|Home | About | Journals | Submit | Contact Us | Français|
The blue bottle experiment is a collective term for autoxidation reactions catalysed by redox indicators. The reactions are characterized by their repeatable cycle of colour changes when shaken/left to stand and intricate chemical pattern formation. The blue bottle experiment is studied based on calculated solution-phase half-cell reduction potential of related reactions. Our investigation confirms that the reaction in various versions of the blue bottle experiment published to date is mainly the oxidation of an acyloin to a 1,2-dicarbonyl structure. In the light of the calculations, we also propose new non-acyloin reducing agents for the experiment. These results can help guide future experimental studies on the blue bottle experiment.
Autoxidation of reducing agents catalysed by redox indicators have been reported in the literature since 1946 [1,2]. The most notable reaction is the ‘blue bottle experiment’ , an oxidation of glucose catalysed by methylene blue under an alkaline condition . For bulk reaction in a flask, methylene blue is reduced to colourless (leuco) form by the aldose sugar when left to stand and is oxidized to blue form by atmospheric oxygen when shaken. The cycle can be repeated many times before the reactants run out or the solution turns brown due to side reactions . For thin-layer reaction in a Petri dish, dot and line patterns of the oxidized indicator develop over a period of time [6–8]. Figure 1 shows the general reaction framework and figure 2 shows patterns and colours observed in different variations of the reaction.
The blue bottle reaction and its analogues are popular chemical demonstrations due to their visual appeal and simplicity, and the majority of the research on this topic is published in the Journal of Chemical Education [1–18]. Fundamental research and non-education applications of the reactions have also been discussed elsewhere [19–26].
In essence, there were attempts to test new reducing agents [7,11,16] and indicators [7,8], to elucidate the mechanism and kinetics of the reactions [5,11] and to model the pattern formation [6,19–23,26]. Pattern formation may be comparable to the Belousov–Zhabotinsky  and the Briggs–Rauscher  reactions. However, the blue bottle model is relatively simple and versatile because it requires only a few reactants and many alternative reagents can produce similar results.
Despite numerous reports, the understanding of the reaction is advanced incrementally by mostly trial-and-error experiments. A number of papers reported only one new reducing agent  or indicator [9,10,17] or pattern formation in one specific system . The reports can also be conflicting or incomplete. For example, in 2012, Anderson et al.  suggested that gluconate is not a major product and the reaction may produce hydrogen peroxide, but later work  as late as 2014 still discusses the gluconate compound as the main product; in 1974, Chen  reported the use of indophenol as a dye for the experiment, but, in 2016, Rajchakit & Limpanuparb  failed to reproduce it. Experimental reports usually mention only a structure of a dye in its solid form but do not explicitly show oxidized or reduced form(s) of the compound [7,17] and there was no experimental identification of the products in all cases except one .
In this first density functional theory (DFT) investigation of the blue bottle experiment, we aim to propose a theoretical framework to resolve discrepancies in the current literature and guide future experimental studies. The manuscript is structured as follows: Methodology describes reactions and computation approach; Results and discussion presents the main results based on reduction potentials, and preliminary experimental evidence, detailed computational/experimental results are given as the electronic supplementary material; Concluding remarks and future work are discussed at the end of the paper.
Figure 1 shows that there are three main groups of reactions in the blue bottle experiment: oxygen reduction reactions (ORRs), oxidation/reduction of redox indicators and oxidation of reducing agents. It is natural to characterize these redox reactions in terms of standard half-cell potential, in aqueous solution at 298.15K. Because is a ‘per electron’ quantity, it conveniently allows quick comparison and helps with our prediction whether a compound can possibly be used in a blue bottle reaction. By considering the potentials, it is equivalent to the consideration of Gibbs energy. A reaction is spontaneous provided that the cell potential made by combination of reduction potentials of two half-reactions,
is positive. In other words, a necessary but not sufficient condition for the combination of ORR, oxidation/reduction of dyes and oxidation of reducing agents to make up a blue bottle experiment is
as shown in figure 1. All the discussions that follow use the same potential comparison process as a thinking framework.
We include representative compounds reported in the blue bottle literature and possible reagents to explore alternative redox indicators/reducing agents and to gain mechanistic insight of the reaction. Table 1 lists the oxidized and reduced structures of all compounds in this study. If oxidized/reduced form(s) of the compounds are not explicitly mentioned in the literature, we do our best to propose them.
Reactions reported in the current literature and our proposal for dyes and reagents are studied as follows (structurally similar compounds are grouped together):
Different procedures [32–50] exist for computation of accurate especially in one-electron case and/or families of structurally similar compounds [32–36,42,44–46,48]. Since all species in our study are simply closed-shell singlet organic molecules, however, a gas-phase DFT optimization followed by a free energy of solvation calculation has been proven successful in many cases [40,43,47–48]. This approach was also included and tested in recent reviews [49–51].
Gas-phase geometries were obtained at B3LYP/6-311++G** level and were confirmed to be a minimum point on the potential energy surface by frequency calculation. Solvation was treated by SMD model  on the gas-phase structure. Some compounds in our study have a number of rotamers and diastereomers. We try to use the lowest energy structure as a representative. However, the difference due to these stereoisomers is expected to be small (1kcalmole−1 of electron is approximately 0.04V). All output files are provided in the electronic supplementary material. (Additional calculation at B3LYP/6-31G* (gas phase) and MP2/cc-pVTZ (solution phase) were also completed on selected compounds for the preparation of initial structures for B3LYP/6-311++G** and for benchmarking, respectively.)
All calculations were performed using the Q-Chem 4.4 developer version . The half-cell reduction potential was directly calculated from these equations:
For a half-cell reaction:
For a chemical structure :
where is the standard Gibbs energy of the half reaction, n is the number of electrons in the half reaction, F is the Faraday constant 96485Cmol−1, G is the standard Gibbs energy, H is the standard enthalpy, T is 298.15K, S is the standard entropy, ΔGsolv is the free energy of solvation from SMD and standard state correction, ε0 is the electronic energy (EB3LYP or EMP2=EHF+EMP2 correlation as applicable) and Hcorr is the total enthalpy correction to ε0.
In solution phase, 1M reference state is used with the exception of water where 55.34M is used [54–58]. The correction for these are 3.02mhartree and 3.80mhartree, respectively. We do not use reference potential and ignore electrons in calculation.
Most variations of the blue bottle experiments take place in a alkaline solution with an exception of ascorbic acid system. The reactions are therefore considered in acid and alkaline conditions separately. Hydronium ion ([H3O+]=1M, pH=0) and water are used in the calculation instead of hydrogen ion (H+) for reactions under acidic condition. Similarly, hydroxide ion ([OH−]=1M, pH=14) is used for reactions under alkaline condition. Examples are provided in table 1 to illustrate reactions under acidic and alkaline conditions.
Reduction potentials of many redox reactions are pH dependent due to deviation from standard condition. The deviation in terms of Gibbs energy (RT ln Q) is expressed in the last term of the Nernst equation,
where Q is the reaction quotient. Since pH is not exactly 0 or 14 and the concentration of reactants are generally lower than 1M, the ln Q consideration may be employed for detail analysis, especially when Ecell is close to zero.
The mean unsigned errors for solution-phase and gas-phase EO of 52 selected reactions obtained at B3LYP/6-311++G** and MP2/cc-pVTZ are 0.86V and 1.10V, respectively. Figure 3 shows satisfactory linear relationships between EO obtained by the two methods (high R2 value but slope values slightly greater than unity). These benchmarking results confirm that B3LYP/6-311++G** yields acceptable results at a relatively small computational cost .
Table 1 shows half-cell reduction potentials of all possible reactions in the blue bottle experiment. For comparison purpose, figure 4 shows the reduction potentials in acidic and alkaline conditions for the three groups of compounds, respectively. The first set of data is four- and two-electron oxygen reduction potentials on the first and fourth column of figure 4. The values are far from the literature values [30,31] but the trend that acidic potentials are higher than alkaline potentials is still preserved. The second set of data is the dye reduction potentials. Figure 4 shows that all dyes including indophenol may be conveniently oxidized by either two- or four-election ORR. The third set of data is the reducing agent reduction potentials. (Refer to Methodology section for grouping of reactions. Some compounds, for example, glucose, belong to two groups of reactions.)
In general, the reduction potentials are quite similar for reactions in the same group under the same condition, and acidic potentials are higher than alkaline potentials. As reduction potentials of dyes and reducing agents are overlapping in figure 4, there are combinations of dyes and reducing agent that may or may not work in the blue bottle experiment.
Positive cell potential from combination of the compounds can be found in figure 4 if the half-cell potentials decrease from left to right. For example, in alkaline condition, the classical blue bottle experiment may proceed via two- or four-electron oxygen reduction (approx. 4.7V), with methylene blue as a catalyst (3.70V) and glucose as a reducing agent (3.30V). In acidic condition, the green version of the blue bottle experiment may proceed via two- or four-electron oxygen reduction (6.5–6.8V), with methylene blue as a catalyst (5.18V) and ascorbic acid as a reducing agent (5.10V).
As our reduction potential considerations here are thermodynamic, the negative prediction (non-spontaneity for large negative value of ) should be valid but the positive prediction (combination of dye and reducing agent make a blue bottle reaction for positive or close to zero value of ) requires further verifications. To produce repeatable cycle of colour change, the rate of reduction of dye by reducing agent must be slower than the oxidation of dye by oxygen  and the direct oxidation of reducing agent by oxygen  should be minimal compared with the dye-catalysed reaction. Additional catalysts similar to the green version of the experiment [8,16] may be needed to make the reaction occur but it is beyond the scope of this study.
The following findings are made based on the calculated results and information in the literature. To support our claims, preliminary experiments to test some reducing agents were also carried out (see the electronic supplementary material).
Half-cell reduction potentials of oxygen, redox indicators and reducing agents have been investigated using DFT calculations. The results help us better understand the blue bottle reaction and guide us to focus the experiments only on a certain number of representative compounds and only for reaction that lead to a positive cell potential. The use of alternative reducing agents can help avoid side reactions that make the solution brown after a number of cycles  and increase solubility of the reducing agent in water which is a known issue for benzoin . Possible future computational investigation includes prediction of pKa [39,56] and stability of intermediates and activated complexes of reduction reactions and prediction of the colour of redox dyes .
We thank Dr Andrew T. B. Gilbert for his guidance in Q-CHEM installation, Ms Punchalee Montriwat, Mr Rattha Noorat, and MUIC Science Division technicians for support in preliminary experimental work, Dr Yuthana Tantirungrotechai for helpful suggestions, and MUIC's Information Technology Section for the computing facility.
The data supporting the findings of this study are available in the electronic supplementary material.
T.L., the principal investigator and corresponding author, directed the research and wrote the manuscript. P.R. and C.A. assisted with calculation and experiment.
We declare we have no competing interests.
T.L. received funding from the Mahidol University International College (MUIC 001/2016) and the Institute for the Promotion of Teaching Science and Technology (IPST 08/2557).