SELFIES and the future of molecular string representations

Artificial intelligence (AI) and machine learning (ML) are expanding in popularity for broad applications to challenging tasks in chemistry and materials science. Examples include the prediction of properties, the discovery of new reaction pathways, or the design of new molecules. The machine needs to read and write fluently in a chemical language for each of these tasks. Strings are a common tool to represent molecular graphs, and the most popular molecular string representation, Smiles, has powered cheminformatics since the late 1980s. However, in the context of AI and ML in chemistry, Smiles has several shortcomings-most pertinently, most combinations of symbols lead to invalid results with no valid chemical interpretation. To overcome this issue, a new language for molecules was introduced in 2020 that guarantees 100% robustness: SELF-referencing embedded string (Selfies). Selfies has since simplified and enabled numerous new applications in chemistry. In this perspective, we look to the future and discuss molecular string representations, along with their respective opportunities and challenges. We propose 16 concrete future projects for robust molecular representations. These involve the extension toward new chemical domains, exciting questions at the interface of AI and robust languages, and interpretability for both humans and machines. We hope that these proposals will inspire several follow-up works exploiting the full potential of molecular string representations for the future of AI in chemistry and materials science.

The Highest Oxidation State of Rhodium: Rhodium(VII) in [RhO3 ]. Angew Chem Int Ed Engl 2022;61:e202207688. [PMID: 35818987 PMCID: PMC9544489 DOI: 10.1002/anie.202207688] [Citation(s) in RCA: 0] [Impact Index Per Article: 0]

Although the highest possible oxidation states of all transition elements are rare, they are not only of fundamental interest but also relevant as potentially strong oxidizing agents. In general, the highest oxidation states are found in the electron-rich late transition elements of groups 7-9 of the periodic table. Rhodium is the first element of the 4d transition metal series for which the highest known oxidation state does not equal its group number of 9, but reaches only a significantly lower value of +6 in exceptional cases. Higher oxidation states of rhodium have remained elusive so far. In a combined mass spectrometry, X-ray absorption spectroscopy, and quantum-chemical study of gas-phaseR h O n + (n=1-4), we identifyR h O 3 + as the1 A 1 ' trioxidorhodium(VII) cation, the first chemical species to contain rhodium in the +7 oxidation state, which is the third-highest oxidation state experimentally verified among all elements in the periodic table.

Identification of a Two-Coordinate Iron(I)-Oxalate Complex

Exotic oxidation states of the first-row transition metals have recently attracted much interest. In order to investigate the oxidation states of a series of iron-oxalate complexes, an aqueous solution of iron(III) nitrate and oxalic acid was studied by infrared free liquid matrix-assisted laser desorption/ionization as well as ionspray mass spectrometry. Here, we show that iron is not only detected in its common oxidation states +II and +III, but also in its unusual oxidation state +I, detectable in both positive-ion and in negative-ion modes, respectively. Vibrational spectra of the gas phase anionic iron oxalate complexes [FeIII (C2 O4 )2 ]- , [FeII (C2 O4 )CO2 ]- , and [FeI (C2 O4 )]- were measured by means of infrared photodissociation spectroscopy and their structures were assigned by comparison to anharmonic vibrational spectra based on second-order perturbation theory.

Exploring the Limits of Transition-Metal Fluorination at High Pressures

Fluorination is a proven method for challenging the limits of chemistry, both structurally and electronically. Here we explore computationally how pressures below 300 GPa affect the fluorination of several transition metals. A plethora of new structural phases are predicted along with the possibility for synthesizing four unobserved compounds: TcF₇ , CdF₃ , OsF₈ , and IrF₈ . The Ir and Os octaflourides are both predicted to be stable as quasi-molecular phases with an unusual cubic ligand coordination, and both compounds formally correspond to a high oxidation state of +8. Electronic-structure analysis reveals that otherwise unoccupied 6p levels are brought down in energy by the combined effects of pressure and a strong ligand field. The valence expansion of Os and Ir enables ligand-to-metal F 2p→M 6p charge transfer that strengthens M-F bonds and decreases the overall bond polarity. The lower stability of IrF₈ , and the instability of PtF₈ and several other compounds below 300 GPa, is explained by the occupation of M-F antibonding orbitals in octafluorides with a metal-valence-electron count exceeding 8.

Beyond Oxides: Nitride as a Ligand in a Neutral IrIX NO3 Molecule Bearing a Transition Metal at High Oxidation State

Theoretical calculations utilizing relativistic ZORA Hamiltonian point to the conceivable existence of an IrNO₃ molecule in C_3v geometry. This minimum is shown to correspond to genuine nonavalent iridium nitride trioxide, which is a neutral analogue of cationic [IrO₄ ]⁺ species detected recently. Despite the presence of nitride anion, the molecule is protected by substantial barriers exceeding 200 kJ mol^-1 against transformations leading, for example, to global minimum (O=)₂ Ir-NO, which contains metal at a lower formal oxidation state.

An Objective Alternative to IUPAC's Approach To Assign Oxidation States

The IUPAC has recently clarified the term oxidation state (OS), and provided algorithms for its determination based on the ionic approximation (IA) of the bonds supported by atomic electronegativities (EN). Unfortunately, there are a number of exceptions and ambiguities in IUPAC's algorithms when it comes to practical applications. Our comprehensive study reveals the critical role of the chemical environment on establishing the OS, which cannot always be properly predicted using fix atomic EN values. By identifying what we define here as subsystems of enhanced stability within the molecular system, the OS can be safely assigned in many cases without invoking exceptions. New insights about the effect of local aromaticity upon OS are revealed. Moreover, we prove that there are intrinsic limitations of the IA that cannot be overcome. In this context, the effective oxidation state (EOS) analysis arises as a robust and general scheme to derive an OS without any external guidance.

Valence Interconversion of Octahedral Nickel(II/III/IV) Centers

Three oxidation states (+2, +3, +4) of an octahedral nickel center were stabilized in a newly prepared RhNiRh trinuclear complex, [Ni{Rh(apt)₃ }₂ ]ⁿ⁺ (apt=3- aminopropanethiolate), in which the nickel center was bound by six thiolato donors sourced from two redox-inert fac-[Rh^III (apt)₃ ] octahedral units. The three oxidation states of the octahedral nickel center were fully characterized by single-crystal X-ray crystallography, as well as spectroscopic, electrochemical, and magnetic measurements; all three were interconvertible, and the conversion was accompanied by changes in color, magnetism, and Jahn-Teller distortion.

Cesium's Off-the-Map Valence Orbital

The Td -symmetric [CsO4 ]+ ion, featuring Cs in an oxidation state of 9, is computed to be a minimum. Cs uses outer core 5s and 5p orbitals to bind the oxygen atoms. The valence Cs 6s orbital lies too high to be involved in bonding, and contributes to Rydberg levels only. From a molecular orbital perspective, the bonding scheme is reminiscent of XeO4 : an octet of electrons to bind electronegative ligands, and no low-lying acceptor orbitals on the central atom. In this sense, Cs+ resembles hypervalent Xe.