Data Mining the C−C Cross‐Coupling Genome

Abstract

The speed and precision of machine‐learning (ML) techniques in determining quantum chemical properties has resulted in a considerable computational speed up in comparison to traditional quantum chemical methods, and now allows a desired property of thousands of molecules to be assessed virtually instantaneously. The large databases that result from employing ML can, in turn, be mined with the goal of uncovering relationships that may be missed through more commonly used small scale screening procedures. Due to its prominent place in chemistry, catalysis represents a particularly fruitful playground, where drawing connections between the quantum chemical properties of catalysts and their overall catalytic performance may lead to the identification of new, highly functional species. In this spirit, we previously trained ML models to predict the performance of 18000 prospective catalysts for a Suzuki coupling reaction using molecular volcano plots. Here, we apply concepts from big data to probe a type of “C−C cross‐coupling genome” that explores results from many different named cross‐coupling reactions. The use of interactive dimensionality‐reducing data‐clustering maps facilitates the identification of relationships between the thermodynamics of different catalysts and the chemical properties of their constituent metal and ligands. Analyzing large numbers of species in this manner leads to the identification of not only unexpected catalysts that have thermodynamically ideal profiles to catalyze C−C cross‐coupling reactions, but also reveals a wealth of interesting chemical trends regarding the influence played by different metals and ligands, as well as their unique combinations.

Design of an Accurate Machine Learning Algorithm to Predict the Binding Energies of Several Adsorbates on Multiple Sites of Metal Surfaces

“…The use of ML algorithms [3] is rapidly increasing in the field of chemistry [4] and catalysis. [5] ML methods can be used to develop atomistic potentials, [6] to search for active motifs in intermetallic structures [7] or to address reaction network complexity, [8] holding great potential in combination with data repositories of electronic structure calculations. [9] ML algorithms have been used to predict binding energies of adsorbates on surfaces but have been limited to predict the binding energy of one or just a few similar adsorbates of the same family, i. e. oxygen-based for instance, while modifying the surfaces (mostly on different facets of metals).…”

Simulating solvation and acidity in complex mixtures with first-principles accuracy: the case of CH$_3$SO$_3$H and H$_2$O$_2$ in phenol

“…Fortunately, data-driven techniques provide a framework that can also be used for unbiased structural characterization. [48][49][50][51][52][53][54] Here we use the sketch-map dimensionality reduction algorithm. [55][56][57] Similar to multi-dimensional scaling, 58 sketch-map tries to find a low-dimensional representation of a set of configurations, matching the distances between high-dimensional sets of features that describe each structure, and those between their projections.…”

Methoxycyclization of 1,5‐Enynes by Coinage Metal Catalysts: Is Gold Always Superior?

“…Subsequently, this led to the development of a computational toolkit based on molecular volcano plots capable of tackling problems of high relevance to homogeneous catalysis including creating unified pictures of different reaction classes, [36,37] addressing issues surrounding product selectivity, [38] probing substrate scope, [39] and establishing theoretical turnover frequencies [40] . Some of the latest development have concentrated on incorporating concepts from big‐data and machine‐learning and the applications of these ideas to important chemical problems [41–44] . The purpose of this contribution is to use molecular volcano plots to understand the fundamental features that make gold catalysts so effective for catalyzing the methoxycyclization of 1,5‐enynes, and to propose design strategies that might be experimentally employed to make silver and copper catalysts more active.…”