The Kinetics of Autoxidation in Wine

Abstract

The kinetics of autoxidation in wine begins with Fenton (1876) who observed that tartaric acid could be oxidized in the presence of iron without peroxide if left in air. Rodopulo (1951) demonstrated that iron tartrate complexes added to wine promoted more extensive oxygen consumption than the molar equivalent of inorganic ferrous or ferric salts. The role of iron complexes in the activation of oxygen, the formation of reactive oxygen species and the initiation of autoxidation are crucial for understanding wine oxidation kinetics. Mechanisms based on hydroxyl radicals versus the ferryl species are likely to have different oxidation products of wine components based on pH effects. The ferryl ion, hydroxyl radical, and tartaric acid radical are proposed as key intermediates in the proposed general mechanism for hydrogen peroxide formation and the autoxidation of wine components. A quantitative kinetic description is presented for the autoxidation of tartaric acid and extended to other acid components as potential ligands. This chapter explores the theoretical considerations of iron complexes formation, oxygen activation, an autoxidative mechanism, and experimental measurements of tartaric acid oxidation as the basis of autoxidation in wine.

1. Introduction

Wine is an interesting chemical reaction system, in part due to its tartaric acid content. The oxidation of wine is known to be autoxidative [1], stimulated by Fe and Cu ions [2], and is thought to involve Fenton chemistry [3], but neither the rate nor the extent of oxygen consumption can be predicted from a knowledge of pH, metal, phenolic or organic component concentrations. Autoxidation is a spontaneous reaction in air and a radical chain reaction sequence [4, 5].

What is known is that the rate of oxygen consumption in wine is relatively slow in the natural pH range between 3.0 and 4.0. The underlying tartaric oxidation with Fe(II) and oxygen can describe the autocatalytic radical chain reaction sequence, with a distinct initiation stage, a faster, accelerating oxygen consumption propagation stage and a termination stage due to the complete consumption of oxygen and/or Fe(II). It appears that during the propagation stage, ferryl ion, hydroxyl radicals, and/or potentially tartaric acid radicals that are the origins of generating hydroxyethyl radicals when ethanol is present. These hydroxyethyl radicals would lead to an array of selective downstream reactions leading to the collective aged composition of wine. The commonly accepted formation of acetaldehyde would only be one possible fate of this selective radical.

2. Autoxidation and kinetics of tartaric acid

In the study of wine oxidation, the original work by Fenton should be considered as it involves major components found in wine and wine ageing: tartaric acid, iron, and oxygen. It is well known that Fenton used hydrogen peroxide, with Fe(II), to drive the oxidation of tartaric acid. However, he also qualitatively describes the oxidation of ferrous tartrate with air [6], without the addition of hydrogen peroxide, which is now described as autoxidation. It is this reaction and its kinetics with air that sparked further exploration and discussion of pH dependences and autocatalytic kinetics [7, 8]. The study of tartaric acid, iron, and oxygen kinetics under wine-like conditions adds yet another dimension to our understanding of this famous reaction.

Wine-like conditions, in this study [9, 10], restrict the pH to an experimental range from 2.5 to 4.5, while constraining reactants: tartaric acid to 4 g/L (26.7 μM), Fe(II) to 5 mg/L (89.5 μM), and oxygen to ∼8.5 mg/L (265 mM). In addition, wine is generally stored in the dark or in darkened bottles to prevent photooxidation. Fenton indicates that the colorimetric response from this fundamental reaction does not appreciably happen without light [11], however with current spectrophotometric instruments, the reaction can be followed without the acceleration from light. The work presented here will also explore a special condition where Fe(II) is equimolar to oxygen, 265 μM, which leads a deeper understanding of the chemical reaction, component limitation and the underlying kinetics.

The time course measurements of dissolved oxygen consumption and Fe(III) formation (Figure 1) show the autocatalytic nature of tartaric acid oxidation with three distinct phases: initiation, propagation, and termination. The initiation phase clearly shows a kinetic pH dependence. The work by Smythe [8], proposed the kinetic importance of pH and the Fe(II)-tartrate complex. Tartaric acid, a dicarboxylic acid, exists as three species in this pH range: tartrate (R--), bitartrate (RH-), and tartaric acid (RH₂). With varying pH, the tartrate species concentration changes [12, 13, 14, 15], thus changing the free Fe(II), and more importantly Fe(II)-tartrate concentration which correlates with a kinetic role in the initiation. The free Fe(III) and Fe(III)-tartrate ligand(s) concentration also changes across this pH range which increases the intricacy of the reaction mechanics and the intermediate species.

Figure 1.

Oxygen consumption and Fe(III) formation at pH 2.5 to 4.5. Initial conditions at 89.5 μM Fe(II) to initiate the autocatalytic reaction in air-saturated 26.7 mM tartaric acid at pH 2.5 (), 3.0 (), 3.5 (), 4.0 (), 4.5 ().

The same time courses (Figure 1) show a distinct 1:1 molar relationship between oxygen consumption and Fe(III) formation. It would be expected that Fe(III) formation correlates with Fe(II) consumption. This relationship between iron and oxygen must be maintained when developing a kinetic mechanism for the reaction.

The propagation and termination phases vary with pH as it does with initiation. Increased pH reduces the initiation period, the rate of propagation and the extent to which oxygen is consumed. The termination phase will elucidate additional kinetic considerations when elevated Fe(II) concentrations (265 μM) at pH 2.5 and pH 3.0 (Figures 2 and 3) are evaluated. At these levels, the Fe(III) formation terminates as the oxygen concentration is depleted, however beyond this point Fe(III) is consumed, apparently returning to Fe(II). This would indicate a secondary reaction with Fe(III) and an intermediate, thus driving the conversion back to Fe(II).

Figure 2.

Oxygen consumption at pH 2.5 to 4.5 with 256 μM Fe(II). Autocatalytic reaction in air-saturated 26.7 mM tartaric acid at pH 2.5 (), 3.0 (), 3.5 (), 4.0 (), 4.5 (). Reproduced from [10], with the permission of AIP Publishing.

Figure 3.

Fe(III) formation at pH 2.5 to 4.5 with 256 μM Fe(II). Autocatalytic reaction in air-saturated 26.7 mM tartaric acid at pH 2.5 (), 3.0 (), 3.5 (), 4.0 (), 4.5 (). Reproduced from [10], with the permission of AIP Publishing.

Hydrogen peroxide has been proposed as a reactant for tartaric acid oxidation [6] and an intermediate in wine oxidation [1, 16]. The simultaneous measurements of hydrogen peroxide at elevated Fe(II) concentrations (265 μM) also show a pH dependency (Figure 4). The lower pH levels have a measurable hydrogen peroxide formation with a peak concentration roughly developing as oxygen is depleted. This peak timing can be clearly seen in modeling fitting figures (Figure 5). It appears that a more rapid propagation rate leads to a greater peak hydrogen peroxide concentration. In turn, this peak hydrogen peroxide concentration has a 1:2 molar ratio with the decomposition of Fe(III) (Figures 3 and 5). Such a decomposition has been previously explored [17], however oxygen does not appear to be regenerated from the decomposition of hydrogen peroxide in this reaction.

Figure 4.

Hydrogen peroxide formation. Autocatalytic reaction initiated with 265 μM Fe(II) in air-saturated 26.7 mM tartaric acid at pH 2.5 (), 3.0 (), 3.5 (). Reproduced from [10], with the permission of AIP Publishing.

Figure 5.

Fitted consumption and formation curves with proposed mechanism. Dissolved oxygen () and predicted (), Fe(III) () and predicted (), and hydrogen peroxide () and predicted () time traces modeled with dissolved oxygen, Fe(III), and hydrogen peroxide time traces at 265 μM initial Fe(II) in air-saturated 26.7 mM tartaric acid at (a) pH 2.5, (b) pH 3.5 and (c) pH 4.5. Reproduced from [10], with the permission of AIP Publishing.

3. Reaction mechanism

Chemical mechanisms for the Fe(II) and oxygen reactions, Fe(II) and hydrogen peroxide reactions, and/or the oxidation of tartaric acid has been explored by several researchers [7, 8, 18, 19, 20, 21, 22, 23]. However, these researchers have not had simultaneous measurements of iron, oxygen, and hydrogen peroxide, and the corresponding constraints that come with these time course measurements. These constraints specifically are: pH dependency on all autocatalytic phases, 1:1 molar oxygen consumption to Fe(III) formation during all autocatalytic phases, and 1:2 molar hydrogen peroxide consumption to Fe(III) consumption during termination. The three simultaneous curves and the constraints have led to the proposed comprehensive mechanism in Figure 6.

Figure 6.

Mechanism used for the modeling of the oxidation of tartaric acid (RH₂).

The chemical reactions associated with k₁/k₋₁ and k₂ describe the two-electron transfer to oxygen to produce hydrogen peroxide in the initiation phase of the overall tartaric oxidation reaction. These reactions consider the Fe(II) speciation and utilize Fe(II)-tartrate as the oxygen activating species. As the pH increases the concentration of Fe(II)-tartrate increases [12, 13, 14], thus leading to a shorter lag time during the initiation phase.

Reactions associated with k₃/k₋₃, k₄, k₅/k₋₅, k₆, k₇, k₈, k₉, and k₁₂ describe the propagation phase. This scheme describes two alternative propagation pathways; one that utilizes a ferryl (FeO²⁺) intermediate and another that utilizes a hydroxyl radical (•OH) intermediate. The ferryl intermediate provides an opportunity to explore pH and iron dependency in the propagation phase as it has pK_a ∼ 4.7 [24]. This pK_a allows for pH varying species concentration in the range of wine pH, whereas the hydroxyl radical will be pH independent. The kinetic modeling work described below will only use the ferryl pathway to fit the Fe(III), oxygen, and hydrogen peroxide simultaneous measurements; the pathway described by hydroxyl radical, k₁₂, will not be explored here.

The oxidation of tartaric acid leads to the formation of dihydroxymaleic acid (R) [6, 11]. The reactions associated with k₈ and k₉ produce dihydroxymaleic acid while continuing to propagate the oxidation cycle by regenerating hydrogen peroxide or regenerating the tartaric radical respectively.

Finally, the reactions described by k₁₀ and k₁₁ describe termination. The reaction associated with k₁₀ terminates the propagation cycle by producing dihydroxymaleic acid, while also regenerating Fe(II) in the process. On the other hand, k₁₁ terminates the tartaric radical by allowing for a dimerization to occur (RR). This scheme attempts to provide a mechanism that describes the various constraints provided by the experimental measurements. However, it should be recognized that these reactions may not produce distinct isolatable species as proposed, but reactions may happen within an iron-ligand complex and/or multiple iron-ligand complexes.

3.1 The chain reaction in fenton autoxidation

The stoichiometry indicates that autocatalytic propagation reaction can be described as follows:

$$\mathrm{Fe}\left(\mathrm{II}\right)+{\mathrm{O}}_2+{\mathrm{H}}_2{\mathrm{O}}_2+2\left(\mathrm{R}{\mathrm{H}}_2\right)\to \mathrm{Fe}\left(\mathrm{III}\right)+{\mathrm{H}}_2{\mathrm{O}}_2+2{\mathrm{H}}_2\mathrm{O}+\mathrm{R}+\mathrm{RH}•$$E1

(proton needed on the right to balance the charge not shown)

In this reaction, H₂O₂ is regenerated, one electron is taken from Fe(II) and three remaining electrons for O₂ reduction are taken from the substrate: two molecules RH₂ are oxidized with the generation of dihydroxymaleic acid, R, and a tartaric radical, RH•. The generation of a radical will result in the formation of peroxy species RHOO• that would undergo the following oxidation,

$$\mathrm{Fe}\left(\mathrm{II}\right)+\mathrm{RH}{\mathrm{O}}_{2^{\bullet }}\to \mathrm{Fe}\left(\mathrm{III}\right)+{\mathrm{H}}_2{\mathrm{O}}_2+\mathrm{R}$$E2

(again, a proton needed on the left to balance the charge)

which maintains the correct 1:1 oxidation stoichiometry and regenerates H₂O₂. The alternative process with correct stoichiometry is the dimerization of radicals in (Eq. 1) before peroxy radical formation. In addition, other processes such as catalytic reduction of Fe(III) in (Eq. 1) by oxidation of RH•, which would violate 1:1 Fe(II)/O₂ stoichiometry, are possible; however, it appears such processes play only minor role at low pH, as experimental data indicate.

As written, in one cycle of the propagation reaction in (Eqs. 1) and (2), two H₂O₂ molecules are generated for each hydrogen peroxide entering the cycle. That means that in one cycle not only H₂O₂ is regenerated, but one additional molecule H₂O₂ is formed. This would result in the unlimited exponential growth of hydrogen peroxide in the system, unless some termination/dissipation processes stop the growth. Such processes are chain termination reactions, of which one is the radical dimerization reaction; another is oxidation of RH• by Fe(III) in (Eq. 1). The relative rates of chain multiplication and dissipation/termination define the condition of the exponential growth. The presence or absence of the propagation phase (or its limited form at high pH) observed in the autoxidation reaction can be related to the exponential growth condition in the kinetics using linear stability analysis.

In order to explore the condition of exponential growth we consider a simplified reduced description of the system, keeping track of only most important variables: hydrogen peroxide p1 (h), tartaric radicals p2 (r), oxygen p3 (o), and Fe(III), p4 (f). For hydrogen peroxide and tartaric radicals we have:

$$\begin{array}{c}\dot{h}=V_0-k_{11}h+k_{12}r\\ \dot{r}=k_{21}r-k_{22}h-k_{00}{r}^2\end{array}$$E3

Here, V₀ is the rate of hydrogen peroxide production by the initiation phase, in the reaction of oxygen and Fe(II); k₁₁ is combined rate of conversion of hydrogen peroxide to hydroxyl radical and to ferryl complexes, and also decomposition of hydrogen peroxide by Fe(II); k₁₂ is the rate of regeneration of hydrogen peroxide by the reaction of tartaric radicals with oxygen; k₂₁ is the rate of generation of tartaric radicals (it may not be exactly same as k₁₁, but close to it); k₂₂ is the rate of tartaric radicals removal due to oxidation by Fe(III) (and generation of dihydroxymaleic acid), reaction that competes with oxidation and generation of hydrogen peroxide; and k₀₀ is the rate of radical dimerization (these should not be confused with the actual rate constants, k₁₁ and k₁₂, in mechanism in the previous section).

As seen, we do not account for all intermediates involved, counting only the initial and final products, partially on the basis that those intermediates are formed on a very short time-scale, such as conversion of hydroxyl radical to tartaric radical, or formation of peroxy-radicals in the reaction of tartaric radicals with oxygen, compared to slow reaction of formation of hydrogen peroxide. The cumulative rates correspond to rate-limiting reactions; for example, formation of hydrogen peroxide from tartaric radical is defined by the proton-coupled electron transfer to peroxy-radical RHOO•. These rates themselves depend on the condition of the reaction, such as pH, and concentration of the substrates; some of them changing significantly in the reaction (oxygen, Fe(II)/Fe(III)), and some do not such as tartaric acid which is in excess. At any given condition, we can assume specific values of these reaction rates and ask what is the kinetics of the system?

The stability of the kinetic system is defined by the linearized equations. Given the current state of the hydrogen peroxide and radicals concentrations, one can ask how a small variation of these concentrations would change the state of the system in time. For small concentrations it is sufficient to consider only a linear part of the system, defined by the kinetic matrix K_ij = k₁₁, k₁₂, k₂₁, k₂₂; (or with a modified coefficient k₂₂, k₂₂ + 2k₀₀r, due to radical dimerization part). The kinetic solution is bi-exponential, the two rates are given by the eigenvalues of the kinetic matrix, and found from the following equation:

$$\begin{array}{l}det\left(\lambda -\widehat{K}\right)=0\\ \left(\lambda -k_{11}\right)\left(\lambda -k_{22}\right)-k_{12}{k}_{21}=0\end{array}$$E4

The populations are changing as combination of two exponentials:

$$p_i\left(t\right)=c_{i1}{e}^{-{\lambda }_{1}t}+c_{i2}{e}^{-{\lambda }_{2}t}$$E5

where c_i are some constants. When the product $k_{12}{k}_{21}=0$ the two eigenvalues are ${\lambda }_1=k_{11}$ and ${\lambda }_2=k_{22}$. The two rates describe bi-exponential relaxation of hydrogen peroxide and tartaric radicals to their equilibrium values. However, when $k_{12}{k}_{21}>0$, one eigenvalue becomes larger (remaining positive), another becomes smaller and may become negative. In this case the negative eigenvalue gives rise to the exponential growth (and propagation phase of the reaction). The condition for a negative eigenvalue and exponential growth is

$$k_{11}{k}_{22}<k_{12}{k}_{21}$$E6

As we already mentioned, the rate k₂₁ is essentially the same as k₁₁, thus the condition is k₂₁ > k₂₂, that is the rate of generation of radicals is higher than their dissipation. As the determinant of a matrix is a product of its eigenvalues, $det\widehat{K}={\lambda }_1{\lambda }_2$ and one eigenvalue is always positive, the condition of one negative eigenvalue is equivalent to:

$$det\widehat{K}<0$$E7

which is equivalent to condition found in (Eq. 6). The rate coefficients, k_ij of the kinetic matrix K, are themselves functions of the conditions of the reaction, which change with time, thus, the above condition may or may not be satisfied at a given every stage of the reaction.

When the exponential growth of hydrogen peroxide and tartaric radicals begins, the dissipation/termination processes get activated and stationary concentrations will be established, until oxygen and Fe(II) diminish. The stationary (maximum) values are found from (Eq. 3) at the stationary conditions $\dot{h}=0$, $\dot{r}=0$; assuming relatively small rate of initiation V₀, the stationary (maximum) values of hydrogen peroxide and tartaric radicals are:

$$\begin{array}{l}{h}_{\mathit{ss}}\simeq \left(k_{12}/k_{11}\right)r_{\mathit{ss}}+V_0/k_{11}\\ r_{\mathit{ss}}\simeq \left(k_{12}-k_{22}\right)/k_{00}+V_0/\left(k_{12}-k_{22}\right)\end{array}$$E8

here we assumed k₂₁ to be about the same as k₁₁, and thus in the exponential phase k₁₂ > k₂₂.

At low pH, the condition of exponential growth appears to be satisfied up to a very low concentration of oxygen; eventually, of course, it breaks down, as k₁₂ – rate of regeneration of hydrogen, for which oxygen is needed, diminishes, and rate k₂₂ – rate of removal of radicals, bypassing peroxidation, is increasing as Fe(III) increases. The initial lag before fully developed propagation stage is due to the absence of hydrogen peroxide initially, and the incubation period is simply an accumulation of hydrogen peroxide in the system; the exponential multiplication of the initially produced hydrogen peroxide results in the development of the chain reaction that is stabilized by various radical termination process. This fully developed and stabilized chain reaction is what forms the propagation stage of the reaction.

The transition to propagation stage therefore involves a competition between the chain multiplication of radicals and their dissipation. The negative eigenvalue in the kinetic coefficients is a signature of a condition when chain multiplication exceeds that of dissipation.

4. Kinetic modeling

Given the proposed scheme (Figure 6), tartaric acid speciation, and iron speciation [12, 13], the experimental Fe(III), oxygen, and hydrogen peroxide simultaneous measurements were fitted using kinetic modeling software, Kintecus [25]. The fitting curves for pH 2.5, 3.5, and 4.5 are shown in Figure 5 and the resulting kinetic constants are shown in Table 1. The modeling provided a reasonable fit for all conditions, especially the highly complex pH 2.5 and 265 μM Fe(II) case where Fe(III) decomposes with a similar timing to hydrogen peroxide.

k1/k−1	2.4E+01	M^−1
k2	1.4E+01	M^−1 s^−1
k3/k−3	7.8E+01	M^−1
k4	8.4E+01	s^−1
k5/k−5	6.2E+02	M^−1
k6	4.9E+05	s^−1
k7/ k−7	7.9E+01	M^−1
k8	1.1E+02	M^−1 s^−1
k9	6.6E+03	M^−1 s^−1
k10	9.4E+00	M^−1 s^−1
k11	1.2E+00	M^−1 s^−1

Table 1.

Estimated kinetic constants for 265 μM initial Fe(II) in air-saturated 26.7 mM tartaric acid at pH 2.5.

The ability of the model to fit experimental data across pH provides directional confidence. It should be recognized that the k values are not constant across pH and further examination of species and pH dependent reactions is required, however the robustness of the model and k values within a single pH can be evaluated by making predictions and examining the resulting fit. Figure 7 shows two predicted curves against experimental measurements where the initial condition of hydrogen peroxide concentration is changed by adding 2.65 μM and 26.5 μM. The predicted and actual measurements are nearly identical, which speaks to the power of the modeling and mechanism.

Figure 7.

Oxygen consumption dependence on addition of H₂O₂ and predictions. Time traces with 0 μM (), 2.65 μM (), and 26.5 μM () hydrogen peroxide added at initiation (t = 0); reaction of 265 μM Fe(II) in air-saturated 26.7 mM tartaric acid at (A) pH 2.5. Fitted trace with 0 μM () added hydrogen peroxide. Predicted traces with 2.65 μM (), and 26.5 μM () added hydrogen peroxide. Reproduced from [10], with the permission of AIP Publishing.

5. Discussion

The characteristic sigmoid shape of oxygen consumption in the autoxidation of tartaric acid is observed at all pH conditions. The initiation step is due to oxygen activation to hydrogen peroxide by a Fe(II)-tartrate complex. It is rapid with a time scale of minutes to hours. This reaction step is pH sensitive, slower at lower pH (2.5) than higher (4.5). The extent of peroxide formation is limited to the pool of the Fe(II) state, free and complexed. Initiation is the critical feature of autoxidation, both in these solutions and in wine.

The propagation step is due to the free Fe(II) and hydrogen peroxide oxidation of tartaric acid to produce a tartaric radical which then goes further to form dihydroxymaleic acid, regenerating Fe(II) and consuming additional oxygen as hydrogen peroxide is reformed.

The termination step cannot be explained by the direct decomposition of residual hydrogen peroxide by Fe(III) since oxygen formation is not observed. This termination step may also involve intermediates and end products such as the tartaric acid radical and dihydroxymaleic acid, which adds to the complexity of tartaric oxidation and the accuracy of this model to explain the termination as observed at low pH.

The mechanism proposed here, (Figure 6), is slightly different from that previously presented [10]. The difference lies in the propagation steps leading to the exponential growth of one hydrogen peroxide leading to two. The current version does not as accurately fit the loss of Fe(III) observed at the end of oxygen consumption, in the 2.5 pH cases.

The reactivity of dihydroxymaleic acid and the initiation of subsequent radical chain reactions makes describing and interpreting of the termination stage in this system complicated; however, it can be linked to termination of the chain propagation conditions (a negative eigen-value of the kinetic matrix) of our pseudo-first order reaction analysis of Section 3.

5.1 The role of the acid

The unique properties of tartaric acid in these autoxidation reactions can be seen if other major organic acids are substituted for it in the same reaction medium. Figure 8 shows the individual oxidation of malic, citric and succinic acids at pH 2.5 and 4.5. At low pH none of these acid show any significant propagation stage while at high pH malic and citric acid show a slow propagation stage but still much slower than tartaric.

Figure 8.

Wine acids and oxygen consumption. Autocatalytic reaction initiated with 265 μM Fe(II)n air-saturated 26.7 mM tartaric acid (), malic acid (), succinic acid (), citric () and hydrochloric acid () at (a) pH 2.5 and (b) pH 4.5. Reproduced from [10], with the permission of AIP Publishing.

The propagation kinetic phase is associated with a chain reaction that requires regeneration of radicals with multiplicity factor greater than one. In our scheme, two hydrogen peroxide molecules are regenerated in the chain propagation reaction with tartaric acid (at low pH) for each hydrogen peroxide entering the cycle. Thus multiplicity factor is two. Figure 9 illustrates the importance of multiplicity factor for a chain reaction. As seen in the figure, upon addition of hydrogen peroxide to citric and succinic acids, the same amount of oxygen is consumed, with ratio 1:1. Here one acid radical is produced per hydrogen peroxide, which reacts further with oxygen to produce peroxy species, but no addition radical or hydrogen peroxide is regenerated. Thus multiplicity factor is zero.

Figure 9.

Wine acids and oxygen consumption initiated with hydrogen peroxide. Autocatalytic reaction initiated with 265 μM Fe(II) and addition of 26.5 μM hydrogen peroxide in air-saturated 26.7 mM tartaric acid () offset to 1 hour to align with: malic acid (), succinic acid (), and citric acid () at pH 2.5. Reproduced from [10], with the permission of AIP Publishing.

Indeed, the reaction overall oxygen consumption is given by the sum of probabilities of the chain, where the multiplicity factor q (probability of radical generation):

$$\text{Oxygen consumption}=1+\mathrm{q}+{\mathrm{q}}^2+{\mathrm{q}}^3+\cdots =1/\left(1-\mathrm{q}\right)$$E9

For malic acid, interestingly, four oxygen molecules are consumed for each hydrogen peroxide added, with ratio 1:4. It can be interpreted as a reaction with multiplicity factor q = 3/4, but still less than one, needed to ignite the chain reaction. The overall number of radicals and oxygen consumed is summed to four, as observed. The factor 3/4 is clearly related to the chemical nature of malic acid that has one OH group out of 4 possibilities on C2 and C3 carbons, and 3/4 occupied by hydrogens.

For tartaric acid the multiplicity factor is obviously greater than one and the overall oxygen consumption is as much as 10 in this trial, and the chain could run without stopping until all oxygen or Fe(II) is consumed. That is what we observe indeed for tartaric acid at low pH.

By comparison in case of succinic and citric acids, the overall probability of oxygen consumption equals to 1.

5.2 The role of ethanol

The addition of ethanol to this system (Figure 10), shows that even at concentrations of 26.5 mM and both pH of 2.5 and 4.5, the propagation reaction is not established, even when oxygen and Fe(II) are available. This effect, due to ethanol at 100 times the Fe(II) concentration is similar across pH. This can be interpreted as a competition of ferryl ion or hydroxyl radical for ethanol over tartaric acid or the depletion of tartaric radicals in the proposed cycle due to hydroxyethyl radical formation rather than a solvent dielectric effect.

Figure 10.

Ethanol and tartaric acid. Autocatalytic reaction with 0 mM (), 0.265 mM (), 2.65 mM (), 26.5 mM (), 265 mM (), and 2.56 M or 15% (v/v) () ethanol with 265 μM Fe(II) in air-saturated 26.7 mM tartaric acid at (a) pH 2.5 and (b) pH 4.5.

This indicates that the hydrogen peroxide is limiting the propagation stage but the external addition overcomes this limitation in rate (Figure 11). The extent of reaction when hydrogen peroxide is available, is independent of pH. In the absence of ethanol, 100% of the oxygen would be consumed within 15 minutes and this result is also pH independent.

Figure 11.

Ethanol and tartaric acid initiated with hydrogen peroxide. Autocatalytic reaction with 0 mM ethanol and 0 μM H₂O₂ (), 26.5 mM ethanol and 0 μM H₂O₂ (),0 mM ethanol and 26.5 μM H₂O₂ (), 26.5 mM ethanol and 26.5 μM H₂O₂ () with 265 μM Fe(II) in air-saturated 26.7 mM tartaric acid at (a) pH 2.5 and (b) pH 4.5.

5.3 The role of sulfur dioxide

The addition of sulfur dioxide at 46.9 μM (30 mg/L) prevents significant development of the propagation stage, at pH 2.5 and 4.5. This is due to the reaction of bisulfite with the hydrogen peroxide formed during initiation, preventing it from accumulating to the level required for a significant propagation reaction to develop. This reaction is known to be rapid [26] at pH 2.5 and 4.5; it is essentially complete within seconds. When the addition is made during the propagation stage, the reaction is terminated even when oxygen and Fe(II) are still present. The implication for wine making is that sulfur dioxide is acting on the hydrogen peroxide production at the point of its formation in the initiation step, essentially preventing the downstream chain reactions that would normally occur (Figure 12).

Figure 12.

Oxygen consumption with SO₂ additions. Autocatalytic reaction with 0 μM () and 30 mg/L SO₂ () at pH 2.5, and 0 μM () and 30 mg/L SO₂ () at pH 4.5 at initiation (t = 0); reaction of 265 μM Fe(II) in air-saturated 26.7 mM tartaric acid.

5.4 The role of copper

The addition of copper(II) as CuSO₄ displays very different responses at pH 2.5 and 4.5. At pH 2.5 the addition at the beginning of the reaction prevents the development of a propagation stage, suggesting that it is reacting with hydrogen peroxide formed in the initiation stage. This can be overcome by a late addition of hydrogen peroxide wherein the reaction quickly goes to completion – data not shown [9]. A similar result occurs when the addition is in mid-propagation, causing subsequent termination due to its reaction with hydrogen peroxide. In contrast, at pH 4.5 the addition at the beginning results in an enhanced initiation reaction, presumably due to the action of Cu(II)-tartaric complex providing additional oxygen activation and diminished free Cu(II) availability. A late addition of hydrogen peroxide allows the reaction to go to essentially the same level of completion as when no addition is made [9]. The oxidation of tartaric acid in the presence of Cu(II) is known to occur [27] (Figure 13).

Figure 13.

Oxygen consumption with copper additions. Time traces with 0 μM (), 7.87 μM Cu(II)SO₄ () at propagation midpoint, and 7.87 μM Cu(II)SO₄ () at initiation (t = 0); reaction of 265 μM Fe(II) in air-saturated 26.7 mM tartaric acid at (a) pH 2.5 and (b) pH 4.5.

5.5 Autoxidation of wine

The connection between the classic tartaric acid oxidation studies of Fenton, Warburg, Wieland and Franke, and Smythe and those occurring in wine comes from investigations of the oxidation of dihydroxymaleic and tartaric acids in the presence of iron (II) salts in wines [28]. Rodopulo [29] described oxidation and rearrangements resulting from tartaric oxidation into intermediates of dioxytartaric, dioxosuccinic, mesoxalic acids and glycolaldehyde and likely final products as glyoxalic and oxalic acids. He noted that in the presence of air dihydroxymaleic spontaneously oxidizes to dioxosuccinic acid, which in the absence of oxygen will react with tartaric acid forming two dihydroxymaleic acids. His opinion was that wines in an anaerobic state would contain dioxosuccinic while those with exposure to air would likely contain traces of glyoxylic and oxalic acids. A key contribution of his work was the demonstration that the addition of a ferrous tartrate precipitate to wine caused further oxygen consumption than that of ferrous sulfate, suggesting the importance of the Fe(II)-tartrate complex in wine oxidation. More accessible descriptions of this finding can be found elsewhere [30, 31]. Baraud [32] further investigated the oxidation of tartaric and dihydroxymaleic acids in wine-like conditions and tried to identify all of the products, including one of the unknown intermediates from these reactions. The relationship between these components has recently been summarized by Duca [27] and referred to as the Baraud cycle.

The proposed (Figure 6) incorporates the formation of ferryl ions and radicals, such as the hydroxyl and the tartaric radical, that are expected to be able to extract the α-hydrogen from ethanol to form the 1-hydroxyethyl radical in the presence of even small concentrations of ethanol. Hydroxyethyl radical was identified as the most important radical in beer by Andersen and Skibsted [33] and shown to be the active intermediate in the oxidation of linoleic acid to (E)-2-nonenal, a key impact volatile in “staling” character of beer [34]. It is now known to be the central radical responsible for the selective oxidation of humulones and hop acid components rather than the hydroxyl radical during the oxidation of beer [35].

Hydroxyethyl radical has been shown to react selectively with the flavonols quercetin and kaempferol, but not with epicatechin [36], and while others found that most flavonoids with a C_n-C_n+1 double bond and caffeic acid were all reactive with the hydroxyethyl radical [37]. The hydroxyethyl radical has been found in a wine held at 55°C, [38] and its selective reaction with some phenol and thiol entities in model wine has been demonstrated [39]. It is known to react with glutathione [40] as well and cinnamates in wine conditions [41]. We propose hydroxyethyl to be the major and more selective oxidizing agent, whose reactivity is likely to be determining the identity and concentrations of the downstream radical products in wine. For this reason we expect there to be little involvement of dihydroxy phenols and tannin in such reactions. As such, hydroxyethyl radical will have a selective influence on specific phenols and glutathione in determining the oxidation outcomes in wine.

The action of sulfur dioxide is to interact with hydrogen peroxide concentrations at the point of initiation and to prevent propagation. The role of ethanol is to divert ferryl ion and/or other radicals into hydroxyethyl radicals and the subsequent radical chain reactions are likely influenced more by reduced glutathione levels and the cinnamates and flavones (but not due to their dihydroxy patterns). As such many of the subsequent reactions associated with oxidation may have little if anything to do with the dissolved oxygen concentration or the quantity of oxygen it has been exposed to.

The autoxidation sequence in wine can be classified into at least 3 periods, those reactions that take place within hours and days, those resulting from that which continue to interact in the days and weeks after, and those that continue to react and rearrange in the subsequent months and years. The first period is the activation of oxygen, the autoxidation described here and the generation of tartaric acid and related radicals. The second period would be the further reactions associated with more stable and long-lived radicals not necessarily in the presence of oxygen and would be selective radical transfer reactions between different wine components, typically not ethanol. The third period would be the long-term aging reactions. These reactions would be disturbed or interspersed with periodic and/or slow diffusional delivery of oxygen, generally in bottles. It is common for wines to be exposed to some oxygen within the winery during transfers, aging and bottling and there can be abrupt increases in concentration of dissolved oxygen or slow diffusional delivery such as though barrels and porous bottle closures.

There appears to be some confusion around the role of certain wine components involved in the initiation reactions compared with those involved in the propagation and termination reactions as well as the time scales involved. It is expected that subsequent radical reactions will continue after the first stage reaction is completed. These will include redox reactions, polymerization reactions and condensation reactions but all would involve relatively stable radicals and are not expected to require additional oxygen. This makes attempts to correlate the extent of product formation in wine with the rate, the extent of oxygen consumption, or the initial wine composition of major components likely to be unsuccessful.

Existing reaction pathways that have been proposed for wine oxidation use mostly free Fe(II) ions as the initial reactive species and all suggest the formation of the hydroxyl radical as the high oxidation state radical [42, 43]. Most of these proposed pathways have a coupled oxidation of a dihydroxy phenol for Fe(II) regeneration but none account for pH or the fact that almost half of the total Fe(II) in wine is in the Fe(II)-tartrate complex form. None of these proposed pathways can be used to describe the observed autoxidation of tartaric acid in wine conditions. Most propose the formation of acetaldehyde from ethanol as the major oxidation product. The indiscriminate and almost instantaneous reactions with hydroxyl radicals should result in acetaldehyde (and glyceraldehyde) formation directly coupled to oxygen consumption and in a ratio of products proportional to their initial concentrations. Formation of acetaldehyde involving dihydroxy phenols does occur but only at elevated temperatures, 50°C [1]. Such formation has not been shown to be related to either the extent or rate of oxygen consumption at ambient or wine storage temperatures.

The role of other transition metal complexes in the initiation and propagation reactions needs to be investigated further. Here, there are effects due to the presence of malic acid and copper (II) which will vary between wines, especially before and after the malolactic fermentation and as a result of copper additions during winemaking. We expect their effects to be related to the concentrations of their complexes.

Lastly, the recovery of wines to an initial state after exposure to oxygen was reported to take several days [16]. This might be interpreted as being due to a slower return of Fe(II) for further initiation and or propagation reactions due to certain wine constituents that are absent in these model solutions studies. This deserves further research attention.

6. Conclusions

The autoxidation of tartaric acid in the presence of Fe(II) has been demonstrated in solution over the pH range 2.5 to 4.5. A mechanism that can describe the observations is proposed and fitted to a kinetic model. Preliminary estimates of the rate constants are presented. The effects on these reactions due to additions of ethanol, sulfur dioxide and copper at wine-like conditions are described. The implications of this radical chain reactions sequence to describe the autoxidation of wine and idea of radical chain propagation in wine are discussed.

Acknowledgments

This work was supported in part by Treasury Wine Estates (REC), by NIH grant GM054052 (AAS) and the Stephen Sinclair Scott Endowment (RBB).

Conflict of interest

The authors declare no conflict of interest.