Silicon-based fluorescent platforms for copper(ii) detection in water

The potential of silicon-based fluorescent platforms for the detection of trace toxic metal ions was investigated in an aqueous environment. To this aim, silicon chips were first functionalized with amino groups, and fluorescein organic dyes, used as sensing molecules, were then covalently linked to the surface via formation of thiourea groups. The obtained hybrid heterostructures exhibited high sensitivity and selectivity towards copper(ii), a limit of detection compatible with the recommended upper limits for copper in drinking water, and good reversibility using a standard metal–chelating agent. The fluorophore–analyte interaction mechanism at the basis of the reported fluorescence quenching, as well as the potential of performance improvement, were also studied. The herein presented sensing architecture allows, in principle, tailoring of the selectivity towards other metal ions by proper fluorophore selection, and provides a favorable outlook for integration of fluorescent chemosensors with silicon photonics technology.


Introduction
Several techniques are currently employed for detecting toxic metal ions at low concentrations in aqueous and biological environments, including Inductively Coupled Plasma Atomic Emission Spectrometry (ICP-AES), 1 Inductively Coupled Plasma Mass Spectrometry (ICP-MS) 2 and Atomic Absorption Spectrometry (AAS). 3 These techniques provide excellent detection limits but involve high-cost instrumentation, time-consuming sample preparation and highly trained personnel. Colorimetric and uorescence sensor devices are fast-growing technologies showing remarkable advantages over conventional techniques, such as fast response times, non-destructive analysis and remote operation, while attaining competitive performances in terms of detection limits, sensitivity, selectivity and reversibility by rational device design. [4][5][6][7][8] Fluorescent probes, which have found widespread use in biomedical applications, 9,10 are being increasingly studied for real-time and remote environmental monitoring such as detection of toxic metals in water and biological media. [11][12][13][14][15] In a molecular-type approach, a uorescent probe, which represents the sensitive part of the uorescent device, basically consists of a recognition (binding) moiety linked to or included into a uorescent species (usually an organic dye) whose emission properties, such as peak wavelength and quantum yield, are modied by the interaction with the analyte. The chemical nature of both the binding and uorescent units can be tailored to optimize the probe for any specic analyte and/or application.
Most uorescent probes are designed to work in solution. However, depending on the application, an improvement in the sensing performance can be obtained by covalently anchoring uorescent probes to different substrates, including biomolecules, 16,17 silica-based nanocomposites [18][19][20][21] and metal-organic frameworks. [22][23][24] Graing uorescent probes to silicon-based planar platforms represents a promising route towards sensor integration with silicon photonics technology and realization of Lab-on-Chip devices. 25,26 To this aim, the sol-gel technique based on organosilane derivatives 27,28 is one of the most suitable strategies for fabricating robust, low-cost and functional silicon-based platforms. In particular, the use of 3-aminopropyltriethoxysilane (APTES) as organosilane derivate, allows to insert amino groups capable to covalently link uorophores on silicon surfaces, via the formation of amides, ureas, thioureas and imines as linking functional groups. [29][30][31][32] Although step-by-step silanization of silicon substrates has already been investigated for the fabrication of uorescent platforms, 33-35 the potential of these platforms for the detection of trace metal ions in aqueous environment is yet to be demonstrated.
A fundamental study on silicon-based uorescent platforms for detection of trace metal ions in aqueous environment is herein presented. Fluorescein isothiocyanate (FITC) was covalently linked to APTES-prefunctionalized silicon substrates via thiourea formation to realize silicon-based on-chip devices integrating both recognition and uorescent units. On the one hand, FITC is highly suitable to serve as prototypical uorescent sensing unit because of its remarkable photophysical properties; 36-38 on the other hand, thiourea groups have been demonstrated effective in binding metal cations such as Cu II , Cd II , and Hg II in water, in fact enabling metal-ion uorescent sensors with competitive performances. 22,[39][40][41][42][43] Selectivity to copper(II), high sensitivity and low limit of detection were reported. Successful surface regeneration was achieved using ethylenediaminetetraacetic acid (EDTA) chelating agent, yielding insight into the mechanism responsible for metal binding.
Among heavy metal ions, copper plays a crucial role in different biological processes, 44 being essential for the functioning of many enzymes such as C oxidase (COX) and dopamine b-hydroxylase. 45 Its accumulation in the human body can be harmful and even lead to genetic disorders such as Wilson's disease, and neurodegenerative disorders including Alzheimer's and Parkinson's disease. [46][47][48] Consequently, EPA (US Environmental Protection Agency) and WHO (World Health Organization) have set the maximum copper level in drinking waters to 1.3 and 2 mg L À1 , respectively. Therefore, the search for innovative metal-ion sensing devices represents still a challenge in the environmental eld.

Results and discussion
Reagents and silicon surface modication steps are illustrated in Fig. 1. (100)-Oriented silicon chips were rst coated with ultrathin silica layers by (i) the sol-gel technique by using tetraethoxysilane (TEOS) as silica precursor (Scheme S1, ESI †) and (ii) the spin-coating deposition method (Step 1). Subsequently, amino groups were introduced on the Si@SiO 2 surface by using APTES (Step 2). Finally, FITC was covalently linked to the Si@SiO 2 @APTES surface via thiourea formation (Step 3).
Freshly prepared samples were investigated following a multivariate approach, namely, Contact-Angle (CA) measurements, Atomic-Force Microscopy (AFM), Reectance and Fluorescence Spectroscopy. This way, it was possible to monitor how each functionalization step modied the morphological and optical properties of the silicon surface. 49 Surface wettability tests were performed by water CA measurements. Data revealed a decreasing surface wettability upon amino-modication by APTES and subsequent FITC linkage. In fact, the mean CA, which was found to be 38.3(0.5) in the Si@SiO 2 chip, increased up to 54.4(0.7) aer silanization with APTES 50 and then up to 72.1(0.9) upon functionalization with FITC 51 (Fig. 2). These results could be easily rationalized as a decreased surface hydrophilicity due to the functionalization of the silica layer (bearing hydroxy groups, -OH) with the amino (-NH 2 ) groups of APTES and then with FITC, which contains hydrophobic aromatic rings (Fig. 1b). Measurements carried out on different points of the sample surface conrmed the homogeneous deposition of the layers. This conclusion was supported at the nanoscale by AFM height images, showing lack of topographic reliefs and root mean square (RMS) values of the surface roughness lower than 1 nm (Fig. S1, ESI †).
VIS reectance spectroscopy was used to evaluate the thickness of the layers deposited through the various functionalization steps. Measurements were performed in hemispherical geometry to collect both specular and diffuse reectance. Results are summarized in Fig. 3. Spectral reectance ratio R i (l)/ R iÀ1 (l) was calculated aer fabrication Step i. Silica deposition resulted in 10-15% decrease in reectance across the visible Step 1: surface silica layer deposition bearing hydroxy groups via the sol-gel method using TEOS (left); Step 2: amino-functionalization of the silica surface with APTES (centre, amino groups highlighted in blue).
Step 3: functionalization with FITC (right), leading to the formation of thiourea groups (highlighted in red). spectrum ( Fig. 3a), as the silica layer acted as an antireection coating on silicon. Knowing the index of refraction of silica from previous work done with similar sol-gel precursor concentrations and physical deposition parameters, 52 it was possible to calculate the thickness of the silica layer with high degree of accuracy. The estimated value of 28(1) nm implies that reectance was very sensitive to any increase in thickness caused by subsequent functionalization steps. In the actual experimental setup, this sensitivity resulted to be $1 nm. In the inset of Fig. 3a, the reectance ratio R/R 0 , where R (R 0 ) is the reectance of the Si@SiO 2 structure (bare silicon chip) at 700 nm, is plotted against the SiO 2 layer thickness.
The high sensitivity of the ratiometric reectance to the thickness of the APTES and FITC layers was readily deduced from the slope of the black line at the intersection with the blue and red vertical bars, depicting the APTES and FITC layers, respectively. Fig. 3b and c show that reectance remained practically unchanged upon functionalization of APTES and FITC, allowing to conclude that these functionalization steps were well-controlled and involved the deposition of monolayers. Indeed, a very small spectral dip was observed in the reectance ratio measured aer FITC deposition (Fig. 3c). This feature was found to be compatible with the optical absorption of a dense monolayer of FITC in the visible.
Furthermore, combined continuous-wave/time-resolved uorescence spectroscopy was applied to check for surface functionalization with FITC and gain information on the xanthene-centred uorescence efficiency upon FITC linkage to the silicon surface. The uorescence spectrum of the sensing device is shown in Fig. 4a. The main emission peak, centred at 540 nm (red curve), was readily assigned to the FITC xanthene unit, while comparative measurements on Si@SiO 2 bare chip (grey curve) allowed for attributing the short-wavelength shoulder, centred at 420 nm, to the silica uorescence spectrum. For reference, emission spectra were also collected for FITC dissolved in water and ethanol (Fig. 4b). FITC graing resulted in loss of uorescence efficiency, as deduced from the stretched exponential decay of the uorescence intensity with reduced lifetime as compared to that of the nearly monoexponential decay observed in solution (Fig. 4c).
A preliminary screening of the sensing performances was carried out on several metal cations (Al III , Cu II , Zn II , Ag I , Cd II , Hg II , Pb II ) in MOPS buffer solutions at pH ¼ 7.2, at a xed molar concentration of 1.3 Â 10 À4 mol L À1 (Fig. 5a). Remarkably, among the investigated ions, only Cu II resulted in a neat uorescence turn-off behaviour (see also the uorescence spectra in Fig. 5b) with a signicant response on a short time scale of the order of tens of seconds. Sensitivity toward Cu II was determined  by uorescence-intensity titration curves as a function of Cu II concentration. In the low concentration regime below 5 ppm, the decrease in uorescence intensity measured for increasing concentration was tted with a linear decay function, yielding a signal loss coefficient of 0.15 ppm À1 or, equivalently, 15% reduction at 1 ppm level (Fig. 5c). Stabilization of uorescence quenching was observed for Cu II concentration values larger than 15 ppm. Finally, regeneration tests carried out in EDTA water solution showed stabilization of the uorescence signal recovery level at $60% already aer the second regeneration cycle (Fig. 5d), thereby demonstrating good reversibility of the surface-analyte interaction.
Several Cu II sensing mechanisms could in principle be envisaged for the present hybrid platforms. Fluorescence sensing of Cu II by a rhodamine B derivative linked to silica nanoparticles through a thiourea group was previously shown to yield Cu II complexes exhibiting an optical absorption band partly overlapping with the uorophore emission band, thus acting as uorescence quencher via photoinduced Forster's resonance energy transfer (FRET). 18 However, no additional absorption bands were detected by ratiometric reectance aer device dipping into a Cu II aqueous solution (Fig. 3d).
Redox processes involving Cu II reduction to Cu I , thiourea oxidation to a disulde moiety and concomitant formation of Cu I -thiourea complexes were also demonstrated experimentally in solution. [53][54][55] In the present uorescent platforms, however, surface oxidation with formation of disulde (-S-S-) bonds would presumably cause irreversible modication of the surface, in contrast with regeneration tests by using EDTA. Thiourea-induced copper(II) reduction was therefore discarded as a possible metal-ion sensing mechanism.
Thiourea groups may indeed not be directly involved in Cu II sensing. Experiments performed on FITC dilute solutions in ethanol/water, where thiourea groups are not present (Fig. S2, ESI †), showed that Cu II can selectively interact with FITC, leading to a decrease in both absorbance and uorescence intensity without formation of new optical absorption bands. FRET-type processes were therefore ruled out. A new FITC-Cu II interaction mechanism was in turn suggested, which involves the formation of a uorescein lactonic species where p-conjugation interruption leads to vanishing of the xanthene-centred optical transition moment 56 and, hence, to uorescence quenching. This mechanism (Scheme S2, ESI †) was further supported by mass and tandem mass spectrometry data (Fig. S3, ESI †). A model-based analysis of the uorescence response for varying Cu II concentration is also reported (Fig. S4, ESI †).
It is worth discussing on the limit of detection (LoD) of the reported uorescent chips. LoD is dened as the analyte concentration that produces a response signal equal to a given threshold level, which is usually set to three standard deviations (s) of the signal (F/F 0 ) from its mean value. Although step-bystep characterization conrmed a good quality of the planar platforms, surface disorder appeared to result in non-negligible sample-to-sample uctuations of the uorescence intensity (Fig. 5b), whereas statistical noise and readout noise of the photodetection apparatus were found to be negligible. The resulting standard error bars of data points in Fig. 5c (3s ¼ 0.63 at [Cu II ] ¼ 2.1 ppm) currently set the Cu II LoD to 4.1 ppm, a value that is close to current limits for copper in drinking waters. The large sensitivity value of 0.15 ppm À1 could be exploited to greatly improve the LoD of the uorescent silicon platforms upon reducing chip-to-chip signal uctuations. 18 Another point of interest is represented by the uorescence efficiency that can be achieved upon uorophore linkage to the silicon substrate. The strong reduction reported for the FITC uorescence lifetime upon surface graing (180 ps against the 2.6 ns value found in water solution, Fig. 4c) hints to a strong sensitivity of FITC uorescence to surface loading, possibly arising from the sizable absorption-emission spectral overlap, which causes uorescence self-quenching through FITC-to-FITC photoinduced energy transfer. 37 This issue, oen encountered in bioanalytical applications of uorescence, opens up to the use of organic dyes with shorter critical (Foerster's) energy transfer distance and, hence, less sensitive to uorescence self-quenching.

Conclusions
Silicon-based uorescent chips for the detection of metal ions in water were reported, where the xanthene-based uorophore, FITC, was covalently linked to an amino-silanized silicon surface via thiourea formation. These solid-state hybrid platforms exhibited selectivity towards copper(II) with good detection limit, competitive sensitivity, and regeneration capability using a metal-chelating agent. An original FITC-Cu II reaction mechanism involving the formation of a lactonic uorescein species, exhibiting disappearance of the fundamental optical transition moment, is proposed as a possible cause of uorescence quenching. The sensing platform architecture, where recognition/uorescent units are integrated on a silicon chip via a layer-by-layer functionalization approach, is very versatile owing to the possibility of tuning the selectivity to other metal ions or different types of analytes by changing the uorophore or the anchoring group. This represents a viable strategy towards silicon-integrated uorescent devices for remote detection of a range of metal ions in water. In the present solid-state planar platforms with high dye loading, spatial uniformity and quantum efficiency of the uorophore are important issues that require further investigations before determining matrix effects, such as the role of pH and interfering ions, in real water samples.

Silica synthesis and deposition
The silica sol precursor was prepared by mixing TEOS, EtOH and distilled water under stirring at room temperature (RT). Here, F 0 refers to the initial fluorescence intensity (at 0 ppm) before the regeneration test started.
Subsequently, HCl was added to the sol and the mixture was maintained under stirring at 60 C overnight. The molar ratio of the starting solution was TEOS : H 2 O : EtOH : HCl ¼ 1 : 5 : 6 : 0.065. A diluted solution was prepared by mixing freshly prepared TEOS solutions with a proper amount of EtOH (volume ratio 1 : 10) in a closed vessel at RT. Silica lms were deposited on 100-oriented silicon substrates cut into $15 Â 15 mm 2 pieces for morphological and optical characterization, and $7 Â 7 mm 2 pieces for sensing measurements. Prior to coating, the silicon substrates were cleaned with water soap (in ultrasonic bath at 50 C for 15 min), distilled water, acetone, and nally rinsed with isopropanol and dried in a nitrogen ow. Ultrathin ($30 nm) silica lms were obtained by spin coating at 7000 rpm for 30 s. Aer deposition, samples were dried at RT for 24 h.

Silica@APTES functionalization
The functionalization of the ultrathin silica lms on silicon substrates with amino groups was done by immersion of the substrates in a 1 mM APTES/hexane solution for 10 s. The lms were then rinsed with hexane followed by acetone to remove the excess of APTES.

Silica@APTES@FITC functionalization
A FITC/EtOH mixture was prepared by dissolving 8 mg of FITC in 10 mL of EtOH. Silica@APTES lms on silicon substrates, vide supra, were then immersed in the FITC solution for 30 min. The reaction was performed at RT in dark conditions. Upon completion of the reaction, the FITC-functionalized silicon substrates were rinsed with EtOH and acetone to remove the excess of FITC, and then kept in 20 mL of milli-Q water for 20 min before characterization measurements.

Morphological characterization
A NT-MDT Solver-Pro atomic-force microscopy (AFM) instrument was used to study the topography of the sample surface. AFM measurements were performed at 0.5-1 Hz scan speed in semicontact mode in air. Topography image analysis and calculation of surface roughness were performed by using WSxM 5.0 Develop3.2 soware. 57 The measurements were performed on at least three different points of the same sample to assess the uniformity of the layers.

Surface wettability experiments
Water contact angle (CA) measurements were performed at 22 C by using a Kruss drop shape analyzer (DSA 30S) and analyzed by the Kruss Advance soware. A sessile drop method was used to measure the contact angle of a 1 mL distilled water drop. The measured CA was the average between the le and right contact angles. The measurements were performed on at least three different points of the same sample to assess the uniformity of the layers.
Optical spectroscopy and uorescence lifetime measurements Reectance spectroscopy measurements were performed under sample direct illumination in a dual-beam spectrophotometer (Agilent Technologies Cary 5000 UV-Vis-NIR) equipped with a diffuse reectance accessory. Fluorescence spectra for samples in air and in aqueous 3-(N-morpholino)propane sulfonic acid (MOPS) buffer in a quartz cuvette were measured under 355 nm irradiation of the samples by a passively Qswitched powerchip laser (Teem Photonics PNV-M02510) operating in pulsed regime (350 ps pulses, 1 kHz repetition rate). The uorescence was spectrally resolved using a single-grating spectrometer (Princeton Instruments Acton SpectraPro 2300i) and acquired by a thermoelectrically cooled Vis CCD camera (AndorNewton EM ). Fluorescence lifetime were measured in air under sample irradiation at 350 nm by an optical parametric amplier (Light Conversion TOPAS-C) pumped by a regenerative Ti:Sapphire amplier (Coherent Libra-HE), delivering 200 fs-long pulses at 1 kHz repetition rate. The uorescence was spectrally dispersed using a single-grating spectrometer (Princeton Instruments Acton SpectraPro2300i) and acquired by a Vis streak camera (Hamamatsu C1091). Pump uence was kept below 50 mJ cm À2 per pulse in all uorescence experiments to prevent sample degradation.

Sensing measurements
Sensing performances towards different metal cations were assessed by uorescence spectroscopy with the silicon-chip sensors placed in a 1 cm-thick quartz cuvette. The cuvette was then lled with 1500 mL of aqueous MOPS buffer (70 mM, pH ¼ 7.2) using fresh distilled water puried by a Milli-Q system (Millipore), and 60 mL of 0.1 mM water solution of various perchlorate salts (Al III , Cu II , Zn II , Ag I , Cd II , Hg II , Pb II ) was added. Sensitivity towards Cu II ions was evaluated through uorescence titration experiments in the 0-120 mL range. A waiting time of 3 min was set before each measurement to ensure that the surface-analyte interaction had reached equilibrium. Surface regeneration aer sensing tests was accomplished by sample sonication in 3 mL of EDTA solution (0.1 M) for 10 min. Three generation cycles were performed on each sample.

Mass spectrometry
Mass spectra were recorded using a triple quadrupole QqQ Varian 310 MS mass spectrometer using the atmosphericpressure Electrospray Ionization (ESI) technique. FITC solutions (20 mL) were injected into the ESI source by a Rheodyne® model 7125 injector connected to a HPLC Varian 212 LC pump, with a 50 mL min À1 methanol ow. Experimental conditions: Dwell time 2 s, needle voltage 3000 V, shield voltage 600 V, source temperature 60 C, drying gas pressure 20 psi, nebulizing gas pressure 20 psi, detector voltage 1600 V. Mass spectra were recorded in the 100-600 m/z range. Collision-Induced Dissociation (CID) tandem mass (MS/MS) experiments were performed using argon as the collision gas (1.8 psi). Collision energy was varied from 20 to 40 eV.