Molecularly imprinted polymer SiO₂ nanostructures combined with anthocyanin from purple sweet potato peel for on-site colorimetric sensing of benzoic acid in milk

Abstract

Benzoic acid is an important preservative-related marker in dairy products, and its rapid on-site detection remains challenging because conventional analytical methods generally require laboratory-based instrumentation and time-consuming procedures. In this study, a sustainable and portable colorimetric sensing platform was developed for benzoic acid detection in milk by integrating molecularly imprinted SiO₂ nanostructures with anthocyanin extracted from purple sweet potato peel. SiO₂ nanoparticles were synthesized using a modified Stöber sol–gel method and subsequently functionalized through a surface molecular imprinting polymer (MIP) strategy using benzoic acid as the template molecule and a tetraethyl orthosilicate-(3-aminopropyl)triethoxysilane (TEOS-APTES) silane system to form MIP@SiO₂ recognition sites. Anthocyanin served as a natural pH-responsive chromogenic probe, converting the selective rebinding of benzoic acid into measurable optical signals. The sensing mechanism was attributed to the synergistic effect of benzoic acid recognition within the imprinted cavities and the microenvironment-induced structural transformation of anthocyanin. The developed MIP@SiO₂/anthocyanin system exhibited a concentration-dependent colorimetric response toward benzoic acid, with a limit of detection of 27.2 mg L⁻¹. The sensor showed good selectivity against common milk-related interferents, including urea, melamine, sodium chloride, and calcium chloride dihydrate. Practical applicability was demonstrated in spiked milk samples, achieving satisfactory recoveries of 96.0–99.0% using smartphone red green blue (RGB) analysis and 96.32–97.20% using the portable device, with low relative standard deviations. Furthermore, a compact detection device integrating a white LED, AS7341 multispectral sensor, Arduino microcontroller, and bluetooth transmission enabled portable optical readout without requiring a conventional spectrophotometer. These results demonstrate that the proposed MIP@SiO₂/anthocyanin sensing platform provides a low-cost, sustainable, and field-deployable strategy for rapid benzoic acid screening in milk.

Molecularly imprinted polymer SiO₂ with anthocyanin for benzoic acid detection.

1. Introduction

Milk is a highly consumed food matrix because of its balanced nutritional composition, but its high market demand and complex physicochemical nature also make it vulnerable to quality deterioration, adulteration, and unauthorized preservative addition. Among chemical preservatives, benzoic acid and its salts are widely used in acidic foods because of their antimicrobial activity; however, their occurrence in milk is analytically and regulatory important because benzoic acid may be naturally present at low levels in fermented or dairy-related products, while excessive or intentional addition can raise food-safety concerns. Previous monitoring studies have reported measurable levels of benzoic acid in commercial milk samples, confirming the need for reliable surveillance in dairy matrices.¹ In addition, benzoic acid has been specifically investigated in liquid milk using advanced spectroscopic platforms, further demonstrating that dairy products are a relevant target matrix for benzoic acid detection.^2,3 Although benzoic acid is generally regarded as safe within permitted limits, adverse responses have been reported in sensitive individuals when exposure exceeds safety thresholds, including asthma, urticaria, metabolic acidosis, and convulsions.⁴ Therefore, developing rapid, selective, and field-deployable methods for benzoic acid monitoring in milk remains an important task for food quality control and consumer protection.

Several analytical methods have been developed for benzoic acid and related preservative determination in foods, including high-performance liquid chromatography (HPLC), ion chromatography, gas chromatography (GC), derivative UV-visible spectrophotometry, fluorescence-based assays, and surface-enhanced Raman (SER) scattering. For example, solid-phase extraction coupled with ion chromatography has been proposed for benzoic acid determination in milk, while reverse-phase high performance liquid chromatography (RP-HPLC) methods have been applied for simultaneous quantification of preservatives in cheese.^5,6 Gas chromatography-mass spectrometry (GC-MS), and UV-visible derivative approaches have also been reported for preservative analysis in different food matrices.^7–9 More recently, SERS-based methods have enabled rapid and sensitive detection of benzoic acid in milk and processed foods.¹⁰ Despite their analytical strength, these techniques often require costly instrumentation, trained operators, laboratory-based workflows, or complicated substrate preparation. As a result, their direct use for decentralized, routine, or on-site screening remains limited. This limitation has encouraged the development of simple optical, paper-based, smartphone-assisted, and portable colorimetric methods that can translate chemical recognition events into visual or digital signals.^11,12

Colorimetric sensing is particularly attractive for food-safety applications because it offers simple operation, low instrumental requirements, rapid response, and the possibility of naked-eye or smartphone-based signal readout. However, colorimetric assays often suffer from insufficient selectivity in complex food matrices, especially when the target molecule has weak visible absorption or when structurally unrelated interferents coexist. Therefore, an effective colorimetric sensor should combine a selective recognition element with a stable and responsive optical transducer. In this context, MIPs provide an appealing route to improve selectivity because they are synthetic receptors prepared in the presence of a target template. After template removal, complementary recognition cavities remain in the polymeric network, allowing preferential rebinding of the target molecule. Recent reviews have highlighted the advantages of MIPs in food analysis, including chemical stability, reusability, low cost, and suitability for rapid or on-site analytical systems.^13–15

Silica-based MIP materials are especially promising for optical sensing because silica nanoparticles possess high surface area, tunable surface chemistry, mechanical robustness, and good compatibility with sol–gel functionalization. The integration of MIPs with silica materials has recently been reviewed as a powerful strategy for electrochemical and optical probes in food safety and biomedical analysis.¹⁶ The use of silane monomers such as tetraethyl orthosilicate and aminopropyltriethoxysilane enables the formation of silica-based imprinted layers containing functional groups capable of interacting with carboxylic acid targets through hydrogen bonding and electrostatic interactions. Such surface-imprinted structures can improve mass transfer, reduce template entrapment, and increase accessibility of recognition sites compared with bulk imprinting. Nevertheless, most MIP-based benzoic acid sensors still rely on instrumental readouts such as SERS, fluorescence, or electrochemical systems, which may restrict their use in low-resource or field-screening settings.^17,18 Thus, coupling MIP recognition with a natural, visually responsive chromogenic probe could provide a more practical and sustainable strategy for benzoic acid screening.

Anthocyanins are water-soluble natural pigments widely found in fruits, vegetables, and agricultural by-products. Their optical properties are strongly dependent on pH because anthocyanin molecules undergo reversible structural transformations among flavylium cation, quinoidal base, carbinol pseudobase, chalcone, and other equilibrium forms. This pH-responsive behavior produces distinct color changes, making anthocyanins attractive as green colorimetric indicators for food freshness monitoring, intelligent packaging, and visual chemical sensing.^19–21 Purple sweet potato is a particularly valuable anthocyanin source because it contains acylated anthocyanins with relatively high color stability, while its peel represents an underutilized agricultural by-product. Recent studies have shown that purple sweet potato anthocyanins can serve as pH-sensitive colorants in freshness indicator labels, active films, and smart packaging systems.^22–24 However, most anthocyanin-based systems have been designed for freshness monitoring through volatile amines or general pH variation, rather than for selective detection of a specific preservative molecule in milk. This creates an opportunity to integrate anthocyanin color transduction with a target-specific MIP recognition layer.

In this work, we developed a hybrid colorimetric sensing platform based on molecularly imprinted SiO₂ nanostructures combined with anthocyanin extracted from purple sweet potato peel for on-site benzoic acid detection in milk. SiO₂ nanoparticles were synthesized using a modified Stöber sol–gel method, followed by surface imprinting with benzoic acid as the template and a TEOS-APTES silane system to form MIP@SiO₂ recognition sites. A non-imprinted polymer (NIP) counterpart, NIP@SiO₂, was prepared under identical conditions without the template for selectivity comparison. The anthocyanin extract acted as a natural chromogenic probe, while the imprinted SiO₂ material provided selective recognition toward benzoic acid. The sensing mechanism is based on the selective rebinding of benzoic acid within MIP@SiO₂ cavities, which modifies the local microenvironment of anthocyanin and induces a concentration-dependent optical response. In addition to UV-Vis spectrophotometric validation, smartphone RGB analysis and a portable device integrating a white LED, AS7341 multispectral sensor, Arduino microcontroller, and bluetooth transmission were employed for digital on-site signal acquisition. According to the current manuscript, the proposed anthocyanin/MIP@SiO₂ system achieved a detection limit of 27.2 mg L⁻¹ and satisfactory recoveries of approximately 96–99% in spiked milk samples, supporting its potential as a low-cost and portable screening tool for benzoic acid monitoring in dairy products. The overall process is showed in Fig. 1.

Fig. 1. Overview of the benzoic acid detection.

Compared with previously reported benzoic acid sensing methods, including SERS substrates, microfluidic paper-based chips, and quantum-dot systems,²⁵ the present approach offers three integrated advantages: selective molecular recognition through MIP@SiO₂, sustainable colorimetric transduction using anthocyanin from purple sweet potato peel, and portable/smartphone-compatible optical readout. This combination addresses key limitations of conventional laboratory methods and provides a practical pathway toward rapid, visual, and on-site detection of benzoic acid in milk.

2. Experimental section

2.1. Chemicals

The following chemicals were procured from Sigma-Aldrich (St. Louis, MO, USA): (3-aminopropyl)triethoxysilane (APTES). Merck (Darmstadt, Germany) supplied hydrochloric acid (HCl), sodium hydroxide (NaOH), acetic acid (CH₃COOH), ammonium hydroxide (NH₄OH) and sodium acetate (CH₃COONa). Xilong Scientific (China) provided benzoic acid (C₆H₅COOH), ammonia solution (NH₃, 25 wt%), calcium chloride dihydrate (CaCl₂·2H₂O), urea (CH₄N₂O), melamine (C₃H₆N₆), and sodium chloride (NaCl). Tetraethyl orthosilicate (TEOS) was obtained from Zhanyun (China), while absolute ethanol (>99.7%) was supplied by Ghtech (Vietnam). All chemicals were of analytical reagent grade and used as received without further purification. Deionized water was used throughout the experiments.

2.2. Characterization

Fourier transform infrared (FT-IR) spectra were recorded using an FT/IR-6600 Type A spectrometer (Jasco, Japan). UV-visible absorption spectra were obtained with a Jasco V670 UV-vis spectrophotometer (Japan). X-ray diffraction (XRD) patterns were collected on an Ultima IV diffractometer (Rigaku, Japan). The surface morphologies were characterized by scanning electron microscopy (JSM-IT200, JEOL, Japan).

2.3. Extraction of anthocyanin from purple sweet potato peel

Purple sweet potatoes were cultivated in Lam Dong, Vietnam. The harvested tubers were thoroughly washed, their peels were taken, and dried using a desiccator at 60 °C for a few hours. The dried samples were subsequently pulverized into a fine powder, which was then stored in airtight, light-protected containers at room temperature until further use. Subsequently, 1.00 g of the powdered sample was accurately weighed and transferred into a 100 mL beaker. A solvent mixture consisting of 40 mL of ethanol and 60 mL of distilled water was added to the sample. The mixture was homogenized briefly and subjected to ultrasound-assisted extraction using an ultrasonic bath for 30–35 min at ambient temperature. Following sonication, the extract was filtered through a Buchner flask to remove solid residues. The filtrate was then transferred into centrifuge tubes and subjected to centrifugation at an appropriate speed for 15 min to further clarify the extract (Fig. 2). The resulting supernatant containing anthocyanins was collected and stored under dark conditions at low temperature prior to further analysis.²⁶

Fig. 2. Illustration of anthocyanin extraction process.

2.4. Quantification of anthocyanin in purple sweet potato peel

The total monomeric anthocyanin content in PSPP extracts was determined using the pH-differential method described by Giusti and Wrolstad. Samples are diluted with aqueous pH 1.0 and 4.5 buffers and absorbance measurements are taken at the wavelength of maximum absorbance of 530 nm and at 700 nm.^27,28

Total monomeric anthocyanin content:

1

Absorbance calculation:

A = (A_λmax − A₇₀₀)_pH1.0 − (A_λmax − A₇₀₀)_pH4.5 2

MW = molecular weight of reference anthocyanin (cyanidin-3-glucoside = 449.2 g mol⁻¹), DF = dilution factor, ε = molar extinction coefficient (26 900 L mol⁻¹ cm⁻¹), l = path length of cuvette (1 cm).

2.5. Portable device fabrication

The device was manufactured using fused deposition modeling (FDM) additive manufacturing. The three-dimensional model was designed in SketchUp Pro 2021 and subsequently prepared for printing using Ultimaker Cura (v4.8.0), which generated the corresponding toolpath file for fabrication on an FDM Delta 300 3D printer. The structure was uprinted using 1.75 mm black poly(lactic acid) (PLA) filament.

The final printed system consists of a light-shielding enclosure to minimize ambient light interference during measurements and a dual-chamber configuration. One chamber accommodates the electronic control unit, comprising an Arduino microcontroller, a 9 V battery as the power source, and an HC-05 Bluetooth module enabling wireless communication with a smartphone. The second chamber integrates the analytical components, including a cuvette holder for sample positioning, a white LED as the excitation source, and an AS7341 multispectral sensor equipped with eight visible-spectrum channels for optical signal acquisition (Fig. 3).

Fig. 3. Illustration of portable device for benzoic acid detection.

After mechanical assembly, the Arduino microcontroller was programmed via USB using the Arduino IDE (v2.3.2). Detailed firmware implementation and AS7341 sensor technical specifications are provided in the SI.

2.6. Synthesis of SiO₂ NPs

SiO₂ NPs were synthesized using a modified Stöber sol–gel method. Initially, ethanol (100 mL) and distilled water (100 mL) were mixed in a beaker under continuous stirring to form a homogeneous solution. Subsequently, NH₄OH was added as a catalyst to obtain an alkaline solution with a pH of 10–11 to promote the hydrolysis and condensation of the silica precursor.²⁹ The reaction mixture was stirred for 1 h to ensure complete homogenization. 40 mL of TEOS was then added dropwise under vigorous stirring to initiate the sol–gel process. The reaction was maintained at 900 rpm on a magnetic stirrer for 5 h to allow controlled nucleation and growth of silica nanoparticles. Following the reaction, the colloidal suspension was aged for 24 h to enhance particle condensation and structural stability. Instead of centrifugation, the resulting silica sol was subjected to freeze-drying (lyophilization) for particle recovery. Prior to lyophilization, the suspension was optionally pre-frozen at −20 °C (or lower) to ensure complete solidification. The frozen samples were then dried under vacuum using a freeze dryer, allowing sublimation of the solvent and yielding dry silica nanoparticles.³⁰ The obtained SiO₂ powder was further dried in an oven at 60 °C to remove residual moisture and improve stability (Fig. 4).

Fig. 4. Synthesis of SiO₂ NPs.

2.7. Synthesis of MIP@SiO₂ and NIP@SiO₂

The synthesis of MIP@SiO₂ and NIP@SiO₂ was carried out via a sol–gel approach as illustrated in Fig. 5. Initially, a homogeneous silane precursor solution was prepared by mixing tetraethyl orthosilicate (TEOS, 5 mL) and 3-aminopropyltriethoxysilane (APTES, 10 µL) under magnetic stirring, after which the mixture was equally divided to ensure identical reaction conditions for both imprinted and non-imprinted materials.

Fig. 5. Synthesis process of MIP@SiO₂ and NIP@SiO₂.

For MIP@SiO₂ synthesis, 500 mg of SiO₂ particles were dispersed under continuous stirring, followed by the addition of 2.7 mL of the silane precursor and 2 mL of ammonia solution, which acted as a catalyst to promote hydrolysis and condensation reactions. Subsequently, 25 mL of 1 mM benzoic acid was introduced as the template molecule, and the reaction was allowed to proceed for 3 h at room temperature to form a molecularly imprinted silica layer on the SiO₂ surface.

The NIP@SiO₂ was prepared under identical conditions, except that no template molecule was added. After polymerization, both materials were purified by repeated washing with ethanol (three times) and deionized water (two times) to remove residual reagents, while the MIP@SiO₂ was additionally subjected to template extraction to generate specific recognition cavities. Finally, the products were dried at 60 °C prior to further characterization.

2.8. Colorimetric sensing of benzoic acid with MIP@SiO₂ and NIP@SiO₂

The colorimetric sensing of benzoic acid was performed using MIP@SiO₂ and NIP@SiO₂ as a selective recognition material in combination with anthocyanin as a chromogenic indicator. The absorbance changes were monitored using a UV-Vis spectrophotometer.

In accordance with the sensing mechanism, benzoic acid was first allowed to interact with the imprinted polymer prior to signal generation. For each measurement, a reaction mixture with a total volume of 3.0 mL was prepared by initially adding 500 µL of MIP@SiO₂ suspension, followed by an appropriate volume of benzoic acid standard solution to obtain final concentrations of 0, 50, 100, 200, 300, 400, 500, and 1000 mg L⁻¹. The mixture was briefly vortexed and incubated at room temperature for several minutes to facilitate the selective binding of benzoic acid into the imprinted cavities. Subsequently, 300 µL of anthocyanin solution was introduced, and distilled water was added to adjust the final volume to 3.0 mL. The resulting mixture was vortexed again and further incubated for 10–15 min to allow the establishment of a localized acidic microenvironment within the binding sites, leading to a shift in the anthocyanin equilibrium and a corresponding change in absorbance.

After incubation, the absorbance of the solution was recorded using a UV-Vis spectrophotometer and the portable device, respectively. The solution was also taken the images by the smartphone for color analysis.

The smartphone imaging procedure includes the smartphone model of iPhone 13, the use of standard camera settings with no additional image effects, white laboratory lighting, a fixed sample-to-camera distance of 10 cm, a vertical imaging angle, an A4 paper background, center-region selection as the region of interest, and Adobe Photoshop for RGB extraction and normalization.

2.9. Selectivity analysis of benzoic acid

All measurements were conducted in a final reaction volume of 3.0 mL in glass vials. In accordance with the proposed sensing mechanism, 500 µL of MIP@SiO₂ suspension was first introduced, followed by the addition of potential interferents (urea, melamine, sodium chloride, and calcium chloride dihydrate; 100 mg L⁻¹ each) in the absence or presence of benzoic acid. The mixtures were vortexed briefly and incubated at room temperature for several minutes to facilitate selective recognition and binding of benzoic acid within the imprinted cavities. Subsequently, 300 µL of anthocyanin solution was added as a chromogenic probe, and the final volume was adjusted to 3.0 mL with distilled water. The resulting solutions were vortexed and further incubated at room temperature for 10–15 min to allow equilibration of the system, enabling interaction between bound benzoic acid, the imprinted polymer matrix, and the anthocyanin indicator, leading to the formation of a localized acidic microenvironment and a measurable colorimetric response. Finally, the absorbance spectra were recorded using a UV-Vis spectrophotometer over the selected wavelength range. This procedure was employed to evaluate the sensing performance of the system in complex matrices and to assess potential interference effects arising from coexisting species.

2.10. Stability, repeatability, and reproducibility

The stability, repeatability, and reproducibility of the MIP@SiO₂-based sensing system were systematically evaluated using UV-Vis spectrophotometric measurements to ensure its reliability for benzoic acid detection.

The stability of the material was assessed over time by recording its sensing performance at predetermined intervals of 0, 7, 14, and 28 days. At each time point, an aliquot of the suspension was used, and the absorbance was measured under identical conditions using a UV-Vis spectrophotometer. The relative change in absorbance over time was used to evaluate the stability and retention of the sensing activity of MIP@SiO₂.

Repeatability was investigated by performing multiple measurements under identical experimental conditions using the same batch of MIP@SiO₂. Specifically, the sensing response was evaluated at two concentrations of benzoic acid, namely 0 mg L⁻¹ (blank) and 300 mg L⁻¹. For each concentration, the measurements were repeated at least three to five times, and the absorbance values were recorded. The precision of the method was assessed by calculating the relative standard deviation (RSD) of the obtained absorbance values.

Reproducibility was examined by comparing the sensing performance of independently prepared batches of MIP@SiO₂ under the same experimental conditions. Two different batches of MIP@SiO₂ were tested at benzoic acid concentrations of 0 and 300 mg L⁻¹. The absorbance responses were recorded and compared to evaluate batch-to-batch variation. A low deviation between the responses indicates good reproducibility of the synthesis and sensing performance.

All measurements were carried out using a UV-Vis spectrophotometer at the selected analytical wavelength, and the results were expressed as mean values with standard deviations. These studies collectively confirm the reliability, consistency, and practical applicability of the MIP@SiO₂-based colorimetric sensor for benzoic acid detection.

3. Results and discussion

3.1. Characteristics

The SEM images of SiO₂, MIP@SiO₂, and NIP@SiO₂ at different magnifications (1, 5, and 10 µm) are presented in Fig. 6. The bare SiO₂ particles (Fig. 6a–c) exhibit an irregular and loosely aggregated morphology with rough and porous surfaces, which is characteristic of amorphous silica synthesized via the Stöber sol–gel method. The particles appear fused into secondary agglomerates due to the drying process, while maintaining a non-crystalline and disordered surface structure. In contrast, MIP@SiO₂ (Fig. 6d–f) displays a rougher and more heterogeneous surface morphology. The molecular imprinting process introduces irregular cavities and porous domains on the polymer-coated silica surface, resulting in increased surface roughness and structural heterogeneity. These features indicate the successful formation of imprinted recognition sites following template removal. The NIP@SiO₂ sample (Fig. 6g–i) shows a comparatively smoother and more homogeneous surface than MIP@SiO₂. In the absence of template molecules during polymerization, a more uniform polymer layer is formed, leading to fewer surface defects and cavities. Similar morphological differences between MIP and NIP materials have been widely reported in silica-based molecular imprinting systems. Overall, the SEM results confirm successful surface modification of SiO₂ and the formation of imprinted cavities in MIP@SiO₂. The higher surface roughness and porosity observed in MIP@SiO₂ are expected to provide a larger number of accessible binding sites, thereby enhancing its recognition and sensing performance toward benzoic acid.

Fig. 6. SEM images of (a–c) SiO₂, (d–f) MIP@SiO₂ (with benzoic acid), and (g–i) NIP@SiO₂ with scale bar of 1 µm, 5 µm, and 10 µm.

All three samples in Fig. 7 exhibited similar characteristic absorption bands of the silica-based structure, indicating that the silica framework was maintained after surface modification and imprinting. The broad absorption in the high-wavenumber region around 3300–3500 cm⁻¹ can be assigned to the stretching vibration of surface silanol groups and adsorbed water. The absorption band at approximately 1630 cm⁻¹ is attributed to the bending vibration of adsorbed water molecules on the silica surface. The strong absorption peak at 1051 cm⁻¹ corresponds to the asymmetric stretching vibration of Si–O–Si, while the peak at 800 cm⁻¹ is related to the symmetric stretching vibration of Si–O–Si. In addition, the band at 440 cm⁻¹ is assigned to the bending vibration of Si–O–Si. These characteristic peaks confirm the successful formation of silica nanoparticles and the preservation of the silica network in both NIP@SiO₂ and MIP@SiO₂.

Fig. 7. FTIR spectra of SiO₂, MIP@SiO₂, NIP@SiO₂.

Compared with bare SiO₂, the spectra of NIP@SiO₂ and MIP@SiO₂ show no remarkable shift in the main Si–O–Si bands, suggesting that the imprinting process did not significantly alter the fundamental silica structure. The high similarity between the spectra of MIP@SiO₂ and NIP@SiO₂ also indicates that both materials were synthesized under comparable conditions, with the main difference being the presence of benzoic acid during the imprinting step. Furthermore, no distinct absorption band corresponding to the carboxyl C O vibration of benzoic acid was observed in the MIP@SiO₂ spectrum, suggesting that the template molecules were effectively removed during the washing process. Therefore, the FTIR results confirm the successful preparation of MIP@SiO₂ while maintaining the silica-based structure required for benzoic acid recognition.

The X-ray diffraction patterns of SiO₂, NIP@SiO₂, and MIP@SiO₂ exhibit a broad diffraction halo centered at approximately 2θ ≈ 23°, which is characteristic of amorphous silica synthesized via the Stöber sol–gel method (Fig. 8). The absence of sharp diffraction peaks across the scanned range confirms that all materials lack long-range crystalline order and retain an amorphous structure after modification. The position of the main halo remains unchanged for all samples, indicating that the average Si–O–Si network structure is preserved during both non-imprinted and imprinted polymer formation. This observation suggests that the polymer layers are formed on the surface of SiO₂ without altering the bulk framework of the silica core. Differences among the samples are reflected primarily in the shape of the amorphous halo. The pristine SiO₂ displays a relatively well-defined broad peak, consistent with uniform short-range ordering. The NIP@SiO₂ sample shows a comparable profile, indicating that polymer formation in the absence of a template proceeds in a relatively homogeneous manner. In contrast, MIP@SiO₂ exhibits a slightly broadened and less defined halo, suggesting increased structural disorder. This behavior can be attributed to the imprinting process, in which the presence of benzoic acid during polymerization introduces heterogeneous interactions with functional monomers and the silica surface. These interactions disrupt the regular arrangement of polymer chains and promote the formation of irregular binding cavities, resulting in a more disordered structure. Consequently, MIP@SiO₂ demonstrates a higher degree of amorphous character compared to NIP@SiO₂. The slight differences in the low-angle region (around 10–15°) may also reflect variations in surface structure and polymer distribution, although no distinct crystalline phases are observed. Overall, the degree of structural disorder follows the trend:MIP@SiO₂ > NIP@SiO₂ ≈ SiO₂

Fig. 8. XRD patterns of SiO₂, NIP@SiO₂ and MIP@SiO₂ (without benzoic acid).

This trend confirms that the imprinting process increases amorphous character due to the formation of specific binding sites and irregular polymer networks. Importantly, the preservation of the main amorphous peak at ∼23° across all samples verifies that the silica core remains intact, while the observed broadening in MIP@SiO₂ provides structural evidence of successful molecular imprinting.

3.2. Quantification of anthocyanin in purple sweet potato peels

Anthocyanin is a natural water-soluble flavonoid pigment commonly found in purple or dark-colored fruits and vegetables. Its color change under different pH conditions is mainly attributed to the ionization and structural transformation of functional groups on the anthocyanin ring, especially phenolic –OH and methoxy –OCH₃ groups. As illustrated in Fig. 9, anthocyanins exist predominantly in the acidic form under low pH, giving a red or pink color. When the pH increases, deprotonation occurs, and the structure gradually converts into neutral and basic forms, resulting in purple, blue-green, or yellowish colors. This reversible structural transformation explains why anthocyanin can be used as a natural pH-sensitive colorimetric indicator in this sensing system.

Fig. 9. Proposed pH-dependent structural transformation and color change mechanism of anthocyanin.

The total monomeric anthocyanin content of the purple sweet potato extract was determined using the pH differential method as described by R.E. Wrolstad et al.²⁸ The absorbance measurements were conducted at the visible maximum wavelength (λ_max = 530 nm) and corrected at 700 nm to eliminate turbidity effects. The calculated absorbance difference between pH 1.0 and pH 4.5 buffers was used to quantify anthocyanins, as polymerized or degraded pigments do not exhibit the same reversible structural transformation between this range, expressed as cyanidin-3-glucoside equivalents (Fig. 10).

Fig. 10. UV-Vis spectra of anthocyanin under pH 1 and pH 4.5.

The final anthocyanin concentration obtained for the extract was 131 mg L⁻¹. This value falls within the expected range reported for purple sweet potato extracts in the literature, which typically varies depending on cultivar, extraction solvent, and processing conditions. The presence of acylated anthocyanins in purple sweet potato contributes to enhanced stability, which may explain the relatively consistent absorbance behavior observed during measurement.

3.3. Mechanism of sensing

The sensing mechanism of the anthocyanin/MIP@SiO₂ system was investigated based on UV-Vis spectra under different mixing conditions, the pH-dependent behavior of anthocyanin, and the pH change of the sensing solution before and after benzoic acid addition. Fig. 11 shows the mechanism for benzoic acid detection.

Fig. 11. Selective rebinding of benzoic acid in MIP@SiO₂ cavities locally acidifies the anthocyanin microenvironment, stabilizing the anthocyanin form and producing concentration-dependent optical signals.

The sensing mechanism of the MIP@SiO₂ and anthocyanin toward benzoic acid was elucidated by analyzing the UV-Vis absorption spectra of the individual components and their corresponding mixtures. As shown in Fig. 12a and b, distinct spectral features are observed for SiO₂, MIP, NIP, benzoic acid, anthocyanin, and their combinations, providing insight into the interaction mechanism governing the colorimetric response.

Fig. 12. (a) UV-vis absorbance spectra (250–800 nm), (b) sample images of (1) SiO₂, (2) MIP@SiO₂, (3) NIP@SiO₂, (4) benzoic acid, (5) anthocyanin, (6) anthocyanin + benzoic aicd, (7) anthocyanin + MIP@SiO₂, (8) anthocyanin + NIP@SiO₂, (9) benzoic acid + MIP@SiO₂, (10) benzoic acid + NIP@SiO₂, (11) anthocyanin + benzoic acid + MIP@SiO₂, (12) anthocyanin + benzoic acid + NIP@SiO₂, and (c) enlarged UV-Vis spectra of the selected MIP@SiO₂ and NIP@SiO₂ systems in the presence and absence of benzoic acid. (Condition: MIP@SiO₂ or NIP@SiO₂: 166.7 mg L; anthocyanin extract: 10% (v/v); benzoic acid: 30 mg L; total reaction volume: 3.0 mL).

Pure SiO₂, MIP@SiO₂, and NIP@SiO₂ exhibit negligible absorbance in the visible region, indicating that these materials do not directly contribute to the optical signal. Benzoic acid shows a characteristic absorption band in the ultraviolet region around 270–280 nm, corresponding to π–π* transitions of the aromatic ring. However, this absorption does not extend into the visible region, suggesting that benzoic acid alone cannot produce a colorimetric response detectable by the naked eye.

Anthocyanin, on the other hand, displays a broad absorption band extending from the ultraviolet to the visible region, with significant absorbance in the range of approximately 280–350 nm and a tail extending toward longer wavelengths. This behavior is attributed to its conjugated flavylium structure, which is highly sensitive to environmental conditions, particularly pH. In acidic conditions, anthocyanin predominantly exists in the flavylium cation form, exhibiting stronger absorbance and a more intense color. As the pH increases or the microenvironment changes, structural transformations occur, leading to variations in absorbance intensity and spectral profile.

When anthocyanin is added to the selective binding of benzoic acid within the imprinted in MIP@SiO₂, a noticeable change in absorbance intensity is observed, indicating an interaction between the two species. Benzoic acid, being a weak acid, can influence the local pH of the solution and stabilize the flavylium cation form of anthocyanin. This results in a modification of the electronic structure of anthocyanin and a corresponding change in its absorption characteristics. However, this interaction alone is relatively moderate and lacks selectivity.

The introduction of MIP@SiO₂ significantly enhances this effect. In the presence of MIP@SiO₂, anthocyanin exhibits a change in absorbance due to interactions with surface functional groups, particularly the amino groups introduced by APTES. These interactions slightly alter the local environment of anthocyanin, but the effect remains limited in the absence of the target molecule. In contrast, when benzoic acid is present together with MIP@SiO₂ and anthocyanin, a pronounced increase in absorbance intensity is observed across the UV region, along with subtle changes in the spectral profile. This enhancement can be attributed to the selective binding of benzoic acid within the imprinted cavities of MIP@SiO₂.

The imprinted sites, formed during synthesis in the presence of benzoic acid, provide complementary functional groups and spatial configuration that facilitate strong interactions through hydrogen bonding and electrostatic attraction between the carboxyl group of benzoic acid and the amino groups of the polymer matrix. This selective rebinding concentrates benzoic acid in proximity to the anthocyanin molecules, effectively amplifying the local acidity and altering the microenvironment. As a result, the equilibrium of anthocyanin shifts toward its protonated flavylium form, leading to increased absorbance intensity and observable color change.

In comparison, the NIP@SiO₂ system shows a weaker response under identical conditions. Although some non-specific interactions between benzoic acid and the polymer surface may occur, the absence of well-defined binding cavities limits the extent of interaction. Consequently, the spectral changes in the anthocyanin + benzoic acid + NIP system are less pronounced than those observed with MIP@SiO₂, confirming the role of molecular imprinting in enhancing selectivity and sensitivity.

As shown in Fig. 12c, the MIP- and NIP-containing systems were further compared based on the benzoic-acid-induced absorbance changes relative to their corresponding backgrounds. At 336 nm, the anthocyanin-related response was evaluated as:

The experimental conditions were as follows: MIP@SiO₂ or NIP@SiO₂: 166.7 mg L; anthocyanin extract: 10% (v/v); benzoic acid: 30 mg L; total reaction volume: 3.0 mL.ΔA_MIP = A₁₁ − A₇ = 1.46687 − 1.45342 = 0.01345ΔA_NIP = A₁₂ − A₈ = 1.59281 − 1.46572 = 0.12709

At this relatively low benzoic acid concentration, the difference between MIP@SiO₂ and NIP@SiO₂ responses is limited and close to the baseline variation of the anthocyanin system. The anthocyanin + MIP@SiO₂ and anthocyanin + NIP@SiO₂ spectra are nearly identical before benzoic acid addition, which is reasonable because the imprinted cavities are not actively involved without the target molecule. After benzoic acid addition, the two systems exhibit distinguishable spectral responses, indicating different interactions between benzoic acid and the imprinted/non-imprinted matrices.

It should be noted that molecular imprinting primarily provides target recognition sites rather than necessarily producing a higher absolute absorbance change. The optical signal is mainly generated by the pH-responsive transformation of anthocyanin, while MIP@SiO₂ contributes by modifying the interaction environment of benzoic acid. Therefore, a larger ΔA value for NIP at a specific low concentration (30 mg L⁻¹) does not directly indicate superior sensing performance of NIP. The imprinting contribution should be considered together with the concentration-dependent response and selectivity results.

The pH-dependent UV-Vis spectra of anthocyanin further support this mechanism, as shown in Fig. 13. Anthocyanin showed a strong absorption band around 520–530 nm under strongly acidic conditions, corresponding to its red flavylium cation form. As the pH increased, the absorbance in this region decreased, and the solution color gradually faded. Under alkaline conditions, a new broad absorption band appeared around 600–620 nm, accompanied by a visible color change toward green and yellow tones. This behavior confirms that anthocyanin is highly sensitive to pH and can respond to changes in the chemical environment of the sensing system.

Fig. 13. pH-dependent UV-Vis spectra of anthocyanin from pH 1 to pH 13.

The pH measurement also confirmed that benzoic acid altered the sensing medium, as shown in Fig. 14. The pH of the mixture containing MIP@SiO₂, anthocyanin, and water was approximately 6.19 before benzoic acid addition. After benzoic acid with a concentration of 1000 mg L⁻¹ was added, the pH decreased to approximately 4.52. This pH decrease shifted the anthocyanin equilibrium toward a more acidic form, resulting in a visible change from pale purple/pink to a lighter pink color.

Fig. 14. pH measurement of the sensing solution at concentrations of 0 and 1000 mg L⁻¹.

In the concentration-dependent UV-Vis spectra in Fig. 15a, increasing benzoic acid concentration caused a gradual increase in absorbance, especially in the UV region. This response can be attributed to the increasing amount of benzoic acid in the system, together with the pH-induced structural transformation of anthocyanin. Therefore, the observed peak variation is associated with both benzoic acid recognition by MIP@SiO₂ and the pH-sensitive optical behavior of anthocyanin.

Fig. 15. Concentration-dependent (a) image, (b) optical response of the anthocyanin/MIP@SiO₂ sensing system toward benzoic acid obtain from UV-Vis. (c–e) Corresponding calibration curve showing the linear relationship between R, G, B signal intensity and benzoic acid concentration (0–1000 mg L⁻¹) from RGB analysis via the smartphone.

Overall, the sensing mechanism can be explained by the combination of molecular imprinting and pH-responsive color change. MIP@SiO₂ offers recognition sites for benzoic acid, while anthocyanin converts the chemical change in the sensing environment into a measurable optical signal. This combined effect enables the colorimetric detection of benzoic acid using the anthocyanin/MIP@SiO₂ system.

3.4. Detection of benzoic acid by smartphone

The colorimetric detection performance of the anthocyanin/MIP@SiO₂ sensing system toward benzoic acid was evaluated using RGB-based analysis. As the benzoic acid concentration increased from 0 to 1000 mg L⁻¹, the recorded RGB intensity values showed an increasing trend, indicating a concentration-dependent optical response. This result is consistent with the gradual fading of the solution color observed visually after the addition of benzoic acid.

Linear calibration curves were established for the red, green, and blue channels. Among the three channels, the red channel exhibited the best linear relationship with benzoic acid concentration, with the regression equation y = 171.10013 + 0.07223x and a coefficient of determination R² = 0.99967. The green channel also showed good linearity, with the equation y = 132.7161 + 0.07383x and R² = 0.99131. In comparison, the blue channel presented a lower correlation, with the equation y = 143.07858 + 0.03897x and R² = 0.93141.

The higher linearity of the red channel suggests that it is the most suitable RGB signal for quantitative benzoic acid detection. Although the green channel showed a slightly higher slope, its correlation coefficient was lower than that of the red channel. Therefore, the red channel was selected as the optimal analytical signal for subsequent RGB-based quantification. These results confirm that the anthocyanin/MIP@SiO₂ system can provide a measurable and concentration-dependent colorimetric response toward benzoic acid.

Table 1 presents a comparative evaluation of the proposed anthocyanin/MIP@SiO₂ sensing platform against recently reported analytical methods for benzoic acid detection in food matrices.

Table 1. Comparison between different sensors for benzoic detection.

Methods	LOD (mg L−1)	Linear range (mg L−1)	Samples	Ref.
Microfluidic paper-based chip platform	500	500–4000	Soy sauce, dried radish, chili sauce, 21 commercial food samples	11
Optical test strip	73.6	100–700	Soy sauce and oyster sauce	31
MIP-SERS	250	500–2500	Vinegar	32
Inhibitive effect on tyrosinase enzyme	109	50–700	Soya sauce	33
Anthocyanin/MIP@SiO2	27.2	50–1000	Milk	This work

For the smartphone method, the red channel was selected because it provided the highest coefficient of determination among the RGB channels, with the calibration equation y = 171.10013 + 0.07223x and R² = 0.99967. The LOD was calculated according to LOD = 3s/m, where m is the slope of the calibration curve, and s is the standard deviation of the blank response. Using the red-channel slope of 0.07223 RGB units per mg L⁻¹ and s ≈ 0.655, the calculated LOD was 27.2 mg L⁻¹.

Overall, the developed system demonstrates superior analytical sensitivity, with a limit of detection (LOD) of 27.2 mg L⁻¹, which is notably lower than most conventional approaches listed. Specifically, microfluidic paper-based chip platforms exhibit relatively high detection limits (500 µg mL⁻¹), reflecting limited sensitivity despite their advantages in simplicity and low cost. Optical test strips improve analytical performance to some extent (73.6 µg mL⁻¹), but still remain insufficient for low-level detection requirements in complex food matrices. Similarly, MIP-SERS based systems, although offering enhanced selectivity through surface-enhanced Raman scattering, report an LOD of approximately 250 µg mL⁻¹, which is substantially higher than the present work. Enzyme inhibition-based assays provide moderate sensitivity (109 mg L⁻¹) but are often constrained by enzyme instability, susceptibility to environmental conditions, and limited operational robustness.

In contrast, the proposed anthocyanin/MIP@SiO₂ platform achieves a markedly lower detection limit while maintaining applicability in real sample analysis (milk), highlighting its practical relevance for food safety monitoring. This improved performance can be attributed to the synergistic integration of molecularly imprinted recognition sites with the highly responsive pH-sensitive anthocyanin transduction system, which amplifies subtle binding events into measurable optical signals. In addition, the use of a natural chromogenic probe and a portable readout strategy (smartphone/AS7341-based detection) enhances the system's suitability for on-site and resource-limited settings, addressing key limitations of instrument-dependent techniques such as SERS and chromatography. Collectively, the results confirm that the proposed method not only achieves competitive or superior sensitivity but also provides a more sustainable, portable, and cost-effective alternative for benzoic acid screening in food samples.

3.5. Detection of benzoic acid by portable device

Fig. 16 illustrates the multichannel spectral response of the AS7341 sensor toward increasing concentrations of benzoic acid (0–1000 mg L⁻¹). This trend confirms that the interaction between benzoic acid and the MIP-modified anthocyanin system induces measurable variations in the optical absorption profile, which are effectively captured by the multispectral sensor.

Fig. 16. Relationship between signal channels from AS7341sensor and benzoic acid concentration.

Notably, the response magnitude and sensitivity vary significantly across different wavelength channels. Channels in the visible region, particularly F2 (445 nm) and F3 (480 nm), exhibit the most pronounced signal changes and strongest linear trends with concentration. In contrast, longer wavelength channels such as F7 (630 nm), F8 (680 nm), and especially F10 (NIR) show weaker responses and reduced sensitivity.

For the AS7341 portable device, a separate LOD has been determined using the calibration equation. For example, S = 0.7066 + 0.000026241x and R² = 0.80523 of F2 (445 nm), where S is the signal from device, and LOD value have now been reported separately with s = 0.001 and LOD = 114.3 mg L⁻¹.

The linear regression results presented in Table 2 further quantify these observations. All channels were fitted using a weighted linear model (y = a + bx), and the obtained slopes confirm channel-dependent sensitivity differences. Among them, F2 (445 nm) demonstrates the highest correlation metrics (R² ≈ 0.85), indicating superior analytical performance for quantitative determination. In contrast, other channels show markedly lower slope values and poor goodness-of-fit statistics, confirming its limited analytical contribution for benzoic acid quantification in this sensing system.

Table 2. Linear regression parameters for AS7341 spectral channels.

Parameter	F1 415 nm	F2 445 nm	F3 480 nm	F4 515 nm	F5 555 nm	F6 590 nm	F7 630 nm	F8 680 nm	F9 clear	F10 NIR
Intercept, a	0.2721	0.7066	0.53577	0.63728	0.66827	0.67375	0.64624	0.46264	0.63096	0.14058
Slope, b	1.35526 × 10−4	2.62410 × 10−5	8.18045 × 10−5	7.72811 × 10−5	4.67525 × 10−5	3.80400 × 10−5	2.28177 × 10−5	4.65295 × 10−5	1.25758 × 10−5	3.46651 × 10−5
R 2	0.74286	0.80523	0.71052	0.67726	0.41424	0.41907	0.27549	0.28485	0.51190	0.29062

Importantly, the consistency between Fig. 16 and Table 2 highlights the robustness of the multispectral detection strategy. The gradual increase in signal intensity across concentration levels, combined with high linear regression coefficients in selected visible channels, demonstrates that the AS7341 sensor can reliably transduce chemically induced color variations into stable electrical signals. This multi-channel redundancy further enhances measurement reliability, allowing the selection of optimal wavelengths to improve sensitivity and reduce noise in practical applications.

Overall, these results confirm that the proposed portable optical system provides a strong analytical foundation for benzoic acid detection. The wavelength-dependent response behavior not only validates the underlying colorimetric sensing mechanism but also emphasizes the importance of spectral channel optimization in achieving high-performance portable food safety monitoring.

3.6. Selectivity of analysis method

To assess the selectivity of the MIP@SiO₂-based sensor, UV-Vis absorption spectra were recorded in the presence and absence of benzoic acid under various interfering conditions.

As shown in Fig. 17a, benzoic acid exhibits a characteristic absorption band in the 250–350 nm region, with a maximum centered at approximately 270–300 nm, attributable to its aromatic π–π* transitions. In the presence of common potential interferents (urea, melamine, NaCl, and CaCl₂·2H₂O), the spectral profiles remain highly similar, with only marginal variations in absorbance intensity and no observable wavelength shifts. The near overlap of these spectra indicates that the coexisting species do not significantly perturb the optical response of benzoic acid.

Fig. 17. Absorbance intensity of sensor at 326 nm with different adulterants in the presence (a) and absence (b) of benzoic acid.

In the absence of benzoic acid (Fig. 17b), the UV-Vis spectra of the individual interferents show negligible absorbance changes compared with the MIP@SiO₂-anthocyanin system, confirming that these species do not generate a measurable optical response within the sensing platform.

Overall, the minimal spectral interference observed across all tested species demonstrates the high selectivity of the MIP@SiO₂ toward benzoic acid. This behavior is attributed to the presence of molecularly imprinted recognition cavities, which provide complementary spatial arrangement and functional groups for benzoic acid binding. Specifically, hydrogen bonding and electrostatic interactions between the carboxyl group of benzoic acid and the amino functionalities of APTES govern the selective recognition process. In contrast, structurally dissimilar molecules such as urea and inorganic salts lack the necessary geometric and chemical complementarity, resulting in negligible binding affinity and optical response. These findings confirm that the proposed MIP@SiO₂ sensor is highly suitable for the selective detection of benzoic acid in complex matrices such as milk.

3.7. Real samples

The milk sample used in this study was cow milk purchased from a local supermarket. Before analysis, the sample was pretreated by protein precipitation, centrifugation, and filtration.

According to Standard for Fermented Milks (CXS 243-2003), the maximum permitted level of benzoic acid in applicable fermented milk products is 300 mg kg⁻¹, corresponding to approximately 309 mg L⁻¹. The 30 mg L⁻¹ level was chosen as a near-LOD challenge, while 150 and 300 mg L⁻¹ represent intermediate and high fortification levels. The 300 mg L⁻¹ level is close to the benchmark permitted for certain flavoured fermented dairy products.

The analytical performance of the proposed colorimetric sensing platform for benzoic acid detection in milk samples is summarized in Table 3. The method exhibits satisfactory accuracy and precision for both detection strategies, with recovery values ranging from 96.00 to 99.00% for the smartphone-based system and from 96.32 to 97.20% for the portable device. The narrow recovery ranges across the investigated concentration levels (30–300 mg L⁻¹) indicate strong quantitative reliability of the MIP@SiO₂ recognition layer.

Table 3. Benzoic acid detection in milk.

Methods	Added (mg L−1)	Found (mg L−1)	Recovery (%)	RSD (%, n = 3)
Smartphone	30	29.70	99	0.25
	150	144	96	1.53
	300	294	98	2.08
Portable device	30	28.90	96.32	0.15
	150	145.80	97.20	0.55
	300	291.60	97.20	2.11

A slight decrease in recovery at higher analyte concentrations, particularly in the smartphone-based readout (96.00% at 150 mg L⁻¹), may be attributed to partial saturation of imprinted binding sites and/or matrix-induced effects from the milk sample, which can modulate local diffusion dynamics and signal transduction efficiency.

Precision analysis reveals consistently low relative standard deviation (RSD) values (≤2.11%, n = 3), confirming excellent repeatability. The portable device shows marginally improved precision at lower concentrations, suggesting enhanced signal stability in the low-concentration regime.

The analytical response is governed by the selective adsorption of benzoic acid within the molecularly imprinted cavities of MIP@SiO₂, which induces localized pH variations that shift the anthocyanin equilibrium system, resulting in measurable changes in absorbance intensity. The strong agreement between spiked and measured concentrations, combined with low RSD values, demonstrates the robustness of the proposed sensing strategy and supports its applicability for reliable determination of benzoic acid in complex food matrices such as milk.

3.8. Stability, repeatability, and reproducibility of MIP@SiO₂

The stability, repeatability, and reproducibility of the MIP@SiO₂-based sensing system were systematically evaluated to verify its reliability for practical benzoic acid detection.

Long-term stability was assessed by monitoring the absorbance response over 28 days (Fig. 18a). The MIP@SiO₂ material was stored as dry powder at ambient temperature in a plastic sample container in a cupboard. The signal gradually decreased from ∼0.18 on day 0 to ∼0.07 after 28 days. A minor decline was observed within the first 7 days, followed by a more pronounced reduction after 14 days, indicating progressive attenuation of sensing performance.

Fig. 18. (a) Stability of the MIP@SiO₂, (b) repeatability, reproducibility.

This decrease may be attributed to particle aggregation in aqueous media, partial blockage or structural deformation of imprinted cavities, and reduced accessibility of active binding sites. In addition, prolonged exposure to aqueous conditions may induce subtle surface restructuring, thereby weakening host–guest interactions with benzoic acid. Despite this gradual deterioration, the system retains a detectable response after 14 days, indicating moderate storage stability suitable for short-to mid-term applications.

Repeatability was evaluated by repeated measurements at 0 and 300 mg L⁻¹ under identical conditions. As shown in Fig. 18b, both concentrations exhibit minimal variation, reflected by small error bars. The absorbance remains stable at ∼1.00 for 0 mg L⁻¹ and increases to ∼1.08–1.12 at 300 mg L⁻¹. The low dispersion of data confirms excellent intra-assay precision and indicates robust control over sample preparation and measurement procedures.

Reproducibility was examined using two independently synthesized batches of MIP@SiO₂. Both batches display nearly identical responses at 0 mg L⁻¹, confirming consistent baseline behavior. At 300 mg L⁻¹, only a slight variation is observed, with batch 2 showing marginally higher absorbance, suggesting minor batch-to-batch differences in imprinting efficiency or surface accessibility. Nevertheless, the deviation remains within acceptable experimental limits, confirming good synthesis reproducibility.

Overall, the sensing platform demonstrates satisfactory stability over a practical storage period of 14 days and excellent repeatability and reproducibility. These results validate the robustness of the MIP@SiO₂-based colorimetric system for reliable detection of benzoic acid.

4. Conclusions

This study reports a highly selective and robust MIP@SiO₂-based colorimetric sensor for benzoic acid detection. The integration of molecular imprinting with a silica support enables well-defined recognition sites, leading to a strong and selective optical response even in complex matrices. The sensor demonstrates good analytical performance, including high selectivity, satisfactory repeatability, and reliable batch-to-batch reproducibility.

Although a gradual decrease in signal is observed during 28 days of storage, the system retains measurable activity, confirming moderate operational stability suitable for practical use. The results highlight that the MIP@SiO₂ architecture effectively enhances recognition fidelity and transduction stability compared with conventional MIP systems.

Overall, the proposed platform provides a simple, cost-effective, and reliable strategy for benzoic acid monitoring in food samples, with strong potential for further development toward practical on-site analytical applications.

Author contributions

Nguyen Ngoc Nghia: methodology, writing – original draft. Phung Thanh An: conceptualization, writing – original draft. Bui Ngoc Phuong Linh: validation, writing – original draft. Nguyen Huu Hieu: supervision, writing – review & editing.

Conflicts of interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability

The data supporting this article have been included as part of the supplementary information (SI). Supplementary information is available. See DOI: https://doi.org/10.1039/d6ra07035c.