A silica nanoparticle supported fluorescence “turn-on” fluoride ion sensing system with tunable structure and sensitivity

Abstract

Amino-silane modified silica nanoparticles (SNP-APTMS and SNP-TMSPEDA) doped with silyl-ether protected fluorescein isothiocyanate (FITC-OSMDBT) were synthesized via a sol–gel method in a simple three-step reaction to give fluoride ion probes SNP-TMSPEDA-FITC-OSMDBT (sensor A) and SNP-APTMS-FITC-OSMDBT (sensor B). The sensors were characterized using FT-IR, EDX, TGA, and fluorescence spectroscopy. It was observed that, at fluoride ion and FITC-OSMDBT mole equivalent titration ratio of x : x, the emission spectra of both sensors were similar when x = 1.0; however, at x = 3.0, 6.0, and 9.0, sensor B displayed 3 fold the emission intensity of sensor A. This phenomenon was attributed to non-radiative emission energy transfer mechanisms, which were controlled by the FITC-OSMDBT loading. The steric hindrance difference present on N-[3-(trimethoxysilyl) propyl] ethylenediamine (TMSPEDA) and (3-aminopropyl)trimethoxysilane (APTMS) ensured that different amounts of FITC-OSMDBT were loaded on each sensor. At x = 3.0, 6.0, and 9.0, the emission intensity increased linearly with [F⁻] in DMSO; thus, from the working curve of sensor A, the fluoride ion detection range of sensor B could be calculated and vice versa. The structure of the sensing system being proposed is simple, sensitive to F⁻, and may prove useful with respect to the development of fluoride anion sensors with structures which can be easily modified to produce sensors with varying detection ranges.

---

1. Introduction

Fluoride ions, like many other anions, play a fundamental role in a wide range of chemical and biological processes.^1–4 They have unique chemical properties and are widely used in toothpaste and pharmaceutical agents for the prevention of dental caries and enamel demineralization while wearing orthodontic appliances, and treatment of osteoporosis.⁵ Fluoride is also important in industrial applications and transformations, especially in steel making and aluminum refining, and it is a well established reagent in organic synthesis. Novel applications of fluoride have been discovered in the fields of ion batteries and in F-PET imaging. However, it benign health effect is up to 1 ppm in living organisms, and it seems to be toxic at higher doses.^6,7 High concentration of fluoride in the environment and drinking water has been related to the occurrence of several types of pathologies in humans, such as osteoporosis, neurological and metabolic dysfunctions, and recently cancer.⁸ This obvious health concerns have prompted scientists to develop analytical techniques which can accurately detect and quantify fluoride ions.

Among several widely used fluoride-sensing and -detecting techniques, including the electrode method,⁹ F NMR analysis,¹⁰ and colorimetric (UV) and fluorescence sensing, electrochemical systems are the most well-established. However, this approach has major disadvantages associated with the need for fragile instrumentation and time-consuming manipulations.¹¹ In addition, F NMR spectroscopy can be used reliably to detect only micromolar levels of fluoride. Moreover, neither the electrochemical nor the NMR approach can be miniaturized for use in studying biological processes in vivo.¹² To this end, other detection techniques, particularly those that are based on fluorescence and color change, have been intensely studied. So far, they have shown to have great promise considering their high selectivity, sensitivity, response time, and ease of use in both intra- and extra-cellular environments. These types of sensor usually consist of three moieties: fluorophore/chemophore, spacer and receptor. Their optical signal changes upon binding to F⁻ and their common adopted sensing strategy involves supramolecular interactions such an anion–π interactions, hydrogen bonding and Lewis acid/base interactions.^13–16 Several fluorescent fluoride chemosensors and chemodosimeters works have been developed and reported in literature.^17–23 Fluorescence chemodosimeters are reaction-based: they utilize F⁻ promoted cleavage reactions and hence, they are more selective towards F⁻ than chemosensors;¹² besides, they work better in organic and/or aqueous solutions.²⁴ Thus, they are preferred to fluorescence chemosensors.

For several years now, scientists have focused on developing ultra sensitivity and selective, and highly responsive fluorescent fluoride chemodosimeters using mainly organic materials to detect fluoride ions in organic media.^18–22 But, with the challenge of sensing F⁻ in organic media seemingly solved, attention has now been shifted to sensing F⁻ in aqueous media. Recent reports indicate that a lot of progress has been made in this field as well with some groups reporting fluoride fluorescent chemodosimeters with amazing sensitivity, selectivity, and response time values.^24,25 Another area that has received a lot of attention is the use of organic–inorganic materials to recognize and sense fluoride. Receptors immobilized on inorganic materials such as SiO₂, Al₂O₃, and TiO₂ have many advantages such as organic–inorganic hybrids nanomaterials can be recycled through suitable chemical treatment, and functionalized nanomaterials combined with fluorophores display highly selective and sensitive fluorescence or absorption changes because of their large surface area and well-defined pores.¹² A few examples of such F⁻ detection systems exist in literature: a method of detecting fluoride in water based on the specific reaction of fluorhydric acid with an MCM-41 solid functionalized with fluorescent or colorimetric signaling unit has been reported.²⁶ Furthermore, a novel covalently bonded luminescent hybrid material and its spectrophotometric anion-sensing properties were described in literature.²⁷

Despite the advances in fluorescent fluoride sensor development, the fact that most of these sensors are developed to be a ‘one-solution-fit-all’ kind of material, with the best sensitivity, selectivity, and response time, makes their use both good and bad: For example, drinking-water regulatory bodies (e.g. EPA) need highly selective and sensitive F⁻ sensor with a narrow detection range since they only have to check if the concentration of F⁻ is below or above 2 ppm; however, for laboratory reactions and industrial applications (e.g. steel-making, batteries, and toothpaste manufacturing), higher concentrations of fluoride are used, and hence, a sensor with a wider detection range is preferred. It therefore seems economically unwise to focus on developing only ultra sensitive and selective sensors which on the average is bound to be expensive. To the best of our knowledge, no work has been done to investigate organic–inorganic fluorescence “turn-on” F⁻ sensing systems having a structure which can be easily modified to vary it detection range. Most of the chemodosimeters reported in literature make use of organic fluorescent dyes to transduce the chemical interaction between it and F⁻ into optical signals which are detected by a UV-vis and/or fluoro-spectrophotometer. The spectra change of the dye serves as the basis on which it sensitivity and selectivity, and response time are evaluated.

Herein, silyl-ether protected fluorescein isothiocyanate (FITC-OSMDBT) molecules were covalently immobilized on silica nanoparticles (SNP) surfaces separately functionalized with two different amino-silane molecules to form sensors A and B. The effect of FITC-OSMDBT percentage (%) loading on the emission spectra of each sensor in the presence of F⁻ was investigated separately, and after which the two results were correlated to show the relationship between their emission intensity increment and detection range. Using this correlation, we hope to show that, by carefully controlling the FITC-OSMDBT loading using different coupling agents, the detection range of the sensor may be varied extensively. The selectivity and response time of both sensors in the presence of Cl⁻, Br⁻, I⁻, and NO₃⁻ were also determined and compared with that of F⁻.

2. Experimental

2.1. Materials

Silicon dioxide nanoparticles (SNP) from Sigma-Aldrich with a particle size range of 10.0–20.0 nm were used for this experiment. 3-Aminopropyltrimethoxysilane (APTMS) and N-[3-(trimethoxysilyl)propyl]ethylenediamine (both from Sigma-Aldrich, 97.0%) were used to modify the SNP and graft fluorescein 5(6) isothiocyanate (FITC) molecules (Sigma-Aldrich, ≥ 90.0%). The chemical structures of the amino-molecules are illustrated in Table 1. The other chemicals used included triple distilled water, anhydrous acetonitrile (99.8%), anhydrous toluene (Acros, 99.8%), anhydrous DMSO (Fisher, 99.9%), anhydrous DCM (Acros, 99.8%), absolute ethanol (Merck KGaA, 99.5%), tetrabutylammonium fluoride hydrate (98%), tetrabutylammonium nitrate (97%), tetrabutylammonium iodide (98.0%), tetrabutylammonium chloride hydrate (98.0%), tetrabutylammonium bromide (≥98.0%), triethylamine (≥99.0%), and tert-butyldimethylsilylchoride (TBDMSCl), (97.0%) (all from Sigma-Aldrich). Unless otherwise stated, all the chemicals used were of analytical grade and were used as received without further purification.

Table 1 Chemical structure of the alkoxysilane molecules used to modify the surface of SNP

Chemical name	Designation	Chemical structure
3-Aminopropyltrimethoxysilane	APTMS
N-[3-(Trimethoxysilyl)propyl]ethylenediamine	TMSPEDA

---

2.2. Characterization

Transmittance infrared spectra of the various attached functional groups were collected using a Varian FTS 2000 Fourier transform infrared (FT-IR) spectrometer. The weight change after each synthesis step was analyzed via TGA measurements using a Perkin-Elmer TGA-7 instrument, a standard platinum crucible, and a sample size of 20.0 mg. The samples were heated at a rate of 10.0 °C min⁻¹ from room temperature to 1000.0 °C under N₂ flowing at 50.0 ml min⁻¹. Elemental composition analysis was done with dispersive X-ray spectrometry (EDX) with a scanning electron electrode at 20 kV. Lastly, surface area and pore diameter changes were measured via BET and BJH analysis.

The anion solution (0.1–9.0 mole equiv. with respect to the amount of FITC-OSMDBT anchored on SNP) was titrated against different concentrations of sensors A and B and mixed for 5.0 min. Their absorption and emission changes were then measured using a Cary 100 a UV-Visible spectrophotometer and a Hitachi a F-4600 fluorescence spectrometer respectively.

2.3. Synthesis of silica nanoparticle supported fluorescence “turn-on” sensor

2.3.1. Silanization of SNP with amino-functional trimethoxysilanes. The silanization method used is essentially similar to that reported in literature,²⁷ except that, the ratio of SNP to TMSPEDA (or APTMS) was changed. Briefly, 0.2 g of SNP was mixed with 30.0 ml of TMSPEDA (or APTMS) solution (2.0% v/v in toluene) and agitated under reflux at 110.0 °C for 24.0 h in N₂ atmosphere. The modified silica nanoparticles were isolated and purified by centrifugation/redispersion processes in toluene (for 30.0 min at 16 000 rpm, 6 times) to remove unreacted and loosely bound TMSPEDA (or APTMS). Finally, the purified solid was cured at 110.0 °C for 12.0 h in a conventional oven after which it was cooled in a desiccator. SNP-TMSPEDA (or SNP-APTMS) was used to denote the cured solid.

2.3.2. Immobilization of FITC on silanized SNP. 0.5 g of SNP-TMSPEDA (or SNP-APTMS) was dissolved in 30.0 ml of absolute ethanol and stirred for 10.0 min, after which a calculated amount of FITC [1.5 fold excess relative to the amount of grafted TMSPEDA (or APTMS)] was added to the suspension. The resulting mixture was stirred for 24.0 h in the dark at room temperature. The red product was collected and washed (for 20.0 min at 16 000 rpm, 4 times) copiously with absolute ethanol until the filtrate was clear of color. Finally, the purified solid was oven dried at 80.0 °C for 1 h and was denoted as SNP-TMSPEDA-FITC (or SNP-APTMS-FITC).

2.3.3. Silylation of phenol groups on FITC with TBDMSCl. A mixture of SNP-TMSPEDA-FITC (or SNP-APTMS-FITC) (0.2 g), anhydrous DCM (10.0 ml), and triethylamine (78.0 mg) in a beaker was stirred under nitrogen atmosphere for 30.0 min while the beaker was immersed in a mixture of cracked ice and water. 0.6 mg of TBDMSCl in 5.0 ml DCM was added drop-wise into the reaction mixture after which it was stirred for an additional 18.0 h. The light yellow product was filtered and washed several times with acetonitrile and water. The purified sample was then dried under vacuum for 48.0 h at room temperature. The final product was denoted as SNP-TMSPEDA-SFITC-OSMDBT (sensor A) [or SNP-APTMS-FITC-OSMDBT (sensor B)].

3. Results and discussion

3.1. Surface group analysis

Sensors A and B possessing FITC and Si–O as fluorescent unit and fluoride sensing site respectively were synthesized as shown in Scheme 1. FT-IR analysis was used to follow the synthesis reaction of each sensor. In both cases, certain relevant peaks were identified and used to varify each reaction step. Fig. 1 shows the peaks of sensors A and B as well as that of their intermediate products. After the silanization reaction, the silanol peak at ∼956.0 cm⁻¹ on SNP vanished, while peaks belonging to –CH₂ (∼2924.0 cm⁻¹) and –NH₂ (∼1543.0 cm⁻¹) groups appeared.²⁷ This confirmed the condensation reaction between the silanol groups and alkoxysilanes. The N–H stretching peaks for both APTMS and TMSPEDA was overshadowed by the much broader and intense –OH peak at ∼3500.0 cm⁻¹, the sharp and strong Si–O–Si stretching peak at ∼1080.0 cm⁻¹ was observed in all the samples, which indicated that the main structure of SNP did not change.

Scheme 1 Synthesis route to sensor A and sensor B.

Fig. 1 FTIR spectra of SNP, sensor A, sensor B, and their intermediate products.

New peaks could be seen after the FITC immobilization reaction: the peaks from ∼1620.0 cm⁻¹ to ∼1303.0 cm⁻¹ belong to the aromatic rings of FITC, while that from ∼914.0 cm⁻¹ to ∼902.0 cm⁻¹ belong to the aromatic –CH bending vibrations.²⁸ The lactone peak was over shadowed by the benzene ring peaks. Likewise, the thione (–C=S) peak was masked by the much dominant Si–O–Si peak. The C–H stretching vibrations at ∼2939.0 cm⁻¹ became more prominent after the silylation reaction which indicated the presence of CH₃ groups belonging to the TBDMSCl. In like manner, the Si–O–Si stretching peak at ∼1080.0 cm⁻¹ also increased in intensity. Furthermore, strong fluorescence was not observed when solutions of both sensors were irradiated with 520.0 nm light, which suggested that FITC-OSMDBT had been successfully protected. Base on these results, it was assumed that the moieties were covalently bonded to SNP.

EDX analysis showed weight% and atomic% of various elements (C, O, Si and S) in sensors A, and B; but, only elements O and Si were present in SNP (Table 2), which confirmed that the modification reaction was successful. Furthermore, the weight% and atomic% of elements C, Si, and S varied strongly among the sensors: sensor A had a higher weight% and atomic% of Si and S than sensor B, while sensor B had a higher C content. The higher S content meant that, sensor A had a higher FITC loading on its surface than sensor B, and thus more FITC-OSMDBT molecules as further indicated by it high Si content. The higher C content in sensor B reflects it TMSPEDA content.

Table 2 EDX for SNP, sensor A, and sensor B

sample	Weight%				Atomic%
	Carbon	Oxygen	Silicon	Sulfur	Carbon	Oxygen	Silicon	Sulfur
SNP	—	69.08	30.92	—	—	79.68	20.32	—
Sensor A	33.7	48.91	17.05	0.32	43.31	47.17	9.37	0.15
Sensor B	51.90	40.94	6.93	0.23	60.57	35.87	3.46	0.10

---

The effect of surface modification on the surface area and pore diameter of SNP, sensor A, and sensor B were studied via nitrogen adsorption–desorption isotherms and Barrett–Joyner–Halenda (BJH) pore diameters analysis as shown in Fig. 2. Results from Fig. 2 (shown in Table 3) revealed that, before modification, SNP had a Brunauer–Emmett–Teller (BET) surface area of 674.65 m² g⁻¹ and a pore volume of 0.77 cm³ g^−1. In contrast, sensors A and B had lower BET surface areas (98.81 and 161.05 m² g⁻¹, respectively) and pore-volume (0.18 and 0.24 cm³ g⁻¹, respectively). The BJH pore diameter was narrow for all the samples: 4.58 nm for SNP, and 5.97 and 7.19 nm for sensors A and B respectively. The smaller surface area and pore diameter of the sensors compared to SNP suggest that the modification reaction was successful, and also the higher surface loading on sensor A compared to sensor B.

Fig. 2 (a) Nitrogen adsorption–desorption isotherms and (b) Barrett–Joyner–Halenda (BJH) pore diameters of SNP, sensor A, and sensor B.

Table 3 Barrett–Joyner–Halenda pore diameter (d_BJH), Brunauer–Emmett–Teller surface area (S_BET), and pore volume (V_p) of SNP, sensor A, and sensor B

Material	dBJH (nm)	SBET (m^2 g^−1)	Vp (cm^3 g^−1)
SNP	4.58	674.65	0.77
Sensor A	5.97	98.8	0.18
Sensor B	7.19	1161.05	0.24

---

3.2. Determination of FITC-OSMDBT loading density via TGA

Theoretically, the amount of FITC-OSMDBT immobilized on each sensor should determine the concentration of F⁻ it can detect, since it is this part of the sensor which reacts with F⁻ to induce fluorescence change. The amount of FITC-OSMDBT per gram of each sensor was determined using the weight loss due to TBDMSCl calculated from TGA analysis. The thermogram of sensor A (Fig. 3) indicates that, the pure SNP lost moisture and underwent dehydroxylation in a single step process between 120.0 °C and 1000.0 °C, just as reported in literature,²⁹ with an accompanying weight loss of about 4.14%. For the SNP-APTMS material, it first suffered a loss of about 5.84% due to moisture and then, as expected, decomposed in two steps: the first loss of about 8.24% (associated with the loss of APTMS) occurred between 120.0 °C and 470.0 °C, while the second loss was about 5.17% (associated with the loss of intraglobular hydroxyl groups) occurred between 470.0 °C and 1000.0 °C. It has been reported that temperatures above 650.0 °C are required to remove all intraglobular hydroxyl groups;³⁰ hence, it was assumed that all loses above 650.0 °C was due only to dehydroxylation. For the SNP-APTMS-FITC sample, the weight loss due to moisture was about 3.94% after which it decomposed, as expected, in three steps: the first weight loss of about 1.79% occurred between 100.0 °C and 230.0 °C, the second ∼8.44% occurred between 230.0 °C and 450.0 °C, and the third ∼10.89% occurred between 450.0 °C and 1000.0 °C. The first loss was attributed to APTMS because it has a lower boiling point compared to FITC (92.0 °C and 400.0 °C respectively). Lastly, the SNP-APTMS-FITC-OSMDBT material also suffered an initial loss of 4.14% due to moisture; but, it decomposed in only two steps afterwards instead of the expected four. The first was between 120.0 °C and 630.0 °C, and the second was between 630.0 °C and 1000.0 °C. The weight loss at this first temperature range was 20.22%, which was higher than the 10.22% (1.79% + 8.44%) loss recorded for APTMS and FITC in sample SNP-APTMS-FITC. Assuming that no leaching occurred during the heating process, the OSMDBT amount was estimated as 10.00% (20.22–10.22%); hence, the FITC-OSMDBT amount was 0.66 mmol g⁻¹.

Fig. 3 TGA results showing the thermogram of (a) sensor A and (b) sensor B, and their intermediate products.

The thermogram of sensor B and it intermediate products suggested that the organic moieties on sensor B underwent similar degradation steps as those on sensor A, with the exception of the amount of weight loss. The SNP is seen to undergo the same two step decomposition steps: the first due to moisture loss and the second due to dehydroxylation. Similarly, the SNP-TMSPEDA material decomposed in a two step reaction after the initial moisture loss. The loss due to TMSPEDA (which occurred between 130.0 °C and 450.0 °C) was 13.29% compared to the 8.24% recorded for APTMS. This was expected since TMSPEDA is bulkier than APTMS. However, the FITC amount on SNP-TMSPEDA-FITC was lower than that on SNP-APTMS-FITC: the mole per gram was 0.058 : 0.084 respectively. This suggested that, the weight of immobilized TMSPEDA was higher than that of APTMS; but, the APTMS molecules were greater in number than TMSPEDA. Lastly, SNP-TMSPEDA-FITC-OSMDBT underwent two decomposition steps after it initial moisture loss of 3.28%. The amount of OSMDBT immobilized was calculated to be 7.40%; hence, the FITC-OSMDBT amount was 0.49 mmol g⁻¹. Table 4 puts the results into perspective.

Table 4 TGA results of the number of moieties loaded on sensor A and sensor B and their intermediate products

Material	Moiety (X)	Reaction step	X (mmol g^−1)	(X^a nm^−2)
a Based on the SNP surface area of 674.65 m^2 g^−1.
SNP	OH	—	2.40	2.14
SNP-APTMS	APTMS	Silanization	0.82	0.73
SNP-TMSPEDA	TMSPEDA		1.33	1.18
SNP-APTMS-FITC	FITC	FITC immobilization	0.84	0.75
SNP-TMSPEDA-FITC	FITC	FITC immobilization	0.58	0.52
Sensor A	TBDMSCl	Silylation	0.66	0.60
Sensor B			0.49	0.40

---

3.3. Absorbance change and fluorescence turn-on sensing for F⁻ in DMSO

The TGA results clearly indicated that sensor A has higher FITC-OSMDBT loading than sensor B. Next up, we investigated how the varying FITC-OSMDBT loading amounts affected the emission and absorption spectra of each sensor in the presence of different amounts of F⁻. The ‘effect’ on each sensor for a given [F⁻] was quantified in terms of emission intensity change and then compared to find any useful correlation between them. This correlation was then related to the [F⁻] to find the detection range of each sensor.

To begin with, low concentrations of F⁻ (0.0 to 1.0 mole equiv.) were titrated against each sensor solution containing 1.0 mole equiv. of FITC-OSMDBT in DMSO. The resulting fluorescence (Fig. 4) and absorbance spectra (Fig. 5) indicated that, both sensors exhibit very weak absorption and emission intensity in the absence of F⁻; however, with the addition of the fluoride anions, the absorbance and fluorescence intensity of both sensors gradually increased, and then leveled off at [F⁻] = 50.0 mM (1.0 mole equiv. F⁻). The fluorescence ‘turn-on’ signaled the successful reaction between F⁻ and Si–O. Working curves established by plotting the increase in emission intensity at 542 nm vs. [F⁻] (insert of Fig. 5b) indicated that, the emission intensity increment for both sensors were nearly the same at every [F⁻]. Which suggested that at low [F⁻], the emission spectra of sensor A and B were independent of FITC-OSMDBT loading amount. Also, the emission intensity increment of both sensors varied linearly with [F⁻] between 10.0 mM to 40.0 mM (insert of Fig. 5a); thus, this concentration range was chosen to be the detection range of both sensors. An interesting emission spectra pattern was observed when 3.0 mole equiv. F⁻ was titrated against three solutions of each sensor containing 3.0, 6.0 and 9.0 mole equiv. of FITC-OSMDBT as depicted in Scheme 2.

Fig. 4 Fluorescence spectra of (a) sensor A (16.7 mg ml⁻¹) and (b) sensor B (20.4 mg ml⁻¹) in the presence of different amounts of TBAF in DMSO.

Fig. 5 UV-vis absorption spectral changes in (a) sensor A (16.7 mg ml⁻¹) and (b) sensor B (20.4 mg ml⁻¹) upon titration with TBAF (0.0 to 1.5 mole equiv.) in DMSO.

Scheme 2 Schematic illustrations of the titration of 3.0 mole equiv. F⁻ against sensor solutions containing 3.0, 6.0, and 9.0 mole equiv. FITC-OSMDBT molecules. The insert shows their emission intensity values measured at 542 nm.

For sensor A, all 3 titrations produced different emission intensity values which increased consecutively in the order n, 2n, and 3n for the 3.0, 6.0, and 9.0 mole equiv. respectively as shown in the insert of Scheme 2. We therefore assumed that, for every 3 mole equiv. FITC-OSMDBT available for F⁻ detection, only one reacted with 1.0 mole equiv. F⁻ and showed emission intensity corresponding to 1.0 mole equiv. F⁻. The fact that the value of n (54 983.85 a.u.) was approximately equal to that produced by 1.0 : 1.0 mole equivalent titration between fluoride ion and FITC-OSMDBT (54 189.50 a.u.) meant that the above assumption might be true. Besides, the emission intensity increment varied linearly with [F⁻] in all three titration, a fact which also supported the assumption. Based on this linear relationship, the detection range for the 3.0, 6.0, and 9.0 mole equiv. FITC-OSMDBT solutions was calculated as to be: 0.0–50.0 mM, 0.0–100.0 mM, and 0.0–150.0 mM respectively.

Sensor B, on the other hand, produced approximately the same emission intensity value (m) for all three titrations (insert of Scheme 2), and this value was approximately equal to 3n. Unlike sensor A, this observation meant that all 3.0 mole equiv. FITC-OSMDBT were able to detect all 3.0 mole equiv. of F⁻; and thus, the detection range of sensor B was assumed to be ≈3.0 folds that of sensor A for solutions of both sensors containing the same equiv. amount of FITC-OSMDBT. The close correlation between 3n and m corroborates the earlier assumption we made regarding [F⁻] and intensity in sensor A. The detection range of sensor B was calculated to be 0.0–150.0 mM for all three titrations.

To further test the validity of our assumption, the above experiment was repeated by titrating 3.0, 6.0, and 9.0 mole equiv. F⁻ solutions against their corresponding sensor solutions containing 3.0, 6.0, and 9.0 mole equiv. FITC-OSMDBT respectively. Their emission spectra (insert of Fig. 6a) clearly shows the big difference in emission intensity between the sensors: To our surprise, Sensor A produced emission intensity values in the order n, 2n, and 3n respectively, just like before. However, the emission values of sensor B increased in the order m, 2m, and 3m respectively as shown in the Fig. 6a. A comparison between the emission values revealed that, m ≈ 3n, 2m ≈ 6n, and 3m ≈ 9n: A result which confirms our earlier assumption. We could therefore conclude that for a given solution of sensor A and sensor B containing equimolar amounts of FITC-OSMDBT, the F⁻ detection limit of sensor B is thrice (3×) that of sensor A. Also, the emission intensity values of m, 2m, and 3m varied linearly with their corresponding [F⁻]. The detection range of sensor B was predicted from the working curve constructed for sensor A, shown in the insert of Fig. 6b. From this curve, the F⁻ detection range values were calculated as follows: For sensor A; 0.0–50.0 mM, 0.0–100.0 mM, and 0.0–150.0 mM respectively; and sensor B; 0.0–150.0 mM, 0.0–300.0 mM, and 0.0–450.0 mM respectively. A repeat of the experiment with other mole equiv. titration ratios x : x (x = 2.0, 4.0, and 5.0) of F⁻ and FITC-OSMDBT failed to give emission intensity values which could correlate the concentration and detection range of both sensors. Fig. 7a shows how erratic the emission intensity values for both sensors changed with concentration.

Fig. 6 (a) Fluorescence intensity increment [F − F₀] in sensor A [(A) 50.0 mg ml⁻¹, (B) 100.0 mg ml⁻¹, and (C) 150.0 mg ml⁻¹] and sensor B [(D) 61.2 mg ml⁻¹, (E) 122.4 mg ml⁻¹, and (F) 183.7 mg ml⁻¹] at 542 nm upon titration with 150.0 mM, 300.0 mM, and 450.0 mM F⁻¹ respectively. The insert shows the emission spectra for sensor A and sensor B at F⁻ : FITC-OSMDBT equiv. mole titration ratio of x : x (x = 3.0, 6.0, and 9.0). The spectra were recorded 15.0 min after addition of F⁻. And (b) Plot of emission intensity (λ_em = 542 nm) verses TBAF concentration using values from A, B, and C.

Fig. 7 (a) Fluorescence intensity increment [F − F₀] in sensor A [(A) 33.0 mg ml⁻¹, (B) 66.1 mg ml⁻¹, and (C) 83.3 mg ml⁻¹] and sensor B [(D) 40.8 mg ml⁻¹, (E) 81.6 mg ml⁻¹, and (F) 102.1 mg ml⁻¹] at 542 nm upon titration with 100.0 mM, 200.0 mM, and 250.0 mM F⁻¹ respectively. The spectra were recorded 15 min after addition of F⁻. (b) Emission intensity increment (λ_em = 542 nm) in sensor A (16.7 mg ml⁻¹) and sensor B (20.4 mg ml⁻¹) at different times after addition of F⁻ (50.0 mM) in DMSO.

In literature, a well known mechanism reported for F⁻ sensing with similar sensors involves an irreversible reaction between F⁻ and the Si–O bond of the sensors which generates a transducing fluorophore.^24,25,33 Our sensor showed a similar pattern, so, it was assumed that a similar reaction took place (as shown in Fig. S1†). This mechanism is, nonetheless, insufficient to explain the unusual pattern observed in sensors A and B. Besides, the TGA results indicated that sensor A has a higher percentage loading of FITC-OSMDBT than sensor B, hence as reported in literature,³¹ the former should exhibit greater fluorescence intensity than the latter; but, it did not.

A detailed study focused on the orientation, size, and distance dependence of energy transfer mechanisms amongst the conjugated parts of the receptor before and after it reaction with fluoride anions is necessary to uncover the complete picture. We are currently working on another paper which may aid us in this direction. With that said, we made the following assumptions regarding the probable underling mechanism based on the results obtained in our studies and some reported results in literature.

The absorption and emission spectrum of as-prepared SNP-TMSPEDA-FITC was studied with and without fluoride anions. The spectra obtained (as shown in Fig. S2†) were identical, clearly indicating that F⁻ did not have any influence on the spectra of SNP-TMSPEDA-FITC. The result was however different for SNP-TMSPEDA-FITC^a generated after SNP-TMSPEDA-FITC-OSMDBT titration with fluoride ions (shown in Fig. S1†). As shown in Fig. S2,† the absorption and fluorescence spectra of SNP-TMSPEDA-FITC^a showed a bathochromic shift and reduced intensity.

The bathochromic shift and reduced intensity may be explained by considering the entities present in the titration solution. As noted earlier, F⁻ has no effect on the spectrum of SNP-TMSPEDA-FITC and by extension on that of SNP-TMSPEDA-FITC^a as well; however, it appears unreacted SNP-TMSPEDA-FITC-OSMDBT molecules do. The FT-IR spectra of TMSPEDA-FITC^a and as-prepared SNP-TMSPEDA-FITC-OSMDBT (shown in Fig. S3†) appear to be similar which suggest the presence of unreacted SNP-TMSPEDA-FITC-OSMDBT molecules on the receptor even after the cleavage reaction with F⁻. An example of changes in the lifetime of dye molecules as well as the rate of their nonradiative energy transfer while in the vicinity of other entities it interacts with have been reported in literature.³² Perhaps, the “new” emission spectrum of SNP-TMSPEDA-FITC^a overlapped with the absorption spectrum of the conjugated segment in unreacted SNP-TMSPEDA-FITC-OSMDBT or vice versa to cause non-radiative emission energy transfer (nEET) interactions between the two species. Hence, the observed reduction in fluorescence intensity and bathochromic shift.

Furthermore, nEET is known to be governed by coulombic or/and dipole interactions.³² In DMSO, SNP-TMSPEDA-FITC^a is dianionic,²⁵ and Si–O bonds are reported to have a dipole moment.³⁴ These two factors could have therefore induced energy transfer interactions between the two entities and lead to nEET.

With a higher density of SNP-TMSPEDA-FITC-OSMDBT molecules, sensor A is likely to have more unreacted SNP-TMSPEDA-FITC-OSMDBT than sensor B. Also, the closer proximity of its SNP-TMSPEDA-FITC-OSMDBT molecules, as indicated by the TGA data, makes sensor A more prone to higher rates of nEET according to eqn (1) (ref. 32) because of it smaller R value. This might explain why sensor A exhibited weaker fluorescence intensity than sensor B, despite having more FITC-OSMDBT molecules.

where: k_DA is the non-radiative energy transfer, k_rad is the radiative rate, R the center to center separation distance between the FITC-OSMDBT molecules. R_F is the Forster radius.

3.4. Response time and selectivity performance of the sensing systems

The response time of sensor A and sensor B towards F⁻ was investigated by measuring their emission increment with time after the addition of F⁻. The results, as shown in Fig. 7b, indicated that the reaction between the sensors and the fluoride anions reached equilibrium quite fast (∼4 min), which suggested that both sensors were well dispersed in solution, and thus, allowed the Si–O active sites to be readily accessed by fluoride anions. The results proved that both sensors are highly responsive toward fluoride anions.

The next property of the sensors was investigated using their selectivity towards F⁻ over other anions. The emission spectra of sensors A and B in the presence of other anions including Cl⁻, Br⁻, I⁻, and NO₃⁻ were recorded and the results compared to that of F⁻. From the results shown in Fig. 8, it was clear that only F⁻ induced the most prominent emission intensity change, whereas, addition of the other anions under the same conditions produced almost no change in emission intensity. At 1.0 : 1.0 (F⁻ : FITC-OSMDBT) equiv. mole titration ratio, the selective of the two sensors was almost the same (Fig. 8a); A pattern which is similar to that observed with their detection limit value. However, at 3.0 : 3.0 equiv. mole titration ratio, sensor B showed better selectivity than sensor A (Fig. 7b): it intensity value was 2.6 folds that of sensor A. Nonetheless, the results show that both sensors exhibit good selectivity towards F⁻ over other anions, a feature which was attributed to the affinity of F⁻ for Si–O.²⁵ The bright yellow color of given off by sensor B solution in the presence of F⁻ under a hand-held UV lamp, as shown in Fig. 9, shows the highly selectivity nature of the sensor. Table 5 puts the performance of the two sensors into perspective.

Fig. 8 (a) Selectivity of sensor A (16.7 mg ml⁻¹) and sensor B (20.4 mg ml⁻¹), and (b) sensor A (50.0 mg ml⁻¹) and sensor B (61.2 mg ml⁻¹) towards F⁻ over Cl⁻, Br⁻, I⁻, and NO₃⁻ (at a concentration of 50 mM respectively) in DMSO.

Fig. 9 Color change in sensor B (20.4 mg ml⁻¹) after the addition of 50.0 mM tetrabutylamonium salts (from left to right: Cl⁻, I⁻, Br⁻, NO₃⁻, and F⁻).

Table 5 Performance of sensor A vs. sensor B

Sensor	Detection range^a (mM)	Selectivity^b
a Based on a fluoride ion: FITC-OSMDBT equiv. mole titration ratio of 3.0 : 3.0.b Based on intensity values from Fig. 7b: 53 784 a.u. was used as the basis for calculation.
A	0.0 ∼ 50.0	53 784.0 a.u.
B	0.0 ∼ 150.0	140 943.0 a.u.
% Difference	—	162.0

---

4. Conclusions

We have successfully synthesized an easy to make fluorescence “turn-on” highly responsive, and selective silica nanoparticle (SNP) supported F⁻ sensing system having a structure which can be easily modified to vary it sensitivity towards F⁻ in DMSO. FT-IR, TGA, EDX, fluorescence spectroscopy analysis data were used to show the successfully synthesis of the sensing system. In this paper, APTMS and TMSPEDA modified SNP were used to show two variations of the sensing system structure and their sensing properties. The amino-silane molecules helped to vary the parentage (%) loading of silyl-ether protected fluorescein isothiocyanate (FITC-OSMDBT) on each sensor, which in turn, helped to regulate the emission spectra of each sensor through non-radiative emission energy transfer (nEET) mechanisms. At 1.0 : 1.0 (F⁻ : FITC-OSMDBT) mole equiv. titration ratio, the emission intensity and detection range values of both sensors were the same; however, at equiv. mole titration ratio x : x (x = 3.0, 6.0, and 9.0), the emission intensity increment in sensor B was found to be thrice that of sensor A. A linear correlation was established between the emission intensity values and detection range of both sensors. From this linear correlation, the detection limit of sensor B was found to be thrice that of sensor A. Finally, at x : x (x = 2.0, 4.0, and 5.0) mole equiv. titration ratio, no correlation could be made between the emission increment and F⁻ concentration. The reason why this phenomenon was peculiar to x : x (x = 3.0, 6.0, and 9.0) molar titration ratios appears to be due to non-radiate emission energy transfer mechanisms; the details of this underlining phenomenon is under investigation. Nonetheless, the current results prove that this type of sensor architecture may prove useful for the development of F⁻ sensing systems with specific detection limits.

Acknowledgements

This research work was supported by the National Research Foundation of Korea NRF-Grants funded by the Ministry of Science, ICT and Future Planning _2001-003090) and the Ministry of Education (2009-009816), Republic of Korea.

References

1. R. M. Duke, E. B. Veale, F. M. Pfeffer, P. E. Kuger and T. Gunnlaugs-son, Chem. Soc. Rev., 2010, 39, 3936–3953.
2. E. Galbraith and T. D. Jams, Chem. Soc. Rev., 2010, 39, 3831–3842.
3. R. Martinez-Manez and F. Sancenon, Chem. Rev., 2003, 103, 4419–4476.
4. A. N. Swinburne, M. J. Paterson, A. Beeby and J. W. Steed, Chem.–Eur. J., 2010, 16, 2714–2718.
5. C. R. Wade, A. J. Broomsgrove, S. Aldridge and F. P. Gabbai, Chem. Rev., 2010, 110, 3958–3984.
6. S. Jagtap, M. K. Yenkie, N. Labhsetwar and S. Rayalu, Chem. Rev., 2012, 112, 2454–2466.
7. R. Wu, S. S. Qian, F. Hao, H. Cheng, D. Zhu and J. Zhang, Environ. Sci. Technol., 2011, 45, 6041–6048.
8. C. Costamagna, A. Bosia and D. Ghigo, Curr. Med. Chem., 2010, 17, 2431–2441.
9. P. Cosentino, B. Grossman, C. Shiek, S. Doi, H. Xi and P. Erbland, J. Geotech. Eng., 1995, 121, 610–617.
10. P. Konieczka, B. Zygmunt and B. Namiesnik, Bull. Environ. Contam. Toxicol., 2000, 64, 794–803.
11. K. Itai and H. Tsunoda, Clin. Chim. Acta, 2001, 308, 163–171.
12. Z. Ying, Z. F. Jun and Y. Juyoung, Chem. Rev., 2014, 114, 5511–5571.
13. Z. Xu, S. K. Kim, S. J. Han, C. Lee, G. Kociok-Kohn, T. D. James and J. Yoon, Eur. J. Org. Chem., 2009, 18, 3058–3065.
14. Z. Q. Lui, M. Shi, F. Y. Li, Q. Fang, Z. H. Chen, T. Yi and C. H. Huang, Org. Lett., 2005, 7, 5481–5484.
15. X. Y. Lui, D. R. Bai and S. N. Wang, Angew. Chem., 2006, 118, 5601–5604.
16. Y. Kubo, M. Yamamoto, M. Ikeda, M. Takeuchi, S. Shinkai, S. Yamaguchi and K. Tamao, Angew. Chem., Int. Ed., 2003, 42, 2036–2040.
17. W. C. Anthony, Fluorescent Chemosensors for Ion and Molecule Recognition, American Chemical Society, 1993, vol. 538.
18. T.-H. Kim and T. M. Swager, Angew. Chem., Int. Ed., 2003, 42, 4803–4806.
19. Y. Kubo, M. Yamamoto, M. Ikeda, M. Takeuchi, S. Shinkai, S. Yamaguchi and K. Tamao, Angew. Chem., Int. Ed., 2003, 42, 2036–2040.
20. Q.-Y. Chen and C.-F. Chen, Tetrahedron Lett., 2004, 45, 6493–6496.
21. S. J. M. Koskela, T. M. Fyles and T. D. James, Chem. Commun., 2005, 7, 945–947.
22. Y. Kubo, T. Ishida, A. Kobayashi and T. D. James, J. Mater. Chem., 2005, 15, 2889–2895.
23. S. Y. Kim and J.-I. Hong, Org. Lett., 2007, 9, 3109–3112.
24. P. Sokkalingam and C.-H. Lee, J. Org. Chem., 2011, 76, 3820–3828.
25. F. Zheng, F. Zeng, C. Yu, X. Hou and S. Wu, Chem.–Eur. J., 2013, 19, 936–942.
26. A. B. Descalzo, D. Jimenez, J. E. Haskouri, D. Beltran, P. Amoros, M. D. Marcos, R. Martinez-Manez and J. Soto, Chem. Commun., 2002, 6, 562–563.
27. P. Vejayakumaran, I. A. Rahmana, C. S. Sipaut, J. Ismail and C. K. Chee, J. Colloid Interface Sci., 2008, 1, 81–91.
28. G. Feng, L. Fabao, Y. Jun and W. Lun, Luminescence, 2008, 23, 392–396.
29. E. Satu, A. Root, M. Peussa and L. Niinisto, Thermochim. Acta, 2001, 379, 201–212.
30. E. F. Vasant, P. Van Der Voort and K. C. Vrancken, Characterization and chemical modification of silica surface, Elsevier, Amsterdam, 1995, ch. 1–6.
31. E. Kim, H. J. Kim, D. R. Bae, S. J. Lee, E. J. Cho, M. R. Seo, J. S. Kim and J. H. Jung, New J. Chem., 2008, 32, 1003–1007.
32. S. Saini, S. Bhowmick, V. B. Shenoy and B. Bagchi, J. Photochem. Photobiol., A, 2007, 190, 335–341.
33. Y. Zhou, J. F. Zhang and J. Yoon, Chem. Rev., 2014, 114, 5511–5571.
34. G. Maroulis, C. Makris, D. Xenides and P. Karamanis, Mol. Phys., 2000, 98, 481–491.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c5ra02158h

---