium-selective membranes, under open circuit potentiometric operation mode, are known to discriminate either of the two ions.46 However, it should be stressed that the selectivity of classical ISEs as well as CWEs can be described by the selectivity coefficient KI,J according to the Nicolskii–Eisenman equation: E ¼ E0 þ

 RT  ln aI þ KI;J aJ ZI =ZJ zI F

(1)

where E is the membrane potential; E0 is a constant; aI and aJ are the activities of the primary and interfering ions, respectively; zI and zJ are the valencies of the primary and interfering ions, respectively. Here the subscript I corresponds to the primary ion, whereas J belongs to the interfering ion.13,40 This suggests that the selective response of the membrane potential depends on both the potentiometric selectivity coefcient KI,J as well as the valency of the respective interfering ion. In the case of the model K+-selective CNT-FET we expect a weaker effect of Ca2+ on the potential of the ion-selective membrane compared to Na+, since the exponent zI/zJ is inversely proportional to the valency of the interfering ion (the potentiometric selectivity coefficient was assumed to be the same for Na+ and Ca2+,47 see eqn (1)). In both cases, Na+ as well as Ca2+, the threshold voltage Vth stays nearly constant over the whole activity range (Fig. 3e), conrming the high selectivity of the ionophore valinomycin towards potassium ions.48 Interestingly, in the case of the interfering ion with the same valency (Na+) as the primary ion (K+) the transconductance gm, which is dened as follows (eqn (2)), changes signicantly (see Fig. 3c and ESI Fig. S2 and S3†). gm ¼

DISD W mCeff VD ¼ L DVEG

(2)

The values for the transconductance were extracted from the linear ts of the transfer curves (red dashed lines, Fig. 3a–d). According to eqn (2) a change in the transconductance can be attributed to either a change in the charge carrier mobility m or it goes along with a change of the effective gate capacitance Ceff, which is mainly given by the in-series capacitances of the semiconductor/electrolyte and the electrolyte/gate-electrode interface. We are assuming that the change in the transconductance goes along with a change of the total capacitance due to changes at the membrane/solution interface. This effect seems to be dominant for the interfering ion with the same valency as the primary ion (see ESI†). In general the interaction

This journal is © The Royal Society of Chemistry 2014

of certain ions with an ionophore depends on the valency of the ion as well as the size, the shape and the binding sites of the ionophore. For biomedical or water-quality applications, sensors have to be developed, which allow the determination of ions, especially for pH and physiological electrolytes (K+, Na+ and Ca2+) in many different kinds of samples. It is especially critical in biological samples such as cell extracts, crude blood samples and growth media or for soil extracts, where the whole composition of the solution has to be considered.49 The selectivity is clearly one of the most important characteristics of such a sensor, as it oen determines whether a reliable measurement in a target sample with a complex background is possible. As K+ is one of the most important cations present in biological systems, we tested the sensitivity and selectivity of the K+-selective CNT-FET in the presence of a high molarity of an interfering ion or in a mixed ionic background. First, transfer curves were recorded in decreasing activities of either KCl or CaCl2 dissolved in deionized water and the threshold voltage Vth has been extracted for each ion activity. Aerwards, decreasing activities of either KCl or CaCl2 in a constant 150 mM NaCl background were analyzed. Fig. 4a depicts the good selectivity of the K+-selective sensor towards different potassium activities over Ca2+ in the solutions without background as well as in the solutions with an ionic background. The response curves Vth vs. log a in deionized water get shied by about 40 mV towards more negative gate voltages by adding a 150 mM NaCl background. This effect can be attributed to electrostatical gating at the CNT/electrolyte interface, caused by the intrinsic ion-sensitivity of the CNT network. With an increasing ionic strength more of the negative charges on the surface of the CNTs get screened due to the formation of an electrical double layer, thus shiing the transfer curves towards more negative gate voltages.10,42–44 No appreciable changes in the lower detection limit could be observed, the K+-selective CNT-FET shows a linear threshold voltage shi of 46.2 mV per dec down to an activity of 10
6 M in deionized water (open symbols) and a linear threshold voltage shi of 47.4 mV per dec down to an activity of 10
5 M for a 150 mM NaCl background (lled symbols). So it is still possible to detect the activity of the primary ion K+ although the background concentration of the interfering ion Na+ is more than 4 orders of magnitude higher. The same measurements (Fig. 4b) were also performed in a mixed ionic background, which equals the maximum concentration range of the cations Na+, Ca2+ and Mg2+ in blood (the exact composition of the mixed ionic background is reported in the Experimental section).38 Again the K+selective CNT-FET shows a linear threshold voltage shi of 43.2 mV per dec down to an activity of 10
5 M for the primary ion K+ and a suppressed response towards different activities of Ca2+. Moreover, if the maximum concentration of KCl in blood (3.7 mM) is added to the mixed ionic background and in addition to it KCl concentrations between 10
1 M and 10
4 M KCl get dissolved in this ionic background, the K+-selective sensor still reveals a linear threshold voltage shi down to an additional KCl concentration of 10
4 M, which corresponds to a K+ activity of 10
2.5 M (Fig. 4c).50 This clearly demonstrates the

Analyst, 2014, 139, 4947–4954 | 4951

View Article Online

Published on 30 June 2014. Downloaded by University of Chicago on 27/10/2014 17:08:23.

Analyst

Paper

Fig. 4 Selectivity of the K+-selective CNT-FET in electrolytes with different ionic backgrounds. (a) Responses of the K+-selective CNT-FET towards different activities of KCl and CaCl2 either in deionized water (open symbols) or in a constant ionic background of 150 mM NaCl (filled symbols). (b) Responses of the K+-selective CNT-FET towards different activities of KCl or CaCl2 either in a mixed ionic background without KCl or (c) in a mixed ionic background with 3.7 mM KCl.

feasibility to detect changes in the concentration of the primary ion, which are at least one order of magnitude lower than the background concentration of the primary ion. The whole results represent an essential step towards a potential biomedical sensor platform: with the membranemodied CNT-FETs it could be possible to detect a small concentration variation of K+ present in a pH-buffered system (e.g. PBS or HEPES), which comes along with a high ionic background, or to monitor changes in the ion content of blood samples. By combining an ion-selective coated-wire electrode (CWEs) with an ion-sensitive CNT-FET as a signal transducer, a multiparametric analysis of complex samples can be easily realized by changing the properties of the modied gate electrode. For example, by incorporating a different ionophore within the ionselective membrane the selectivity of the membrane can be altered, leading to a different type of ion sensor without changing the transducing component, due to the given highstability of the CNT-based active layer. By using the Ca2+-selective ionophore N,N-dicyclohexyl-N0 ,N0 -dioctadecyl-3-oxapentanediamide51 we are able to realize calcium ion sensors. Similar to experiments with potassium-selective membranes, we obtain a shi of the threshold voltage towards more negative gate voltages with a decreasing Ca2+ activity present in the electrolyte, which was to be expected for a positively charged ion (see Fig. 5). The device shows a linear response of +21.7 mV per dec towards its primary ion between 10
1 M and 10
6 M of it diluted in deionized water. According to the Nernst equation the change of the membrane potential is inversely proportional to the valency of the ion and therefore has a theoretical response limit of +29.8 mV per dec for a divalent cation.40 The control measurement with an unfunctionalized Pt-electrode in decreasing activities of CaCl2 dissolved in deionized water again shows a threshold voltage shi towards more positive gate voltages. Good selectivity towards Ca2+ is emphasized by the control measurements with the Ca2+-selective device in solutions containing monovalent or divalent interfering ions. The response towards the interfering ions Na+ (monovalent) and Mg2+ (divalent) is suppressed and shows a slight shi of Vth towards more positive gate voltages for

4952 | Analyst, 2014, 139, 4947–4954

Ca2+-selective response of the CNT-FET modified with a PVCbased Ca2+-selective membrane. The sensor shows a linear response of about 22 mV per dec in the activity range from 10
1 M to 10
6 M. For a better comparison of the data all response curves, despite the one for the primary ion, were shifted to give the same value at an ion activity of 10
4 M. Fig. 5

decreasing ion activities. This is most probably caused by a superposition of the ion-sensitive signal of the CNT network and a relatively small signal at the ion-selective gate-electrode. For the transconductance we observe the same effect as already mentioned for the K+-selective device: the most signicant change in transconductance is noticeable for a divalent interfering ion, here in this case Mg2+ (see ESI Fig. S4†). To prove the working principle of the CNT-FET gated with the membrane-modied electrode, we additionally modied the Pt-gate with an anion-selective polymeric membrane containing the ion-exchanger tridodecylmethylammonium chloride,15 which should result in a shi of the transfer curve towards more positive gate voltages for decreasing anion activities. As expected, the transfer curve shis towards more positive gate voltages with a decreasing anion activity (Cl
) present in the electrolyte. In Fig. 6 the values of Vth are plotted against the logarithm of the ion activity. The maximum possible shi in the membrane potential for a monovalent anion is given by the Nernst limit, which predicts a shi of
59.5 mV per dec at 300 K.40 For chloride (Cl
) we

This journal is © The Royal Society of Chemistry 2014

View Article Online

Paper

Analyst

effective, exible and disposable biosensors this paves the way towards new applications in the above mentioned elds.

Published on 30 June 2014. Downloaded by University of Chicago on 27/10/2014 17:08:23.

Acknowledgements

Fig. 6 By modifying the CNT-FET with an anion-selective membrane the threshold voltage shifts towards more positive gate voltages for decreasing activities of the primary ion. For KCl we obtain a linear response of
51 mV per dec and a lower detection limit of 10
5 M.

obtain a slightly lower linear response of
51.5 mV per dec between 10
1 M and 10
5 M.

4. Conclusion In conclusion we demonstrated the selective sensing of K+, Ca2+ and Cl
in the biologically relevant concentration range by functionalizing the gate-electrode of an electrolyte-gated CNTFET with different polymeric ion-selective membranes. This allows reliable and selective ion detection in a target sample with a complex background, which is especially important for biomedical diagnostics with samples such as cell extracts and growth media or for water-control measurements. All sensors show a good selectivity towards their respective primary ions. The sensing mechanism can be, additionally to the already known ion-sensitive response of the bare CNT-network, attributed to a change in the threshold voltage of the transfer curves due to a selective and ion activity-dependent response of the membrane towards different types of ions, which alters the effective gate potential present at the semiconducting channel. Furthermore it is shown for the K+-selective CNT-FET that the sensors can be successfully used to selectively detect concentrations of primary ions down to a concentration in the mM range even in solutions with an ionic background of 150 mM of its interfering ion or in a more complex background containing different types of interfering ions as well as a low constant concentration of the primary ion. The combination of an ion-selective electrode as a functional element and a CNT-FET as a transducer enables the analysis of complex samples by the subsequent detection and quantication of the activity of different ions, beneting from the fact that the electrode can be exchanged without changing the underlying CNT-FET. Our results underline the possibility to further miniaturize the whole sensor, which is also a signicant advantage from the point of view of ion-selective electrodes, by using the coated-wire principle,33 instead of a conventional ISE setup, in combination with the CNT-FET as a transducer element. Since electrolyte-gated FETs on the basis of a solutionprocessable CNT-network have the potential to be used as cost-

This journal is © The Royal Society of Chemistry 2014

This work (K. Melzer, A. M. M¨ unzer and G. Scarpa) was supported by the International Graduate School for Science and Engineering (IGSSE) at the Technische Universit¨ at M¨ unchen. Financial support from the National Centre of Science (Poland), project N N204 247640, in the years 2011–2014 (A. Michalska and K. Maksymiuk) and from the Foundation for Polish Science Ventures Programme co-nanced by the EU European Regional Development Fund (E. Jaworska) is gratefully acknowledged.

References 1 P. Bergveld, IEEE Trans. Biomed. Eng., 1970, 17, 70–71. 2 K. Maehashi, T. Katsura, K. Kerman, Y. Takamura, K. Matsumoto and E. Tamiya, Anal. Chem., 2007, 79, 782– 787. 3 I. Heller, A. M. Janssens, J. M¨ annik, E. D. Minot, S. G. Lemay and C. Dekker, Nano Lett., 2008, 8, 591–595. 4 E. D. Minot, A. M. Janssens, I. Heller, H. a. Heering, C. Dekker and S. G. Lemay, Appl. Phys. Lett., 2007, 91, 093507. 5 Y. Choi, I. S. Moody, P. C. Sims, S. R. Hunt, B. L. Corso, I. Perez, G. A. Weiss and P. G. Collins, Science, 2012, 335, 319–324. 6 Y. Choi, T. J. Olsen, P. C. Sims, I. S. Moody, B. L. Corso, M. N. Dang, G. A. Weiss and P. G. Collins, Nano Lett., 2013, 13, 625–631. 7 X. Duan, Y. Li, N. K. Rajan, D. A. Routenberg, Y. Modis and M. A. Reed, Nat. Nanotechnol., 2012, 7, 401–407. 8 Y. Huang, C. Wu, C. Chuang, S. Pang, T. Pan, Y. Yang and F. Ko, Anal. Chem., 2013, 16, 7912–7918. 9 X. Duan, R. Gao, P. Xie, T. Cohen-Karni, Q. Qing, H. S. Choe, B. Tian, X. Jiang and C. M. Lieber, Nat. Nanotechnol., 2012, 7, 174–179. 10 T. Haeberle, A. M. M¨ unzer, F. Buth, J. A. Garrido, A. Abdellah, B. Fabel, P. Lugli and G. Scarpa, Appl. Phys. Lett., 2012, 101, 223101. 11 A. M. M¨ unzer, K. Melzer, M. Heimgreiter and G. Scarpa, Biochim. Biophys. Acta, 2013, 1830, 4353–4358. 12 A. M. M¨ unzer, M. Heimgreiter, K. Melzer, A. Weise, B. Fabel, A. Abdellah, P. Lugli and G. Scarpa, J. Mater. Chem. B, 2013, 1, 3797–3802. 13 E. Bakker and E. Pretsch, Anal. Chem., 2002, 15, 420–426. 14 G. A. Crespo, S. Macho and F. X. Rius, Anal. Chem., 2008, 80, 1316–1322. 15 A. Michalska, J. Dumanska and K. Maksymiuk, Anal. Chem., 2003, 75, 4964–4974. 16 C. Zuliani, G. Matzeu and D. Diamond, Electrochim. Acta, 2014, 132, 292–296. 17 C. Jimenez-Jorquera, J. Orozco and A. Baldi, Sensors, 2010, 10, 61–83. 18 P. Bergveld, A. Sibbald and G. Svehla, in Analytical and Biomedical Applications of Ion-Selective Field-Effect Transistors, Elsevier Science, Amsterdam, 1988, pp. 75–99.

Analyst, 2014, 139, 4947–4954 | 4953

View Article Online

Published on 30 June 2014. Downloaded by University of Chicago on 27/10/2014 17:08:23.

Analyst

19 U. Oesch, D. Ammann and W. Simon, Clin. Chem., 1986, 32, 1448–1459. 20 P. L. H. M. Cobben, R. M. Egberink, J. G. Bomer, P. Bergveld, W. Verboom and D. N. Reinhoudt, J. Am. Chem. Soc., 1992, 114, 10573–10582. 21 Y. Miyahara and W. Simon, Electroanalysis, 1991, 3, 287–292. 22 J. Sanchez, A. Beltran, J. Alonso, C. Jimenez and M. Valle, Anal. Chim. Acta, 1999, 382, 157–164. 23 S. D. Moss, J. Janata and C. C. Johnson, Anal. Chem., 1975, 47, 2238–2243. 24 P. T. Mc Bride, J. Janata, P. A. Comte, S. D. Moss and C. C. Johnson, Anal. Chim. Acta, 1978, 101, 239–245. 25 M. Wipf, R. L. Stoop, A. Tarasov, K. Bedner, W. Fu, I. A. Wright, C. J. Martin, E. C. Constable, M. Calame and C. Sch¨ onenberger, ACS Nano, 2013, 7, 5978–5983. 26 G. Scarpa, A.-L. Idzko, S. G¨ otz and S. Thalhammer, Macromol. Biosci., 2010, 10, 378–383. 27 N. Mzoughi, A. Abdellah, Q. Gong, H. Grothe, P. Lugli, B. Wolf and G. Scarpa, Sensor. Actuat. B-Chem., 2012, 171172, 537–543. 28 M. J. Berridge, P. Lipp and M. D. Bootman, Nat. Rev. Mol. Cell Biol., 2000, 1, 11–21. 29 C. C. Cid, J. Riu, A. Maroto and F. X. Rius, Analyst, 2008, 133, 1001–1004. 30 K. Schmoltner, J. Ko, A. Klug and E. J. W. List-Kratochvil, Adv. Mater., 2013, 25, 1–5. 31 K. Melzer, A. M. M¨ unzer, E. Jaworska, K. Maksymiuk, A. Michalska and G. Scarpa, Org. Electron., 2013, 15, 595–601. 32 O. Knopfmacher, M. L. Hammock, A. L. Appleton, G. Schwartz, J. Mei, T. Lei, J. Pei and Z. Bao, Nat. Commun., 2014, 5, 2954. 33 D. M. Pranitis, M. Telting-Diaz, M. E. Meyerhoff and R. R. Schroeder, Crit. Rev. Anal. Chem., 1992, 23, 163–186. 34 H. Tang, F. Yan, P. Lin, J. Xu and H. L. W. Chan, Adv. Funct. Mater., 2011, 21, 2264–2272. 35 S. Casalini, F. Leonardi, T. Cramer and F. Biscarini, Org. Electron., 2013, 14, 156–163.

4954 | Analyst, 2014, 139, 4947–4954

Paper

36 T. Takahashi, K. Tsunoda, H. Yajima and T. Ishii, Jpn. J. Appl. Phys., 2004, 43, 3636–3639. 37 A. Michalska, K. Pyrzy´ nska and K. Maksymiuk, Anal. Chem., 2008, 80, 3921–3924. 38 E. Bakker and M. E. Meyerhoff, in Encyclopedia of Electrochemistry, Wiley-VCH Verlag GmbH & Co. KG, 2007, Vol. 9. 39 P. B¨ uhlmann and L. D. Chen, in Supramolecular Chemistry: From Molecules to Nanomaterials, Wiley CH, 2012, pp. 2539–2579. 40 A. J. Bard and L. R. Faulkner, Electrochemical Methods -Fundamentals and Applications, Wiley, New York, 2001. 41 A. J. Michalska, C. Appaih-Kusi, L. Y. Heng, S. Walkiewicz and E. A. H. Hall, Anal. Chem., 2004, 76, 2031–2039. 42 I. Heller, S. Chatoor, J. M¨ annik, M. A. G. Zevenbergen, C. Dekker and S. G. Lemay, J. Am. Chem. Soc., 2010, 17149– 17156. 43 A. B. Artyukhin, M. Stadermann, R. W. Friddle, P. Stroeve, O. Bakajin and A. Noy, Nano Lett., 2006, 6, 2080–2085. 44 F. Chen, Q. Qing, J. Xia, J. Li and N. Tao, J. Am. Chem. Soc., 2009, 131, 9908–9909. 45 S. M. Sze and K. N. Kwok, Physics of Semiconductor Devices, John Wiley and Sons Inc., New York, 2007. 46 A. Michalska, Anal. Bioanal. Chem., 2006, 384, 391–406. 47 E. Jaworska, W. Lewandowski, J. Mieczkowski, K. Maksymiuk and A. Michalska, Analyst, 2013, 138, 2363– 2371. 48 M. C. Rose and R. W. Henkens, Biochim. Biophys. Acta, 1974, 372, 426–435. 49 E. Bakker, P. B¨ uhlmann and E. Pretsch, Chem. Rev., 1997, 97, 3083–3132. 50 D. A. Skoog, D. M. West, J. F. Holler and S. R. Crouch, Fundamentals of Analytical Chemistry, Brooks/Cole, Cengage Learning, Hampshire, 2014. 51 P. Gehrig and W. Simon, Chimia, 1989, 43, 377–379.

This journal is © The Royal Society of Chemistry 2014