Cleaning Up Micelle Chemistry

For the past decade I have been intrigued by the emergence of micelle technology and its ability to enable many organic reactions in water, but perplexed about why the reactions work as effectively as they do, and have been waiting for more universal reaction conditions. Two recent papers address these issues.

The title of the first paper, “Entropy reduction from strong localization – an explanation for enhanced reaction rates of organic synthesis in aqueous micelles”, summarizes a new hypothesis for why the micelle methodology is surprisingly effective and has helped me think about these reactions in a new way. (M. P. Andersson, Journal of Colloid and Interface Science 2022, 628, 819–828. https://doi.org/10.1016/j.jcis.2022.08.105.)

The first insight that I gained from the paper was just how extraordinary the micelle rate acceleration effect is. A typical micellar Suzuki reaction is performed with 2 weight % surfactant with 0.1-1.0 ml % catalyst at ambient temperature in an aqueous medium that typical organic chemistry substrates are poorly soluble in. By contrast the “standard” reaction conditions have the substrates fully dissolved at elevated temperature and 1-10 mol % catalyst. Therefore, any mechanistic explanation has to account for the often-accelerated rate of micelle reactions despite a 50-fold reduction in reaction volume, an order of magnitude less catalyst and at least an order of magnitude from performing the reaction at ambient temperature!

The second insight that I got from the paper was that the reaction substrates are located in the region of the linker between the relatively large hydrophilic PEG unit at the surface of the micelle and the lipophilic core. (I’d previously naively assumed that the organic substrates were reacting in the lipophilic core!).

The final insight was that any explanation of the rate-explanation needs to be valid across the wide range of reaction types and substrates within each reaction type. This sets a high bar for any proposed mechanism, but also in many ways simplifies the analysis of various proposals since these strict criteria will rule out many possible options.

Increased local concentration inside the micelle has been proposed previously to be the reason for the rate acceleration. However, this paper argues that these explanations have not accounted for any thermodynamic limit to reactant solubility. Through computational calculations the author shows that the absolute maximum solubility limit is not higher in TPGS-750-M micelles than in organic solvents for reaction components typically encountered in the pharmaceutical industry (i.e. solids at room temperature) and therefore the argument of localized concentration cannot be generally applicable. The key exception noted in the paper is for the catalytic species which typically do partition favorably into the micelle leading to high local catalyst concentrations which enable the reduction in loading for the expensive metal component.

Changes in enthalpy (DH in the Gibbs Free Energy of activation from the substrates to the reaction transition state) are ruled out because the energy barrier of the reaction is dominated by electronic effects which are only weakly dependent on the chemical environment. Changes of overall reaction energy have been reported previously for Suzuki reactions as the result of solvent, but these can only account for less than an order of magnitude acceleration at room temperature. This is insufficient to overcome the 50-fold reduction in reaction volume, let alone account for the fact that reactions in micelles often run faster at lower temperatures than their conventional analogs.

The first part of the argument for entropy-based acceleration relates to the fact that micellar conditions are typically applied to coupling reactions. By definition a coupling reaction has fewer components in the reaction transition state than in input reaction components, and therefore there is a loss of entropy on forming the activated species. Any explanation of the micelle acceleration based on entropy must therefore show a reduced loss in entropy on forming the reaction transition state in a micelle compared to the same transformation in bulk organic solvent.

The key calculations in the paper show that the reaction component’s free energies are lowest at the joining interfaces between the three reaction components. In the figures below these are the gray areas between the PEG layer (green), the linker (red) and lipophilic core (blue). This result can be rationalized by the fact that the asymmetric solvation environment enables any molecule to optimize its orientation to minimize its free energy. This includes aryl bromides, boronic acids and most catalyst complexes which have a polar region located on a convex part of their molecular surface.

As a result one can consider all of the relevant reagents to all be located in a two-dimensional region, or in other words as an “adsorption” to the linker region. This is relevant because it is known that adsorption onto solid surfaces leads to a general loss of ~⅓ of the entropy due to loss of one degree of freedom (i.e. three dimensions to two dimensions). Because the translational entropy contribution does not scale linearly with mass, the transition state will lose less entropy upon localization to a surface than the sum of the constituent reactant parts.

The author then performs a series of DFT calculations in order to estimate the reaction rate gain for a general Suzuki reaction from the entropy loss. The underlying assumptions for these calculations are that:

  • The entropy of components in the liquid phase is qualitatively similar to the entropy in the micellar system.
  • The micelle interface can be modeled as a solid surface

At 298 K the pre-factor increase of the micellar catalysis compared to conventional bulk dissolution in organic solvent is calculated to be 2.5 x 104. This very large number is more than enough to outweigh the reduction from volume effects, or minor contributions from solvation. The dynamic micelle boundary will be less effective at reducing the degrees of freedom and localization than a solid surface so real-world effects will be smaller than these computed values, but still significant and general enough to explain why micelles effectively catalyze such a broad range of coupling reactions. This explanation is consistent with observations that the reaction rate is relatively independent of the presence of co-solvents, micellar shape and micellar size. The author also notes how the hypothesis of an enhanced level of pre-organization is consistent with the fact that often reactions in micelles show enhanced stereo- and/or enantioselectivity. Finally, the paper concludes that these mechanistic insights could guide the development of new surfactants.

The second paper in this article is a new surfactant, Savie, recently reported by Bruce Liphutz’s team at UC Santa Barbara which was developed with the intent of replacing the PEG as the hydrophilic with a more biodegradable, non-peroxide forming alternative (J. Am. Chem. Soc. 2023, 145, 7, 4266–4278, https://doi.org/10.1021/jacs.2c13444). Given the closeness in publication date it would have been impossible for the conclusions of the first paper to have influenced the design of this new surfactant, but it helped me understand why the micelle effect holds even when a completely new hydrophilic moiety is introduced.

The structure of Savie and TPGS-750-M are shown below. Both surfactants share Vitamin E as the lipophilic moiety, but notably Savie contains no linker and has replaced the PEG component with a N-functionalized polypeptide derived from N-methylglycine (i.e. sarcosine). The name Savie is derived from Sarcosine + vitamin E)

Polypeptides have been previously investigated for bioconjugation applications and are known to minimize internal hydrogen bonding, and as a result adopt random coil conformations in water (similar to PEG). N-methyl glycine was selected because:

  • Proteinogenic amino acids would require traditional methods of peptide synthesis which would require precisely the types of environmentally unsustainable solvents that micelle technologies are trying to replace
  • N-methylated amino acids can be readily polymerized via N-carboxyanhydrides (NCAs)
  • Sarcosine is already widely used to make biodegradable surfactants.
  • Unlike PEG is it non-immunogenic, does not form hydroperoxides on exposure to air and is fully biocompatible

The synthesis of Savie is readily achieved as summarized below with an excellent E-factor of 3 which is over four-fold better than that required to make TPGS-750-M.

Reaction work-up requires simple filtration to remove precipitated sodium acetate and unreacted NaH followed by removal and recovery of THF under vacuum to afford the desired product as a powder. Because all of the Sar-NCA monomer is consumed during the reaction the average hydrophilic chain length is determined by the ratio of the monomer to initiator. A library of surfactants was prepared and the surfactant with an average chain length of 15 monomers was found to be optimal and named Savie. A commercial supply of Savie will soon be available which should expedite its uptake by the synthesis community.

A few of the properties of Savie make it superior to TPGS-750-M

  • Savie is readily handled powder. TPGS-750-M is waxy solid
  • Savie dissolves in water within seconds but TPGS-750-M can take hours for full dissolution
  • Savie is far more polar than TPGS-750-M due to the greater hydrophilicity of the polyamide
  • Savie has a critical micelle concentration of 0.015 wt % compared to 0.06 wt % for TPGS-750-M
    • Interestingly Savie appears to exist as a binomial distribution of micelle sizes with peaks at ~13 nm and ~270 nm which challenges the narrative that the enabling properties of PEG-surfactants is related to the fact that they form micellar aggregates of 45-60 nm. (This result is consistent with the entropy driven explanation in the first paper in this article but the authors of this second paper only state that ‘This divergence from expectation implies that “new rule” must exist for PSar-containing surfactants vs. those containing PEG chains’)
  • No difference was observed in the tendency of Savie to foam since foaming is predominantly caused by the lipophilic moiety which is vitamin E in both cases
  • Savie is able to be used at higher temperatures than TPGS-750-M suggesting that the H-bonding interactions between the polyamide and surrounding water are much stronger than for PEG
  • Similarly, the stronger H-bonds mean that Savie is less affected by salting out effects which opens up a wider range of suitable reaction conditions, especially where homogeneity is paramount
  • It is simpler to separate Savie from the desired reaction products by a standard aqueous/organic extraction which translates to better recycling of the aqueous reaction medium and less surfactant contamination in the final product

The paper then demonstrates with numerous examples across a wide range of reaction types previously implemented using micelle technology that in most cases Savie outperforms TPGS-750-M. Most notably the authors discovered that in many cases Savie does not need an organic co-solvent to form stable emulsions. In my opinion, eliminating the need for a co-solvent is a major breakthrough that should help promote the routine use of micelle technology because chemists no longer need to perform additional studies to find the best co-solvent and solvent concentration. It should also enable chemists to feel more confident scaling up this chemistry. The photos below for the Suzuki-Miyaura cross coupling examples clearly show the improvement in reaction mixture homogeneity that Savie provides. This final example (6) is performed on gram scale.

The paper also mentions a “nano-to-nano” effect which I was previously unaware of, which is the propensity of micelle technology to stabilize Pd, Cu and Ni nanoparticles and prevent Ostwald ripening. The substrate-laden nanomicelles are in close proximity to the nanorod catalyst particles resulting in unusually mild reaction conditions with excellent conversion at Pd loading levels typically <1000 ppm. Hopefully, this can spur chemists to significantly reduce the quantity of precious metals in these now ubiquitous coupling reactions!

In this article I have briefly summarized both papers and encourage readers interested in more details and examples to read both full articles. Hopefully, we’ll continue to see progress in micelle chemistry and more pharmaceutical companies describing its use for large scale synthesis campaigns.