Pushing Photochemistry into Water: Acceleration of the Di‐π‐Methane Rearrangement and the Paternó‐Büchi Reaction “On‐Water”

Abstract

Two representative organic photoreactions, namely a bimolecular photocycloaddition and a monomolecular photorearrangement, are presented that are accelerated when the reaction is performed “on‐water”, that is, at the water‐substrate interface with no solvation of the reaction components. According to the established models of ground‐state reactions “on‐water”, the enhanced efficiency of the photoreactions is explained by hydrophobic effects (Paternó‐Büchi reaction) or specific hydrogen bonding (di‐π‐methane rearrangement) at the water‐substrate interface that decrease the energy of the respective transition state. These results point to the potential of this approach to conduct photoreactions more efficiently in an ecologically favorable medium.

Organic photoreactions “on water”, namely in aqueous suspension or emulsion, are significantly accelerated in this ecologically favorable medium.

Introduction

Over the last decades, water has developed from a bothersome impurity to a highly attractive medium in organic synthesis. ^[1] First and foremost, water is a safe, non‐toxic, relatively cheap, and easily available solvent and, therefore, may be considered the ideal substitute for the commonly employed organic solvents. Along with these ecological and economical advantages, it has been demonstrated that water has a rather unexpected positive impact on organic reactions. Specifically, it has been shown that organic reactions are significantly accelerated “on water”, that is, in heterogeneous medium with no solvation of the reaction components. ^[2] Hence, the observed reaction rates of Aldol reactions, ^[3] cycloaddition reactions, ^[4] multicomponent reactions, ^[5] polymerization reactions, ^[6] and even metal–organic reactions ^[1] are larger in heterogeneous aqueous medium as compared with the ones obtained in homogeneous solutions in organic solvents. The mechanistic origin of the apparently general phenomenon is still not fully understood and presently debated,[ ¹ , ^2a ] and it certainly depends on the respective specific interactions of the substrates with the surrounding medium. Nevertheless, it is agreed upon that hydrogen bonding, hydrophobic effects and the cohesive energy density have a significant effect on the rate acceleration of organic reactions at the water interfaces,[ ¹ , ² ] especially on reactions with a negative volume of activation. In this context, detailed physico‐chemical studies of structures and dynamics at water‐substrate interfaces contributed to the understanding of reactions “on water” and provided promising lines of explanations, such as the Kobayashi model ^[2] or the Marcus‐Jung model. ^[7]

Although the acceleration of organic reactions “on water” is an established phenomenon, the investigation of photoreactions “on water” has been rather neglected, so far. This lack of knowledge is surprising because photoreactions are also considered ecologically more friendly, specifically as they can be performed without further reagents or catalysts and without heating, and are ideally induced by visible light or even sunlight. In addition, reactions in the excited‐state offer special tools for mechanistic investigations that are not available for ground‐state reactions. Notably, examples along these lines have only been reported sporadically. Thus, it has been shown that the photodimerization of 9‐substituted anthracene derivatives in water‐oil microemulsion has a different regioselectivity than in dichloromethane solution, ^[8] but a rate acceleration has not been reported explicitly in this case. Nevertheless, other photodimerization reactions have been shown to proceed much faster in water than in organic solvents, however, the authors tended to exclude “on‐water” conditions in this case. ^[9] Furthermore, it has been demonstrated that the photoinduced homolytic dissociation of phenol proceeds much faster at the water‐gas interface than within the aqueous solution. ^[10] Likewise, the oxidation of the phenolate ion, ^[11] as well as other photoinduced oxidation reactions, in particular related to atmospheric chemistry, are enhanced at the water‐gas interfaces. ^[12] Even photocatalytic reactions can be performed “on water”. ^[13] Moreover, examples have been provided for photoreactions of water‐suspended solid substrates; however, to the best of our knowledge, these reactions were not observed systematically with respect to the rate acceleration at the solid‐water interface. ^[14] To bridge this obvious gap of knowledge regarding photochemical reactions “on water”, i. e. at water interfaces, we investigated whether representative organic photoreactions may also be significantly accelerated when performed in heterogeneous aqueous solution. Specifically, we chose the well‐established di‐π‐methane rearrangement of dibenzobarrelenes ^[15] and the Paternó‐Büchi reaction ^[16] of benzophenone as model reactions for a monomolecular and a bimolecular reaction, respectively. Herein, we show that, indeed, these reactions are faster “on water” than in organic solvents.

Results

The dibenzobarrelene derivatives 1 b–f were prepared in moderate to good yields from acid‐catalyzed transesterification reactions of the known dimethylester 1 a ^[17] with the corresponding alcohols (Scheme 1). For the preparation of authentic product samples, the substrates 1 b, 1 d–f were irradiated in acetone (λ >280 nm), that also served as external triplet sensitizer, to give the corresponding semibullvalenes 2 b and 2 d–f in good yields (Scheme 1). The known compounds 1 a–1 c, 2 a and 2 c as well as the novel products 1 d–1 f, 2 b, 2 d–2 f were identified and characterized by NMR spectroscopy (¹H, ¹³C, COSY, HSQC, HMBC), melting point, elemental analysis, and ‐ in the case of 1 d–1 f ‐ by mass spectrometry (cf. Supporting Information).

Scheme 1.

Synthesis of the dibenzobarrelenes 1 b–f and their di‐π‐methane rearrangement to the dibenzosemibullvalenes 2 b–f. [a] Ref. [17c], t not reported. [b] Ref. [17b], t and yield of photoreaction not reported.

For the systematic investigation of the “on‐water” photoreactivity the di‐π‐methane rearrangement of the dibenzobarrelene derivatives 1 a–f was induced by direct irradiation with a high‐pressure Hg‐lamp (λ >280 nm) either in acetonitrile solution or as suspension in water under otherwise identical conditions, and the conversion was determined by quantitative HPLC‐UV analysis of the photolysate (cf. Supporting Information). Subsequently, the data of the time‐dependent formation of the photoproducts was analyzed considering a 1st‐order rate law to give the corresponding rate constants of the different derivatives (Table 1). As a general trend, the di‐π‐methane rearrangements of the dibenzobarrelene derivatives 1 a–f were significantly faster in water as compared with the reactions in acetonitrile solution. Specifically, the rate constants in water fall in a range of k_obs =3.4–5.6×10⁻⁴ s⁻¹, whereas in acetonitrile the rates were observed at k_obs =7.5–12×10⁻⁵ s⁻¹ (Table 1). Thus, the photoreaction is faster by a factor of 4–7 in aqueous suspension; however, there is no indication of a direct relationship between the reaction rate and the length of the ester alkyl group. In further control experiments, the photoreaction of 1 a was also exemplarily performed in different media. Hence, in benzene, n‐hexane and DMF solution, rate constants, k_obs , of 1.6×10⁻⁴ s⁻¹, 1.5×10⁻⁴ s⁻¹, and 9.0×10⁻⁵ s⁻¹ were observed, respectively, for the di‐π‐methane rearrangement, which is approx. 2–3 times slower than in water. In methanol and in the crystalline solid state, however, the photoreaction is much slower, as even after 4 h of irradiation the conversion was only 33 % in methanol and 15 % in the solid state (in water: >95 %). The reaction rates of the latter very slow reactions were not determined. In addition, we have shown exemplarily with substrate 1 d that the “on‐water” reaction can be performed on preparative scale to give 2 d in 79 % yield.

Table 1.

Rate constants of the di‐π‐methane rearrangement of dibenzobarrelene derivatives 1 a–f in acetonitrile and in water.

Substrate	kobs /10−5 s−1[a]
	in MeCN	in H2O
1 a	8.4±0.1	34±1
1 b	12±1	54±1
1 c	7.5±0.1	44±2
1 d	8.1±0.1	56±5
1 e	9.9±0.2	55±2
1 f	9.0±0.2	55±5

[a] Irradiated with a high‐pressure Hg‐lamp (λ>280 nm), analyzed with HPLC‐UV.

The known photoinduced Paternó‐Büchi reaction of benzophenone (3) with Z‐cyclooctene (4) ^[18] (Scheme 2) was performed in acetonitrile solution and in water suspension. According to literature protocol, ^[18] the alkene was used in excess in the Paternó‐Büchi reaction. The ¹H NMR‐spectroscopic analysis of the reaction mixture indicated the formation of the bicyclic cis‐ and trans‐oxetanes cis‐5 and trans‐5, and of the pinacol coupling product 6 in essentially the same ratio of 24 : 72 : 4 in water (t=1 h) and 26 : 68 : 6 in acetonitrile (t=4 h). The trans‐oxetane trans‐6 results from the isomerization of the intermediate 1,4‐biradical. ^[18]

Scheme 2.

Paternó‐Büchi reaction of benzophenone (3) with Z‐cyclooctene (4). The values of rates are an average of three runs with the standard deviation.

The progress of the Paternó‐Büchi reaction was followed by direct photometric analysis of the photolysate (Figure 1). Specifically, the conversion was monitored as decreasing absorption of the benzophenone absorption band at 252 nm, as the absorption of the alkene substrate and the photoproducts do not overlap at this wavelength (cf. Supporting Information). The resulting photometric data were used to determine the pseudo‐first‐order reaction rates at the different reaction conditions (Scheme 2, cf. Supporting Information). Accordingly, the photoreaction of benzophenone (3) with the Z‐cyclooctene (4) is much faster in water suspension (k_obs =6.2×10⁻³ s⁻¹) than in acetonitrile solution (k_obs =7.1×10⁻⁴ s⁻¹). To exclude an effect of the alkene concentration on the relative rates of the photoreaction in water and acetonitrile, control experiments were made with a higher concentration of Z‐cyclooctene (4) (cf. Supporting Information). Under these conditions, the reaction rates are essentially in the same range as the ones with lower cycloalkene concentration, and the reaction is also faster in water suspension (k_obs =4.7×10⁻³ s⁻¹) than in acetonitrile (k_obs =6.1×10⁻⁴ s⁻¹), that is, under both conditions this photoreaction is accelerated by a factor of ca. 8 in water (Figure 1).

Figure 1.

Photometric monitoring of the Paternó‐Büchi reaction of 3 (1 mM) and 4 (3 mM) in MeCN (A) and in water (B). Irradiation was performed with a high‐pressure Hg‐lamp. Spectra were recorded after dilution with MeCN (1/49, v/v for A and 1/24, v/v for B).

Discussion

Overall, our studies of the di‐π‐methane rearrangement and the Paternó‐Büchi reaction revealed that organic photoreactions can be performed under “on‐water” conditions and have the same propensity as ground‐state reactions to be accelerated under these conditions. However, it should be emphasized that the quantification of the effect of “on‐water” conditions on photoreactions, especially the comparison with photoreactions in homogeneous photoreactions or ground‐state reactions, has to be assessed very carefully, because inner filter effects and scattering in the heterogeneous medium may interfere with the efficiency of the investigated reaction more than they do in homogeneous solution. To add to that, it is not known – and certainly very difficult to be determined – how much of the employed light does actually reach the water‐substrate interface, that is, where the “on‐water” effect supposedly operates. Unfortunately, these heterogenous conditions do not allow to determine whether the absorbed photon fluxes are the same in each reaction, and they also impede the determination of reliable quantum yields under “on‐water” conditions. But despite all these uncertainties and the impossibility of real quantitative comparison of reaction kinetics in all phases, that need to be assessed in detail in future physical‐chemical studies, it is clear from our results that under otherwise identical, commonly employed conditions of an organic photoreaction (i. e. identical lamp, distance of sample to light source, stirring, reaction time) the reactions on water are significantly faster, i. e. complete after much shorter irradiation times, in direct competition with the ones performed in organic solvent. Thus, even if the physical‐chemical situation at the water interface remains debatable, the results of the photochemical studies show an undeniable effect on the reaction efficiency.

But even with the above‐mentioned analytical limitations, namely a likely reduced efficiency because of limited light harvesting, the results of the Paternó‐Büchi reaction are in agreement with the ones obtained with ground‐state cycloaddition reactions, for example the Diels‐Alder reaction. ^[19] In the latter cases, the detailed processes that lead to the rate‐accelerating effect of water are still not fully clarified and are even considered to be specific for each particular reaction, but it is commonly accepted that a hydrophobic effect operates in the transition state.[ ¹ , ² ] Specifically, in bimolecular reactions the close contact between the reactants in the transition state minimizes the water‐accessible surface of the pair of molecules, as compared with the one of the separated substrates, thus leading to lower energy of the transition state. In addition, evidence was provided that hydrogen bonding between the reactants and water molecules at the water interface perturbs the frontier molecular orbitals and thereby leads to slightly more favorable HOMO‐LUMO interactions, which in turn result in faster reactions. ^[20] Likewise, it has been demonstrated that hydrogen bonding interactions between benzophenone and ROH groups play a crucial role to control the regio‐ and steroselectivity of the oxetane formation, ^[21] thus supporting the proposed influence of hydrogen bonding on this photoreaction. Nevertheless, the rate acceleration of ground‐state cycloadditions is much more pronounced, i. e. by several orders of magnitude,[ ¹ , ² ] than the ones of the photoreactions presented herein. This observation may lead to the conclusion that the transition state of the Paternó‐Büchi reaction is stabilized to lesser extent by water. At the same time, it should be considered that mechanisms of photoreactions are usually more complex than those of ground‐state reactions, as has already been argued in the interpretation of photodimerization reactions in water. ^[9] In the case of the Paternó‐Büchi reaction, the mechanism involves absorption, intersystem crossing (ISC), exciplex formation, 1,4‐biradical formation, ISC, and finally ring closure to the products. ^[16] Moreover, the intermediates along this pathway can also collapse back to the starting materials. As the water interface may have an influence on each of these reaction steps and intermediates, this complexity poses a problem for a detailed mechanistic assessment and requires more detailed photophysical investigation in future studies. Nevertheless, as the formation of the primary intermediate biradical and the preceding exciplex formation have most likely the largest impact on the reaction rate ^[16] it may be proposed that the rate‐accelerating effect of water also operates on the transition state that leads to these intermediates.

The di‐π‐methane rearrangement was chosen as another model reaction because it offers the opportunity to investigate a monomolecular reaction, which has been rarely a focus in “on‐water” reactions.[ ¹ , ² , ¹³ ] Indeed, even this reaction is faster in water suspension than in organic solvents or in the solid state. Especially the observation that the reaction in the solid state is much slower than in water suspension confirms that the rate acceleration is not just a feature of the solid‐state reaction. Instead, the fast photoreaction takes place on the solid‐water interface. For an explanation of this effect the transition state, TS1, of the rate‐determining reaction step has to be considered, which has been calculated to be the formation of the 1,3‐biradical resulting from the initial vinyl‐benzo bridging (Scheme 3). ^[22] Obviously, this transition state TS1 does not have a structure with significantly limited solvent‐accessible surface as compared with the starting material, so that a strong hydrophobic effect can be excluded. This interpretation is supported by the results with different dibenzobarrelene derivatives 1 a–f with varying alkyl substituents, because with a significant hydrophobic effect the rate should have increased with the more lipophilic substrates that carry longer alkyl groups. Instead, the influence of water may be based on hydrogen‐bonding effects. In fact, it has been shown by the calculations mentioned above that intramolecular hydrogen bonding to carbonyl functionalities in the 9 or 10 position lowers the energy of the transition states of the di‐π‐methane rearrangement of dibenzobarrelenes. ^[22a] Based on these literature data, we propose that the reaction is accelerated under “on‐water” conditions in a resembling hydrogen‐bonded substrate, specifically by means of the hydrogen bonding between the ester functionalities and the available, more reactive hydroxy functionalities at the water‐substrate interface, the latter usually referred to as dangling‐OH groups (Scheme 3).[ ¹ , ² ]

Scheme 3.

Proposed transition state of the photoinduced vinyl‐benzo bridging in the di‐π‐methane rearrangement of dibenzobarrelene derivatives (based on Ref. [22a]).

Conclusion

In summary, we provided further evidence that organic photoreactions have the ability to be significantly accelerated when the reaction is performed under “on‐water” conditions and we thus contribute some of the rare examples of this so far rather unexplored field. Although the effects are not as pronounced as the ones reported for ground‐state reactions, the results clearly show the potential of this approach to perform photoreactions more efficiently in an ecologically favorable medium. Especially considering the current renaissance of photoreactions in organic synthesis, ^[23] in particular by photocatalysis with visible light, ^[24] our results should stimulate further investigations of photoreactions under “on‐water” conditions.

Experimental Section

Materials: Commercially available reagents were obtained from Acros Organics (neo‐pentanol, Z‐cyclooctene), Alfa Aesar (dimethyl acetylene dicarboxylate, n‐propanol, n‐butanol, n‐pentanol), Carl Roth (ethanol) and Merck (anthracene, benzophenone). Technical‐grade solvents were distilled prior to use. Absolute ethanol was obtained by addition of hydroxylamine hydrochloride and sodium hydroxide followed by distillation. Column chromatography was carried out with silica gel 60 M (0.0063–0.25 mm) from MACHEREY‐NAGEL GmbH & Co. Dimethyl‐9,10‐dihydro‐9,10‐ethanoanthracene‐11,12‐dicarboxylate (1 a) was synthesized according to the literature protocol. ^[17a]

Equipment: The NMR spectra were recorded on a JEOL ECZ 500 spectrometer (¹H: 500 MHz, ¹³C: 125 MHz, 22 °C). The spectra were referenced to residual signals in DMSO‐d ₆ [δ(¹H)=2.05 ppm, δ(¹³C)=39.5 ppm] or CDCl₃ [δ(¹H)=7.26 ppm, δ(¹³C)=77.2 ppm] and processed with the MestReNova software. The melting points were determined with a BÜCHI 545 (Büchi, Flawil, CH) and are uncorrected. Elemental analysis data were determined by Rochus Breuer (Organic Chemistry I, University of Siegen) on a HEKAtech EUROEA combustion analyzer. Mass spectra (Supporting Information) were recorded on a Finnigan LCQ Deca (U=6 kV; working gas: Helium; auxiliary gas: Nitrogen; temperature of the capillary: 200 °C, injection rate: 10 μL/min). HPLC analyses were performed on a Agilent 1200 Series HPLC‐UV system in an isocratic run with a Restek Raptor C₁₈ column (50 mm×2.0 mm, 5 μm particle size) or a MACHEREY‐NAGEL Nucleodur Gravity C₁₈ column (150 mm×4.6 mm, 3 μm particle size) with different mobile phases. Absorption spectra were recorded with an Analytik Jena SPECORD S spectrometer. Photoreactions were performed in a cuvette (Paternó‐Büchi reactions) or in a quartz glass test tube equipped with a rubber septum upon irradiation with a high‐pressure Hg‐lamp (Heraeus TQ 150, 150 W). Distances between light source and outer limits of the reaction vessel were 10 cm for the di‐π‐methane rearrangement and 4.5 cm for the Paternó‐Büchi reactions.

Methods: Reaction solutions were stirred with a magnetic stirring bar. Reaction temperatures refer to the medium that surrounded the reaction vessel. Solvents were usually removed under reduced pressure at 40–50 °C with a rotatory evaporator. The room temperature (r.t.) was approximately 22 °C. Air‐ and/or water‐sensitive reactions were carried out under inert atmosphere with Schlenk equipment.

General procedure for di‐π‐methane reactions: The photoreactions were carried out in acetonitrile or in water. For this purpose, a solution of the dibenzobarrelenes 1 a–1 f in acetonitrile (c=5.0 mmol/L, v=7.0 mL) and a suspension in water (c=5.0 mmol/L, v=7.0 mL) were irradiated (λ>280 nm). Samples were taken (v=200 μL) before irradiation and then every 5–20 min. For analysis, the samples were diluted with MeCN/H₂O (MeCN sample; 6.0 mL; H₂O sample: 1.0 mL, 1 : 1, v:v), containing the internal standard dimethylisophthalate (c=167 mmol/L). The analytical samples were microfiltered (PES membrane filters, 0,2 μm) and separated and analyzed by HPLC‐UV. In this process, 1 a and 2 a were detected at 210 nm and 1 b‐1 f, 2 b–2 f at 192 nm. The corresponding parameter values are shown in Table S1 (cf. Supporting Information). Triplicate determinations were performed on all samples, and the mean value was used in each case for the evaluation. For the determination of the relationship between mass concentrations and irradiation time, the normalized signal intensities of the respective barrelene and semibullvalene were plotted in logarithmic scale versus the irradiation time. The observed rate constants were obtained by linear regression analysis (confidence interval: 95 %).

General procedure for Paternó‐Büchi reactions: For photoreactions in acetonitrile, an anaerobic solution of benzophenone (3) (1 mM) and Z‐cyclooctane (4) (3–6 mM) in MeCN was prepared. For reactions “on‐water”, a suspension of benzophenone (1 mM) and Z‐cyclooctane (3‐6 mM) was prepared. The samples were irradiated with the full spectrum of a Hg‐lamp (d=4.5 cm) under stirring and analyzed photometrically after specific time intervals. The observed rate constants were obtained by an exponential regression analysis of the time dependent absorbance, considering a pseudo 1^st order rate law.

Conflict of interest

The authors declare no conflict of interest.

Supporting information

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supporting Information

Schulte R., Löcker M., Ihmels H., Heide M., Engelhard C., Chem. Eur. J. 2023, 29, e202203203.

Data Availability Statement

The data that support the findings of this study are available in the supplementary material of this article.