A miniaturized sensing device with ionically imprinted nanostructured films for the ultrasensitive detection of ions


 The detection of ions is essential for a wide range of applications including biomedical diagnosis, and environmental monitoring among others. However, current ion sensors are based on thick sensing films (typically 100 µm), requiring time-consuming preparations, and have a thermodynamic limit to their sensitivity of 59 mV.Log[C]-1. Such configuration hinders the development of high-performance ion sensors due to the inherent limitations of the bulk diffusion of ions inside sensors. Consequently, they cannot be applied for high-precision applications that require high sensitivity. Furthermore, the research of anion monitoring is hampered due to the limited availability of molecular receptors with acceptable performances. We overcome such limitations by using a 300 nm nanostructured sensing film based on a novel nanoporous ion imprinted core-shell silica/gold nanoparticulate sensing film. The novel sensing film was highly selective towards chloride ions when compared to other anions such as nitrate, sulphate and carbonate. Moreover, this nanostructured sensing film exhibited above 3-fold higher sensitivity (-186.4 mV.Log[C]-1) towards chloride ions when compared to commercial devices. Such breakthrough has led to the fabrication of the smallest and most sensitive reported anion sensor working on open circuit potentiometry, with an exceptional selectivity towards chloride ions.


Introduction
The measurement and monitoring of ions in solution is essential for a wide range of applications such as the clinical analysis, physiological monitoring [1][2][3] , water quality evaluation [4] performance monitoring of energy storage devices [5] , corrosion monitoring in structural engineering [6] , or agricultural smart farming [7] among others. Specifically, anions imbalance play a key role in the development of multiple diseases such as chronic kidney disease [8] and can lead to multiple pathological conditions such as acidosis, caused by an accumulation of acid that can be a consequence of an excess of chloride ions, with deleterious effects in the body [9] . In addition, excessive presence of anions such chloride ions in aquatic environments can lead to the acidification of the environment or decrease the anti-microbial capability of treated water. Therefore, its concentration must be tightly regulated [10] . From an industrial perspective, the presence of chloride ions is a strong indicator of environmental degradation processes such as corrosion [6] and the quality and health of the structural engineering materials [11] . However, due to the generally small size of these anions, less than 0.2 nm in the case of chloride ion (Cl -), and their scarce lipophilicity [12] , only a few commercial anion sensors available in the market, and there are limited reported literature on these sensors. Most of these sensors also exhibit a rather poor performance due to ionic interferences and hence low selectivity [13] .
While there is a demand for miniaturized anion sensing devices with super-sensitivity, greater than the Nernst limit of 59 mV.Log[C] -1 for the measurement of anions for high-precision applications, to date, no approaches have been able to fulfil such requirements. Traditionally, the main technology for the development of ion-selective electrodes consists of a plasticized PVC membrane with embedded ionophores, which are specific complexing agents that perform the electrolyte recognition process [14] . However, numerous studies have been focused on the formalization of ion-selective electrodes theory [15][16][17][18] , and concluded that the sensitivity of the devices is thermodynamically limited to 59 mV.Log[C] -1 . [14] In addition, their miniaturization has been proven to be challenging, and typical sensors tend to be thick, with a typical thickness in the range of 100 µm [19] .
Early attempts in the miniaturization of these membranes showed a faster ion equilibration kinetics when they were reduced to 35 µm thick layers [20] . However, no sub-micron level films operating at zero-current conditions have been reported in the literature. Such level of reduction could offer advantages in terms of the time-response of the sensors and allow a better integration in miniaturized systems, especially desirable for portable or wearable technologies. As such, an enhancement in the sensitivity and miniaturization of the ion sensors would be beneficial for portable and wearable applications. Such miniaturization of the sensing materials would also be beneficial to interface this sensor technology with upcoming technologies such as ion-selective field effect transistors [21] . In addition, the development of rapid sensors operating with super-sensitivity beyond the Nernst limit is crucial to characterize the concentration of ions locally around neurons, which could help in the diagnosis of seizures [22] , and could allow the genome sequencing [23] .
Thus far, the performance of commercial and state-of-the-art ion sensors based on ionophorecontaining polymeric films are constrained to the use of relatively thick sensing films, in the range of micrometers, and a restricted sensitivity of 59 mV.Log[C] -1 . Shklovskii et al. [24,25] observed that this sensitivity limit could be surpassed when charged surfaces are employed, offering the possibility of developing a new family of sensors with super-Nernstian sensitivities. This observation was a consequence of the strong interaction between the opposite charges of ions and the membranes. Although this work was based on the study of macroions, recently Sivakumarasamy et al. [26] demonstrated the suitability of this concept to develop an ion-sensor based on specific sites on a 25 nm 0D transistor using a silica surface for the simultaneous detection of multiple cations in serum. However, the costly fabrication methods, based on nanolithography and their sensing mechanism, requiring complex nanofluidic channels, led to challenging data analysis, due to the necessity of modelling the response of the sensor in the presence of multiple electrolytes. This has not only led to a large uncertainty on the measurements and sensing data but also makes them an impracticable technology commercially. Moreover, this device configuration could not be used for the determination of negatively charged ions such as chloride ions. Since then, multiple ionic exchange resins have been developed for the adsorption of multiple ions, limited to cations such as zinc [27] , nickel [28] and lead [29,30] . However, these materials lacked specificity, limiting their suitability as components of ion sensors, and in general there is a lack of advances in the field of anion-selective materials. In fact, the discovery of anions sensitive materials or materials for the selective detection of certain anions represents a major challenge in analytical chemistry, electrolyte monitoring in patients, diagnosis, and anion measurement for environmental evaluation.
The exploitation of imprinted technology for the fabrication of sensing materials represents a promising alternative in this field, allowing the development of tailor-made sensing electrodes with reduced dimensions. The material imprinting process is based on the synthesis of sensitive species by direct polymerization of functional monomers in the presence of a template. Thus, specific recognition elements can be developed virtually for any compound [31,32] . To date, this approach has been applied mostly to the detection of cations, including copper [33][34][35] , cadmium [36,37] , mercury [38] and lead [39] , and rare earth metal ions such as europium [40] and yttrium [41] . However, due to the weak interaction of the anions with the chelating species commonly found, thus far there has been a limited development in this field.
Although similar imprinting approaches have been applied to the fabrication of sensing materials for the detection of relatively small anions such as phosphate and nitrate ions [42,43] for potential applications in environmental monitoring, its potential for the specific interaction with single anions such as chloride ions has yet to be explored.
Within the present work, a miniaturized device with a nanometer-size thick sensing film was developed by exploiting the ion imprinting technology with a novel material functionalization and grafting methods. The new device structure reported here is based on the use of a porous core-shell configuration with gold nanoaggregates and a shell of nanoporous silica containing positively charged amine groups, which enhanced the sensitivity of the sensors and greatly reduced the preparation time. These Au nanoparticles were synthesized to be specific towards chloride by condensing silica monomers in the presence of HCl, and using gold nanoaggregates as the core material. Such material transformed the typical signal transduction mechanism of the traditional sensors based on the bulk diffusion of electrolytes, into a surface effect through the adsorption of ions. Such novel materials and sensing device configuration reported here enabled the development of the smallest Ion sensor working on Open Circuit potential with an unprecedented sensitivity and selectivity.

Structure of the nanostructured sensing layer
The first step in the development of the nanostructured sensors was the characterization of the fabricated gold nanoclusters. Such gold nanoclusters were synthesized by a simple reduction method using citric acid. Gold was used for the transduction of the electrochemical signal due to its strong interaction with chloride ions [44] and its low charge transfer resistance [45] . A solid shell of nanoporous silica using tetraethyl orthosilicate (TEOS) as a precursor was then integrated, enabling the adsorption, increasing the specificity of the sensor, and allowing the attachment to the gold contact. In addition, the amine group present in aminopropyl triethoxysilane (APTES) was used as the recognition element towards chloride ion species.
This functional group was proven to be versatile, enhancing the interaction with the anions after being positively charged using an acidic solution of HCl.
The specificity of this material was accomplished by the synthesis of the TEOS and gold precursors in the presence of HCl, which contains the target ion Cl -, allowing the formation of a specific pore structure in this material. The structure of this recognition element, using a porous silica as the receptor located on the outer shell of the sensor, was confirmed by high resolution transmission electron microscopy (HRTEM) (Figure 1.a) and Figure 1  A porous structure was formed within the ionic imprinted silica/gold, which was consistent with the previous work in silica-based nanoimprinted particles [46] . The pore radius here observed was confirmed by BET, being in the range of 1.4 nm (Figure 2.a)), and a total surface area of 108 m 2 g -1 . Such surface area was calculated from the isotherms obtained using BET and the associated BET equation (Figure S.2). This pore size is comparable to previous ion imprinted systems based on silica nanoparticles for dissolved metals [47,48] .
The spherical architecture of the ionically imprinted nanoparticles with a size in the range of 118±29 nm after the synthesis was further corroborated by SEM (Figure 2.b) and Figure S.3).
The nanoparticle grafting on the gold electrodes was then carried out by a directional attachment of APTES via EDC/NHS chemistry. A successful anchoring of molecularly imprinted polymeric nanoparticles has been achieved in the literature by previous functionalization with primary amines and posterior reaction with the substrates [49] . However, this process had to be modified in our nanoimprinted material since the direct use of the EDC reagent could modify the amine groups present in the specific recognition sites, leading to a loss in sensitivity. Thus a novel fabrication process was developed. Here, the gold electrodes were first functionalized using a self-assembled monolayer of 16-mercaptohexadecanoic acid.
A directional attachment of APTES using the standard 1-Ethyl-3-(3dimethylaminopropyl)carbodiimide and N-Hydroxysuccinimide (EDC/NHS) chemistry was then employed, grafting the molecules and exposing the silane groups that could interact with the nanoparticles. Finally, the nanoparticles were deposited. The chemical characterization of the nanoimprinted material was performed using FTIR (Figure 2.c)) which revealed the typical spectrum of APTES, with a peak in the 1390 cm -1 range, indicative of the -CN stretch.
The 1032 and 778 cm -1 peaks were also detected due to the presence of the Si-O-Si structure of silica [50] . This 2-step approach for the fabrication of the sensing devices could be applied to the production of an array of up to 5 electrodes onto a glass substrate (Figure 2.d)).

Demonstration of the sensing performance of the ultra-thin Imprinted ion sensor
After grafting the nanoparticles onto 50 nm thick gold electrodes with an APTES linker, a 290 nm thick nanostructured film was obtained as measured by a stylus profilometer (Figure S.4). To the best of our knowledge, this is the smallest reported potentiometric ion sensor based on Open Circuit Potential (OCP) readings. After the fabrication of the full device, a homogeneous film was obtained, with the presence of regularly distributed domains rich in ion imprinted nanoparticles containing gold. This surface morphology of the film consisting of ion imprinted nanoparticles, formed small aggregates in the range of 500 nm as observed by SEM equipped with EDS elemental mapping (Figure 3.a)).
The attachment of the ion imprinted nanoparticles to the electrodes using the covalent bond with the APTES linker here reported was analyzed by quartz microbalance. This technique can be used for the determination of weight changes over an electrode in the nanogram range, confirming the success of the functionalization due to the changes in the mass on the electrodes in Figure S.5)). Within the first step of the fabrication, where the APTES group was directionally grafted after the deposition of the self-assembled film, a gain of 188±20 ng.cm -1 was observed. The post-modification with silica/gold nanoparticles increased the total weight by 294±3 ng.cm -1 , being consistent with the two-step approach for the nanofilm fabrication. This technique could additionally be employed for the determination of the total amount of adsorbed chloride ions on the surface of the ion imprinted porous silica/gold nanoparticles. Such test was used to confirm the relevance of using the chloride ion templates during the synthesis. In this case, the differences in the mass of the sensors during the conditioning in DI and after their immersion in 0.1 M KCl solution was recorded, and normalized by dividing the result by the initial weight of the films in DI water.
Here, an absorption ratio of 8.1 Δngion.ngeq -1 was obtained when templated gold/silica ionically imprinted nanoparticles were employed. On the contrary, when no HCl was employed, also called non-templated ion imprinting nanoparticles, a significantly lower absorption with a maximum of 3.4 Δngion.ngeq -1 was obtained as shown in Figure 3.b).
Consequently, these devices could not be employed for the sensing. Such lower absorption could be attributed to the reduced formation of nanopores and the low presence of charged functional groups.
A final test of the electrochemical performance was subsequently performed on the sensing device. The OCP of the ion sensor was used as the measurement method, similar to the case of the commercial polymeric based sensing films, which tends to require a low energy consumption. Contrary to the standard commercial polymer-based ion-selective sensors, the devices here employed did not require a long pre-conditioning process. The commercial sensors typically need to be subjected to a highly concentrated solution in the range of 0.1 M of the target analyte for at least 24 h to reach equilibrium before their usage [51] . Consequently, they require time-consuming preparation for such preconditioning process. However, the present ion sensing device reported here could reach the equilibrium with the solution shortly after the synthesis, and only 46 mins were required to reach the equilibrium (Figure 3.c)). The signal stability of the ion sensors was in the range of 1.1 mV.h -1 , being significantly lower than the standard commercialized devices for anions, in the range of 5 mV [52] , and was stable for at least 13 h under constant contact with the solution. Thus, the ion sensing devices could be applied to the continuous measurement of electrolytes due to the rapid nature of the adsorption reaction, offering a prompt monitoring of analytes.
When using metal salts that contained chloride ions such as NaCl, KCl or CaCl2, a super-Nernstian response in the range of -186.4 mV.log[Cl -] -1 was achieved within the linear range of the plots, with a limit of detection in the range of 10 -4 M (Figure 3.d)). This value overcomes the traditional Nernst sensitivity limit of -59 mV.Log[C -] -1 , representing a new record in the performance of these anion sensors, specifically chloride ions. Here, the sensitivity was calculated by measuring the slope of the responses within the linear range of the measurements (Figure 3.e)). These sensors could be re-used, giving a reproducible response towards Cl -. This reproducibility of the measurements was additionally proven using 3 different chlorinated salts. The ion sensor showed a similar response when measuring chloride ions from NaCl, KCl and CaCl2 as shown in Figure 3.e). Additionally, the ion sensing device showed an unprecedented selectivity towards chloride ions when compared to other halogens such as F -, Brand anions like SO4 2-, NO3 -, and HCO3 -. These electrolytes were specifically tested due to their essential roles in the clinical diagnostic for diseases resulted from electrolytes imbalance [53][54][55] .  (Figure 3.f). As such, the selectivity of the devices was demonstrated. Such selectivity was also evaluated by calculating the selectivity coefficient of the silica/gold ionically imprinted nanoparticles, being higher than the previously reported work in the literature, especially in the case of Brand Fions [56] (Table S.  Only a significant change in the potentiometric signal can be observed when Clis added. The amount of template ions (Cl -) employed during the synthesis of the Ion imprinted silica/gold nanoparticles played a crucial role in the performance of the final device. Since Clwas used for the creation of a specific pore by its interaction with the aminated APTES molecules, its concentration and stoichiometry with respect to APTES had to be tightly mV.Log[Cl -] -1 . Although such sensitivity was higher than the one achieved when low amounts of HCl template were employed, it was significantly lower than the optimal sensitivity at 1 mM HCl. Under optimal conditions, the chloride ions were adsorbed by a combination of electrostatic interactions and hydrogen bonds [31] . However, when high concentrations of template were used compared to APTES, the pores did not present the optimal conditions for such interactions. Consequently, the sensitivity of the devices was reduced ( Figure S.6).

Sensing mechanism
Since our nanostructured sensing film is based on nanoporous core/shell silica/gold nanoparticles, its sensing mechanism is a consequence of the adsorption of ions on the surface of the nanomaterials. This mechanism has been proven by multiple simulations studied on molecularly imprinted materials. Here, this particular molecular structure, forming a specific pore with a specific disposition of functional groups and ligands, in combination with the stoichiometry employed, allowed the formation of a highly sensitive and selective sensing film. The ions present in solution in a solvated form, surrounded by a water shell (Figure 4.a)) can then interact with the amine groups at the surface of the nanoporous core/shell silica/gold nanoparticles (Figure 4.b) and Figure 4c)). Ion imprinted silica/gold nanoparticles on the sensing layer can address the thickness and sensitivity limitations of conventional membrane-based ion-selective electrodes by offering specific binding sites directly exposed to the sample. Consequently, the sensitivity of the electrodes will be triggered by surface effects, due to the adsorption of chloride ions onto the sensitive silica-gold nanoparticles, instead of the traditional bulk diffusion. This new mechanism allows the sensitivity to go beyond the traditional Nernst limit and permit the size of the sensing electrodes can be reduced. Gold nanoclusters were here employed due to their low charge transfer and electrical resistances and their affinity towards chloride ions [44,45] .
These gold nanoaggregates provided support for the formation of the selective silica shell during the synthesis of the films as depicted in Figure S.7. In addition, such gold nanoaggregates improved the conductivity of the nanoparticles, since the low conductivity of silica could hinder the electrochemical transduction of the sensors. Such effect of improved conductivity has been reported in the case of molecularly imprinted sensors [57] . In the present work, the use of gold was fundamental for improving the sensitivity rates of the ion sensors by improving the signal transduction. When no gold was employed during the sensitivity of the devices towards Clions was -66±7 mV.Log[Cl -] -1 when using NaCl salts for the calibration. This value was significantly lower than the value reported when gold was used as the core material for the ionically imprinted nanoparticles, which was in the range of -186 mV.Log values. This effect was reported by Hamza et al. [58] , who described the higher absorption rate of uranium ions in the presence of quaternary amine groups. As such, the involvement of the quaternary amine groups from the APTES in the sensing was confirmed, which allowed the potentiometric detection of chloride ions in the electrolytes. This device configuration could overcome the Nernst sensitivity limit.

Conclusions
We have demonstrated for the first time the application of ionically imprinted nanomaterials in the form of ultra-thin (e.g. 300 nm thick) nanostructured films with silica/gold core-shell nanoparticles as sensing element for the detection of Cl -. The device showed a superior sensitivity of -186.4 mV.Log[C -] -1 . This sensitivity surpassed the limit of the current state-ofthe-art chloride ion-selective electrodes, which tend to have a thermodynamical limit of 59 mV.Log[C -] -1 under Open-Circuit potential measurements and tend to have a relatively thick sensing layer (typically above 100 µm). Therefore, the novel anion sensing device developed here exhibited a 3-fold higher sensitivity, with a detection limit in the range of 10 -4 M. These sensors were proven to offer a stable signal, with a low drift in the range of 1.1 mV h -1 and did not require the current lengthy and laborious pre-conditioning process in the commercial anion sensing devices. Such significant enhancement was possible due to the adsorption of the electrolytes to the specific surface interactions with the ion imprinted nanopores on the silica/gold core-shell nanoparticles and the use of positively charged functional groups. This proposed new sensing mechanism based on the adsorption of ions on the silica/gold core-shell nanoparticles has enabled the miniaturization and fabrication of fully functional ultra-thin sensing films with sub-micrometer dimensions, which is not possible using the current commercial plasticized PVC technology which relying on the bulk diffusion of ions inside the sensing films. These anion sensors also showed an unprecedented selectivity towards chloride ions, as demonstrated by calibrating the response of the sensors towards some of the most common interferences found in these sensors, including HCO3 -, NO3and SO4 2-. In addition, they remained stable up to a pH of 11. Such breakthrough would open up the opportunity to the fabrication of tailor-made nanostructured thin films for the selective detection of challenging ionic species, especially anions with high specificity and selectivity, which is critical for a wide range of currently important technologies, including environmental remediation, drug development and energy storage among others.

Fabrication of thin film conducting electrode
A 50 nm thin gold film was deposited onto a glass substrate by plasma sputtering (Q150RES, Quorum technologies, UK) using standard conditions. Here, a methacrylate mask was first pre-fabricated using a laser cutter (PLS6.75, Universal Laser Systems, Austria). This mask was designed for the deposition of up to 5 electrodes after positioning them onto 7.5x2.5 cm glass slides and subsequently to be sputtered with gold. Each individual electrode had a radius of 5 mm.

Fabrication of sensing electrode and device
The fabrication steps and procedures of the sensing materials and device are summarized in Figure 5 and the details are presented below.

Gold NP synthesis
The ionically imprinted material consisted of a gold nanocluster aggregate integrated with a nanoporous silica shell. These gold nanoparticles were synthesized by the reduction of tetrachloroauric acid (TCA) with citric acid as reported elsewhere [59] . Briefly, a 5 mL solution

Chemical analysis and composition characterization of the sensing layers
Chemical analysis was performed using FTIR (L160000A Perkin Elmer, US) and EDS mapping of the elemental composition was carried out on the sensing layers ( Figure 5.f). The surface area and pore size distribution of the ion imprinted silica/Au nanoparticles were determined using BET Surface analyzer (Quantachrome, NOVATouch, UK). Functionalization of Au electrode: This is achieved by introducing a self-assembled monolayer of 16-mercaptohexadecanoic acid, containing a -SH group that can strongly bind to the gold film, and a carboxyl group(-COOH). e) Linkages: The free -COOH group of the functionalized Au electrode in d) was used to guide the directional attachment of the APTES linker that bridge the electrode with the siloxane groups on the core shell silica-gold synthesized in b) in order to produce f) via the use standard EDC/NHS chemistry; f) A 300 nm thick nanostructured sensing film is finally produced specific for the detection of chloride ions.

Sensing performance characterization
The relative ion intake of the sensing device in Figure 5.f) was determined using a quartz microbalance (Q-sense, Biolin Scientific, Switzerland), and compared to the non-templated control samples (without including HCl in the precursor solution). The electrochemical analysis and sensing performance was carried out by measuring the Open Circuit Potential (OCP) of the sensors upon being subjected to different solutions with anions Cl -, SO4 2-, NO3 -, HCO3at different concentrations (10 -4 , 10 -3 , 10 -1 , 1, 10, 40, 80 and 160 mM) and studying the specificity towards interference ions using a standard 3-electrode configuration, with a Ag/AgCl reference and a platinum counter electrode. The sensitivity of the electrodes towards Clwas also repeated by using imprinted nanoparticles with no gold nanoparticles as the core material. The selectivity of the devices was determined by using the matching potentials method. As a verification of the selectivity, the OCP potential of the sensors was recorded while concentrations of 10 mM of the different anions here employed were added.
For the determination of the effects of pH on the sensing performance, NaOH and H2SO4 were added to DI water to change the pH values. 5 different pH were tested (3, 5, 7, 9 and 11), and the sensitivity of the devices using NaCl solutions at the concentrations above mentioned was determined.        was tailored using NaOH and H2SO4. The Nernst sensitivity limit has been indicated.