Detection of congeners and contaminants in commercial and homemade alcoholic beverages under varying temperatures using metal-phenolic film-coated quartz crystal microbalances

Abstract

Consumption of cheap, adulterated alcohol containing high congener levels (e.g., aldehydes, esters, fusel alcohols, methanol) takes a major socioeconomic toll, highlighting the need for on-site analysis. Handheld chemical profiling of spirits is feasible via Raman spectrometry, but its applicability is limited by dark packaging as well as unverified accuracy for low-alcoholic products and homemade distillates. To address this shortcoming, we examine the resonance behavior of metal-phenolic film-coated quartz crystal microbalances in response to the saturated vapor of five commercial or homemade fermented and distilled beverages (beer, white wine, red wine, whiskey and grape brandy). Validated across 270 trials, the sensor demonstrated superior repeatability in chemical characterization of the drinks at mild and room temperatures, maintaining minimal error of ~ 0.2–3.6%. Competitive vapor sorption leads to non-linear quantitative detection upon injection of 0.25–2.5 mL methanol into the alcoholic drinks. Under optimal temperature of the beer and grape brandy (6–16 °C), however, the sensor demonstrates a linear relationship with increasing methanol levels. It achieves a detection limit of ~ 0.13 g/100 mL, a value 7.4 times lower than the current EU permissible dose. This outcome emphasizes the resilience of our technology in point-of-use quality assessment of beverages, counteracting counterfeit alcohol trading.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1038/s41598-026-61378-w.

Introduction

Alcohol exhibits a multifaceted nature, since it is a foundational cultural staple¹, part of daily routine in ancient Mesopotamia and Egypt², and a modern asset driving industrial innovations^3–6. Yet, its immoderate consumption causes rapid physiological impairments by altering brain function within minutes, reducing body temperature via vasodilation and inducing anxiety due to the accumulation of the toxin acetaldehyde^7–10.

Currently, alcohol is a contributing factor to several societal challenges, including global cancer incidence (4.1% of cases in 2020)¹¹, child abuse and sexual harassment¹², loss of workplace productivity¹³ and environmental degradation¹⁴. The “alcohol harm paradox” exacerbates these impacts, as lower-income people suffer disproportionately higher rates of alcohol-related illness, hospitalization and death compared to wealthier groups, despite similar or lower consumption¹⁵– partly due to the purchasing of cheap, unrecorded alcohol¹⁶. Attempts to curb health risks caused by illicit alcohol are actively blocked by the beverage industry, which undermines evidence-based policies¹⁷. Therefore, the invention of point-of-use screening technology for alcoholic products is vital to ease this worrying situation.

The sensory profile of fermented and distilled beverages, which determines their quality¹⁸, is based on volatile and non-volatile compounds formed during raw material selection, fermentation, distillation and maturation. While primarily composed of ethanol and water, these drinks also contain congeners and trace contaminants (e.g., aldehydes, esters, methanol, higher alcohols, heavy metals) that are toxic at high concentrations¹⁹. Real-time monitoring is feasible via portable chemoresistive metal oxide sensors²⁰ or chiral nematic liquid crystal devices²¹, but both are designed predominantly for methanol detection. Spatially offset Raman spectroscopy (SORS) enables complex chemical profiling of spirits, but it suffers from low signal-to-noise ratio through dark or amber containers²². These background interferences are not insurmountable and can be mitigated through hardware optimization or by using longer-wavelength NIR lasers coupled with background-filtering algorithms or convolutional neural networks^23,24. Nevertheless, such solutions present notable shortcomings: advanced optics and sensitive detectors raise equipment costs, analyzing dark packaging requires longer exposure times that hinder rapid testing, while background filtering frequently suppresses authentic Raman signals, thereby degrading trace compound detection limits. Moreover, because the SORS method is validated only for high ethanol concentrations²², its applicability to the complex multicomponent matrices of beers, wines and homemade distillates (often produced in Southeastern Europe) remains unverified.

Metal-phenolic film-coated quartz crystal microbalances (MPF-QCMs) are a good alternative for high-resolution, point-of-use alcohol analysis²⁵. The selectivity of metal-organic adsorbent endows MPF-QCMs with linear sensor signal upon gas sorption, enabling accurate detection of trace methanol in whiskey (2.7 µL/100 mL)²⁵. By adjusting the film’s thickness and chemistry during synthesis, these devices detect methanol concentrations up to 1000 times below regulatory standards under equilibrium vapor pressure²⁶, rendering external analyte-delivery equipment unnecessary. The operation principle of a future portable MPF-QCM is related to the incorporation of the actual electronic device into a miniature corps equipped with a sensor oscillator, a frequency counter, an ohmmeter and an alcohol container²⁵. Injecting 10 mL of a given beverage into the container triggers automated displays of resonance frequency and dynamic resistance within ~ 5–10 min, confirming or denying the presence of excessive amounts of harmful congeners. These electrical parameters ensure the key technical innovation of QCM methods, since alcohol profiling can be performed based on both mass loading and/or viscoelasticity changes, accurately identifying the type of adhering molecules without the limitations of conventional spectrometry. Furthermore, the acoustic rigidity and hydrophobicity of metal-phenolic films ensure negligible cross-sensitivity to temperature and humidity (~ 5.6 Hz/°C/%RH)²⁷. However, routine use of MPF-QCMs is still questionable unless they can differentiate spirits and low alcohol-by-volume drinks at their specific serving temperatures²⁸.

This research addresses a critical literature gap by demonstrating that the temperature, ethanol content and chemical composition of five reference beverages directly affects the response of MPF-QCMs, validating their capability to assess the quality of traditional alcoholic products. We also reveal that tuning the mixing order of ferric chloride and phenol influences the sensitivity of quartz resonators.

Results

Characterization and electrical performance in air of two 5 MHz MPF-QCMs

The one-step synthesis (mixing in direct or reverse order of saturated aqueous FeCl₃·6H₂O and C₆H₅OH solutions) yielded metal-phenolic films with morphology and elemental composition presented in Fig. 1; Table 1.

Fig. 1.

Morphological features of two metal-phenolic films, prepared by spin-coating ~ 0.3 mL of two stock solutions onto two unpolished 5 MHz QCMs. The solutions are obtained by mixing 5 mL of saturated aqueous FeCl₃·6H₂O and 15 mL of saturated aqueous C₆H₅OH in direct (left image) or reverse (right image) order. The images are taken at an accelerating voltage of 20 kV and a working distance of 15.37 mm.

Table 1.

Elemental analysis in the black spot areas of metal-phenolic films, identified via scanning electron microscopy (SEM).

Type of QCM	Chemical element (at%)
	C	O	Si	Cl	Fe	Cr	Au
QCM FeCl 3 -PhOH	4.8	58.9	30.5	0.4	4.3	0.2	0.9
QCM PhOH-FeCl 3	22	25	37.5	0.3	3.9	0.9	10.4

By implementing this swift one-step approach, stable iron phenolate coordination networks are formed, producing homogeneous, cross-linked films with flaky features, morphologically identical to standard tannic acid-based coatings²⁹. Both films shown in Fig. 1 have similar morphology, which appears to stem from the unpolished quartz surfaces and the absence of NaOH in the stock solutions. Alkaline admixtures increase pH, thereby phenol hydroxyl groups remain unprotonated and negatively charged, leading to electrostatic repulsion that prevents the agglomeration of dissolved or suspended chains²⁹. Although the SEM images depict black spots initially resembling micron-scale voids (see Fig. 1), energy dispersive spectroscopy reveals these features are actually residual FeCl₃ microparticles (see Table 1).

To elucidate the chemical states within the metal-phenolic films, X-ray photoelectron spectroscopy (XPS) was performed and the corresponding spectra are illustrated in Figs. 2, 3 and 4.

Fig. 2.

Photoelectron scan surveys of two metal-phenolic films, prepared by spin-coating ~ 0.3 mL of two stock solutions onto two unpolished 5 MHz QCMs. The solutions are obtained by mixing 5 mL of saturated aqueous FeCl₃·6H₂O and 15 mL of saturated aqueous C₆H₅OH in direct (left image) or reverse (right image) order.

Fig. 3.

C1s photoelectron core level of two metal-phenolic films, prepared by spin-coating ~ 0.3 mL of two stock solutions onto two unpolished 5 MHz QCMs. The solutions are obtained by mixing 5 mL of saturated aqueous FeCl₃·6H₂O and 15 mL of saturated aqueous C₆H₅OH in direct (left image) or reverse (right image) order. Shirley-type background subtraction method is employed and the curve fitting is performed using a Gaussian/Lorentzian (GL) sum function.

Fig. 4.

O1s photoelectron core level of two metal-phenolic films, prepared by spin-coating ~ 0.3 mL of two stock solutions onto two unpolished 5 MHz QCMs. The solutions are obtained by mixing 5 mL of saturated aqueous FeCl₃·6H₂O and 15 mL of saturated aqueous C₆H₅OH in direct (left image) or reverse (right image) order. Shirley-type background subtraction method is employed and the curve fitting is performed using a Gaussian/Lorentzian (GL) sum function.

Contrary to expectations, the photoelectron surveys fail to detect iron and chlorine species. The deconvoluted O1s spectra further support the absence of iron oxide and iron hydroxide complexes, eliminating their characteristic signatures usually observed at ~ 529–530 eV and ~ 533 eV^26,30. In the C1s region, the binding energy of ~ 285 eV is assigned to the hydrocarbon benzene rings of phenyl groups, while the signals at 286–288 eV correspond to phenol hydroxyl groups and quinone-type carbonyl functionalities^31,32. These results cast doubt on the formation of metal-phenolic films, which may be proven by the static contact angles of water droplets measured on both 5 MHz QCMs (the films make quartz resonators hydrophobic^25,26, as shown in Fig. 5.

Fig. 5.

Static contact angle of a 10 µL water droplet on QCM FeCl₃-PhOH (left image), QCM PhOH-FeCl₃ (central image) and an uncoated 5 MHz QCM (right image). The wettability tests are performed with three droplets placed on different areas across the surface at standard psychrometric conditions (T_amb ~21 ± 1 °C; RH_amb ~35 ± 5%). The standard deviation (SD) of contact angle values is minimal (SD ~ 1–2 °), indicating high surface uniformity.

Due to oxygen-driven hydrogen bonding, stock solutions with high phenol content (1:3 FeCl₃-to-phenol ratio) generate surfaces with water contact angles ~ 83–86 °, just below hydrophobicity²⁶, that are nevertheless much less wettable than conventional hydrophilic uncoated QCMs (see Fig. 5). This is strong evidence for the successful formation of metal-phenolic films on the used QCMs, where the unpolished surfaces enhance the intrinsic tendency to wetting, but ensure high gas sensitivity due to the large number of sorption sites²⁶. The lack of FeCl₃ signal in XPS implies this compound is trapped too deeply for the typical 1–10 nm probe depth to capture. It might be that the unreacted ferric chloride is confined beneath a thin coating and that the surface primarily consists of phenolic ligands (due to their lower surface energy blooming to the air-coating interface) obscuring the metal beneath it, even though energy dispersive spectroscopy detects both.

The film deposition decreases the resonance frequency of 5 MHz MPF-QCMs in air by ~ 300–975 Hz, corresponding to thicknesses of d ~ 17–60 nm, based on the Sauerbrey equation²⁶. The small 1 Ω change in their dynamic resistance confirms the “acoustic rigidity” of metal-phenolic films, meaning that the coating deposition-induced dissipation of oscillation energy is negligible and the Q-factor remains the same as that of the uncoated quartz crystals, which is a prerequisite for unchanged nominal (designed) sensitivity.

MPF-QCMs as exploratory tools for profiling alcohol vapors of fermented and distilled beverages

In conjunction with QCM measurements, we performed chemical profiling of the selected alcoholic beverages via gas chromatography at licensed laboratories (NIIS Sofia and FebaLab F&B Analysis) and the outcome is summarized in Table 2.

Table 2.

Chemical composition of the tested alcoholic drinks. TTA means total titratable acidity.

Alcoholic beverage	Ethanolcontent (vol%)	Acetaldehyde (mg/L)	Esters (mg/L)	Higher alcohols (mg/L)	MeOH (mg/L)	TTA (g/L)	ρ (g/mL)	Heavy metals (mg/L)
								Fe	Cu
Beer	4.35	343	15.5	70.2	27.5	1.66	1	0.11	0.04
White wine	10.86	333.6	56	157	4.1	5.5	0.99	2.5	0.4
Red wine	13.31	153.3	331.7	376.6	253.8	4.9	0.99	4.3	0.1
Whiskey	40.02	25.2	80	649.3	20.6	0.12	0.95	0.04	0.1
Grape brandy	53.5	100	481.1	2116.8	186.7	0.96	0.92	0.5	1.4

Data from the accredited institutions show a high content of acetaldehyde in the beer, white wine and red wine (above the threshold of 20–50 mg/L), suggesting incomplete yeast fermentation or ethanol oxidation and low quality of the considered commercial brands³³. Background levels of this compound in the whiskey and grape brandy fall within normal limits. The concentration of higher alcohols (i.e., 1-butanol, 2-butanol, 1-propanol, 2-methylpropanol, 2-methyl-1-butanol and 3-methyl-1-butanol) in the high alcohol-by-volume distillates surpasses the reference range, while the low-alcoholic beverages comply with standards. The analysis reveals elevated ester profiles in both the red wine and grape brandy, but methanol content in all drinks is far below the maximum tolerable levels³⁴. As alcohol-by-volume decreases, the density of alcoholic drinks increases, with total acidity remaining within typical ranges³⁵. Distinguishing fermented from distilled alcoholic beverages and assessing their quality based on specific indicators is of practical relevance for point-of-use safety testing, providing a robust tool against the distribution and consumption of counterfeit alcohol – the core motivation of this research.

The effectiveness of the two 5 MHz MPF-QCMs in characterizing drink chemical profiles is evaluated by their respective sensor responses, shown in Fig. 6.

Fig. 6.

Resonance frequency shifts of QCM FeCl₃-PhOH and QCM PhOH-FeCl₃ induced by the saturated vapor of commercial and homemade alcoholic beverages at (a) 6 °C, (b) 16 °C and (c) 22 °C. Each column is the average of three measurements, but error bars are omitted for temperatures above 6 °C, as the measurement uncertainties are too small to be depicted (0.2–3.6%/~0.05–14.4 Hz; see the supporting spreadsheet file). Due to space limitations, representative results are visualized in the article as column charts, while the individual sorption-desorption curves are included as Figures S1-S30 in the supporting information (SI) file.

A few tendencies are apparent in Fig. 6. Increasing the temperature of the drinks amplifies the sensor signal of both film-coated quartz resonators, but QCM FeCl₃-PhOH appears more efficient at differentiating alcoholic products based on their ethanol content. The frequency shifts caused by the high alcohol-by-volume spirits (i.e., whiskey, grape brandy) are nearly twice as large as those recorded for the beer, white wine and red wine, regardless of the temperature (see Fig. 6). It is noted, however, that the lightest alcoholic drink (beer) induces a larger sensor response than the stronger red wine, a trend also visible when comparing white wine to red wine and whiskey to grape brandy (see Table 2; Fig. 6). Attributed to variations in acetaldehyde, higher alcohols and heavy metal concentrations, this interesting result is interpreted later in the article. Another significant finding is that at mild and room temperatures (16–22 °C), the gas sensitivity of QCM PhOH-FeCl₃ declines and converges, lowering the differentiation of individual alcoholic beverages compared to QCM FeCl₃-PhOH. This means that producing metal phenolic films via reverse-order mixing (adding 5 mL of FeCl₃·6H₂O to 15 mL of C₆H₅OH) is unsuitable for alcohol sensing applications. The reasons for the poor performance of QCM PhOH-FeCl₃ are likely related to the growth behavior of the metal-organic films, which depends on the sequence of reagent addition²⁹. Despite having an identical chemical composition (see Figs. 2, 3 and 4) as reported elsewhere²⁹, the mixing order of phenol and FeCl₃ determines whether the reaction proceeds via coordination complex formation or oxidative polymerization. These distinct reaction pathways dictate the sorptive properties of the resulting coatings by altering their porosity and surface area.

Because the QCM PhOH-FeCl₃ configuration demonstrates low detection capabilities, the experiments were optimized by reducing the measurement cycles from 450 to 270 and selecting QCM FeCl₃-PhOH as the sole sensor to evaluate subtle variations of methanol content in alcoholic drinks, as illustrated in Figs. 7 and 8.

Fig. 7.

Resonance frequency shifts of QCM FeCl₃-PhOH induced by the saturated vapor of commercial and homemade alcoholic beverages containing sub-threshold contamination levels of methanol at (a) 6 °C, (b) 16 °C and (c) 22 °C. Each column is the average of three measurements, but error bars are omitted for temperatures above 6 °C, as the measurement uncertainties are too small to be depicted (0.08–4.1%/~1.3–10.2 Hz; see the supporting spreadsheet file). Due to space limitations, representative results are visualized in the article as column charts, while the individual sorption-desorption curves are included as Figures S31a-S90a in the SI.

Fig. 8.

Dynamic resistance shifts of QCM FeCl₃-PhOH induced by the saturated vapor of commercial and homemade alcoholic beverages containing sub-threshold contamination levels of methanol at (a) 6 °C, (b) 16 °C and (c) 22 °C. Each column is the average of three measurements, but error bars are omitted for temperatures above 6 °C, as the measurement uncertainties are too small to be depicted (0.1–2%/~0.03–0.6 Ω). Due to space limitations, representative results are visualized in the article as column charts, while the individual sorption-desorption curves are included as Figures S31b-S90b in the SI.

At refrigerated temperatures of 6 °C, methanol addition in the grape brandy triggers frequency downshifts, whereas for the other beverages, the resonance frequency first moves upwards (towards the baseline in air) and then decreases in accordance with methanol content. The constant dynamic resistance values (fluctuations of ± 1 Ω) suggest that the process is dominated by chemisorption, leading to the formation of individual adsorption complexes²⁵. Increasing the temperature to 16–22 °C substantially enhances the sensor response induced by the original alcoholic products. In most cases methanol addition attenuates the signal independently of its concentration (see Fig. 7). In line with the observed patterns in chilled conditions, a bidirectional change in resonance frequency (upshift-downshift) is also noted in the presence of methanol. Here, the chemisorption is replaced by physisorption (formation of adsorbate monolayers), as evident by the increased dynamic resistance (see Fig. 8). This occurs although the multicomponent alcohol compounds do not penetrate the bulk of the metal-phenolic film, in contrast to saturated methanol vapors²⁶. The observed resonance behavior of the QCM FeCl₃-PhOH indicates a deviation from linearity of frequency shifts, which questions its practical utility. This concern is further examined in Fig. 9.

Fig. 9.

Maximum-sensitivity calibration curves of QCM FeCl₃-PhOH to methanol injection in beer and grape brandy at refrigerated and mild temperatures. Each data point is the average of three measurements, but error bars are omitted at 16 °C, as the measurement uncertainties are too small to be depicted (1.7–3.6%/~2.2–10.2 Hz).

Although in most experiments the resonance response of QCM FeCl₃-PhOH is described by a third order polynomial fit, the sensor signal becomes linear with respect to methanol content in the beer and grape brandy at refrigerated temperatures. The coefficient of determination R² = 0.91–0.93 shows that the statistical model accounts for a substantial proportion of the variance, demonstrating strong predictive capability for the trends in the observed data (see Fig. 9). At 16 °C, R² = 0.36 suggests a weak linear fit and reflects the transition to a quasilinear resonance response. The experimental results documented in Figs. 6 and 7-8 are statistically significant with p < 0.05 (except for Figs. 6a and 8b), which undisputedly reveals the potential of MPF-QCMs as an analytical tool for alcohol profiling.

According to Fig. 9, the sensitivity of QCM FeCl₃-PhOH to adulterated beer is 5.5 Hz/mL at 6 °C and 18 Hz/mL at 16 °C. For denatured grape brandy, the sensitivity is 4.2 Hz/mL at 6 °C. Considering a short-term stability of ± 1 Hz/s (worst case scenario for MPF-QCMs) and a signal-to-noise ratio 3:1^25–27, the minimum variation in methanol concentration that the sensor system can resolve (the detection limit) depends on the specific drink and temperature. For the tested beer, the detection limits are 0.55 mL at 6 °C and 0.17 mL at 16 °C, while for the grape brandy, the limit is 0.71 mL at 6 °C. These concentrations are ~ 1.8–7.4 times lower than the maximum tolerable amount of MeOH in 100 mL beverages (10 g MeOH in 1 L EtOH or ~ 1.26 mL MeOH in 100 mL EtOH)³⁴.

Rather than using principal component analysis, we plotted the dependencies among resonance frequency, dynamic resistance and temperature directly. This preserves variances independent of maximum mathematical spread, clearly distinguishing the beverages and revealing the effect of methanol contamination. In addition, homemade distillates possess inherent complexity and further work is required to propose a universal QCM approach (for example, by optimizing the dynamic range of MPF-QCMs) to handle this concern and juxtapose the outcome with well-established chromatographic and spectroscopic methods. Given that the saturated vapors of various interfering substances, pure alcohols, water and their blends are extensively documented in existing literature^25–27, this research excludes alcohol/water baseline controls. Undoubtedly, such controls will be mandatory when the proposed sensor configuration transitions from research to commercialization, as calibration curves for various alcoholic beverages must be built to ensure the device is usable in real-world scenarios. Currently, our work serves as a proof-of-concept for the efficiency of MPF-QCMs in evaluating authentic fermented and distilled beverages, advancing knowledge in the field of alcohol sensors.

Our device outperforms liquid crystal-based sensors (detection limit of 1wt.% or 1.26 mL)²¹ and some portable NIR spectrometers (typical detection limit within 0.5–1% v/v)³⁶, but it appears less sensitive compared to SORS. However, this method reports a lowest detectable methanol concentration of 0.025%, but ref.²² fails to explain how this value is determined. As another benefit of MPF-QCMs, their measurement platform relies on passive vapor delivery for analyzing alcoholic beverages, thereby overcoming problems related to varying transparency of containers²² and light’s color and intensity²¹. The non-linearity observed in Figs. 7 and 8 might enable accurate quantification of alcoholic drinks by leveraging the excellent repeatability of frequency responses (error within ~ 0.2–3.6%; see the supporting spreadsheet file), provided robust calibration curves are used³⁷. Figures 7, 8 and 9 imply a probable beverage-specific temperature range where the signal induced by changes in methanol content becomes linear – a topic meriting future research.

Discussion

The organoleptic properties of alcoholic drinks, encompassing their aroma, flavor, chromaticity and viscosity, are governed by the chemical composition of raw materials and fermentation byproducts¹⁸. Ethanol acts as the primary solvent matrix, but its concentration dictates the solubility and headspace volatility of aroma compounds, often attenuating fruity esters while accentuating woody and spicy notes at higher alcohol-by-volume levels³⁸. Elevated ethanol concentrations enhance the solubility of lipophilic constituents, facilitating the retention of esters and higher alcohols within the liquid phase³⁹. Diluting ethanol decreases ester solubility, which increases their volatility. Apart from ethanol, temperature predominantly influences the solubility and volatility of congeners. Ethanol concentration determines the equilibrium solubility, but high temperatures significantly increase it due to an endothermic heat of solution that concomitantly enhances the volatility of the compounds⁴⁰. As temperature increases, molecules acquire sufficient kinetic energy to overcome intermolecular attraction forces, vaporizing at rates governed by their respective boiling points⁴¹.

The low volatility of chilled alcohol minimizes the matrix effect of minor volatile compounds, ensuring that the sensitivity of the QCM FeCl₃-PhOH is dominated by ethanol vapor. This yields the highest outputs for grape brandy (53.5 vol%) and whiskey (40 vol%). However, regardless of the lowest alcohol-by-volume, beer induces a larger sensor response compared to both white wine and red wine (see Fig. 6a). This is attributed to the highest acetaldehyde concentration (readily vaporizing due to its boiling point of 20.2 °C) and negligible content of heavy metals in the drink. Heavy metals interact with volatile aroma compounds to form heavier, stable and non-volatile metal-organic coordination complexes⁴². This explains why the heavy-metal-rich red wine with lower acetaldehyde concentration induces the weakest signal (see Table 2; Fig. 6).

At mild and room temperatures, white wine triggers the largest resonance frequency shifts among all of the tested low-alcoholic beverages, driven by the notable volatility of acetaldehyde at such temperatures. Given the comparable acetaldehyde levels in beer and white wine (~ 330–340 mg/L), the 2.5 times higher ethanol content in white wine increases the mass loading on the sensor surface., whereas the halved acetaldehyde content in red wine results in a weaker signal (see Fig. 6b and c). It is curious that these psychrometric conditions favor the sorption of whiskey over grape brandy, leading to a lower sensor signal triggered by this fruit spirit. This likely occurs because of the six-fold higher methanol content in the grape brandy compared to whiskey (see Table 2). The higher volatility of methanol increases the concentration of lighter, polar molecules in the vapor phase, which modifies the gas-phase composition and impedes the sorption of other heavier compounds.

It is practically important to evaluate how the resonance characteristics of QCM FeCl₃-PhOH are affected by the gradual addition of small volumes of methanol to the alcoholic drinks. As shown in Figs. 7 and 8, the injection of methanol diminishes the resonance response compared to the signals from the original products, a pattern consistent across all beverage types. One may conclude this is due to the reduced molecular weight of alcohols. Such a hypothesis fails to account for why the sensor signal remains weak as methanol content increases moderately (up to 0.5 mL) and why it eventually rises, but still remains lower than the signal induced by reference alcoholic drinks. The non-linear sensor behavior is primarily due to competitive adsorption, surface site saturation and/or complex matrix effects. The QCM FeCl₃-PhOH is more impacted by the high ethanol background and methanol competes for active sites (e.g., ferric groups²⁶), and displaces the larger adsorbing species that dominate the resonance response⁴³. Once the concentration of methanol molecules is enough to saturate the reactive adsorption centers, the major components (i.e., ethanol/water) are moved away, facilitating the preferential reaction between methanol and the metal-phenolic film²⁶. The opposite directions of resonance frequency and dynamic resistance shifts (i.e., frequency decrease is accompanied by resistance increase) unequivocally confirm the physical basis of our findings (see the SI pdf and spreadsheet files).

Materials and methods

Materials

Ferric (III) chloride hexahydrate (FeCl₃·6H₂O), phenol (99 wt% C₆H₅OH) and methanol (99.5 wt%) were delivered by Valerus Ltd. (Sofia, Bulgaria). AT-cut unpolished quartz resonators with 25.4 mm diameter, circular gold electrode structure and 5 MHz resonance frequency were supplied by Piezoquartz Ltd. (Sofia, Bulgaria). Commercial beer (Ariana, Zagorka JSC), white wine and red wine (Muscat & Dimiat, Trakia Estate 2024; Sakar Mavrud, Villa Bassarea 2011), and whiskey (100 Pipers, Chivas Brothers Ltd.) were purchased from Kaufland (Sofia, Bulgaria). Homemade grape brandy/rakiya (Merlot variety, 2024) was distilled and provided by local spirit producers.

Synthesis and deposition of metal-phenolic films

Preceding work revealed that metal-phenolic coatings, deposited from an aqueous solution of saturated ferric chloride and phenol in a 1:3 volume ratio, exhibit the highest sensitivity to alcohols combined with the lowest cross-sensitivity to psychrometric conditions^26,27. We employed the same mixture configuration with two modifications: the residence time of ~ 0.3 mL from the stock solution on the upper electrode of two nominally identical 5 MHz QCMs (denoted as QCM FeCl₃-PhOH and QCM PhOH-FeCl₃) was reduced from 2 h²⁵ to 30 min to optimize film thickness²⁶. The metal-phenolic film on the first QCM was spin-coated from a stock solution prepared by adding 15 mL of C₆H₅OH to 5 mL of FeCl₃·6H₂O, while the film on the second QCM was spin-coated by reversing the order of addition, since recent studies indicate that the sequence dictates the self-assembly and formation kinetics of the films²⁹.

Surface characterization

The morphology, chemistry and surface wettability of the metal-phenolic films were examined via scanning electron microscopy (SEM), energy dispersive spectroscopy (EDS), X-ray photoelectron spectroscopy (XPS) and contact angle measurements. Top-view SEM images were recorded through a LYRA3 GM field emission scanning electron microscope (TESCAN, Brno Czech Republic) at a magnification of 5 kX. Elemental analysis was performed with an energy dispersive spectrometer Bruker at 12.6 keV with an EDAX detector having a 1.3 mm² active area. X-ray photoelectron lines were recorded via an ESCALAB MK2 spectrometer using achromatic AlKα radiation at 1486.6 eV. The photoelectron core levels were corrected by subtracting a Shirley-type background, and quantified by the peak area and Scofield’s photoionization cross-sections. The chemical states were defined by deconvoluting the high-resolution spectra with XPSPEAK41 software. Тhe wetting state of metal-phenolic films was determined by measuring the static contact angles of 10 µL distilled water droplets via an optical system OCA 15EC (DataPhysics, Germany).

Sensor platform

Because the sensory attributes of alcoholic drinks, such as the volatilization of aromas, flavor perception and mouthfeel⁴⁴, are influenced by serving temperature²⁸, we performed the experiments using the setup illustrated in Fig. 10.

Fig. 10.

Setup for vapor profiling of alcoholic beverages. Nomenclature: 1 – QCM200 analytical instrument (SRS, USA); 2 – digital thermometer; 3–5 MHz MPF-QCM; 4 – styrofoam vessel; 5 – refrigerant gel pack; 6 – thermocouple; 7 – glass jar with drink-under-test; 8 – drink-under-test; 9 – sensor oscillator QCM25 (SRS, USA); 10 – box fixing the oscillator and MPF-QCM horizontal with respect to the drink-under-test; 11 – personal computer.

To encompass the optimal drinking temperatures of the tested beverages (4–22 °C for the specified beer, wines and spirits), the experimental trials were carried out at 6 °C, 16 °C and 22 °C. The used sensor platform was identical to the one described in ref²⁶., but the lowest temperature of 6 °C was maintained by first refrigerating 100 mL of the drinks and transferring them one at a time into a glass jar (height: 80 mm, diameter: 105 mm) placed in a firm contact with a frozen gel pack (see Fig. 10). The measurements were taken by mounting one of the MPF-QCMs in a quartz crystal holder. After ~ 40–60 s, the sensor readings in air equilibrated to Δf ~ ± 1 Hz and ΔR ~ ± 1 Ω and remained stable for another 180 s. The assembly was then embedded into the sheet-iron lid of the glass container, which exposed the film-coated upper electrode to the saturated vapor of the drink-under-test. As natural convective heat transfer between the side walls of the glass jar and the ambient air occurred, the drink’s temperature rose and settled at 6 °C (slightly higher than that during refrigeration). This procedure was repeated for the second MPF-QCM and later at 16 °C and 22 °C. These psychrometric conditions were obtained by removing the gel pack, setting the air conditioner and waiting for two hours to equalize the system’s temperature, verified with a digital thermometer. The described method was carried out in three distinct measurement cycles on both 5 MHz MPF-QCMs.

Based on the outcome and following comparative analysis, each beverage was gradually supplemented with methanol in incremental steps (0.25–0.5-1–2.5 mL). All tests were repeated in triplicate using only QCM FeCl₃-PhOH. Methanol volumes were selected to detect this dangerous adulteration at sub-threshold contamination levels or slightly above the tolerable dose³⁴. This approach simulates realistic scenarios of counterfeit alcohol distribution and illicit market practices.

Statistical analysis

The relative standard deviation (RSD), also known as the coefficient of variation (CV), was determined to evaluate the repeatability of sensor signals. After collecting the resonance frequency values in Microsoft Excel, the software was used to compute the means (µ) and standard deviations (σ) of individual datasets, providing the RSD = σ/µ. A Single-factor Analysis of Variance (ANOVA) test, performed using the same software, defined the statistical significance of observed variations among the datasets.

Conclusions

When a pair of 5 MHz QCMs, functionalized with metal-phenolic films deposited from a 1:3 FeCl₃·6H₂O to C₆H₅OH mixture, was exposed to saturated vapors of five alcoholic beverages, one of the sensors successfully detected and differentiated the ethanol concentration in each drink. It was found that at mild and room temperatures, the sensor performance of the QCM PhOH-FeCl₃ degraded and this device failed to distinguish between the chemical composition of the samples due to the reverse-order mixing of reagents prior to spin-coating. The QCM FeCl₃-PhOH, however, proved efficient in assessing the quality of alcoholic products based on congener content. Contamination of drinks with marginal or sub-acute amounts of methanol led to non-linear frequency response (bidirectional up- and downshifts). At refrigerated temperatures of 6 °C, the signal became linear with respect to the increasing methanol concentration in beer and grape brandy. Such an outcome presumes an optimal alcohol temperature for specific fermented and distilled beverages that suppresses chemical background adsorption and enables linear point-of-use sensing of hazardous denaturants, addressing the societal challenges associated with illegal alcohol.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

K.E. conceived the idea to study how the type and temperature of alcoholic drinks affects the resonance response of metal-phenolic film-coated QCMs. He organized and supervised the research, processed the raw data, contrived the scientific concept of the article and wrote its first and final versions. Y. L. performed all experiments and formulated a hypothesis on the underlying mechanism regulating the thermal sensitivity of tested alcohols. Both authors approved the scientific content of the manuscript prior to submission.

Funding

declaration.

This research was performed thanks to the financial support of Bulgarian National Science Fund under grant № KP-06-Н77/3/28.11.2023.

Data availability

All data needed to evaluate the conclusions in this paper are presented duly. Additional data related to this research may be requested from the authors.

Declarations

Competing interests

The authors declare no competing interests.