A microfluidic distillation system for rapid separation of sorbic acid from complex food matrices

1 Introduction

In processed foods, sorbic acid (SA) is widely used to delay microbial spoilage without markedly disturbing product formulation. Its antimicrobial activity is mainly attributed to the undissociated form of the molecule. SA first enters microbial cells in its undissociated form. In the cytoplasm, dissociation releases protons and drives intracellular acidification, disturbing the conditions that support normal metabolism and continued cell growth (Lambert and Stratford, 1999). Such effectiveness, however, does not remove the need for careful control. Although SA is generally permitted for food preservation, excessive intake cannot be regarded as risk-free. Concerns arise when SA intake exceeds the level normally expected from regulated food use. Excessive exposure has been reported in connection with asthma-like reactions, pseudoallergic skin responses, and disturbances in fatty acid metabolism (Stopforth et al., 2005; Lück and Jager, 1997; EFSA Panel on Food Additives and Nutrient Sources added to Food, 2015).

Accurate SA quantification is necessary for food regulation and consumer protection. Although GC-MS, HPLC-PDA, and UV–Vis spectroscopy are routinely used for SA analysis (So et al., 2023; Destanoğlu, 2024), direct measurement is easily affected by pigments, proteins, lipids, sugars, and organic acids in processed food matrices. Steam distillation therefore remains an important pretreatment method. Acidification shifts SA toward a form that can be transferred with steam; during distillation, the analyte is carried into the condensate while most matrix constituents remain in the original food sample (Montaño et al., 1995; Kishi and Yamada, 2007; AOAC International, 2023).

Steam distillation is effective for removing matrix interferences, but the conventional format is slow, equipment-heavy, and difficult to move beyond the laboratory. A typical operation requires bench-scale glassware, controlled conditions, and experienced personnel (Wang et al., 2016). The time burden is especially limiting. Large liquid volumes must be heated and condensed, so one pretreatment cycle often takes 4–6 h; some extraction-based procedures may take several days (Chemat and Strube, 2015). Such a workflow fails to meet the demands of on-site inspection, rapid screening, and routine high-throughput control in food-production settings.

Miniaturizing the distillation process offers a practical route away from the slow and equipment-heavy nature of conventional pretreatment. In a microfluidic configuration, evaporation, vapor transport, condensation, and product capture are co-localized within a single compact architecture. This integration reduces sample demand and compresses handling time while maintaining separation performance. Importantly, micro-distillation has progressed beyond proof-of-concept operation in clean standards, with increasing attention to performance under practical, compositionally complex samples (Giordano et al., 2018). Brasil et al. (2020) described a diffusive chip in which ethanol was separated from biochemical substrates before measurement, showing that miniaturized distillation can still function in chemically crowded samples. Food-safety applications have followed naturally because many regulated additives and contaminants require separation from complex matrices before detection. Chip-based distillation has been applied to formaldehyde, sulfur dioxide, propionic acid, and methanol, all of which require matrix removal before reliable analysis in complex food samples (Hsu et al., 2019; Dayao et al., 2019; Lu et al., 2023; Wang et al., 2016).

Micro-distillation is highly sensitive to thermal placement within the chip. The hot zone must drive vapor generation, the channel must sustain directed transport, and the cold region must condense and collect the distillate efficiently, without heat leakage that blurs the temperature gradient and erodes separation. Numerical simulation is useful at this stage because the flow path and temperature field can be checked before fabrication; earlier distillation models considered reflux ratio, evaporation rate, and feed-stage position to narrow the experimental search space (Stanisch et al., 2015; Wende et al., 2021). This thermal-design issue is closely related to preservative analysis. Microfluidic methods have been introduced for benzoic acid, propionic acid, sodium dehydroacetate, methylparaben, salicylic acid, and SA (Liu et al., 2018; Lu et al., 2023; Chen et al., 2022; Ko et al., 2023; Domingues et al., 2025; Wu et al., 2020). SA analysis becomes less straightforward in real foods, where pigments, proteins, lipids, sugars, organic acids, and other matrix constituents can remain close to the analyte. A separation step is therefore needed before detection to protect the final signal from sample-derived interference.

A poly(methyl methacrylate) (PMMA)-based microfluidic distillation chip was developed here to solve the sample-cleanup problem that still limits SA analysis in real foods. The main novelty lies in adapting micro-distillation to the physicochemical and matrix challenges of SA rather than simply transferring an earlier chip format to a new preservative. Previous micro-distillation systems from the group were developed for formaldehyde, sulfur dioxide, methanol, or propionic acid, each requiring different vapor-generation, condensation, and collection conditions. These earlier chips could not be directly applied to SA analysis without modification because SA requires different vapor residence, condensation control, and matrix-removal conditions before HPLC–PDA detection. In particular, commercial food matrices contain sugars, proteins, lipids, pigments, and organic acids that can alter analyte release and interfere with direct instrumental analysis. In contrast, SA analysis in real foods requires efficient recovery from matrices containing sugars, lipids, proteins, pigments, and organic acids while still producing a distillate compatible with official-method HPLC–PDA detection. To address this need, the present chip incorporates a buffer channel for vapor deceleration, a protected sample outlet to reduce vapor leakage, and an optimized condensation channel coupled with a 120 °C/4 °C thermal partition. These structural and operational features define the present platform as an SA-oriented matrix-removal device rather than a generic micro-distillation adaptation. Rather than relying on bench-scale steam distillation, the device confines vapor generation, transport, condensation, and distillate collection within a compact thermally controlled platform. Steam pressure, distillation time, and channel width were optimized as the main variables affecting separation efficiency. After optimization, the system was evaluated against the official analytical method using SA standards and commercial food samples, allowing its practical performance to be judged in terms of accuracy, recovery, and applicability to food safety testing. Practically, the chip reduces sample consumption and pretreatment time before official-method HPLC analysis of SA in real food products.

2 Materials and methods

2.1 Micro-distillation chip fabrication

Figure 1a shows the architecture of the micro-distillation chip. The device consisted of three stacked layers: a 1.5 mm-thick PMMA capping layer, a 6 mm-thick PMMA body layer, and a 0.3 mm-thick aluminum-foil adhesive layer. As shown in Figures 1b,c, the body layer incorporated the sample reservoir, buffer channel, condensation microchannel, and distillate collection zone. The completed chip measured 170 mm × 60 mm × 7.8 mm.

The sample reservoir, buffer channel, condensation microchannel, and distillate collection zone were patterned in the 6 mm-thick PMMA body layer. The condensation microchannel had a rectangular cross-section with a depth of 1.0 mm and an optimized width of 2.5 mm. The buffer channel and vapor-transport channel were fabricated with the same depth to maintain a continuous vapor pathway between the sample reservoir and condensation region. Owing to the thermal nature of CO₂ laser ablation, the channel bottom showed a slightly rounded profile at the lower corners; nevertheless, the effective hydraulic section was reported as rectangular for dimensional reproducibility.

The PMMA layout was drafted in AutoCAD 2020 and transferred to the substrate by CO₂ laser ablation using a 10.6 μm CO₂ laser. Ablation was performed at a laser power of 12 W, a scanning speed of 100 mm/s, and a single-pass cutting mode. After laser machining, loose PMMA debris was removed with compressed air, and the patterned substrates were rinsed sequentially with deionized water and ethanol before nitrogen drying. The PMMA cover and body layers were then aligned under a stereomicroscope and bonded by hot pressing. The aligned layers were preloaded at 0.6 kg/cm² while the temperature was raised to 98 °C, after which the pressure was increased to 1.5 kg/cm² and maintained for 15 min. The bonded chip was cooled naturally to room temperature under pressure to minimize distortion of the laser-machined channels.

The lower side of the PMMA body layer was sealed with a 0.3 mm-thick aluminum-foil adhesive layer. Before sealing, the PMMA surface was cleaned with ethanol and dried under nitrogen. The aluminum foil adhesive was aligned with the PMMA body layer and laminated gradually from one end to the other to avoid air entrapment. Rolling pressure was applied across the entire chip to bring the aluminum-foil adhesive into close contact with the channel network and reservoir edges. Each chip was examined before distillation to confirm that the foil seal remained intact and that no channel blockage or leakage was visible. The aluminum-foil layer, besides sealing, facilitated heat distribution over the distillation zone.

The present chip was modified from the group’s earlier micro-distillation designs, with emphasis on improving vapor control and condensation stability (Hsu et al., 2019; Dayao et al., 2019; Lu et al., 2023; Wang et al., 2016). The added buffer channel reduced steam velocity before condensation, increasing the residence time for cooling and limiting bubble carryover into the extract chamber. A side cover at the sample outlet further reduced SA vapor leakage and helped stabilize the analytical response.

The previously reported formaldehyde, sulfur dioxide, and propionic acid chips offered useful engineering references, although their vapor actuation modes, channel architectures, and condensation strategies could not be transferred directly to SA separation. For example, the formaldehyde platform relied on a ten-layer PMMA assembly with distinct vapor-generation, cooling, and collection regions (Hsu et al., 2019). By contrast, the present SA chip uses a more compact three-layer PMMA/aluminum-foil configuration. The aluminum sealing layer provides close thermal contact, while the integrated vapor pathway shortens mass transport without adding extra chip modules. Its separation mechanism differs from that of the sulfur dioxide device, where carrier gas moved vapor along a serpentine distillation column. In this system, steam supplies the transport force. The short buffer channel then serves as a damping section, reducing vapor acceleration and stabilizing the flow before condensation. A further distinction lies in the evaporation scheme. Unlike the propionic acid chip, which used an on-chip micro-evaporator chamber, the SA platform relies on an external steam source connected to a compact sample reservoir–buffer channel–condensation channel layout. This configuration reduces the chip complexity while providing sufficient vapor residence and condensation efficiency for SA separation.

Because food additives vary in boiling point, density, polarity, and chemical behavior, chip geometry and thermal layout must be matched to the target analyte to achieve efficient separation. In the present design, the major steps of distillation are assigned to separate functional chambers rather than confined to a single continuous space. This simple compartmentalized layout allows the sample reservoir, vapor pathway, condensation channel, and collection zone to be adjusted independently, making it easier to identify suitable operating conditions for SA separation.

2.2 Distillation module

Figure 2a illustrates the custom distillation module used to operate the micro-distillation chip. The module integrated two temperature controllers, a power supply, a heating unit, and a cooling unit. Heat was generated by a 15 W flexible heater (TSA(C)0400050hR37.5, King Lung Chin Co., Ltd., Taichung, Taiwan) attached to a solid copper block measuring 52 mm × 30 mm × 10 mm; this unit operated up to 120 °C. Temperature reduction was achieved with a commercial 10 W thermoelectric module (9504/071/150 B, Ferrotec Taiwan Co. Ltd., Hsinchu, Taiwan), which reached 0 °C and measured 110 mm × 40 mm × 10 mm. The module was enclosed in an ABS casing printed with a Kingssel K3040 3D printer (Mastech Machine Co., Ltd., New Taipei City, Taiwan), giving an overall size of 160 mm × 100 mm × 180 mm.

2.3 Experimental details

Only commercially available reagents were used. Acetonitrile (CH₃CN) and methanol (CH₃OH) were supplied by Honeywell International Inc (USA). Citric acid (C₆H₈O₇), sodium chloride (NaCl), sodium citrate (Na₃C₆H₅O₇) and tartaric acid (C₄H₆O₆) were purchased from Nihon Pharmaceutical Co. Ltd (Japan). Silicone oil was purchased from Cen Hao Technology Co. Ltd. (Taiwan), whereas sorbic acid (SA; C₆H₈O₂) and sodium hydroxide (NaOH) were purchased from Sigma-Aldrich (USA) and MilliporeSigma Inc. (USA), respectively. No additional purification was performed. All aqueous solutions were prepared using 18.2 MΩ cm deionized water.

The 50% (v/v) acetonitrile working solution was prepared by mixing acetonitrile with an equal volume of deionized water. The alkaline medium for SA dissolution was 0.1 M NaOH, prepared by dissolving 0.4 g of sodium hydroxide in 100 mL of deionized water. Citrate buffer was prepared as a stock solution (7 g citric acid and 6 g trisodium citrate in 1.0 L deionized water); before use, the stock was diluted 10-fold, adjusted to pH 4.5, and membrane-filtered. Tartaric acid was prepared at 15% (w/v) by dissolving 15 g in 100 mL of deionized water. SA calibration standards were generated by dissolving 2.5 g of SA in 0.1 M NaOH and diluting the solution with 50% acetonitrile to obtain 50–2500 mg/L. For HPLC-PDA analysis, solvent A was 5 mM citrate buffer at pH 4.5, and solvent B was methanol/acetonitrile at 1:2 (v/v). All solvents were filtered through a 0.45 μm polyvinylidene fluoride (PVDF) membrane and degassed before use.

Real food samples were prepared by homogenizing 5 g of each product. A 0.1 g aliquot of the homogenate was then dispersed in 0.6 mL of deionized water and used as the SA sample solution. After completion of the micro-distillation step, the collected distillate was analyzed by HPLC-PDA according to the Taiwan Food and Drug Administration official method, TFDA No. MOHWA0020.03. The same food samples were processed by conventional benchtop steam distillation and then analyzed by HPLC-PDA according to the TFDA procedure for verification.

For each run, the heating and cooling zones were maintained at 120 °C and 4 °C. Steam was generated by raising the heating plate to 300 °C. The steam inlet tubing was then coupled to the SA distillation chip, after which the prepared SA solution was loaded into the sample reservoir. After sample loading, the inlet of the sample reservoir was sealed with heat-resistant tape to prevent vapor leakage. After the chip was mounted on the distillation device (Figure 2b), the steam-generator pressure valve was opened and adjusted to control the steam flow rate. Steam from the generator entered the sample reservoir and promoted the formation of SA vapor. Vapor leaving the sample reservoir passed across the condensation zone and entered the extract reservoir, assisted by steam flow and gravity. Each run was terminated after 30 min, yielding approximately 2 mL of distillate for subsequent analysis. The HPLC-PDA measurement was performed according to the TFDA official method using a 10 µL aliquot of this distillate.

SA was quantified by HPLC–PDA (Shimadzu LC-20AT) using an ACE C18-AR reversed-phase column (5 μm, 4.6 mm × 250 mm). The mobile phase consisted of 5 mM citrate buffer as solvent A and methanol/acetonitrile at 1:2 (v/v) as solvent B, delivered at 1.0 mL/min. PDA detection was performed at 260 nm.

Additional validation experiments were carried out to examine precision and matrix effects. For intra-day precision, SA-spiked food samples were prepared at three concentration levels, and five independent replicates at each level were analysed on the same day. Inter-day precision was examined by repeating the same procedure on three separate days. Precision was expressed as relative standard deviation (RSD, %). Device-to-device reproducibility was evaluated using independently fabricated micro-distillation chips. SA-spiked samples were prepared at 100, 500, and 2000 mg/L and processed under the optimized operating conditions. Each chip was used for one complete micro-distillation–HPLC–PDA run, and the variation among chips was expressed as relative standard deviation (RSD, %). This experiment was included to determine whether CO₂-laser patterning, PMMA hot-press bonding, and aluminum-foil sealing introduced measurable chip-to-chip variation in SA recovery. The matrix effect was assessed by comparing the HPLC–PDA response of SA standards prepared in matrix-matched distillates with that of SA standards prepared in solvent at the same concentrations. The matrix effect was calculated as shown in Equation 1:

M a t r i x e f f e c t % = S l o p e o f m a t r i x − m a t c h e d c a l i b r a t i o n c u r v e S l o p e o f s o l v e n t − b a s e d c a l i b r a t i o n c u r v e − 1 × 100 % (1)

A matrix effect close to 0% indicated negligible signal suppression or enhancement, whereas positive and negative values indicated matrix-induced enhancement and suppression, respectively.

3 Results

3.1 SA numerical simulation

Before fabricating the chip, ANSYS Fluent was used to map the internal flow field and temperature distribution of the PMMA micro-distillation design. This pre-fabrication simulation verified whether the heating and condensation regions could reach the target operating temperatures while allowing the chip layout to be refined before prototyping, thereby reducing material use, repeated fabrication trials, and development time. Figure 3a shows the simulated flow field. Steam from the generator entered the microchannel at high velocity and reached the sample reservoir, where the resulting fluid motion disturbed the sample surface and enhanced SA vaporization. The SA-containing vapor then entered the condensation channel, where it was directed toward the extract reservoir. Figure 3b shows clear thermal separation across the chip: the heating region remained near 120 °C, whereas the condensation region stayed close to 4 °C. The simulated profile therefore matched the planned operating conditions of the device.

The numerical results showed that the micro-distillation design could provide the vaporization and condensation conditions needed to separate SA from homogenized food samples before HPLC-PDA analysis. Thermocouple measurements gave average temperatures of 118.2 °C at the sample reservoir and 4.0 °C at the distillate outlet. The measured values differed by less than 3% from the simulated temperatures of 117.9 °C and 4.3 °C, supporting the reliability of the model for predicting the chip’s thermal behavior.

3.2 SA optimization experiment

Distillation efficiency was used to evaluate device performance. Steam pressure, distillation time, and channel width were systematically varied to determine the operating conditions that maximized SA recovery. During method optimization, distillation efficiency was defined operationally as the chip-level recovery of SA. After micro-distillation, the SA concentration measured in the collected distillate was divided by the theoretical concentration of the prepared standard solution loaded into the device, and the resulting ratio was expressed as a percentage. This reference value denotes the nominal concentration of the loaded SA standard; it is not a concentration obtained from the conventional steam-distillation procedure. The distillation efficiency was calculated using Equation 2:where C_microdistillation is the SA concentration measured by HPLC–PDA after micro-distillation, and C_standard is the theoretical concentration of the prepared SA standard solution before distillation.

D i s t i l l a t i o n e f f i c i e n c y % = C _ m i c r o d i s t i l l a t i o n / C _ s t a n d a r d × 100 % (2)

Increasing steam pressure enhanced SA vaporization and improved distillation efficiency (Figure 4). The response reached its optimum at 4 bar, after which further pressure increase provided no additional improvement in recovery. At higher pressure, the vapor stream crossed the condensation channel too rapidly, leaving less time for cooling and reducing the amount recovered in the extract reservoir. A maximum recovery of approximately 98% was obtained at 4 bar; this pressure was therefore used in subsequent experiments.

As shown in Figure 5, extending the distillation time improved efficiency at the early stage. After about 30–35 min, the curve approached a plateau, suggesting that most of the recoverable SA had already vaporized and been collected as condensate. Distillation efficiency reached 97.0% at 30 min and increased only slightly to 98.1% at 35 min. Because the additional 5 min produced only a small increase in recovery, 30 min was selected as the operating time for subsequent SA distillation experiments.

Distillation efficiency increased as the condensation channel widened from 1.5 to 2.5 mm (Figure 6), suggesting that the narrower channels restricted vapor passage and condensate collection. Efficiency decreased when the channel was widened further. This decrease was not attributed simply to faster vapor velocity, because a larger cross-sectional area would reduce mean velocity under constant-flow conditions. In the present pressure-driven steam system, excessive widening likely reduced hydraulic resistance and altered the vapor distribution within the cooled region. The larger channel also weakened vapor–wall contact, lowering heat-transfer efficiency and reducing complete condensation of SA-containing vapor. A channel width of 2.5 mm therefore provided the best balance between vapor transport and wall-assisted condensation, delivering approximately 97.6% distillation efficiency.

3.3 Micro-distillation system validation

Method validation was performed using SA standards ranging from 50 to 2500 mg/L. After distillation by each pretreatment method, SA concentrations were determined by HPLC–PDA. The micro-distillation workflow closely matched the conventional steam-distillation method across the tested samples (Figure 7; R ² = 0.9964). The relative accuracy was 95.68% ± 3.8%, further confirming that the chip-based distillation process can provide reliable SA pretreatment. Precision and matrix effects were further examined to complete the validation of the micro-distillation–HPLC–PDA workflow (Table 1). For intra-day precision, five independently prepared replicates were analysed at 100, 500, and 2000 mg/L; the resulting RSD values ranged from 2.1% to 4.8%. Inter-day precision was tested at the same three concentrations over three separate days, giving RSD values of 2.8%–5.6%. Chip-to-chip reproducibility was assessed with independently fabricated devices at the same three SA concentrations. The corresponding RSD values were 5.2%, 4.1%, and 2.4% at 100, 500, and 2000 mg/L, respectively. Such low variation shows that differences introduced during fabrication and assembly had only a modest effect on SA recovery. The gradual decrease in RSD at higher concentrations further reflects a more robust distillation response once the analytical signal moved away from the low-concentration boundary. The precision results show that SA could be measured consistently within a single day and remained stable when the assay was repeated on different days. Matrix effects stayed within −5.2% to +3.4%, suggesting that micro-distillation removed most interfering components and left only a small influence on the final signal. Together, the precision and matrix effect data confirm that the chip-based pretreatment could reduce the matrix-derived interference while maintaining reproducible SA recovery from the complex food samples.

TABLE 1. Validation performance of the micro-distillation–HPLC–PDA workflow for SA determination.

Table note: RSD, relative standard deviation. Intra-day precision was obtained from five independently prepared replicates analyzed within the same day. Inter-day precision was assessed over three separate days. Matrix effect was calculated by comparing the slope of the matrix-matched calibration curve with that of the solvent-based calibration curve. Device-to-device reproducibility was evaluated using independently fabricated micro-distillation chips and expressed as RSD.

Practical applicability was assessed by quantifying SA in ten commercially purchased food samples from a wholesale supermarket in Taiwan (Table 2). Each sample was processed using the micro-distillation system, and SA content was determined by HPLC–PDA according to TFDA Official Method No. MOHWA0020.03. A reference workflow at CAAPIC (NPUST) was used to analyze the same samples via benchtop steam distillation, separation, and official-method detection. Five independent replicates were run per sample; each prepared from a freshly homogenized portion.

TABLE 2. Evaluation and comparison of SA detection results from two methods using ten commercially available food samples.

N.D., non-detectable. Relative accuracy was calculated by comparing the micro-distillation–HPLC–PDA, result with the CAAPIC, reference value and was not calculated when the CAAPIC, result was N.D., The jam result obtained by micro-distillation was close to the method detection limit and was therefore regarded as a trace-level response near the detection boundary rather than as a confidently quantified value.

For commercial food matrices, CAAPIC comparison results are not used to estimate absolute recoveries, but rather to describe relative accuracy. Specifically, the SA concentrations obtained after chip-based micro-distillation and HPLC-PDA analysis are compared with values obtained from the CAAPIC reference workflow (following the TFDA official method) using Equation 3:where C_microdistillation is the SA concentration measured after micro-distillation followed by HPLC–PDA, and C_CAAPIC is the SA concentration reported by the CAAPIC reference workflow for the same commercial food sample. Relative accuracy was not calculated when the CAAPIC result was N.D. Relative error was calculated using Equation 4:

R e l a t i v e a c c u r a c y % = C _ m i c r o d i s t i l l a t i o n / C _ C A A P I C × 100 % (3)

R e l a t i v e e r r o r % = C _ m i c r o d i s t i l l a t i o n − C _ C A A P I C / C _ C A A P I C × 100 % (4)

To avoid mixing validation concepts, each metric was used with a specific analytical meaning. Recovery refers to the fraction of SA retrieved from standard or spiked samples after micro-distillation. Relative accuracy applies to real food matrices and expresses the closeness of the chip-based result to the CAAPIC reference value. Relative error records the signed percentage difference between the two workflows, whereas method agreement describes their overall concordance across the tested concentration range.

Against the CAAPIC reference method, the micro-distillation workflow gave a tightly grouped set of results across the ten commercial foods listed in Table 2. In samples with reportable reference concentrations, relative accuracy extended from 92.05% for orange juice to 107.07% for peanut butter; none departed from the official value by more than ±8%. The peanut butter result sits just above full recovery and therefore reflects a small positive bias, not poor agreement. A reasonable explanation lies in the sample itself: its lipid- and protein-rich structure can modify SA liberation, vapor-phase transfer, and repartitioning during steam-assisted separation.

The jam sample was interpreted more cautiously. Micro-distillation gave only a faint response, 62 ± 2 mg/L, while the CAAPIC reference method classified the sample as N.D. This signal sat just above the 60 mg/L method detection limit, leaving too little analytical separation to support confident quantification. For this reason, the result was described as a trace-level response at the detection boundary, rather than as a definitive positive value. In this low-concentration region, even modest differences in the sampled portion, uneven preservative distribution within the jam, or small variations in analyte transfer during pretreatment can shift the outcome from a weak detectable signal to an N.D. report.

Soy sauce and apple juice were assigned N.D. in both workflows. To check whether these negative results reflected the actual samples or possible matrix-related loss, spike recovery tests were carried out in the same two matrices. After the full micro-distillation–HPLC–PDA sequence, the SA-spiked soy sauce and apple juice samples (100 and 500 mg/L) recovered 96.4% and 105.6%, respectively. The recovery results stayed within the expected analytical range, with no evidence of marked matrix suppression or appreciable SA loss during pretreatment. For this reason, the N.D. assignments are better attributed to undetectable SA, or to concentrations below the method’s detection capability, rather than to matrix-derived artifacts.

The method reached a detection limit of 60 mg/L. This value should be read in the context of the chip’s intended function: it serves as a matrix-removal pretreatment unit coupled to HPLC–PDA, not as an independent sensing device. Its sensitivity is comparable to that of a previously reported propionic acid micro-distillation–HPLC–PDA workflow, which gave a detection limit of 50 mg/L and a quantitation limit of 96 mg/L (Lu et al., 2023). A direct comparison with formaldehyde or sulfur dioxide micro-distillation chips is less informative because those systems involved different analytes and detection chemistries. For practical SA analysis, however, the present performance is adequate for commercial food matrices, especially when considered alongside the small sample requirement of 0.1 g, the 30 min pretreatment time, and the close agreement with the TFDA/CAAPIC reference workflow. These results position the chip as a useful pretreatment platform for SA screening and quantification in compositionally complex foods.

Because micro-distillation performance depends strongly on analyte volatility, vapor transport, condensation residence, and matrix composition, the earlier formaldehyde, sulfur dioxide, and propionic acid chips were treated as design references rather than directly interchangeable platforms for SA analysis. Table 3 positions the SA micro-distillation chip against earlier micro-distillation platforms and helps clarify where the present design differs. Earlier micro-distillation studies provide useful benchmarks for the present SA platform. Hsu et al. (2019) applied a ten-layer PMMA chip to formaldehyde analysis, obtaining approximately 98% distillation efficiency within 10 min with either AHMT-based detection or paper-based colorimetric readout. For sulfur dioxide, Dayao et al. (2019) selected a three-layer PMMA design in which carrier gas assisted vapor transport; the average distillation efficiency reached 90.5% after 25 min. In the propionic acid system reported by Lu et al. (2023), the evaporation chamber was integrated on chip, and the distillation efficiency reached nearly 97% after 15 min. The present SA system takes a more compact route, using a three-layer PMMA/aluminum foil architecture and reaching approximately 98% distillation efficiency under 4 bar steam pressure, 30 min distillation, and a 2.5 mm condensation channel. Its relevance extends beyond a simpler chip structure, as the workflow was benchmarked by official-method HPLC–PDA analysis across ten commercial food matrices, including lipid-rich, sugar-rich, and acidic samples. This comparison shows that the contribution is not merely an analyte substitution but an SA-directed vapor-buffered architecture coupled with matrix removal for complex food analysis.

TABLE 3. Comparison of the present SA micro-distillation system with previously reported micro-distillation platforms.

4 Conclusion

A PMMA micro-distillation device enabled efficient separation of SA from food matrices. Design optimization supported by numerical simulation established a robust thermal partition—120 °C at the heater and 4 °C at the condenser—while minimizing thermal coupling between zones. The simulation–experiment deviation remained below 3%, supporting the reliability of the model for design decisions. Consequently, chamber-layout iteration can be streamlined, with reduced material consumption during early prototyping. Experimental results indicated that distillation efficiency depended strongly on steam pressure, distillation time, and channel width. The highest distillation efficiency was achieved with a steam pressure of 4 bar, a distillation time of 30 min, and a channel width of 2.5 mm. The micro-distillation system was applied to commercial food samples before HPLC–PDA analysis. The method achieved a 60 mg/L detection limit and showed relative accuracies of 92%–107% for quantifiable commercial samples, with deviations within 8% against the CAAPIC/TFDA reference workflow. Trace-level and N.D. results were interpreted conservatively with support from matrix-specific spike recovery tests. More importantly, the system addresses a key limitation in microfluidic SA analysis by integrating matrix-removal pretreatment directly into a compact chip-based distillation format. The key contribution of this study is the development of a sorbic acid–targeted micro-distillation chip, rather than the direct application of earlier formaldehyde, sulfur dioxide, or propionic acid devices to a new analyte. Several structural elements were deliberately redesigned for this purpose, including the buffer channel, protected outlet, shortened 2.5 mm condensation path, and sharply divided 120 °C/4 °C thermal zones. Together, these features stabilized vapor transport, reduced condensate loss, and improved condensation control during sorbic acid isolation. The work also extends beyond device modification by demonstrating the chip as a practical pretreatment module for commercial food samples with markedly different compositions prior to official-method HPLC–PDA measurement. This integration links microfluidic distillation with regulatory preservative analysis and provides a compact, matrix-tolerant route for sorbic acid cleanup before instrumental quantification.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Angeliki Tserepi, National Centre of Scientific Research Demokritos, Greece

Reviewed by: Makoto Tsunoda, The University of Tokyo, Japan

Qiangyuan Zhu, Zhejiang University, China