Selective adsorption of 1,3-dimethyltrisulfane (DMTS) responsible for aged odour in Japanese sake using supported gold nanoparticles

Gold (Au) nanoparticles (NPs) supported on SiO2 (Au/SiO2) were prepared by a practical impregnation method and applied as an adsorbent for 1,3-dimethyltrisulfane (DMTS), which is responsible for an unpleasant odour in drinks, especially Japanese sake. Compared with a conventional adsorbent, activated carbon, Au/SiO2 selectively reduced the DMTS concentration in Japanese sake without decreasing the concentrations of other aromatic components. DFT calculations revealed that the selective adsorption of DMTS occurred through the formation of a stable intermediate. The size of the supported Au NPs was controlled by the preparation conditions and determined from TEM observations and XRD measurements, and the size was ranged from 2.4 nm to 30 nm. Au/SiO2 having Au NPs with a diameter of 2.4 nm adsorbed DMTS the most efficiently. Smaller Au NPs showed better DMTS adsorption capabilities because larger amounts of Au atoms were exposed on their surfaces in the size range of this study. Langmuir-type monolayer adsorption and one-to-one binding of Au–S are proposed to occur based on an adsorption isotherm experiment. Even though significant differences of the fruity aroma score were not observed in the sensory evaluation between Au/SiO2 and activated carbon for this less aromatic Japanese sake, Au/SiO2 selectively decreased the DMTS concentration in the instrumental analysis.

Many factors affect DMTS formation. For example, the fermentation temperature, which affects rice dissolution and yeast viability, has a large influence on DMTS formation during the storage of sake 8,9 . The sulphur content in the rice grain used to brew sake also affects the DMTS level after storage 10 . Suppression of DMTS formation by the addition of inorganic ions such as Mg 2+ and SO 4 2− to the fermentation mash of sake 11 and beer 12 has been reported. Copper, which is used in the distillation apparatus for whisky, the boiling kettles for beer, or the bottles for Japanese sake, causes the level of DMTS to both increase and decrease [13][14][15][16] . Controlling such brewing parameters will be effective in reducing DMTS formation; however, this control may not always be possible according to the priorities of the product design. For example, lower fermentation temperature, which suppresses the formation of DMTS after storage of sake, decreases the quantity of sake produced, because alcohol fermentation is decelerated at a lower temperature. The content of sulphur in rice grains varies according to not only the genetic background but also cultivation area. It is sometimes difficult to control DMTS only by brewing conditions; therefore selective adsorbents to remove DMTS in sake is required.
In the practical brewing process of Japanese sake, activated carbon has been traditionally used to reduce unpleasant flavours, perform decolouration, and adsorb impurities. However, activated carbon adsorbs a broad range of molecules into its micropores 2 . The use of activated carbon sometimes results in the production of an unscented sake. Although the DMTS concentration in Japanese sake can be decreased by less than 50% by activated carbon 2 , esters, including EH, which makes "ginjoka", are also effectively adsorbed by activated carbon due to their hydrophobicity. Therefore, there is a demand for the selective removal of off-flavours without a decrease in the concentration of other components.
In this study, we first introduced noble metal nanoparticles (NPs) supported on metal oxides such as SiO 2 to adsorb DMTS in a model solution. Because noble metals and sulphur compounds are known to have a high affinity, supported metal NPs have been applied in the adsorptive desulfurization of fuels to ppm levels 17,18 . Very few examples of noble metal adsorptive desulfurization of beverages exist 19 . The Ag supported on zeolite reduced the sulphur concentrations of dimethylsulfane, dimethyldisulfane, and DMTS in immature whisky. The performance of adsorbents was limited because a selective adsorption mechanism has not been clarified. Here we have investigated the mechanism of DMTS adsorption on the Au NPs with various diameters in model solution with various DMTS concentrations to develop the effective adsorbents. Since the organoleptic thresholds of sulphur compounds are extremely low, i.e., lower than ppb (≈ μg L −1 ) concentrations, we have carried out deep desulphurization of Japanese sake containing those concentration levels. Moreover, Au and SiO 2 , which are permitted for use in Japanese sake by laws, were applied to the selective adsorption of DMTS in stored Japanese sake without concomitant removal of esters.

Results and Discussion
DMTS adsorption from a model solution by supported noble metals. A model solution containing 4.7 mg L −1 of DMTS in ethanol was used to screen the adsorbents. Although this concentration is much higher than the threshold concentration of DMTS in Japanese sake (0.18 μg L −1 ), we chose this condition for the initial screening experiments to enable accurate analysis by normal FID-GC. Au, Pt, Pd, and Ru NPs supported on SiO 2 , whose secondary particle sizes were approximately 100 µm, with a metal/sulphur atomic ratio of approximately 6 (all 1 wt%) were examined for adsorption at room temperature without stirring and shaking (Table S-1). All metals except for Ru showed adsorption abilities. Au/SiO 2 adsorbed almost all the DMTS within 24 h (Table S-1, entry 1). Au NPs, whose particle size was ranged from 3.0 nm to 10.9 nm, supported on several different kinds of oxide supports with secondary particle sizes of approximately 100 µm were tested (Table S-2, Fig. S-1). The results indicated that the smaller Au NPs showed faster and more extensive DMTS adsorption independent of the support material.
For further studies using the Japanese sake, we chose Au/SiO 2 because the use of Au and SiO 2 in Japanese sake brewing is permitted by law. SiO 2 is commonly used as a clarification agent, and Au is used for decoration as gold foil. TiO 2 could also support Au NPs as small as those supported on SiO 2 . Au/TiO 2 showed a high DMTS adsorption performance, while for the removal of DMTS from Japanese sake, SiO 2 is a more suitable support material than TiO 2 owing to its acidity. The pH value of Japanese sake, which contains many carboxylic acids, is approximately 4.3 20 . Thus, SiO 2 , which is an acidic material, is expected to have no effect on the tastes of Japanese sake. On the other hand, TiO 2 , which possesses amphoteric and hydrophobic properties, adsorbs organic acids and esters, similar to activated carbon. However, the deposition of small Au NPs (<5 nm) on SiO 2 as a support is difficult. Thus, we developed a novel practical preparation method for Au/SiO 2 , and a portion of the adsorption study was previously described 21 . The effect of the Au NPs particle size was explored. Au/SiO 2 having Au NPs with a diameter of 2.4 nm adsorbed 88% of the total DMTS after 24 h and 100% at equilibrium ( Fig. 2 and Table S-3, entry 1). As the size of the Au NPs increased, the amount adsorbed at equilibrium decreased. Au NPs with a diameter of 30.1 nm adsorbed only 1% after 1 day and quickly reached equilibrium. Moreover, gold foil (bulk) did not adsorb any DMTS under these conditions. The effect of temperature on the DMTS adsorption performances was investigated at 40 °C, 25 °C, and 10 °C (Fig. S-2). Au/SiO 2 was suitable for use at room temperature, although the adsorption rate was slightly higher at higher temperature. The average diameter of the Au NPs on SiO 2 was almost the same before and after the adsorption experiments ( Fig. S-3). The average diameters were 3.3 ± 0.9 nm before the experiment and 3.7 ± 1.0 nm after it. These results suggest that Au/SiO 2 was stable in solution. In addition, the amount of Au ions leached into the solution after the adsorption experiment, which is listed in Table S-3, entry 1, was measured by microwave plasma atomic emission spectroscopy (MP-AES). The concentration of Au in the solution after filtration through a membrane filter (0.5 µm) was 4 mg L −1 , namely, 3.2% of the total Au atoms leached into the solution. As mentioned below, the amounts of Au that leached in Japanese sake was below the detection limit of MP-AES because the concentration of DMTS in Japanese sake is lower than that in the model solution, and Au/SiO 2 is stable even in the Japanese sake.
Competitive adsorption between DMTS and EH was carried out under similar conditions (Table 1). EH is a typical compound possessing a pleasant flavour. Using Au/SiO 2 , 100% of the total DMTS was adsorbed within 3 days, while 0% of the EH disappeared (entry 1). Au/Al-MCM-41 also adsorbed DMTS selectively with a slightly faster adsorption rate (entry 2). In contrast, Au/C adsorbed 21% of the EH as well as 100% of the DMTS, possibly due to the hydrophobic character of the pore interior. The Au NPs supported on SiO 2 -based materials selectively adsorbed DMTS from a solution mixture containing EH. The adsorption rate and amount were not reduced even with the coexistence of EH.
To understand the adsorption mechanism, an adsorption isotherm was created from several different initial concentrations of DMTS (Table S-4). This experiment was carried out with the Au NPs having a mean diameter of 3.5 nm supported on SiO 2 , where approximately 40% of the Au atoms were located on the surface. When the standard condition (DMTS = 4.7 mg L −1 ) was used (entry 1), the apparent Au/S atom ratio was 5.9, but the actual ratio to surface Au atoms was approximately 2.4; virtually 100% of the total DMTS was adsorbed and removed from the solution. When the initial amount of DMTS was doubled (entry 2), namely, the exposed Au/S ratio was approximately 1.2, 29% of the DMTS remained in solution at equilibrium. When the initial concentration of DMTS was increased to 28.1 mg L −1 (entry 4), 36% of the DMTS was adsorbed; under this condition, the Au/S and surface Au/S ratios were 0.96 and 0.4, respectively. In the solution with a DMTS concentration above 9.4 mg L −1 , the Au NPs seemed to be fully covered by DMTS molecules. Figure 3 shows the adsorption isotherm plotted from the adsorbed amounts and concentration of DMTS at equilibrium listed in Table S-4, suggesting the occurrence of monolayer Langmuir-type adsorption. A Langmuir plot is given by Eq. 1.
In this equation, x is the amount adsorbed, C is the concentration of DMTS at equilibrium, a is the amount adsorbed at saturation, and K is the adsorption constant. The Langmuir plot derived from the adsorption isotherm (inset of Fig. 3) was well fitted by Eq. 1. The adsorption constant, K, and the amount adsorbed at saturation, a, were determined to be 64200 L mol −1 and 0.346 μmol, respectively, with 0.52 mg of Au under this condition.  The surface coverage, which corresponds to the ratio of DMTS adsorbed at saturation, can be calculated based on the Langmuir equation (Table 2). The coverage can also be calculated based on the number of atoms assuming that the proportion of Au atoms on the surface is 40% for an icosahedral Au NP and that the Au/S ratio for adsorption is 1. Interestingly, the coverage values are very close in each case, which means that the assumptions of monolayer Langmuir-type adsorption and one-to-one Au-S adduct formation are reasonable 22 .
DFT calculation of DMTS adsorption on a Au 24 cluster. DFT calculations were performed for DMTS adsorption on a Au 24 cluster model system. The details of the calculation conditions are summarized in the Supplementary Information. The calculated reaction pathway is shown in Fig. 4. Based on this reaction pathway, the two S-S bonds in DMTS completely dissociated on the surface of the Au 24 cluster, and the first and second activation barriers of this reaction are 10.2 and 103.3 kJ mol −1 , respectively. Consequently, two CH 3 S groups and a S atom are generated on the Au 24 cluster. A stapler-like Au-S-Au-S-Au bonding structure is formed on Au 24 in the final structure (FIN) of the reaction pathway. This stapler structure is a well-known bonding mode for gold cluster complexes. Additionally, the adsorption energies of DMTS and EH on Au 24 were compared. The structures and adsorption energies are summarized in Fig. S-4. The adsorption energies of these two substrates are almost   Instrumental analysis and sensory evaluation of Japanese sake. The utility of the supported Au NPs as adsorbents were examined with Japanese sake. Table 3 shows the results of instrumental analysis of DMTS and the primary flavour compounds in filtered Japanese sake. A control sake with aged odour contained 0.25 μg L −1 of DMTS, which is higher than the threshold value (0.18 μg L −1 ). The control sample (entry 1), also contained 1.  Table 1. Other aroma compounds, such as 3-methylbutyl acetate (another "ginjoka" component), ethyl acetate, and 3-methylbutan-1-ol, all of which are basic flavour components of sake, were hardly affected by the Au NPs. After the adsorption experiment listed in Table 3, entry 2, the concentration of Au in the Japanese sake was below the detection limit of MP-AES. The influence of supported Au NPs on sake flavour was also investigated by sensory evaluation (Table 4). Concerning the evaluation terms of odour, we examined four attributes; fruity aroma, aged odour, sulphur smell, caramel-like smell. DMTS is considered to contribute to sulphur smell and aged odour of Japanese sake. Fruity aroma is mainly attributed to EH and 3-methylbutyl acetate, and caramel-like smell in aged Japanese sake is mainly caused by sotolon. The adsorbents significantly decreased the aged odour and sulphur smell. For example, the sulphur smell score was 1.67 for the control (entry 1), 0.17 for both Au NP adsorbents (entries 2 and 3), and 0.33 for activated carbon (entry 4). The results well reflected the instrumental analysis (Table 3). However, significant differences were not observed in the fruity aroma in spite of the differences in EH levels between samples. This may be because the concentration of EH was too low. Actually, the sensory scores of fruity aroma were below 1 (1 = weak) for all samples. It is needed to use sake samples with higher level of EH to investigate the influence on fruity aroma. Similarly, the scores of caramel-like smell were low and not significantly different among samples. Sotolon is formed through a long period of storage (more than several years) 2 , and the test samples were not considered to contain enough level of sotolon to be perceived. We also examined taste attributes (body, dryness, smoothness, and aftertaste), and significant differences were not observed, which means that none of the adsorbents had a strong effect on the taste. It was demonstrated that Au/SiO 2 selectively decreased the aged odour and the sulphur smell without significant changes in other sensory properties.

Conclusion
To remove one of the off-flavours caused by DMTS from stored Japanese sake without concomitant decreases in the concentrations of other aromatic compounds, we investigated the DMTS adsorption performances of noble metal NPs, especially Au NPs, supported on SiO 2 . Smaller Au NPs supported on metal oxide materials showed better DMTS adsorption performances in the size range of this study. Moreover, Au/SiO 2 selectively adsorbed DMTS from a solution mixture containing EH. The mechanism of DMTS adsorption by the Au NPs was analysed using a Langmuir plot and DFT calculations. The Au NPs were covered by a monolayer of DMTS molecules, and one-to-one Au-S adducts were generated. The adsorption energies of DMTS and EH on a Au 24 cluster were almost the same, while the generation of a more stable intermediate from the dissociation of the S-S bonds of DMTS led to the adsorption selectivity. Au/SiO 2 also selectively reduced the DMTS concentration in stored Japanese sake. Significant differences in sensory evaluations were not observed in the fruity aroma for the less aromatic Japanese sake used in this study. The aged odour and the sulphur smell became weaker without significant changes in other sensory properties using Au/SiO 2 . It is expected that Au/SiO 2 is more effective to use for drinks with highly aromatic concentration.

Methods
Preparation of adsorbent. Au/SiO 2 was prepared by an incipient wetness impregnation method recently developed by us 21 . The typical preparation procedure is as follows: a Au-amino acid complex was obtained by mixing a HAuCl 4 •4H 2 O ethanol solution with a NaOH and amino acid ethanol solution. The mixture was left at −18 °C for 12 h, during which the Au-amino acid complex formed as a precipitate. The Au-amino acid complex   Table 3. Instrumental analysis of Japanese sake with aged odour and treated with adsorbents. a 5 g/500 mL adsorbent was added, left for 24 h, and then filtered. b 1 g/500 mL adsorbent was added, left for 24 h, and then filtered. c 0.5 g/500 mL adsorbent was added, left for 1 h, and then filtered. Adsorption experiments with a Japanese sake containing 0.25 μg L −1 DMTS. Typical procedure: A control sake sample was prepared by blending a 4/1 ratio of a sake stored at 40 °C for one month and a sake stored at 15 °C for 5 years. This blend was prepared because the sake stored at 40 °C for one month did not contain enough DMTS to be for this compound to be detected, whereas the sake stored for 5 years at 15 °C contained more DMTS. To this blend was added adsorbent, and the mixture was left at room temperature for 24 h. The adsorbent was removed by filtration using a 0.45 μm membrane filter. The filtrate was analysed by GC and sensory evaluation. Sensory evaluations were performed by 6 well-trained panellists. Detailed information on the instruments, materials, Table S1 and S2, and Fig. S1 and S2 are available in the Supplementary Information.  Table 4. Sensory evaluation of Japanese sake with aged odour and treated with adsorbents. a 1 g/500 mL adsorbent was added, left for 24 h, and then filtered. b 5 g/500 mL adsorbent was added, left for 24 h, and then filtered. c 0.5 g/500 mL adsorbent was added, left for 1 h, and then filtered. A,B Values with different letters are significantly different at p < 0.05 (p is the probability of obtaining the observed results, or more extreme, under the null hypothesis, i.e., no difference between samples) according to a Tukey-Kramer honestly significant difference (HSD) test.