Potassium hydrogen tartrate (≥ 99 %), L-(+)-tartaric acid (≥ 99.5 %), riboflavin (≥ 99 %), sodium sulphide (≥ 99.5 %), and sodium methanethiolate (≥ 90 %) were all purchased from Sigma-Aldrich (Castle Hill, NSW, Australia). Ethanol (99.5 %) was obtained from ChemSupply (Gillman, SA, Australia). Methanol (LC grade solvent), acetonitrile (LC grade solvent), and glacial acetic acid were sourced from Honeywell (North Ryde, NSW, Australia), Sigma-Aldrich (Castle Hill, NSW, Australia), and Ajax Finechem (Taren Point, NSW, Australia), respectively. Solutions and dilutions were prepared using 18.2 Milli-Q water.

All samples were prepared in an anaerobic hood, using nitrogen gas (99.99 % purity, BOC Gas & Gear, Wagga Wagga, NSW, Australia) to ensure oxygen minimisation, and the initial dissolved oxygen concentrations in the tartrate-buffer and wine samples were 0.22 ± 0.09 and 0.32 ± 0.07 mg/L, respectively. Oxygen concentrations were measured with a PSt3 oxygen sensor within a PSt3-oxygen sensor dipping probe and Fibox 4 trace meter (Precision Sensing GmbH, Regensburg, Germany). The tartrate-buffer solutions consisted of potassium hydrogen tartrate (0.011 M) and tartaric acid (0.008 M) added to either Milli-Q water or to 12 % (v/v) aqueous ethanol. The pH of these solutions was 3.2 ± 0.1, measured by a S20 SevenEasy™ pH meter (Mettler Toledo, Port Melbourne, VIC, Australia). The commercial wine used was a de-alcoholised 2024 Chardonnay (0.4 % (v/v) ethanol) labelled as a product of South Australia. It was originally bottled in 750 mL glass bottles sealed with screw cap closures. The general specification of the wine is shown in Table S1, and it was used without modifications of its pH (3.2) or sulphur dioxide concentration ( i.e., free SO ₂ of 27 ± 3 mg/L, Table S2). Eight treatments (described in the following section) were prepared for the wine, and also for the tartrate-buffer, with each treatment consisting of three replicate bottles. Due to limited irradiated incubator capacity, treatments were prepared in four batches of two. For each de-alcoholised Chardonnay batch ( i.e., two treatments), three 750 mL bottles were opened and homogenised inside the anaerobic hood. Either water (266.4 mL) or ethanol (266.4 mL) was mixed with de-alcoholised wine (1,953.6 mL) to provide final alcohol concentrations of 0.4 % (v/v) and 12.4 % (v/v), respectively, as measured using an Alcolyzer Module (Anton Paar GmbH, Graz, Austria). Three portions of 370 mL were then transferred into three separate 375 mL flint glass bottles (Plasdene Glass-Pak, Milperra, NSW) to provide the replicate bottles for a given treatment. Tartrate-buffer solutions (370 mL) were also aliquoted into 375 mL flint glass bottles. A riboflavin concentration of 0.5 mg/L, freshly prepared from an aqueous riboflavin stock solution (203 ± 3 mg/L), was added directly to all samples. This relatively high concentration enabled sufficient analysis time points to be conducted for accurate determination of the rapid riboflavin decay kinetics. It also facilitated the production of significant concentrations of detrimental sulfhydryl aroma compounds within the timeframe of the storage experiment. Each bottle was then sealed using screwcap closures with tin liners (Orora, Hawthorn, VIC, Australia) under a nitrogen gas atmosphere. The final concentrations of riboflavin were 0.50 ± 0.02 mg/L for the tartrate-buffer solutions and 0.56 ± 0.01 mg/L for the Chardonnay wines.

All samples ( i.e., tartrate-buffer solution (T), tartrate-buffer solution with 12 % (v/v) ethanol (TE), de-alcoholised wine with 0.4 % (v/v) ethanol (DA) and de-alcoholised wine with 12.4 % (v/v) ethanol (DA+E) were exposed to different commercial light sources from Philips lighting (Shanghai, China), being: a cool white fluorescent light tube (18 W, 1,350 lm, 4,000 K); cool white LED (8 W, 1,050 lm, 4,000 K), and LED with a yellow filter (10 W, 1,050 lm, 2,200 K) hereafter referred to as ‘yellow LED’ in this study. The emission spectrum of each light tube was measured over the wavelength range of 350–1000 nm (Figure 1A), determined by FieldSpec 4 (Malvern Panalytical Ltd., Worcestershire, UK) coupled to a VNIR (type 512 element silicon) photodiode array detector (Malvern Panalytical Ltd., Worcestershire, UK), and run by ViewSpec Pro version 6.2 software (Malvern Panalytical Ltd., Worcestershire, UK). Samples were positioned to receive a light intensity of 25 μmol/m ²/s from each light source, measured by a LI-COR Biosciences LI-185A photometer (Lincoln, NE, USA). This intensity was chosen to provide a standardised experimental condition and falls within the range of typical illuminance observed for retail shelves in supermarkets and department stores, which has been reported to be approximately 5–30 μmol/m ²/s (Jolley-Rogers et al., 2017). During light exposure, all samples were maintained at 20 ± 2 °C inside a temperature-controlled incubator over a period of hours to 7 days for tartrate-buffer solutions ( i.e., T and TE) and a period of 1 to 14 days for the Chardonnay ( i.e., DA and DA+E). Control samples were stored under the same conditions but covered with aluminium foil to ensure darkness.

During sampling, nitrogen gas was continually blown into the top of the bottles for 5 min to minimise any increase in dissolved oxygen concentrations. Dissolved oxygen concentrations in all samples were maintained below 0.5 mg/L throughout the trial (data not shown). Another set of wine samples in each treatment was also prepared and exposed to either the cool white fluorescent or cool white LED sources for 1 day, or exposed to the yellow LED source or stored in darkness for 14 days without opening the bottles. These wine samples were then used for volatile sulphur compound analysis and sensory evaluation. Experimental factors were coded as shown in Table 1.

Ultrahigh-performance liquid chromatography (UPLC, Waters, Milford, USA), using a Waters Acquity BEH C18 column (2.1 mm, 50 mm, 1.7 µm, Waters, Milford, USA), connected to a photodiode array detector (PDA, Waters, Milford, USA), was used to measure concentrations of riboflavin as outlined in Dias et al. (2012). Free and total sulphur dioxide (SO ₂) concentrations were measured using colorimetric test kits (Thermo Fischer Scientific, Scoresby, VIC, Australia) and a Konelab 20XT automated analyser (also Thermo Fisher Scientific). Total phenolic compounds (TPC) were determined by UV-visible spectrometry at a wavelength of 280 nm (Iland et al., 2004). The volatile forms of H ₂S and MeSH, subsequently referred to as ‘free’ forms within this study, were quantified by gas chromatography (GC, Agilent 7890B GC system, Agilent Technologies Australia, Mulgrave, VIC, Australia) coupled to a sulfur chemiluminescence detector (SCD, Agilent 355 SCD, Agilent Technologies Australia), using a DB-Sulfur SCD column (60 m, 0.32 mm, 4.2 µm, Agilent J&W Scientific, Agilent Technologies Australia). The analysis was conducted as per Kontoudakis et al. (2017), which is based on a modified protocol of Franco-Luesma and Ferreira (2014). Quantitation of the free sulfhydryl compounds was performed using ethyl methyl sulfide as an internal standard with internal standard curves prepared in either tartrate-buffer solution or tartrate-buffer solution with 12 % (v/v) ethanol for analysis of DA and DA+E samples, respectively.

Sensory analysis was performed after the exposure of wine samples to cool white fluorescent lighting for 1 day, cool white LED for 1 day, or yellow LED for 14 days. A control sample ( i.e., stored in darkness) also underwent sensory analysis at day 14. Pivot ^© Profile sensory methodology was utilised as outlined in Pearson et al. (2020). An experienced tasting panel was composed of 11 assessors (four females and seven males, aged 25–65), from Charles Sturt University and the Gulbali Institute (Wagga Wagga, NSW, Australia). All the panellists were familiar with the Pivot ^© Profile sensory method and had all previously been involved in sensory studies that had investigated reductive characters in wine. The sensory analysis was conducted in an open-plan, specialised sensory room at 20 ± 2 °C in two sessions; the four different DA samples ( i.e., F, L, Y and D) were assessed in the early afternoon, and the four different DA+E samples were evaluated in the late afternoon on the same day. A 30 mL volume of wine from a single bottle of each treatment was served in 220 mL ISO XL5 wine glasses labelled with random three-digit codes at ambient temperature, and extra wine was available upon request. Water and crackers were also provided. The panellists recorded differences in appearance, aroma, and palate descriptors when comparing the control pivot ( i.e., dark sample) to all the different treatments ( i.e., F, L, Y and D), according to Thuillier et al. (2015).

Statistical analyses were conducted using IBM SPSS Statistics software (version 27, Chicago, IL, USA). The paired-sample t-test was used to determine significant differences ( p < 0.05) before and after light exposure in each treatment for total phenolic compounds, and free and total SO ₂ concentrations. One-way analysis of variance (ANOVA) tests with post-hoc multiple comparisons were applied to assess the differences amongst rate constants and half-life data across treatments using Tukey’s test for equal variances and Dunnett’s T3, where unequal variances were evident. Correspondence analysis (CA) was performed on the sensory analysis data using Microsoft Office Excel version 2302 (Microsoft, Redmond, Washington, USA) installed with the XLSTAT statistical software (Lumivero, Denver, Colorado, USA). The quoted uncertainty is the standard deviation of three replications for a given treatment.

The emission characteristics of the cool white fluorescent, cool white LED, and yellow LED are shown in Figure 1A. The cool white fluorescent tube emitted wavelengths of light spanning the UV to visible range, with emission peaks at 409, 438, 488, 546, 588, 612 and 708 nm. In contrast, the cool white LED only emitted visible light within the 400–760 nm wavelength range, and had maxima at 451 and 591 nm, while the yellow LED only emitted light with wavelengths above 500 nm, and with a single maximum at 591 nm. As previously mentioned, the photochemical reaction is initiated by the absorption of light energy, typically in the form of photons, and the photon energy (E) can be calculated using E = hc/λ, where h is Planck’s constant, c is the speed of light and λ is the wavelength. Figure 1B shows the relative photon energy spectra for each light source calculated over the wavelength range most relevant to the photoexcitation of riboflavin ( i.e., 350–520 nm). The integrated area under the spectra curves in Figure 1B corresponded to ratios for photon energies of 9:45:1 for cool white fluorescent:cool white LED:yellow LED.

A graph of a person's body AI-generated content may be incorrect.

The use of the tartrate-buffer solution allowed the composition of the sample matrix to be controlled for the preliminary experiments of this study, which focused on the impact of the different light sources and ethanol concentrations on riboflavin degradation. Figure 2A shows the rapid depletion of riboflavin in the tartrate-buffer solutions after only 2 h of exposure to the cool white LED or fluorescent light sources. This occurred regardless of the ethanol concentration of the samples. The yellow LED source led to slower riboflavin loss compared to the cool white fluorescent and cool white LED samples, but after seven days, 30–40 % more riboflavin was lost relative to samples stored in darkness (Figure 2B). As expected, the riboflavin concentrations in samples stored in darkness remained stable throughout the trial ( i.e., D/T and D/TE, Figure 2B).

Exposure of Chardonnay samples to the cool white LED or fluorescent light sources also resulted in riboflavin loss, with more than half the initial riboflavin being degraded within 5 h of irradiation, irrespective of the presence of ethanol. After this, the rate of riboflavin loss decreased (Figure 2C). Conversely, Chardonnay samples exposed to the yellow LED source ( i.e., Y/DA and Y/DA+E, Figure 2D) only had a 10 % loss in riboflavin concentration after 14 days of light exposure, again with no influence of ethanol concentration noted. No changes were observed in samples stored in darkness throughout the storage period ( i.e., D/DA and D/DA+E, Figure 2D).

A graph of different colored lines AI-generated content may be incorrect.

The decrease in riboflavin concentration in the tartrate-buffer solution was best modelled by first-order kinetics ( R ² = 0.985–0.998, Table 2), in agreement with a previous study (Fracassetti et al., 2019). The tartrate-buffer solution exposed to cool white LED had the highest riboflavin decay rate, followed by fluorescent and yellow LED. This order for light sources was consistent with their corresponding photon energy emission over the wavelength range of 350–520 nm (Figure 1B). That is, the cool white LED source emitted a photon energy approximately 5-fold higher than for fluorescent and 45-fold higher than for the yellow LED. The presence of ethanol led to significantly ( p < 0.05) faster degradation of riboflavin when the sample was exposed to light from cool white LED or yellow LED, but not light from cool white fluorescent tube. This difference is possibly due to the unique wavelengths less than 400 nm emitted by the cool white fluorescent tube compared to the other light sources. The half-life for riboflavin decay was approximately 1000-fold longer when the tartrate-buffer solution was exposed to the yellow LED compared to the cool white LED. The presence of ethanol decreased the riboflavin half-life by almost half, particularly when the sample was exposed to either the cool white LED (L/TE vs L/T, Table 2) or the yellow LED (Y/TE vs Y/T, Table 2). A previous study in a non-wine matrix reported that the decay rate of riboflavin in water was higher than in ethanol, and this was attributed to the lower viscosity of water (Ahmad et al., 2015). However, the ethanol concentrations ( i.e., 100 vs 12 % (v/v)) and the pH ( i.e., 7 vs 3.2) reported by Ahmad et al. (2015) were different to those employed in the current study. A previous investigation also showed that the photodegradation of riboflavin in aqueous solutions became marginally slower as the pH increased from 3 to 4, but then rapidly increased from pH 6 to 10 (Ahmad et al., 2004).

It was not possible to calculate the riboflavin decay rates for the Chardonnay samples exposed to yellow LED or stored in darkness, as ≤ 10 % of the riboflavin was degraded. For the remaining treatments, the riboflavin decay rate in Chardonnay appeared to have two distinct kinetic phases: the 0–7 h period, which has been assigned as the 1st phase, and the 7–24 h period assigned as the 2nd phase. The fit of modelled decay to the experimental data for these samples was better using second order ( i.e., R ² > 0.9, Table 3) rather than first order or zero order ( i.e., R ² < 0.9, data not shown) kinetics for both phases. This two-stage second-order kinetic behaviour of the riboflavin decay was in contrast to the decay in the tartrate-buffer solutions that showed single-stage first-order decay (Table 2). In a previous study reporting the photodegradation of riboflavin in Chardonnay wine (13 % (v/v) alcohol), riboflavin was also shown to follow second-order decay, albeit with only one kinetic phase rather than two (Vongluanngam et al., 2024). For the Chardonnay treatments, the samples with higher alcohol and exposure to the cool white LED tube (L/DA+E) had the highest decay rates for both phases, consistent with the order of decay rates observed in the tartrate-buffer solutions treatments ( i.e., L/TE in Table 2). The half-lives calculated from the 1st phase of degradation were 3–5 h for the Chardonnay samples exposed to either cool white LED or fluorescent tube (Table 3). However, when the concentration of riboflavin fell below 0.25 mg/L, coinciding with the transition to the 2nd kinetic phase, the riboflavin half-life for these treatments extended from hours to a day or greater (Table 3).

The photo-induced loss of riboflavin in the Chardonnay was much slower than that observed in the tartrate-buffer solutions for a given light treatment. This was most likely due to the presence of phenolic compounds in the Chardonnay wine, as such compounds have been shown to effectively shield riboflavin from light exposure (Fracassetti et al., 2019; Maujean & Seguin, 1983b). A significant decrease in total phenolic compounds ( p < 0.05) was evident in the Chardonnay samples that had substantial riboflavin loss ( i.e., irradiated with the cool white light sources) (Table S2), with the largest decrease being for the L/DA+E treatment. The fate of these phenolic compounds was not determined, but it may have been a consequence of light-induced polymerisation or degradation reactions (Lan et al., 2021), causing a decrease in their molar absorptivity at 280 nm. A decrease in total SO ₂ concentration was also observed for the cool white light treatments (Table S2); however, there was no significant ( p < 0.05) change in free SO ₂ concentration. The different outcomes for free and total SO ₂ may have been a consequence of weakly bound bisulphite adducts ( e.g., aldehydes and ketones) replenishing some of the free SO ₂ lost due to light-induced effects (Waterhouse et al., 2016).

Free MeSH and H ₂S are known to be key reductive off-aroma compounds generated during the photodegradation of riboflavin in white wine (Vongluanngam et al., 2025). Quantification of these compounds was conducted at two time-points: at bottling and after the irradiation period; i.e., day 1 for samples stored under the cool white LED or fluorescent light sources and day 14 for the samples stored under the yellow LED source or in darkness. At bottling, the concentrations of both H ₂S and MeSH in DA and DA+E (Figures 3A and 3B) were considered low relative to their quoted aroma thresholds, i.e., 1.6 and 3.1 µg/L, respectively (Siebert et al., 2009; Solomon et al., 2010).

After one day of cool white LED or fluorescent light exposure, the free MeSH concentration increased 3–4-fold above its aroma threshold in all wine samples, regardless of ethanol concentration (Figure 3A). No significant differences were observed in the free MeSH concentrations across the cool white lighting treatments, despite the differences in their riboflavin decay (Table 3). For the wine exposed to the yellow LED source for 14 days, the free MeSH concentration increased to twice the aroma threshold level; however, this only occurred in the higher ethanol concentration treatments (samples Y/DA+E in Figure 3A). The reason for this difference is not certain, especially given the similar decay rates of riboflavin under yellow LED (Y/DA and Y/DA+E samples in Figure 1D), but it may be a consequence of ethanol impacting the yield of MeSH from methional (Figure S1). Trends for more MeSH with higher ethanol were also observed under fluorescent and cool white LED, though the differences were not significant, perhaps due to the shorter incubation period (1 day vs 14 days). The wine samples stored in darkness (D/DA and D/DA+E) showed no increase in MeSH concentration in the same period (Figure 3A).

Exposure of wines to the cool white light sources increased the free H ₂S concentration far above the aroma threshold after one day, irrespective of ethanol concentrations (samples L/DA, F/DA, L/DA+E, and F/DA+E in Figure 3B). Although there were no significant differences in H ₂S concentrations between treatments, there were trends to higher H ₂S concentrations with cool white LED exposure and/or lower ethanol concentrations. The large uncertainty associated with the measurement of H ₂S concentrations from wine samples exposed to light has been previously reported (Vongluanngam et al., 2025). Factors contributing to the uncertainty may be the sensitivity of H ₂S to oxygen ingress and/or H ₂S being an intermediate species in a suite of photo-mechanistic pathways rather than a stable terminal product. As expected, the concentrations of H ₂S in samples exposed to yellow LED and control samples remained relatively low (below aroma threshold levels) throughout the 14 days of storage (Figure 3B).

A group of different colored bars AI-generated content may be incorrect.

A sensory panel characterised the wine samples using 22 descriptors for the de-alcoholised Chardonnay (DA) and 20 descriptors for the de-alcoholised Chardonnay with ethanol added (DA+E). Correspondence analysis (CA) was employed to analyse sensory data and produce Figures 4A and 4B. For the DA samples (Figure 4A), 95.86 % of variance was explained by Factor 1 (89.28 %) and Factor 2 (6.58 %), and 97.43 % of variance for DA+E samples (Figure 4B) was described by Factor 1 (85.90 %) and Factor 2 (11.53 %). The distribution of wine samples and descriptors in both Figures 4A and 4B suggested that Factor 1 was predominantly driven by the exposure of samples to wavelengths < 500 nm and associated with the photodegradation of riboflavin. Samples exposed to yellow LED, and samples stored in darkness, clustered on the left-hand side of this axis, were more closely associated with descriptors such as ‘tropical/pineapple’ and ‘stone fruit’. This was the case for both 0.4 % (v/v) (Figure 4A) and 12.4 % (v/v) (Figure 4B) ethanol concentrations. The close proximity of these two treatments in both CA figures was a result of similar descriptors being assigned to them by the sensory panellists. These results were consistent with the lower concentrations of free H ₂S and MeSH in the samples not exposed to cool white light, and therefore, reductive characters were not perceived. Although the sample with 12.4 % (v/v) ethanol and exposure to the yellow LED source accumulated a concentration of free MeSH that was marginally above the reported aroma threshold (sample Y/DA+E in Figure 3A), panellists did not report reductive characters in these samples.

Conversely, samples exposed to cool white LED and fluorescent light, irrespective of the presence of ethanol, were both located on the right-hand side of the Factor 1 axis and were associated with reductive attributes. Interestingly, Chardonnay samples irradiated by cool white LED were projected into the positive aspect of Factor 2 and associated with ‘sulphur/rotten egg’ and ‘reductive’ descriptors. This coincided with trends to higher H ₂S production in the same samples (Figure 3B). Regardless of ethanol concentrations, Chardonnay samples exposed to fluorescent light were located lower on the Factor 2 axis and were more aligned with the ‘cabbage’ attribute. The association of cool white LED and fluorescent light-exposed samples with these reductive sensory features were consistent with the concentrations of MeSH and H ₂S being well in excess of their corresponding aroma thresholds (Figures 3A and 3B). Panellists also rated the light-exposed wines, particularly those exposed to the cool white light treatments, as having less yellow colouration, which was consistent with previous studies (Cáceres-Mella et al., 2014; Vongluanngam et al., 2025).