**Memorial Issue Dedicated to Dr. Howard D. Flack: The Man behind the Flack Parameter**

Editors

**Catherine E. Housecroft Katharina M. Fromm**

MDPI • Basel • Beijing • Wuhan • Barcelona • Belgrade • Manchester • Tokyo • Cluj • Tianjin

*Editors* Catherine E. Housecroft University of Basel Switzerland

Katharina M. Fromm University of Fribourg Switzerland

*Editorial Office* MDPI St. Alban-Anlage 66 4052 Basel, Switzerland

This is a reprint of articles from the Special Issue published online in the open access journal *Chemistry* (ISSN 2624-8549) (available at: https://www.mdpi.com/journal/chemistry/special issues/for Flack).

For citation purposes, cite each article independently as indicated on the article page online and as indicated below:

LastName, A.A.; LastName, B.B.; LastName, C.C. Article Title. *Journal Name* **Year**, *Volume Number*, Page Range.

**ISBN 978-3-0365-1914-2 (Hbk) ISBN 978-3-0365-1915-9 (PDF)**

© 2021 by the authors. Articles in this book are Open Access and distributed under the Creative Commons Attribution (CC BY) license, which allows users to download, copy and build upon published articles, as long as the author and publisher are properly credited, which ensures maximum dissemination and a wider impact of our publications.

The book as a whole is distributed by MDPI under the terms and conditions of the Creative Commons license CC BY-NC-ND.

## **Contents**


#### **Jonathan O. Bauer and Tobias G ¨otz**


## **About the Editors**

**Catherine E. Housecroft** is a Professor of Chemistry in the Department of Chemistry at the University of Basel, Switzerland. Her research group, run jointly with Professor Edwin Constable, is engaged in aspects of materials, coordination, and structural and supramolecular chemistries. Much of the emphasis is upon the design and development of materials which will contribute to mankind achieving the 17 Sustainable Development Goals designated in 2015 by the United Nations General Assembly. Compounds of interest range from coordination compounds which can absorb or emit light to coordination networks which can act as hosts for small molecules.

**Katharina M. Fromm** was raised and educated in Germany, France, and the United States. After receiving a Ph.D. in metal-organic chemistry from the University of Karlsruhe in Germany in 1994, she joined the group of Prof. Joachim Strahle in Tubingen (solid-state chemistry) and Nobel Prize winner ¨ Prof. Jean-Marie Lehn (supramolecular chemistry) for her postdoctoral studies. In 1998, she moved to the University of Geneva for her habilitation, which she received in 2002. After a short intermission at the University of Karlsruhe with an Emmy Noether Stipend II, she was awarded a Swiss National Science Foundation Professorship at the University of Basel, allowing her to expand her research at the University of Fribourg, taking over the chair from Prof. Alexander von Zelewsky. From 2010 to end of 2019, she served as a Research Councilor for the Division "Programs" of the Swiss National Science Foundation, of which she became president in mid-2015 before being elected Vice-President of the Research Council in 2016. After her time with SNSF, she was nominated Vice-Rector for Research and Innovation in January 2020. Her main research interests deal with s-block elements and silver bioinorganic chemistry, although her activities go beyond that and include mechano-responsive polymers, nanocapsules and batteries. In 2013, she was named Fellow of the American Chemical Society (first in Europe), and became a member of the European Academy of Sciences in 2018, the same year she was also announced winner of the Prix Jaubert of the University of Geneva. In 2019, she was elected member of the Swiss Academy of Technical Sciences.

## *Editorial* **Memorial Issue Dedicated to Dr. Howard D. Flack: The Man behind the Flack Parameter**

**Catherine E. Housecroft 1,\* and Katharina M. Fromm <sup>2</sup>**


**\*** Correspondence: catherine.housecroft@unibas.ch

This Special Issue of *Chemistry* is dedicated to Dr. Howard D. Flack (1943–2017), a renowned crystallographer who transformed the way in which, by using single crystal X-ray diffraction, we are able to determine the absolute structure of a crystalline material, and thereby determine the absolute configuration of molecular species within the material. All of us who routinely use cifs are familiar with the 'Flack Parameter', and the younger generations take it for granted. Not all realize that this is a relatively recent implementation [1].

Howard Flack studied at the University of Cambridge, UK, and moved to a position of Maître-assistant in the Laboratoire de Cristallographie at the University of Geneva, Switzerland, in 1972. He remained in Geneva working as a crystallographer for his entire career, and contributed his knowledge, enthusiasm and encouragement to the establishment of crystallographic facilities at other Swiss universities. His invaluable computational and scholarly contributions to single-crystal X-ray diffraction have left an extraordinary mark on the field, as can be appreciated by the bibliography of the Howard Flack's publications contributed to this Special Issue by Constable [2].

As noted above, Howard Flack is probably most associated with the Flack parameter, which is routinely reported for all chiral molecules for which-single crystal X-ray diffraction data are reported in the literature. Watkin and Cooper provide an excellent account of the importance and applications of the Flack parameter, and conclude their paper with some advice on its use [3]. Complementing this, Constable and Housecroft have focused on the development of crystallographic methods within metal coordination chemistry, and demonstrated the role of the Flack parameter in providing a routine method for determining the absolute configuration of coordination compounds [4]. The contributions made by Howard Flack in providing tools for crystallographers to take account of crystal twinning in structure elucidation are represented in the paper by Øien-Ødegaard and Lillerud, which describes the crystal structures of three zirconium-containing metal–organic frameworks determined from twinned crystals [5].

At the heart of a crystal lattice are intermolecular interactions. Tools to analyze structures in terms of these weak interactions are widely available, starting from the molecular structure, although predicting the crystalline assembly is far from straight forward. Aakeröy and coworkers approached this challenge using energy- and informaticsbased prediction models founded on molecular electrostatic potentials, hydrogen-bond energies, propensity and coordination. They have applied these criteria to the crystal structures of twelve pyrazole-based molecules in order to explore the ability to correctly predict supramolecular synthons [6]. Remaining with the theme of crystal engineering, Cern ˇ ý et al. present a review of the crystal chemistry of inorganic hydroborates. This detailed review deals with salts containing hydroborate ions as the only type of anion. The structural patterns that emerge from this investigation should be invaluable to those interested in the design of hydroborate-based materials [7].

**Citation:** Housecroft, C.E.; Fromm, K.M. Memorial Issue Dedicated to Dr. Howard D. Flack: The Man behind the Flack Parameter. *Chemistry* **2021**, *3*, 818–820. https://doi.org/ 10.3390/chemistry3030058

Received: 23 July 2021 Accepted: 23 July 2021 Published: 27 July 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

An octahedral metal center coordinated by three bidentate ligands, e.g., [M(2,2- bipyridine)3] *<sup>n</sup>*+, is the archetypal coordination motif that chemistry undergraduates learn can possess a Δ- or Λ-configuration. Lappin and coworkers have redetermined the singlecrystal structure of Λ-[Co(en)3]Δ-[Co(edta)]2Cl·10H2O at low temperature, and they also report the structure of racemic [Co(sep)][Co(edta)]Cl2·2H2O. The former provides valuable insight into chelate ring conformation and hydrogen bonded interactions in the lattice. The data serve to provide a better understanding of stereoselectivity in ion-pairing and electron transfer reactions [8]. Piguet and coworkers present a detailed investigation of pseudooctahedral spin-crossover [Fe(L)3] 2+ complexes, in which L is an unsymmetrical and sterically demanding α,α'-diimine ligand [9]. The theme of metal coordination chemistry continues with a contribution from McKenzie and coworkers, describing the chemisorption of gaseous NO by the enantiomorphic phases of the cobalt coordination compounds containing a chiral salen ligand [10]. Six-, seven- and eight-coordinate lanthanoid (Ln) metal complexes of types *cis*-[Ln(dppmO2)2Cl2]Cl, [Ln(dppmO2)3Cl]Cl2 and [Ln(dppmO2)4]Cl3 are reported by Reid and coworkers; the dependence of the coordination number and geometry on the Ln3+ radius, and the role of the chloride ions are discussed [11].

Small molecules and their crystal structures are the focus of several contributions to this Special Issue. Bauer and Götz present the crystal structure of chloropentaphenyldisiloxane and highlight intermolecular interactions in monofunctionalized disiloxanes [12]. A new synthetic approach to arylsulfonates has been exemplified by Ngassa and coworkers by the preparation of a series of substituted nitrophenyl-4- -phenylbenzenesulfonates; structural characterizations provide insight into the intermolecular interactions in these species [13]. This structural discussion is complemented by the contribution from Seidel et al., which reports the structural characterization of (*R*)-and *rac*-2-bromo-3-methylbutanoic acid with particular focus on hydrogen-bonded motifs and conformational preferences observed in the solid-state [14]. In a related study, Lehmann and coworkers describe the determination of the absolute configuration of the flavoring agent (+)-γ-decalactone; crystal growth was achieved in situ by cryo-crystallization methods [15]. Rifamycins are an important group of antibacterial agents, and in their contribution to this Special Issue, Frampton et al. report the structural characterization and absolute stereochemistry of a semi-synthetic rifamycin which was prepared using an Alder-Ene addition [16].

A fascinating illustration of the application of Erdmann's anion, *trans-*[Co(NH3)2(NO2)4] −, in the isolation and crystallization of impounded street drugs is provided by Lalancette and coworkers; using salts of cocaine, methamphetamine and methylone, single-crystal structural data were used to calculate powder X-ray diffraction patterns, giving a means for rapid screening of confiscated materials [17].

We are indebted to all the contributors to this Special Issue which provides a timely opportunity to pay tribute to Howard Flack—the man behind the Flack parameter and more.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Editorial* **The Publications of Howard Flack (1943–2017)**

#### **Edwin Charles Constable**

Department of Chemistry, University of Basel, BPR 1096, Mattenstrasse 24a, CH-4058 Basel, Switzerland; edwin.constable@unibas.ch; Tel.: +41-61-207-1001

Received: 17 June 2020; Accepted: 18 June 2020; Published: 2 July 2020

**Abstract:** Howard Flack was a driving force in the development of modern crystallography. Today "the Flack parameter" has entered into the common parlance of crystallography but his influence was far wider. This article provides an overview of his scientific output and a full bibliography.

**Keywords:** crystallography; Howard Flack

#### **1. Introduction**

This short article provides a bibliography of the publications of Howard Flack. This is not the place to provide an appreciation of the contributions of this outstanding crystallographer, but rather to provide a convenient point of reference for his life's work. Excellent overviews of the contribution of Howard Flack are to be found in this volume and elsewhere [1,2]. The short accompanying text groups the output by theme. Published conference abstracts are included for completeness. The bibliography is as complete as I have been able to compile, but as always, the author is responsible for any omissions (and apologizes for any such omissions in advance).

#### **2. The Scientific Output**

#### *2.1. Properties and Structure of Solid State Inorganic Materials*

A number of publications or conference presentations appeared concerning the crystal, mechanical and electronic properties of solid state materials such as MoS2 [3], NbSe2 [3], HfS2 [3], SmAu6 [4], GaxSe1–x [5], TiC [6,7], TiN [6,7], VN [8,9], Al2O3 [10], Mo4Ru2Se8 [11], Mo1.5Re4.5Se8 [11], MNi3Al9 (M = Y, Gd, Dy, Er) [12], TiGePt [13], Nb3Si [14] and Nb3As [14]. Another reported structure of an inorganic material is the "simple" salt, MgSO3·6H2O [15]. One publication from 2003 entitled "Anti-wurtzite reoriented" [16] is noteworthy as it provides a link to the earliest studies of zinc sulfide in which the Friedel pairs were shown to have different intensities [17,18].

Two very early studies were concerned with the properties of crystalline polytetrafluoroethene [19,20].

#### *2.2. Structural Studies on Discrete Species*

Although Flack is not widely thought of as a service crystallographer, a number of small molecule structures determined by him have been reported, including [Cl2(SEt2)2Ir(μ–CI)2IrCl2(SEt2)2] [21], [Cl3(SEt2)Ir(μ–CI)(μ–SEt2)IrCl2(SEt2)2] [22], [Ir(NCS)(NH3)5]Cl2 [23], [Ir(NCS)(NH3)5](ClO4)2 [24], MeOC(Ph) = C(Ph)OH [25], (+)-8β-acetoxy-12-(4-bromobenzoyloxy)-13,14,15,16-tetranorlabdane [26] and other chiral organic [27] and inorganic [28] species.

#### *2.3. Crystallographic Publications*

#### 2.3.1. Disorder

A number of early publications describe types of disorder in anthrone crystals [29–31] and present a general description of X-ray diffraction by such disordered crystals [32]. A body of publications commencing in the 1970s is concerned with new methods for making absorption corrections in diffraction measurements [30,33–37]. The crystallographer's obsession with the "ideal" spherical crystal also attracted his attention at this time [36,38–42].

#### 2.3.2. Improving Data Quality

There is a significant body of work concerned with general aspects of improving the quality of collected crystallographic data [43–48]. Flack also made a number of additional contributions to the use of statistical methods in crystallography [49–53] including a review of the standard work by Shmueli and Weiss [54].

#### 2.3.3. Absolute Configuration, Absolute-Structure Refinement and the Flack Parameter

Probably the most important paper published by Howard Flack was the 1983 work "On enantiomorph-polarity estimation" in which he derived an expression for the determination of the absolute structure of crystals and, therefore, the absolute configuration of any chiral molecules contained therein [55]. Flack included a parameter which he called "x", but which the community now knows as 'The Flack Parameter'. He subsequently published numerous articles on the theory and practice of solving absolute structures [56–82]. Related work dealt with the origins of chirality [83] and general reflections on the use and misuse of symmetry and chiral descriptions in chemistry [84–88].

One additional interesting publication is a review of planar-chiral five-membered metallacycles [89].

#### 2.3.4. Technical Contributions and Software

By its nature, crystallography is a technical and sometimes mathematically demanding and intense subject and Flack made a wide range of contributions to fundamental aspects of crystallography [90–98]. A 2017 book chapter presented a concise state-of-the-art view of chemical crystallography [99].

Flack was also involved in the new knowledge culture, and contributions entitled "Crystallographic publishing in the electronic era" [100], "Internet resources for crystallography" [101] and "World Wide Web for crystallography" [102] attest to this, as does a 1996 workshop "Surfing the Crystallographic Net (CRYSNET) workshops" [103]. Another conference contribution entitled "Is your crystal representative of the bulk?" addressed one of the fundamental issues for a synthetic chemist relying on crystallography [104].

He was a co-author of many software programs including DIFRAC [105] and HUG [106], X-RAY 76 [107] X-RAY 76 (the first System offering all the software needed for an X-ray analysis) and its successor, XTAL [108,109].

Finally, I should note that one of his last publications testifies to the breadth of his engagement with, and commitment to, crystallography in its broadest sense [110]. The web-site http://crystal.flack.ch/ maintains an up-to-date publication list and links to other sources commemorating his work.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The author declare no conflict of interest.

#### **References**


© 2020 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Howard Flack and the Flack Parameter**

#### **David John Watkin \* and Richard Ian Cooper**

Chemical Crystallography Laboratory, Department of Chemistry, University of Oxford, Mansfield Road, Oxford OX1 3TA, UK; richard.cooper@chem.ox.ac.uk

**\*** Correspondence: david.watkin@chem.ox.ac.uk

Received: 25 August 2020; Accepted: 21 September 2020; Published: 23 September 2020

**Abstract:** The Flack Parameter is now almost universally reported for all chiral materials characterized by X-ray crystallography. Its elegant simplicity was an inspired development by Howard Flack, and although the original algorithm for its computation has been strengthened by other workers, it remains an essential outcome for any crystallographic structure determination. As with any one-parameter metric, it needs to be interpreted in the context of its standard uncertainty.

**Keywords:** Howard Flack; Flack parameter; structure analysis; X-ray crystallography

#### **1. Introduction**

The formation of Howard Flack (1943–2017) was as a chemist, but he was also an able mathematician who turned this skill to crystallographic problems (Figure 1). Most of his contributions to crystallography will pass unnoticed by chemists relying on X-ray analysis to robustly characterise new materials and are probably not well known even to many professional structure analysts. He did not grind out an endless stream of papers, but instead produced a small number of carefully written, carefully thought out works in the domains of data acquisition, space group theory and crystallographic twinning.

Much of Flack's output has gone into the infrastructure of crystallography and need not concern structural chemists, but two of his insights have an immediate impact on ordinary structure analyses—correcting the observed data for absorption effects, and evaluating the absolute structure of chiral materials. An excellent technical review of the physics and mathematics behind absolute structure determination can be found in [1].

**Figure 1.** Howard Flack, courtesy of his widow Evelyne.

#### **2. Impact on Service Crystallography**

#### *2.1. Background*

Modern atomic-resolution small molecule X-ray structure analyses produce an enormous number of experimental observations even from relatively simple chemical substances. This wealth of data has enabled two procedures first described by Flack to become universally accepted as part of routine structure analyses.

Unlike specular reflection of light by a mirror, interaction between a crystal and X-rays is an interference phenomenon, which means that 'reflections' (actually diffracted beams) only occur when the crystal is orientated to satisfy Bragg's Law. Every diffracted beam (identified by the Miller Indices *h*,*k*,*l*) has a unique relationship with the unit cell of the crystal and an intensity dependent upon the internal make-up of the sample. Diffraction occurs at the Bragg Angles (θ) given by:

$$\left(\frac{\sin\theta}{\lambda}\right)^2 = \text{hG}^\*\text{h}^\text{t}\tag{1}$$

where G\* is the reciprocal metric tensor (characteristic of the crystal) and h is the row vector of indices. *h*,*k*,*l* are positive and negative integers permutated up until the maximum sin θ = 1. In practice there are experimental limitations below this maximum, but even so, X-ray crystal structure analysis is *data rich*, that is to say, there are very many more observations available than there are parameters describing the structure. For much of the 1960s, 70s and 80s most X-ray data were measured on instruments which observed reflections one at a time (serial diffractometers). Depending upon the space group of the sample, some (and perhaps many) of the permutated indices refer to *equivalent reflections*. Equivalent reflections are ones for which the theoretical intensity should be identical even though the indices are different. Thus, in the space group *Pc* the reflection *h*,*k*,*l* will be equivalent to *h*,−*k*,*l*. Because of the time taken to measure data, it was common practice to only measure one of a group of equivalent reflections, thus, in the example above, halving the data collection time. When more than one equivalent reflections are measured, after correction for experimental geometry and short comings, they are averaged (merged) by crystallographic analysis software to provide a set of *unique* observed reflections. Rmerge and Rint are measures of the self-consistency of groups of equivalent reflections. Current area detector diffractometers measure hundreds of reflections quasi-simultaneously, so that many equivalent reflections can be observed.

In a structure determination, the analyst postulates an atomic model (usually via a *structure solution* program), generally with 9 parameters per independent atom. A structure factor can be calculated from this model, and the parameters optimized to match *Ic* with *Io* (see Appendix A) by the method of nonlinear least squares. In a modern analysis, the observation: parameter ratio is often in the range 10 to 20, equivalent to fitting a straight line (*y* = *mx* + *c*) to between 20 to 40 observations.

Just as the image of an object remains visible when a mirror is rotated about its normal, so diffraction continues to occur as the crystal is rotated about the normal to the diffracting plane. This means that on a suitably engineered instrument, (e.g., a four circle diffractometer), the rotational orientation (ψ) of the crystal can be varied for a given Miller index. Before the widespread use of area detectors, about 25 years ago, only a single measurement was generally made for each unique Miller index. Modern instruments are normally programmed to make between 5 and 20 observations of equivalent reflection at different ψ values. These independent measurement of the same quantity are termed redundancy, or multiplicity of observation, MoO. Measurements on a typical organometallic crystal might take a less than an hour on a laboratory instrument, or just a few minutes at a synchrotron source, yielding a data set of tens of thousands of observations.

#### *2.2. Correction for Absorption*

One problem facing all X-ray structure analysts is 'correcting' their observed data for experimental effects. Some of these, such as the Lorentz and polarisation corrections, have been well characterised analytically since the 1920's. More problematic has been computing corrections for absorption effects. X-rays are attenuated as they pass through a medium according to Beer's Law:

$$I\_{\mathcal{O}} = I\_i \mathbf{e}^{-\mu t} \,, \tag{2}$$

where *Io* is the observed X-ray intensity, *Ii* is the incident intensity, μ is the absorption coefficient (computed from the elemental composition [2]) and *t* is the path length through the sample. Unless the crystal can be ground to a sphere (which is still sometimes done for special purposes), the exponential means that exact mathematical calculations depend upon very accurate measurements of the crystal shape and size—still not easily done even with modern digital microscopes.

For an X-ray beam travelling along the long axis of a prismatic crystal, the intensity of the emergent beam will reduce and then increase again as the crystals is rotated through an angle ψ about a vector perpendicular to the long axis. In the 1960s, methods based on experimental observations were developed for making empirical corrections for absorption by tracking the intensity of a chosen reflection as ψ was systematically varied [3,4]. These methods created a kind of calibration curve for the sample. Flack recognised that these curves could be better represented by a smooth mathematical function [5]. Because of the periodic nature of the curves, one natural function of choice was the Fourier series. Flack's careful experimental verification of the method, which included making thousands of observations on a serial diffractometer, demonstrated the strength of the concept. Today, area detector diffractometers inevitably measure many reflections at several different values of ψ. While these differing ψ values almost never correspond to systematic curve tracking, their great abundance enables them to be used is a similar manner. Blessing replaced the Fourier series by spherical harmonics [6], and this has become the basis of all experimental data corrections up to modern times. It turns out that this empirical method of correcting for absorption can also be used to correct for a range of other experimental problems, and since modern area-detector diffractometers provide a large number of equivalent observations (high MoO), the corrections are generally applied without any special action from the analyst. The robust correction of the observations is an important contributor to a robust determination of the absolute structure of a crystal.

#### *2.3. Determination of Absolute Structure*

However, for most chemists working in chiral chemistry, it is Flack's parameter that will most evidently impact them. It had been known since the 1930's that under specific conditions the diffraction of X-rays from a non-centrosymmetric crystal displays small effects which reflect the symmetry of the crystal, and hence the symmetry of the materials the crystal was built from [7]. Of special interest to the chemist working with chiral materials is the fact that for crystals of these materials, the diffracted intensities of a pair of reflections *h*,*k*,*l* and −*h*,−*k*,−*l* (which are identical for crystals in centrosymmetric space groups) are subtly different [8]. The difference between the members of these Friedel pairs are generally very small and so easily masked by experimental issues. The magnitude of these Friedel differences depends upon the resonant scattering of the elements in the material and the wavelength of the X-rays. Early attempts to determine the absolute structure of a material depended upon the analyst identifying enantiomer sensitive pairs and then carefully remeasuring them. Hamilton's 'R-factor Ratio Test' tried to use all the observable reflections [9], but was difficult to apply reliably. A breakthrough was Roger's proposal that the absolute structure of a material could be estimated, together with a reliability index, by the determination of one additional parameter (which he called η) in the model [10]. Unfortunately, the derivative of this parameter with respect to the calculated structure factor is discontinuous over the physical range of the parameter, meaning that its interpretation was unsound except for truly enantio-pure materials. Flack became interested in this mathematical problem, and within two years had reformulated it in terms of a new parameter (which he called '*x*', see Section 4 for derivation) which had a continuous derivative, and thus could be used in the analysis of crystals which, while containing only chirality-preserving symmetry operations, contained a mixture of domains of two enantiomers—in crystallography called '*twinning by inversion*' [11]. When *x* is zero, the material is enantiopure and the atomic coordinates determined from the data correspond to the actual structure. If *x* is unity, then the atomic model is the inverse of the actual structure—the model is 'inverted'.

#### **3. Twinning by Inversion**

It has been estimated that when solutions contain exactly equal amounts of both enantiomers of a compound form crystals, approximately 90% will form racemic crystals, containing exactly equal numbers of each enantiomer [12]. These may be related by a centre of inversion—the centrosymmetric crystals drawn from the set labelled CA in Figure 2, or by chirality-inverting non-centrosymmetric operations in non-centrosymmetric space groups (crystal classes labelled NA in Figure 2) for example space group *Pc*.

Occasionally, as Pasteur had the good fortune to observe, the two enantiomers preferentially separate out into enantiopure crystals, a phenomenon called spontaneous resolution. The enantiopure constituent molecules restrict the choice of space group to one of the 65 Sohncke groups which contain exclusively chirality-preserving symmetry operations, i.e., rotations and translations (from the crystal classes labelled NC in Figure 2).

In rare cases both enantiomers may crystallize together in a Sohncke space group as independent molecules which are not related by a symmetry operation of the crystal. Approximately 50 examples of this type were identified in Flack's review of chiral and achiral crystal structures in 2003 [13]. Materials exhibiting this phenomenon have subsequently been termed kryptoracemates, as their racemic nature is hidden by the space group symmetry. A recent survey identified 724 examples in the crystallographic literature [14].

In addition to these cases, it is also possible to form a solid solution of enantiomers. The crystal grows without chiral selectivity so that either enantiomer can occupy a molecular site within a crystal structure resulting in a partially disordered solid solution of enantiomers [15].

**Figure 2.** The 32 geometric crystal classes. CA means Centrosymmetric & Achiral, NA means Noncentrosymmetric & Achiral, NC means Non-centrosymmetric & chiral. Note that the symbols in the achiral classes contain either an m, 4 or 6 operator, allowing pairs of enantiomers to be related by a chirality-inverting symmetry operation which is not a centre of symmetry. Figure by courtesy of Evelyne Flack [16].

If the solution contains one and only one enantiomer, it will crystallise in a Sohncke space group. In this case, the determination of Flack's parameter will enable the stereochemistry of the material to be estimated. Note well that the stereochemistry is only estimated. Refinement, by least squares, of all the parameters needed to characterise the material (the positions of the atoms in the unit cell, their motion about their mean positions, and Flack's parameter) provides both values for these parameters and estimates of their standard uncertainties (written u or s.u.), previously called standard deviations. This is really important. For example, the standard uncertainties in atomic positions enable one to

compute the standard uncertainty in the distance between them, which in its turn enables a rational decision to be made as to whether two bond lengths are significantly different. Flack's parameter is also determined with a standard uncertainty, enabling the analyst to assign a level of confidence to the proposed absolute structure.

It is not clear when 'Flack's x parameter' became simply 'The Flack parameter'. For the 20 years following its formulation, the determination of the Flack parameter required extreme attention to experimental details. For first row elements the magnitude of the resonant scattering is small for copper radiation (λ = 1.54 Å) and even smaller for molybdenum radiation (λ = 0.71 Å), and so it was generally held that the absolute structure of light atom materials required copper radiation. It was not uncommon at that time to form a derivative by inserting an atom with larger resonant scattering (such as chlorine) into the molecule. One snag with this approach was that heavier atoms tend to have higher absorption coefficients—though the absorption corrections described above tend to minimize the impact on the data quality. The millennium saw the widespread use of area detectors for the measurement of diffraction data at temperatures close to the boiling point of liquid nitrogen. Working at this temperature reduces the motion of the atoms in the molecules, and greatly improves the quality of the experimental observations. These instruments measure the intensities of very many Bragg reflections quasi-simultaneously rather than measuring the reflections one-by-one. As well as providing the many observations needed for the absorption corrections, these instruments also measured many of the Bijvoet (commonly used as a synonym for Friedel) pairs of reflections needed for the determination of Flack's parameter (see Appendix B). The determination of absolute structure by the determination of the Flack Parameter became common place.

If the enantiopurity of the sample is unquestionably known by other experimental methods, then an estimate of the Flack parameter by any diffraction calculation is likely to be sound. Detailed evaluation of modern data from light atoms structures of known chirality has shown that even analyses using molybdenum radiation are unlikely to point to an incorrect absolute structure [17].

Problems arise if the enantiopurity is uncertain. Unless instructed otherwise, most X-ray analysts search through the sample for the 'best' looking crystal, i.e., one with bright faces, roughly isometric shape, no inclusions, etc. The outcome of the analysis is the absolute structure of that particular crystal. Nothing can be said about the nature of the other crystals in the batch, nor of the solution from which the crystals were grown. Other, non-crystallographic, methods must be used to verify that the absolute structure of the sample crystal is indeed the same as that of the whole batch. Although the crystal used in the X-ray analysis may measure less than 0.1 × 0.1 × 0.1 mm, it is now possible to verify by NMR, chromatography or optical methods that the single crystallographic specimen really represents the bulk sample [18]. Without a references sample or sophisticated simulations such methods cannot directly determine the absolute configuration of enantiomers, however these methods are extremely valuable for determining the enantiomeric excess of samples which can be used as a valuable piece of prior information in a crystallographic absolute structure determination.

Thompson et al. tabulated all the possible outcomes from samples of unknown purity [19]. The bottom line is that for a sample which is not enantiopure, these cases may occur singly or together:


The crystallographer can do little about cases 1 and 2 except pay attention to the morphology of the crystals in the sample. In case 1 crystals may grow, as they did for Pasteur, having small facets which are mirror images in the two enantiomeric crystals. In case 2 it is very likely that the achiral crystals will look quite different to the enantiopure chiral crystals. It is now widely accepted that case 3

is uncommon for organic materials, but may be an issue for inorganic or organo-metallic materials where chirality of the sample may arise from the arrangement of the building blocks in the crystal [20]. Where twinning by inversion does exist, it may be clearly identifiable under an optical microscope, or it may be on a sub-microscopic scale with small zones of each enantiomer distributed throughout the crystal. Case 4 is illustrated by silica. The building block of quartz, SiO4, is achiral but these motifs can be assembled in crystals to form left or right handed spirals. The hand of the spiral sometimes reveals itself as small facets on the surface of well-formed crystals but is extremely rare [21].

#### **4. Interpretation of the Flack Parameter**

The equation defining the Flack Parameter [11] is simply:

$$I\_o \cong I\_c = (1 - \mathbf{x})I\_s^+ + \mathbf{x}I\_s^-,\tag{3}$$

where *x* is a kind of partition or mixing coefficient with extreme values of zero or unity. *Io* is the observed structure amplitude, *Is* is the amplitude computed for a single enantiomer, *Ic* is the amplitude computed for the twin. The crystallographer develops an atomic model representing the structure (i.e., one which can be displayed in a graphics program) and from this can compute structure amplitudes (*Is* <sup>+</sup> and *Is* −) for the structure and its inverse.

*x* can be included as a parameter in the structure refinement, or Equation (3) can be solved for *x* by least squares from all the reflections once *Ic* <sup>±</sup> are known. The first method is that used by Flack in his original work [11]. Sheldrick developed a post-refinement method for estimating the parameter as a one-off computation once the main refinement was completed which he called the "Hole in One" method [22]:

$$I\_o^+ - I\_s^+ \triangleq \propto \left(I\_s^- - I\_s^+\right) \tag{4}$$

Later Hooft et al. [23] re-wrote the expression to use the differences of Friedel pairs and developed a Bayesian statistical post-refinement method:

$$D\_{\mathcal{O}} \cong (1 - 2\chi)D\_{\mathcal{S}\_2} \tag{5}$$

where *Do* = *Io* <sup>+</sup> <sup>−</sup> *Io* <sup>−</sup> and similarly for *Ds*. Parsons included the averages of the observed and of the computed Friedel pairs as a way of reducing the influence of experimental errors—the method now widely known as Parsons' Quotients in Equation (6) [24]:

$$Q\_{\mathcal{O}} \cong (1 - 2\pi) Q\_{\mathcal{S}\prime} \tag{6}$$

where *Qo* <sup>=</sup> *I* + <sup>0</sup> − *I* − 0 /1 2 *I* + <sup>0</sup> + *I* − 0 and similarly for *Qs*.

For a while these post-refinement methods were treated with suspicion because they lacked the estimates of covariance between the atomic parameters and Flack's parameter which classical full matrix refinement provided. Leverage analysis of the full matrix for refinement of the Flack parameter together with the atomic parameters identifies those reflections most influential in determining *x* [25]. Ironically, it turns out that relatively few reflections carry strong information about the Flack parameter—very careful remeasurement of these reflections should reduce the standard uncertainty of *x*. Eventually, an in-depth analysis by Parsons et al. showed that except in marginal cases the covariance was small, and that post refinement methods were generally safe to use [26].

Most crystallographic programs enable you to compute (with greater or lesser ease) the Flack parameter by both direct refinement in the main analysis (Flack's original method), and by one or more of the post refinement methods. Generally, the methods give substantially the same result, except the standard uncertainty of the direct refinement can be about twice as large as post refinement methods. Where there is a significant difference either in the value of *x* itself or in the standard uncertainty, this can sometimes be traced back to problems with the analysis [27]. These problems are masked in the main refinement because the observations are given weights which include both a contribution from

the observed variance of the reflection and a modifier chosen so that the weighted residual, *w F*2 *<sup>o</sup>* <sup>−</sup> *<sup>F</sup>*<sup>2</sup> *c* 2 , is fairly uniform across all the data. This modifier is justified in that it accounts for unidentified short comings in both the data and the model, and yields a linear normal probability plot. Post refinement methods generally give a linear normal probability plot using the unmodified variances.

#### **5. Conclusions**

#### *Practical Advice Concerning the Flack Parameter*

It is crucial to make it clear to the X-ray analyst from the outset that the absolute structure needs to be determined. Although modern instruments routinely yield exceptionally high-quality data, data collection parameters can be further optimized for the determination of absolute structures. Since the X-ray analysis will usually be made on just one single crystal selected from the material submitted, it is important to provide an estimate the sample's enantiopurity. This will warn the analyst about the need to look out for crystals with an anomalous morphology. Where possible, recover the actual crystal used in the X-ray analysis and verify that it represents the bulk sample. It is no longer fashionable to make heavy atom derivatives since it is preferable to avoid unnecessary additional synthetic processes. However, it does not seem to be widely appreciated that the heavy atom does not have to be part of the material under investigation. Samples containing solvent of crystallization where the solvent contains a heavy atom (e.g., CH2Cl2) will generate larger Bijvoet differences.

The determination of a parameter from many thousands of observations by the method of least squares means that as well as getting an estimate of the parameter value, one also gets an estimate of its reliability—the standard uncertainty, e.g., 0.05 ± 0.01, written by crystallographers as 0.05(1).

The standard uncertainty is key to interpreting the value of *x*. If the parameter is determined with a large s.u. then the value of *x* is unreliable and caution should be used when drawing conclusions about the enantiopurity of the sample. Flack and Bernardinelli argued that for a material known to be enantiopure, a standard uncertainty of less than 0.08 is sufficient for the assignment of absolute structure to be well determined [28]. A chirally sensitive HPLC or NMR experiment can establish the ratio of enantiomers in a sample, but not the absolute configuration of the dominant enantiomer. If the enantiopurity is unknown at the time of the crystallographic analysis, the s.u. needs to be below 0.04 to indicate a confident outcome. Even here, there is room for uncertainty. If the refinement gives *x* = 0.10(4), this can be interpreted as *x* lying within 3 s.u. of zero, so the absolute configuration is confirmed. However, it can also be interpreted as meaning that the sample contains up to 10% of the opposite enantiomer in the form of a racemic twin. X-ray crystallography can provide a robust indication of the absolute structure of the bulk of the single crystal used in the experiment, not the absolute structure of the bulk of a batch crystalline sample, and it can only provide an estimate of the enantiopurity (with standard uncertainty).

**Author Contributions:** Conceptualization and production, R.I.C. and D.J.W. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **Appendix A. Observed and Calculated Structure Amplitudes**

If an atomic model exists for a crystalline material, a calculated structure factor can be computed from:

$$F = A + iB$$

and hence:

$$F^2 = A^2 + B^2$$

Note that the structure factor is a complex quantity, having both a magnitude and a phase. The squared term is the calculated structure intensity which is proportional to the observed structure intensity. The magnitude can easily be derived from the intensity, but in general the phases are unknown.

In the older literature, the observed intensity usually referred to the physically measured magnitude squared of a given reflection and was represented by the symbol *Io*. This was converted to an observed squared structure factor by the application of various experiment-dependent correction factors, represented by the symbol *F*<sup>2</sup> *<sup>o</sup>* , and the corresponding term computed from the atomic model was represented by *F*<sup>2</sup> *c* .

Increasingly (and confusingly), following the usage in the macromolecular community, *Io* is used to represent *Fo <sup>2</sup>*, and similarly for *Ic*. We have followed this convention in this manuscript, with the additional extension that *Is* represents the squared structure amplitude computed for an enantiopure material, and *Ic* allows for twinning. See Equation (3).

#### **Appendix B. Friedel and Bijvoet Pairs**

A definition gaining traction is that Friedel pairs are strictly related by inversion (e.g., *h,k,l* and −*h,*−*k,*−*l,*), where as a Bijvoet pair is any symmetry equivalent related by a symmetry operator of the second kind (e.g., *h,k,l* and *h,*−*k,l*) [29].

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **When Stereochemistry Raised Its Ugly Head in Coordination Chemistry—An Appreciation of Howard Flack**

#### **Edwin C. Constable \* and Catherine E. Housecroft**

Department of Chemistry, University of Basel, BPR 1096, Mattenstrasse 24a, CH-4058 Basel, Switzerland; catherine.housecroft@unibas.ch

**\*** Correspondence: edwin.constable@unibas.ch; Tel.: +41-61-207-1001

Academic Editor: Spyros P. Perlepes Received: 24 August 2020; Accepted: 8 September 2020; Published: 12 September 2020

**Abstract:** Chiral compounds have played an important role in the development of coordination chemistry. Unlike organic chemistry, where mechanistic rules allowed the establishment of absolute configurations for numerous compounds once a single absolute determination had been made, coordination compounds are more complex. This article discusses the development of crystallographic methods and the interplay with coordination chemistry. Most importantly, the development of the Flack parameter is identified as providing a routine method for determining the absolute configuration of coordination compounds.

**Keywords:** crystallography; Flack parameter; chirality; coordination chemistry

#### **1. Introduction**

When the authors first studied chemistry, the accepted mantra was that crystallography could not be used routinely to determine the absolute configuration of a compound. The basis for this assertion was subsequently shown to be incorrect, but routine determination of the absolute configuration of a compound by crystallography remained in the realms of the relatively exotic. In 1966, a comprehensive listing identified 54 organic structures, the absolute configurations of which had been determined by crystallographic methods [1]. Over the next few years, more determinations were reported, with an additional 40, 39 and 133 absolute structures identified in 1968 [2], 1969 [3] and 1970 [4], respectively. For organic compounds, classical correlation methods relying on real or virtual chemical transformation were still used to relate the configuration to one of the known absolute configurations. Indirect methods of determining the absolute configuration relied upon esoteric spectroscopic methods or empirical correlations such as the octant rule [5]. In 1983, the situation was changed by a paradigm-shifting publication by Howard Flack [6].

This article is not a detailed account of the crystallographic background, but rather takes the opportunity to survey the impact of crystallographic methods on the investigation of coordination compounds and to appreciate the broader contributions of Howard Flack. Nevertheless, a short introduction to the crystallographic complexities is included. Optical activity and subsequently chirality have had a profound influence on the development of coordination chemistry, from the time of Werner onwards.

#### **2. Chirality**

#### *2.1. Through a Glass, Darkly*

From the earliest documented history, mirrors have excited and fascinated Mankind. In the New Testament of the Christian Bible, the phrase "For now we see through a glass, darkly" refers to a poorly discerned image in a mirror. Mirrors, images in mirrors, and mirror images influenced Western art and literature from the early Renaissance to modern times [7]. The 19th Century C.E. imagination was energized and inspired by mirror images and their relationship to "reality". One of the best known literary works on this theme is the 1871 work *Through the Looking-Glass, and What Alice Found There* by Lewis Carroll (Figure 1) [8].

**Figure 1.** Mankind has long been fascinated by objects and their mirror images. In this illustration by John Tenniel from *Through the Looking-Glass, and What Alice Found There* we see the subtley different reality in the enantiomeric world. (Public domain image. Source https://en.wikipedia.org/wiki/Through\_the\_ Looking-Glass#/media/File:Aliceroom3.jpg).

By the end of the 19th Century C.E., chemists were familiar with the fact that some compounds exhibited the phenomenon of optical activity. Jacobus Henricus van't Hoff [9,10] and Joseph Achille LeBel [11] had independently explained the phenomenon of optical activity as being a consequence of four groups attached to a carbon atom being oriented in space in the form of a tetrahedron. In his paper, "A suggestion looking to the extension into space of the structural formulas at present used in chemistry and a note upon the relation between the optical activity and the chemical constitution of organic compounds", van't Hoff used the word *asymmetric* to describe a carbon atom with four different groups attached. Both van't Hoff and LeBel showed that when four different groups were bonded in a tetrahedral arrangement about a carbon atom, the object and its mirror image were non-superposable. Although it is critical to distinguish between the observable phenomenon (optical activity) and its origin (dissymmetry), it was slowly becoming clear by the end of the 19th Century, that optical activity should be regarded as arising from a molecular dissymmetry rather than an asymmetry associated with a single atomic centre [12–18].

Although most people "knew" what a mirror image was and understood, on some level at least, the consequent left–right inversion, it fell to William Thomson (1824–1907), better known under his later title of Lord Kelvin to make a scientific definition that brought clarity to the concept of mirror images. Kelvin is well known for his works in physical chemistry and thermodynamics, indeed he is commemorated in the S.I. unit of temperature, the kelvin. Nevertheless, for us his critical contribution, which was not widely recognized or adopted at the time, is the introduction of the terms chiral and chirality; "I call any geometrical figure, or group of points, *chiral*, and say it has chirality, if its image

in a plane mirror, ideally realized, cannot be brought to coincide with itself. Two equal and similar right hands are homochirally similar. Equal and similar right and left hands are heterochirally similar or 'allochirally' similar (but heterochirally is better). These are also called 'enantiomorphs,' after a usage introduced, I believe, by German writers. Any chiral object and its image in a plane mirror are heterochirally similar" [19]. The word 'chiral' is derived from the Greek word for hand, χειρ. Kelvin's dismissal of enantiomorphs is a little disingenuous. Not only was this description being used by the United Kingdom chemical community in the 1890s [20,21], but also had a long tradition in the German literature [22–24] being introduced by Naumann in 1856 [25] and treated at length in Schoenflies' 1891 text on crystallography [26]. The IUPAC has subsequently made recommendations on the basic terminology of stereochemistry [27] and Gal has published extensively on the early etymology of stereochemical terms [28–32].

As mentioned above, Kelvin's proposal was widely ignored by the community and according to a Scifinder search, the first subsequent mention of the word chiral in the chemical literature occurred in the early 1920s [33,34] and with no further mention up to 1950, when Raman used the term in the context of quartz crystals [35]. The rehabilitation of the word seems to stem from letters to *Nature* in the late 1950s [36,37]. It is interesting to note that in the first two publications introducing the Cahn–Ingold–Prelog system, the words chiral and chirality are not used [38,39] and only in the third paper do they appear extensively, with the comment "This useful word [chirality] was brought to our attention by Professor K. Mislow, who referred us to Webster's Dictionary (2nd Edition), where *chiral* is defined as *Of, or pertaining to the hand, specifically turning the plane of polarisation of light to either hand*"[40].

#### *2.2. Some Definitions*

In introducing the concept of chirality, Kelvin made the link between molecular dissymmetry and the experimental observable of optical activity. It cannot be stressed strongly enough that chirality is a property of an object as a whole rather than something associated with a portion of that object, such as an "asymmetric atom". The term chirality equates exactly with the term dissymmetry (*dissymmétrie*) originally used by Pasteur [28,29,41–43]. Note that dissymmetric and asymmetric are not equivalent—asymmetric means without symmetry whereas dissymmetric is more specific and means "lacking improper symmetry elements". In simpler language, dissymmetric means chiral. The IUPAC definition of asymmetric is very clear [44]:

*"Lacking all symmetry elements (other than the trivial one of a one-fold axis of symmetry), i.e., belonging to the symmetry point group C1. The term has been used loosely (and incorrectly) to describe the absence of a rotation–reflection axis (alternating axis) in a molecule, i.e., as meaning chiral, and this usage persists in the traditional terms asymmetric carbon atom, asymmetric synthesis, asymmetric induction, etc".*

If an object is chiral, then the object and its mirror image are not identical and cannot be superposed. The IUPAC uses the word superpose rather than the more familiar English word superimpose.

The concepts of chirality and symmetry are closely related. This article is not intended as a primer in group theory, and the requirements for chirality are presented without clarification of the symmetry operations that are involved. The reader is referred to standard texts on group theory for further education and elucidation [45].

For a molecule to be chiral, it must have no symmetry elements of the second kind, such as a mirror plane, a centre of inversion or a rotation–reflection axis. The presence of any one of these symmetry elements precludes chirality. It is not sufficient to only look for a mirror plane—even if no mirror plane is present, the presence of an inversion centre will ensure that an object is not chiral. This reinforces the primary criterion for chirality—that of non-superposability of the object on its mirror image. If an object is superposable on its mirror image, it is described as being *achiral*. Such objects possess a mirror plane, a centre of inversion, or a rotation–reflection axis. And remember, nothing can surpass a good quality model or diagram to determine whether a molecule is chiral. The IUPAC definition of chiral is concise and precise [44]:

*"The geometric property of a rigid object (or spatial arrangement of points or atoms) of being non-superposable on its mirror image; such an object has no symmetry elements of the second kind (a mirror plane,* σ = *S1, a centre of inversion, i* = *S2, a rotation-reflection axis, S2n). If the object is superposable on its mirror image the object is described as being achiral."*

#### **3. The Importance of Chirality in Coordination Chemistry**

#### *3.1. It all Began with Werner*

The observation of different optical forms of chiral coordination compounds played a critical role in the development and acceptance of Werner's coordination theory and the establishment of the octahedral geometry of six-coordinate metal complexes. Werner reported the resolution of salts of *cis*-[CoCl(en)2(NH3)]2<sup>+</sup> [46], *cis*-[CoBr(en)2(NH3)]2<sup>+</sup> [46], *cis*-[Co(en)2(NO2)2] <sup>+</sup> [47], *cis*-[CoCl(en)2(NO2)]<sup>+</sup> [48], *cis*-[CoCl2(en)2] <sup>+</sup> [49], [Co(en)3] <sup>3</sup><sup>+</sup> [50], [Rh(en)3] <sup>3</sup><sup>+</sup> [51] and the preparation by ligand exchange reactions of resolved precursors of optically active salts or by seeding of *cis*-[Co(en)2(H2O)(NH3)]3<sup>+</sup> [46], *cis*-[Co(CO3)(en)2] <sup>+</sup> [52], *cis*-[Co(C2O4)(en)2] <sup>+</sup> [53,54], *cis*-[CoBrCl(en)2] <sup>+</sup> [55], [(en)2Co(NH2)(NO2)Co(en)2] <sup>4</sup><sup>+</sup> [56], [(en)2Co(NH2)(O2)Co(en)2] <sup>4</sup><sup>+</sup> [57] and, finally the "all-inorganic" compound [Co{(OH)2Co(NH3)4}3] <sup>6</sup><sup>+</sup> [58]. For determining the optical rotation, Werner utilized an F. Schmidt and Haensch polarimeter [59], and originally used the *d* and *l* descriptors to describe the compounds with positive and negative rotation of 656.3 nm wavelength linearly polarized light, respectively. He subsequently reported optical rotatory dispersion (ORD) spectra which are simply plots of the variation in optical rotation of linearly polarized light with wavelength [60]. The development of the successful resolution methods is described in the doctoral thesis of Victor King [61]. A number of excellent surveys of this aspect of Werner's work have been published [62–64].

#### *3.2. Non-Crystallographic Approaches to Determining the Absolute Configuration of Metal Complexes*

Although Werner had successfully obtained the large selection of optically pure compounds described above, they were only defined by a relative configuration. Subsequent workers adopted the practice of using the (+)<sup>d</sup> or (−)<sup>d</sup> notation to identify the clockwise or anticlockwise rotation at the sodium D-line (589 nm). The situation regarding absolute configuration was worse than that with organic compounds, as only few transformations between optically active species were known and there was no understanding of rules for retention or inversion at octahedral centres. This meant that it was not possible to use mechanistic principles to correlate the relative configuration in a series of complexes. An example of the complexity of the situation is seen in a classical publication from John Bailar Jr. in which the the reaction of (−)d-*cis*-[CoCl2(en)2]Cl with liquid ammonia at 196 K gave (−)d-*cis*-[Co(en)2(NH3)2]Cl ([α] 298 *<sup>D</sup>* −32◦), whereas reaction with a saturated solution of ammonia in methanol at 298 K gave (+)d-*cis*-[Co(en)2(NH3)2]Cl ([α] 298 *<sup>D</sup>* +31◦) [65].

In the 1930s, Matthieu attempted with only poor success to correlate the relative (or indeed absolute) configuration of complexes [66–79] with the sign of the Cotton effects that they exhibited [80–82]. If a compound under investigation has an absorption band in the region of the optical rotatory dispersion spectrum, an anomalous dispersion effect is seen. This is called the Cotton effect and refers to the change in sign of the optical rotatory dispersion close to an absorption band. Close to the region where light is absorbed, the magnitude of the optical rotation varies rapidly with wavelength, passes through zero at the absorption maximum and continues varies with wavelength but with an opposite sign.

The first successful attempts at determining the absolute configuration of octahedral transition metal complexes were made by Werner Kuhn in the 1930s [83,84]. Kuhn used a coupled oscillator model to calculate the sign of the Cotton effect. Kuhn assigned the absolute configuration of Δ to the (−)d-[Co(C2O4)3] 3– anion [83] and the (−)d-[Co(en)3] <sup>3</sup><sup>+</sup> cation (Figure 2) [84].

**Figure 2.** (**a**) IUPAC recommends the notation Δ and Λ for denoting the absolute configuration of octahedral metal complexes. The Δ (pale blue) and Λ (pink) enantiomers of an [M(bpy)3] *<sup>n</sup>*<sup>+</sup> complex (bpy = 2,2 -bipyridine) are presented here. (**b**) The absolute configuration proposed by Kuhn and Bein for (−)d-[Co(C2O4)3] <sup>3</sup><sup>−</sup> and a representation using the same colour coding as in (a) to show the Δ configuration (only the carbon and the coordinated oxygen atoms of the oxalate ligands are shown) [83]. Figure 2b ©1934 Kuhn, W.; Bein, K. *Z. Anorg. Allg. Chem.* **1934**, *216*, 321–348 reproduced with kind permission of John Wiley and Sons.

#### **4. A Brief History of Crystallography**

#### *4.1. X-rays—The Early Days*

The availability of routine X-ray crystallographic methods for the determination of solid state structures has had a profound effect on the way in which coordination (and other) chemists characterize their compounds. It is necessary to add a small caveat to curb the enthusiasm of the chemist—typically a single crystal structure provides information about the arrangement of atoms, molecules and ions *within the crystal studied* but, in the absence of other methods, gives no information about the bulk material.

X-rays were discovered in 1895 by Wilhelm Conrad Roentgen when he identified a new type of radiation from a Crookes tube [85,86]. Laue demonstrated that the wavelengths of X-rays were commensurate with the spacings between atoms, ions and molecules in crystals and that, therefore, an optical diffraction was to be expected [87–89]. This critical work by Laue was paradigm shifting for chemistry—suddenly, atoms became objects which could be detected and quantified in terms of size and spatial position. By 1913, the Braggs had built their first single-crystal X-ray spectrometer with a gold leaf electroscope as the detector and the crystal could be rotated in the X-ray beam.

A number of excellent reviews on the early history of X-ray diffraction have been published and the reader is referred to these to learn more about this fascinating story [90–96].

#### *4.2. From X-rays to Chemical Crystallography*

Chemical crystallography began when William Henry Bragg and his son William Lawrence Bragg showed that the diffraction data could be solved by ad hoc trial-and-error methods to yield models of the spatial arrangement of atoms and ions in crystalline solids such as NaCl, KCl, KBr, KI, ZnS and diamond [97–105].

#### *4.3. Initial Approaches to the Phase Problem*

X-ray detectors measure the intensity of radiation but not the phase of that radiation. Any phase change due to scattering of radiation is almost the same for all atoms. The underlying physics of X-ray diffraction was developed by Darwin and Bragg from 1914 onwards [106–108]. The diffraction data correspond to the amplitude of the Fourier transform of the unit cell electron density in the unit cell. Friedel pairs of diffraction spots are Bragg reflections which are reflected through the origin and related by Friedel's law, which states that they have equal amplitudes and opposite phases [109].

The electron density can be obtained by Fourier synthesis if the phase is known. The introduction of Fourier analyses [110], and subsequent elaboration by Patterson with his eponymous function, started the transition of crystallography to become a more routine technique and also provided the methods that could be used for phase evaluation of the diffraction data [111,112]. The Patterson method recognizes that although the phase information is needed to locate the peaks in electron density within a unit cell, the magnitudes of the structure factors contain information about the spacings. The Patterson map shows peaks at all positions corresponding to an interatomic vector rather than the position of the atoms. Patterson methods can only be used for relatively small molecules (<50 atoms) and are not appropriate for the ab initio determination of absolute configuration.

The next development was the study of series of compounds which only differed by the replacement of one atom in the structure. The method relied upon the replaced atom being on a special position and having a different scattering factor. Within these constraints, the effect is the modification of the structure factors by values dependent on the phase of the reflection [113]. One of the earliest applications was to the alums, and diffraction data from the compounds AB(SO4)2•12H2O (A = NH4, K, Rb, Cs, Tl; B = Cr, Al) were used to determine the structures using this isomorphous replacement method [114]. Probably the most spectacular early success used the related heavy atom method in an isomorphous series for solving and refining the centrosymmetric phthalocyanine complexes [M(pc)] (M = Ni, Pt) (Figure 3) and consequently, the correct phasing of H2Pc [115–117]. One of the early applications to a chiral molecule was reported in the structure of cholesteryl iodide. In this compound, the iodine is not on a special position and the step-wise work-flow is worth rehearsing to demonstrate the skill and tenacity of the early crystallographers: (1) measure the intensities of all reflections, (2) locate the heavy atoms from the Patterson map, (3) calculate the phases from the heavy atom contributions, (4) use chemical knowledge to choose correct distances and angles, and (5) recalculate phases on all the atoms and repeat the Fourier summations.

**Figure 3.** The projection of the nickel phthalocyanine molecule presented by Robertson in his pioneering work on the crystallography of phthalocyanines [116]. ©1937 Robertson, J.M.; Woodward, I. *J. Chem. Soc.* **1937**, 219–230 reproduced with kind permission of The Royal Society of Chemistry.

#### *4.4. The Bijvoet Method*

Most crystallographers accepted Friedel's law that the Bragg pairs had equal amplitudes and opposite phases and considered that X-ray methods could not be used directly for the determination of absolute configuration without chemical modification or isomorphous replacement. However, a number of physicists and crystallographers remained open-minded. If the X-ray radiation has a wavelength close to the absorption edge of an atom in the compound, a small phase change occurs in the scattered X-rays from these atoms, phenomenologically similar to the Cotton effect. As a consequence, the diffraction pattern is no longer centrosymmetric but has pairs of spots with unequal intensities. This effect was first reported in studies of zinc blende (ZnS) using W Lβ (1.2447–1.3017 Å) [118] or Au Lα<sup>1</sup> (0.8638 Å) [119] radiation, which are close to the Zn K edge (1.2837 Å). However, these results were forgotten by the community for over a decade.

At the end of the 1940s, the Dutch crystallographer Johannes Bijvoet developed a general isomorphous replacement method for non-centrosymmetric structures and successfully solved the structure of strychnine using the sulfate and selenate salts [120–125]. However, the most significant contribution of Bijvoet was in rediscovering that X-ray analysis could determine absolute configurations using the anomalous scattering of X-rays of a wavelength close to an X-ray absorption of an atom in the compound. This effect was observed in sodium rubidium tartrate (Figure 4) using Zr Kα radiation, which is close to the K-edge of rubidium, and in the early 1950 s, Bijvoet published a series of landmark papers, commencing in 1951 with "Determination of absolute configuration of optically active compounds by means of X-rays" [95,126–128].

A further simplification was introduced by Mathieson who, making use of the known absolute configurations that were available from the Bijvoet method, proposed the use of diastereoisomers in which one of the stereochemical centres was absolutely defined. This avoided the need to use X-ray radiation of a wavelength close to an absorption edge in the compound and utilized the heavy atom method with auxiliaries of known absolute configuration such as (*R*) or (*S*)-chloroiodoacetate [129].

**Figure 4.** The absolute configuration of tartaric acid as depicted by Bijvoet after the determination of the structure of sodium rubidium tartrate. [116,127]. ©1951 Bijvoet, J.M.; Peerdeman, A.F.; van Bommel, A.J. *Nature* **1951**, *168*, 271–272, reproduced with kind permission of Springer Nature. The absolute configuration confirmed the random choice selected by Fischer.

In parallel, direct methods were being developed to permit the extraction of the phase information from the crystallographic structure factors and crystallography entered its modern phase [130,131]. Direct methods estimate and then test and select the initial phases and were initially introduced for centrosymmetric space groups and subsequently extended to non-centrosymmetric space groups [132,133].

#### **5. Chirality and Crystallography**

Chirality, crystallography and symmetry are related in intimate and subtle ways that lead to fascination and frustration. The nuances and manifestations of these relationships delighted and occupied Howard Flack through much of his career. In this section, we rehearse a few of the consequences of these relationships for coordination compounds.

#### *5.1. Absolutism*

Chemists tend to talk of *absolute configuration* meaning the spatial arrangement of the atoms of in a chiral molecular entity and denoted by a stereochemical descriptor such as *R*/*S*, *P*/*M*, *D*/*L* or Δ/Λ. In contrast, crystallographers use the term *absolute structure*, introduced in 1984 by Peter Jones [134], to describe spatial arrangement of atoms in a non-centrosymmetric crystal [135]. The term absolute structure refers specifically to the crystalline state, whereas the broader term absolute configuration can be applied to any phase or to solutions.

#### *5.2. Chiral Space Groups and Chiral Molecules*

There are 230 three-dimensional space groups. Of these 22 are chiral space groups comprising 11 enantiomorphic pairs (*P*41-*P*43: *P*4122-*P*4322: *P*41212-*P*43212: *P*31-*P*32: *P*3121-*P*3221: *P*3112-*P*3212: *P*61-*P*65: *P*6122-*P*6522: *P*62-*P*64: *P*6222-*P*6422: *P*4132-*P*4332). It would be tempting to expect that a chiral molecule would crystallize in a chiral space group, but that would be too simple, and would not take the infinite perversity of nature into account. In Section 2.2, we stated that for an object to be chiral, it must possess no symmetry elements of the second kind. Of the 230 space groups, there are 65 (including the 22 chiral groups) which only possess operations of the first kind (rotations, rotation–translations, and translations) and these 65 are known collectively as the Sohncke groups [136]. Enantiopure chiral molecules must crystallize in one of these 65 Sohncke space groups.

Returning to the perversity of nature, achiral molecules can also occur in any of the 230 space groups—the packing of achiral objects may be in a chiral or an achiral manner. A systematic survey of achiral molecules in non-centrosymmetric space groups was reported in 2005 [137]. Similarly, a crystal composed of equal numbers of the two enantiomers of a molecule (a racemate) may occur in any of the 230 space groups, although examples in the Sohncke set are very rare.

#### *5.3. The Flack Parameter*

By the early 1980s, the ratios of crystallographic R or Rw values for structures refined with alternative absolute configurations were being used to assign absolute configurations and absolute structures, although the statistical methods were debated [138,139]. In 1983, Flack published a paper entitled "On Enantiomorph-Polarity Estimation" which revolutionized the field of structure determination of chiral molecules and structures [6]. Rogers had introduced a parameter η which he proposed as a good method for distinguishing between refinements of structures with opposite configurations. Flack pointed out an inherent problem with the use of η and introduced a new parameter that he called *x* and defined in terms of the structure factor for reflection **h** in Equation (1).

$$\left| F(\mathbf{h}, \mathbf{x}) \right|^2 = (1 - \mathbf{x}) \left| F(\mathbf{h}) \right|^2 + \left| F(-\mathbf{h}) \right|^2 \tag{1}$$

If the experimental data and the model used for refinement have the same chirality, the value of *x* is 0, and if they have opposite configurations, then *x* has a value of 1. A value of 0.5 indicates a racemic crystal with equal amounts of both enantiomer. This single publication revolutionized the determination of absolute structures. Flack also recognized that there was no need to formally solve the structures in both configurations and in a note added in proof to the original paper, he notes that "The refinement of ... *x* has now been added as a permanent feature in our implementation of the X-RAY76 system ... *x* is varied automatically in the final stages of refinement with a non-centrosymmetric structure." The direct refinement of the Flack parameter together with other structural parameters is now routine. To date, this paper has been cited 11,161 times. The parameter *x* rapidly, and universally, became known as the Flack Parameter. To date, the Flack parameter does not yet belong to the data indexed by the Cambridge Structural Database (CSD) [140,141].

#### *5.4. Some Musings on Racemates, Spontaneous Resolution and Other Complexities*

Although the introduction of the Flack parameter transformed the mechanics and quantification of the determination of absolute configurations, it is our opinion that a second paper by Flack is equally important. The 2003 article entitled "Chiral and Achiral Crystal Structures" provides the clearest and most comprehensive compilation of the language of chirality and crystallography that we know. He commences by identifying three origins of chirality in crystal structures: (i) the molecular components (ii) the crystal structure itself and (iii) the symmetry group of the structure. The really valuable part of the publication concerns the quantification of a descriptor (racemic, racemate) that is often used loosely or incorrectly by the chemical community [142]. These descriptors simply refer to equimolar amounts of opposite enantiomers with no restriction to phase. If the two enantiomers are not present in equal amounts, the correct description is an enantiomeric mixture.

Particularly important is the identification of a *racemic conglomerate*, obtained when the crystallization of a solution of a racemic compound results in spontaneous resolution and the generation of equal numbers (strictly equal weights) of enantiopure crystals each only containing components with only one chirality.

He then identified the term *ordered racemic crystal structure* or racemic structure to describe a crystal containing an ordered array of equal numbers of the different enantiomers. The term anomalous racemate had been used to describe ordered crystals in which the ratio of the two enantiomers was not 1:1, and Flack proposed a new description of *M:N mixed enantiomeric crystal structure* or M:N enantiomeric structure, where M:N is the ratio of the two enantiomers present. He further proposed the term *disordered mixed enantiomeric structure* to replace pseudoracemate in describing crystals in which each position can be occupied by a molecule of either configuration. This publication is recommended to anyone who wishes to think more deeply about the consequences of chirality in the solid state.

#### **6. Chiral Coordination Compounds**

Although chirality has been so important in the development of coordination chemistry, it is surprising how few monographs or comprehensive reviews exist. The standard works on stereochemistry concentrate on organic compounds [143]. For coordination chemistry, the bibles are von Zelewsky's 1996 work "Stereochemistry of Coordination Compounds" [144] and Hawkins earlier work "Absolute Configuration of Metal Complexes" [145]. A more organometallic-oriented, but extremely useful and relevant, presentation is found in the 2008 book "Chirality in Transition Metal Chemistry" by Amouri and Gruselle [146]. Two earlier reviews complement these to provide an overview of the period at which crystallography was starting to provide information about absolute stereochemistry and a volume of the ACS Symposium Series from 1980 that was dedicated to the "Stereochemistry of Optically Active Transition Metal Complexes" provides an excellent overview of the state of the art at that time [147–149]. Volume 12 of Topics in Stereochemistry was entitled "Topics in Inorganic and Organometallic Stereochemistry"and for the real afficionados, the chapter "Conformational Analysis and Steric Effects in Metal Chelates" provides a masterly overview of conformational effects within chelate rings and is both a *tour de force* and a challenge [150].

#### *6.1. The First Absolute Determination*

The first determinations of the absolute configuration of metal complexes using the Bijvoet method was reported in 1955 by Yoshihiko Saito who studied Λ- and Δ-[Co(en)3]Cl3·0.5NaCl [151]. In a 1974 review, Saito surveyed the literature up to 1972 and reported that, in the intervening 23 years, the absolute configuration of salts of only an additional 53 metal cations had been determined through structural characterization [148].

The list of compounds characterized provides a snap-shot of the contemporary coordination chemistry. The first class included those which required the "pure" Bijvoet anomalous dispersion approach for compounds such as [ML3] *<sup>n</sup>*<sup>+</sup> (L = chelating bidentate ligand) or *cis*-[ML2X2] *n*+ (L = chelating bidentate ligand, X = monodentate ligand). The structural elucidation of the diamines into related polyamines with polymethylene spacers resulted in new classes of chiral complexes in which the chirality arises from the dissymmetric arrangement of the ligand donor atoms about the metal centre (see the general references at the beginning of this section). These latter compounds can be seen as the direct progenitors of macrocyclic chemistry and the grandparents of supramolecular chemistry. Interesting as these compounds are, we return to our main theme when we consider the remaining types of compounds characterized.

We start with a small digression into the consequences of multiple stereogenic centres in a compound. The classical method of resolving a chiral coordination compound, for example a cation C+, is to form salts with a chiral anion A–. Four possible compounds can be formed, of which [(Δ-C)(Δ-A)] and [(Λ-C)(Λ-A)] form a pair of enantiomers as do [(Δ-C)(Λ-A)] and [(Λ-C)(Δ-A)] (Figure 5). All other relationships between the combinations are as diastereoisomers. Enantiomers have identical physical properties (solubility, melting point, NMR spectra) and only differ in their interactions with other chiral agents (for example polarized light). On the other hand, diastereoisomers differ in physical properties as the spatial interactions between, for example, (Δ-C) with (Λ-A) and (Δ-A) will be different (think about putting your left foot into a left shoe and into a right shoe—the thermodynamics of the pairing are different). A typical resolution method might involve treating a racemic mixture of cations (Λ-C) and (Δ-C) with (Λ-A) and hoping that the solubility of the diastereoisomeric salts [(Δ-C)(Λ-A)] and [(Δ-C)(Δ-A)] might be sufficiently large that one selectively precipitates or crystallizes. Salts of this type provided the next class that were studied extensively in this first period of determining the absolute configuration of coordination compounds. Very typically, chiral organic anions, or anions

containing coordinated chiral organic ligands, were used for the precipitation. After Bijvoet established the absolute configuration of tartrate, known chemical transformations allowed correlation to a large number of other "simple" chiral organic compounds. As a consequence, the absolute configuration of the organic component of the anion was generally known when the structure of the diastereoisomeric salt [(C)(A)] was determined and the configuration at the cation followed from the correct assignment of the (known) configuration to the anion.

**Figure 5.** The four possible combinations of a chiral cation and a chiral anion: [(Δ-C)(Δ-A)] and [(Λ-C)(Λ-A)] form a pair of enantiomers (denoted by a red double-headed arrow) as do [(Δ-C)(Λ-A)] and [(Λ-C)(Δ-A)]; all other relationships are as diastereoisomers (denoted by a blue double-headed arrow).

The final class of compounds studied belonged to the next level of structural development in which the ligands themselves are chiral. Consider propane-1,2-diamine (H2NCH2C\*H(Me)NH2, pn) in which C2 (indicated with an asterisk) is a stereogenic centre; as a result the ligand pn is typically encountered as one of the two pure enantiomers *R*-pn or *S*-pn, as the racemate containing equal amounts of *R*-pn and *S*-pn or the enantiomeric mixture *x*(*R*-pn):1–*x*(*S*-pn). We now consider the formation of the octahedral complex [M(pn)3] *<sup>n</sup>*+.

Starting with enantiomerically pure *R*-pn or *S*-pn, four compounds could be obtained: Δ-[M(*S*-pn)3] *<sup>n</sup>*<sup>+</sup> and Λ-[M(*R*-pn)3] *<sup>n</sup>*<sup>+</sup> (a pair of enantiomers) and a second pair of enantiomers Λ-[M(*S*-pn)3] *<sup>n</sup>*<sup>+</sup> and Δ-[M(*R*-pn)3] *<sup>n</sup>*+. All other relationships are diastereoisomeric. Thus, reaction with *R*-pn will give two chemically distinct diastereoisomers Λ-[M(*R*-pn)3] *<sup>n</sup>*<sup>+</sup> and Δ-[M(*R*-pn)3] *<sup>n</sup>*+.

The more perverse readers will now ask what happens with the racemic ligand? In this case, in addition to the four homoleptic complexes already mentioned, we now have the possibility of the heteroleptic complexes Δ-[M(*S*-pn)2(*R*-pn)]*n*<sup>+</sup> and Λ-[M(*R*-pn)2(*S*-pn)]*n*<sup>+</sup> (a pair of enantiomers) as well as Δ-[M(*R*-pn)2(*S*-pn)]*n*<sup>+</sup> and Λ-[M(*S*-pn)2(*R*-pn)]*n*<sup>+</sup> (a second pair of enantiomers). But it gets worse! The two nitrogen donor atoms of each pn ligand are not chemically equivalent as one is attached to the stereogenic C2 and the other to C1. As a consequence, there are also the complexes with a facial or meridional arrangement of the C\*-NH2 donors! Once again, this is not the place to follow the fascinating stereochemistry of "simple" systems like this.

We make one further foray into the world of stereochemical complexity revealed in these early crystallographic studies. Our discussion commenced with the definition of the Δ- and Λ-configuration using [M(bpy)3] *<sup>n</sup>*<sup>+</sup> complexes as an example (Figure 2). Why did we not start with Werner's [M(en)3] *n*+ compounds? Whereas the chelate ring in bpy complexes is planar, that in en complexes is non-planar and chiral. The configuration is denoted by the descriptor δ or λ as defined in Figure 6. Although individual δ or λ chelate rings are enantiomeric, the situation is different in [M(en)3] *<sup>n</sup>*<sup>+</sup> when we have the possibilities of Δ-[M(δ-en)3] *<sup>n</sup>*<sup>+</sup> and Λ-[M(λ-en)3] *<sup>n</sup>*<sup>+</sup> (a pair of enantiomers), Λ-[M(δ-en)3] *<sup>n</sup>*<sup>+</sup> and Δ-[M(λ-en)3] *<sup>n</sup>*<sup>+</sup> (a second pair of enantiomers) and Δ-[M(δ-en)2(λ-en)]*n*<sup>+</sup> and Λ-[M(λ-en)2(δ-en)]*n*<sup>+</sup> (a third pair of enantiomers) as well as Δ-[M(λ-en)2(δ-en)]*n*<sup>+</sup> and Λ-[M(δ-en)2(λ-en)]*n*<sup>+</sup> (a fourth pair of enantiomers), with each pair of enantiomers having a different thermodynamic stability. The early crystallographic studies confirmed that the favoured crystal forms were typically Λ-[M(δ-en)3] *<sup>n</sup>*<sup>+</sup> and Δ-[M(λ-en)3] *<sup>n</sup>*+.

**Figure 6.** The two enantiomeric forms of a five-membered chelate ring involving an en-like ligand are denoted by the descriptors δ or λ.

#### *6.2. The Influence of Flack*

It is in the area of dissymmetric metal complexes that we see the real influence of Flack and his parameter. If a metal complex crystallizes in one of the Sohncke space groups, most crystallographic software will automatically generate the Flack parameter (hopefully close to 0 or 1) to indicate whether the choice of absolute configuration is correct. A Flack parameter value close to 0.5 is an indication that something is amiss, typically an inversion twin. What is remarkable is how often the chemist, as opposed to the crystallographer, does not comment on the absolute configuration of the complex. There are many examples of spontaneous resolution of tris(chelate)metal complexes in the CSD which are not specifically identified in their associated publications.

Today, we have a veritable wealth of chiral systems and different types of chirality that Werner and Pasteur could only have dreamed of. The Flack parameter finds widespread use in the study of novel chiral systems such as helicates, cyclic helicates and knotted systems exhibiting topological chirality and has been used to establish the absolute configuration of molecular trefoil knots assembled in self-sorting processes [152]. The use of the Flack parameter to establish the asymmetric crystallization of coordination networks and metal-organic frameworks is an interesting new development [153–155].

And what of the future? Routine inclusion of the Flack parameter in CIF files should be encouraged, if only to stimulate further discussion of chirality aspects within the inorganic and coordination chemistry communities! The merits and demerits of determining the Flack parameter during the refinement or post-refinement are currently being discussed [156]. Finally, data-mining activities would be dramatically improved by the inclusion of the Flack parameter in the standard data in Crystal Structure Databases. In the course of writing this article, we became aware of the difficulty of answering simple questions such as "How many chiral coordination compounds have been structurally characterized?" and "How many coordination compounds exhibit asymmetric crystallization?".

#### **7. Concluding Remarks**

This article is an attempt to express our thanks to Howard Flack for providing a tool that made our lives as coordination chemists more interesting and rewarding and, at the same time, to place this innovation of the Flack parameter in its historical perspective.

**Author Contributions:** This article was conceived and written jointly by E.C.C. and C.E.H. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work received no external funding.

**Acknowledgments:** As always, we give our thanks to the various library and abstracting services which have aided us in identifying and sourcing material.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Twinning in Zr-Based Metal-Organic Framework Crystals**

#### **Sigurd Øien-Ødegaard \* and Karl Petter Lillerud**

Centre for Materials Science and Nanotechnology, University of Oslo, P.O. box 1126 Blindern, 0318 Oslo, Norway **\*** Correspondence: sigurdoi@kjemi.uio.no

Received: 19 August 2020; Accepted: 14 September 2020; Published: 16 September 2020

**Abstract:** Ab initio structure determination of new metal-organic framework (MOF) compounds is generally done by single crystal X-ray diffraction, but this technique can yield incorrect crystal structures if crystal twinning is overlooked. Herein, the crystal structures of three Zirconium-based MOFs, that are especially prone to twinning, have been determined from twinned crystals. These twin laws (and others) could potentially occur in many MOFs or related network structures, and the methods and tools described herein to detect and treat twinning could be useful to resolve the structures of affected crystals. Our results highlight the prevalence (and sometimes inevitability) of twinning in certain Zr-MOFs. Of special importance are the works of Howard Flack which, in addition to fundamental advances in crystallography, provide accessible tools for inexperienced crystallographers to take twinning into account in structure elucidation.

**Keywords:** MOFs; crystallography; twinning

#### **1. Introduction**

Metal-organic frameworks (MOFs) are porous solids consisting of inorganic nodes linked by organic multidentate ligands (e.g., through strongly coordinating groups such as carboxylates or Lewis bases) to form extended networks [1]. The combination of diverse coordination chemistry and the structural richness of organic ligands enable a vast number of possible MOF structures and network topologies [2]. Due to their remarkable ability to harbor a large range of chemically functional groups in pores that are accessible to guest species, MOFs are studied mainly (albeit not exclusively) for their properties as adsorbents [3,4] and catalysts [5]. In particular, MOFs based on zirconium oxide clusters are frequently reported due to their relatively high stability and structural diversity [6]. These MOFs are normally obtained as single crystals by adding growth modulators (monocarboxylic acids) to the synthesis liquor [7].

The principal method of MOF structure determination is X-ray diffraction methods, and single crystal X-ray diffraction (SC-XRD) in particular [8]. Despite being the most powerful technique to solve crystal structures, a successful SC-XRD experiment normally depends on a relatively large single crystal of high quality.

The phenomenon in which a crystal consists of two or more separate domains, and these domains are related by a symmetry operation that is not present in the space group of the crystal, is known as twinning [9]. Twinned crystals represent a commonly encountered problem in crystallography and will often prevent a successful structure determination. The observed reflections of all the present domains can be interpreted as originating from the same crystal instead of separate entities, thus obscuring the true symmetry (and thus the structure) of the crystal. Crystalline MOFs and related network structures are prone to interpenetration (intergrown separate networks that may occur if the void fraction of the structure is large) and twinning, because twin domain interfaces are enabled by the flexibility of the MOF's building units (i.e., geometrical flexibility of (1) the linkers' points of connectivity due to free rotation around C-C bonds and (2) the possible existence of metal oxoclusters with a different arrangement of coordination sites in twin boundaries).

In our ongoing crystallographic characterization of Zr-based MOFs, several cases of crystal twinning or related problems were discovered, which can be categorized into three groups:


**Figure 1.** Crystals of UiO-67-Me2, showing the presence of twinning by the spinel law.

The following sections describe how these twinning issues has been resolved for specific Zr-MOF samples, and how a correct assessment of twin laws has revealed structural features with significant implication for the MOFs' properties. The strategies described herein to detect and resolve twinning will presumably be useful to researchers encountering the issue while working with MOFs and other crystalline network solids.

#### **2. Materials and Methods**

#### *2.1. General Materials and Synthesis Methods*

All chemicals, unless otherwise stated, were obtained reagent grade from Sigma-Aldrich and used without further purification. The linkers for 1 and 2 (3,3 -dimethyl-4,4 -biphenyldicarboxylic acid (Me2-H2bpdc) and 2 ,3-dimethyl-[1,1 :4 ,1:4,1-quaterphenyl]-4,4-dicarboxylic acid (Me2-H2qpdc)) were prepared using previously published protocols [11]. Crystallization was performed under static conditions using conical glass flasks that were treated with NaOH(aq) 30% wt. overnight then thoroughly rinsed and dried before the reaction. This treatment seems to decrease the MOF's tendency to nucleate rapidly on the glass surface. During crystallization the flasks were capped with loose lids to prevent pressure build-up and accumulation of decomposition products of DMF (mainly formic acid and dimethylamine) and HCl.

Synthesis of UiO-67-Me2 (1): To 13 mL of dimethylformamide (DMF), 56 μL (3.1 mmol) H2O, 241 mg (1.03 mmol) ZrCl4, and 3.79 g (31.0 mmol) benzoic acid was added and stirred until a clear solution was obtained, and subsequently heated to 120 ◦C. 280 mg (1.03 mmol) Me2-H2bpdc was then added, a clear solution was quickly obtained, transferred to a clean conical flask and kept at 120 ◦C for 72 h. The solid crystalline product was isolated and washed, once briefly with DMF at 100 ◦C, once with dry DMF at room temperature and three times with dry 2-propanol. The MOF crystals were kept in dry 2-propanol prior to activation and XRD measurement.

Synthesis of Zr-stilbene dicarboxylate (2): To 50 mL of DMF, 70 μL (3.9 mmol) H2O, 301 mg (1.29 mmol) ZrCl4 and 4.73 g (38.8 mmol) benzoic acid was added and stirred until a clear solution was obtained, and subsequently heated to 120 ◦C. 346 mg (1.29 mmol) trans-stilbene-4,4 -carboxylic acid was then added, a clear solution was obtained and the solution was transferred to a clean conical flask and kept at 120 ◦C for 72 h. Large octahedral crystals were isolated and washed. First briefly with DMF at 100 ◦C, then dry DMF at RT, then three times with tetrahydrofuran (THF), and lastly three times with n-hexane. The crystals were kept in dry n-hexane prior to measurement.

Synthesis of UiO-69-Me2 (3): To 20 mL of DMF, 19 μL (1.03 mmol) H2O, 120 mg (0.52 mmol) ZrCl4 and 1.89 g (15.5 mmol) benzoic acid was added and stirred until a clear solution was obtained, and subsequently heated to 120 ◦C. 218 mg (0.52 mmol) Me2-H2qpdc was then added, a clear solution was obtained, transferred to a clean conical flask and kept at 120 ◦C for 72 h. The solid crystalline product was isolated and washed.

#### *2.2. X-ray Crystallography*

The crystals of 1 and 3 were dried in air at 200 ◦C for 2 h prior to measurement to ensure that the pores were free of physiosorbed solvent and water. The crystals were mounted on MiTeGen polymer loops using a minimum amount of Paratone oil. Data collection for 1–3 was performed on a Bruker D8 Venture diffractometer equipped with a Photon 100 CMOS detector using Mo Kα radiation (λ = 0.71073 Å) at 100 K. 3 was also measured at beamline ID11 at the ESRF synchrotron (Grenoble, France), equipped with a Frelon2 detector, using a wavelength of λ = 0.31120 Å.

Due to the poor chemical stability of 2, it was kept in dry hexane after synthesis. Before measurement, several crystals were placed on a glass slide and the hexane was allowed to evaporate. The crystals were then mounted on MiTeGen polymer loops using a minimum amount of Paratone oil. 2 was also measured at beamline I911-3 at the MAX2 synchrotron (Lund, Sweden) [12].

All frames were integrated, and the reflection intensities were scaled and evaluated using the APEX3 suite from Bruker AXS, consisting of SAINT, SADABS and XPREP [13]. The structures were solved with XT [14] and refined with XL [15], using Olex2 as graphical user interface [16]. Space group determination of twinned MOFs was done using comprehensive tables by Howard Flack [17]. A summary of the crystal data, data collection and structure refinement details can be found in Table 1, and full structural information can be found in the Supplementary Materials.


**Table 1.** Summary of crystallographic data and refinement indicators for reported MOFs 1–3.

#### **3. Results and Discussion**

#### *3.1. UiO-67-Me2*

UiO-67-Me2, consisting of the Zr6 oxocluster and the Me2-bpdc linker, is isostructural to UiO-67, with nearly identical lattice parameters and PXRD pattern. The MOF readily forms large single crystals during synthesis, by the addition of 30 molar equivalents of benzoic acid (in respect to Zr) as modulator. This MOF has been found to have slightly better stability to water, and higher affinity to methane than UiO-67, presumably due to the steric shielding of cluster-adjacent hydrophilic sites by the methyl groups [18].

A significant number of the crystals that were screened had visibly distorted morphology (Figure 1). The crystals were seemingly intergrown, as if one crystal was sprouting out of the facets of its parent. In all of these crystals, the initial indexing resulted in a larger unit cell than expected, a primitive hexagonal cell, closely related to the reduced cell of UiO-67. (Reduced unit cell of UiO-67: a = b = c = 19.0 Å, α = β = γ = 60.0◦. Hexagonal cell: a = b = 19.0 Å, c = 47.0 Å, a = b = 90◦, c = 120◦).

Upon inspection of the reciprocal lattice, it was apparent that the crystals were twinned by reticular merohedry, the twin law being the so-called "spinel law", 2[111]. This twin law translates to a two-fold rotation about the [1 1 1] axis, coinciding with a three-fold rotation-reflection axis (S6 or 3) in the ideal structure. This is a common twinning mode in cubic close packed crystals, as it is associated with stacking error of the close packed layers.

The twin law is easily visualized by reciprocal lattice displays of the reflections of the hk0 layer (Figure 2). The cause for twinning or the exact structure at the interface of this MOF remains unknown, but there are other Zr-based oxoclusters that facilitate such a boundary [19].

**Figure 2.** Schematic display of the hk0 layer of the diffraction of 1, twinned by 2[111]. The lattices of the twin domains are shown in red and blue, respectively.

#### *3.2. Zr-stilbene Dicarboxylate*

Zr-stilbene dicarboxylate (2) has previously been reported in its disordered form (space group Fm-3 m), in which the crystal symmetry is the same as for the UiO MOFs [20,21]. These MOFs consist of the same Zr6 oxoclusters in an fcc arrangement, connected by linear ditopic linkers. However, the stilbene linker (being non-linear) is not compatible with this symmetry, implying that it is randomly disordered over two rotational conformers. The disorder could be dynamic, in which the linkers would rotate freely and continuously throughout the sample because of the low energy barrier of rotation. In this case, one would expect a threshold temperature at which the rotation is no longer feasible, and where the individual linkers adopt a static conformation. Such a static phase could exist with linkers assuming random conformers (preserving the overall Fm3 m disordered symmetry) or ordered conformations resulting in a symmetry change.

Repeated measurements at room temperature and after flash cooling samples to 100 K showed the randomly oriented linker structure (Fm3 m) previously reported [21]. To investigate whether a low-symmetry phase could be obtained, individual scans along [1 0 0] were acquired while slowly cooling the crystal by 5 K/min. The appearance of new diffraction peaks (forbidden in face centered cubic crystals) occurred around 160 K (Figure 3).

**Figure 3.** Oriented diffraction images of **2** along the [1 0 0] direction acquired at room temperature (**left**) and at 100 K (**right**) after slow cooling of the crystal (by 5 K/min).

Solving the structure using the data from the slowly cooled crystal revealed a structure in the primitive cubic space group Pn3 (shown in Figure 4a–c). In this structure, the conformation of the linkers are unambiguous; the space group does not include mirror planes intersecting the linkers as in Fm3 m. When the crystal is brought back to room temperature, the face centered phase is observed again (shown in Figure 4d–f). This reversible phase transition can be used to study the molecular mechanics of these nonlinear linkers, which have also been shown to facilitate linker exchange and lattice expansion in these MOFs [22].

**Figure 4.** Partial crystal structures of **2** in Pn3 (**a**–**c**) and Fm3 m (**d**–**f**) showing the ordered and disordered conformations of the linker.

Inspecting the initial solution of the Pn3 structure, it displayed several warning signs: Very high *R*-values and presence of significant noise in the Fourier difference maps, large negative peaks in particular. Looking at the crystal structure along [1 0 0], one could envision two different orientations of the linker "kink" being present in the crystal (Figure 5). In fact, are separate domains of the crystal with opposite oriented linkers related to each other by a two-fold rotation about the [1 1 0] face diagonal. Thus, the twin law to implement in the crystal structure refinement is the matrix (0 −1 0 −10000 −1). Applying this twin law to the refinement significantly improves the fit, and the twin fractions refine freely to 0.5.

#### *3.3. Interpenetrated UiO-69-Me2*

UiO-69-Me2 (3) was synthesized as single crystals using the linker 2 ,3-dimethyl-[1,1 :4 ,1:4, 1-quaterphenyl]-4,4-dicarboxylic acid (Me2-H2qpdc) [11]. Using 30 equivalents of benzoic acid in respect to ZrCl4 as modulator, single crystals of up to 1 mm could be obtained. The large octahedral single crystals featured a distinct opaque pattern originating from the center of each crystal and out to each face (Figure 6).

**Figure 5.** The twin law of 2 in the Pn3 space group.

**Figure 6.** Single crystals of 3, featuring an opaque pattern from the center and out towards each face.

High quality SC-XRD data was acquired using synchrotron radiation, and the subsequent structure solution revealed a structure composed of doubly interpenetrated fcu networks analogous to the PIZOF materials [23]. Based on analysis of reflection intensities, the structure appears to have the space group Fd3 m, as is observed for the other family of two-fold interpenetrated Zr-MOFs [24]. The structure could indeed be solved in this space group, but the refinement accuracy came out very poor (R1 > 13.5%).

When a structure appears centrosymmetric it could in reality be a twin of two equally large non-centrosymmetric domains [9,25]. Consequently, the structure was solved in the lower-symmetry space group F43m. The difference between Fd3m and F43m is one symmetry element, a two-fold rotation relating the closest neighboring clusters from the two MOF lattices (intersecting 1/8 1/8 1/8). When the symmetry element is present, the two lattices must be identical for the symmetry to be true. However, this MOF feature only partial interpenetration, so the secondary lattice has a lower occupancy coefficient than the main lattice (in this case 0.66). When this coefficient is not equal, the real symmetry must be the lower F43m.

The reduction in symmetry reveals a curious detail: Neighboring clusters from the separate lattices have a clear tendency to orient opposite μ3-O/OH functionality towards each other. The fully occupied lattice tend to orient the μ3-O towards the partially occupied neighboring cluster's OH. This preference would violate the two-fold rotational symmetry of Fd3m. There are two possible settings of the secondary lattice within the main, which requires a twin law (e.g., twinning by inversion) to resolve (Figure 7). This twinning feature contributes to the high apparent symmetry of the data, leading to the initial erroneous space group assignment.

**Figure 7.** The twinning in UiO-69-Me2 (3) can be visualized by displaying the two possible settings of the interpenetrating lattice.

#### **4. Conclusions**

The results reported herein clearly show that twinning should always be considered in the single-crystal structure determination of MOFs and other network structures, in particular in cases where unexpected features arise during structure solution and/or refinement. Details such as partial interpenetration or structural heterogeneity within the sample may easily be overlooked if twinning is not considered. It is likely that several of the many reported MOF structures that display poor refinement values or chemically unlikely features are in fact non-treated twins.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2624-8549/2/3/50/s1, Crystal structure data of compounds **1**–**3**. The crystallographic data for this paper (CCDC) can also be obtained free of charge via www.ccdc.cam.ac.uk/data\_request/cif, by emailing data\_request@ccdc.cam.ac.uk, or by contacting The Cambridge Crystallographic Data Centre, 12 Union Road, Cambridge CB2 1EZ, UK; Fax: +44-1223-336033.

**Author Contributions:** Conceptualization, S.Ø.-Ø.; methodology, S.Ø.-Ø.; validation, S.Ø.-Ø.; formal analysis, S.Ø.-Ø.; investigation, S.Ø.-Ø.; resources, S.Ø.-Ø.; data curation, S.Ø.-Ø.; writing—original draft preparation, S.Ø.-Ø.; writing—review and editing, S.Ø.-Ø.; visualization, S.Ø.-Ø.; supervision, K.P.L.; project administration, K.P.L.; funding acquisition, K.P.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the European Union's Horizon 2020 research and innovation programme under grant agreement number 685727 (ProDIA).

**Acknowledgments:** Knut T. Hylland is acknowledged for providing the organic linkers for the synthesis of compounds **1** and **3**.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Sample Availability:** Not available.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Assessment of Computational Tools for Predicting Supramolecular Synthons**

**Bhupinder Sandhu, Ann McLean, Abhijeet S. Sinha, John Desper and Christer B. Aakeröy \***

Department of Chemistry, Kansas State University, Manhattan, KS 66506, USA; bhupinder.sandhu@bms.com (B.S.); amclean@sco.edu (A.M.); sinha@ksu.edu (A.S.S.); desperj65@gmail.com (J.D.)

**\*** Correspondence: aakeroy@ksu.edu

**Abstract:** The ability to predict the most likely supramolecular synthons in a crystalline solid is a valuable starting point for subsequently predicting the full crystal structure of a molecule with multiple competing molecular recognition sites. Energy and informatics-based prediction models based on molecular electrostatic potentials (MEPs), hydrogen-bond energies (HBE), hydrogen-bond propensity (HBP), and hydrogen-bond coordination (HBC) were applied to the crystal structures of twelve pyrazole-based molecules. HBE, the most successful method, correctly predicted 100% of the experimentally observed primary intermolecular-interactions, followed by HBP (87.5%), and HBC = MEPs (62.5%). A further HBC analysis suggested a risk of synthon crossover and synthon polymorphism in molecules with multiple binding sites. These easy-to-use models (based on just 2-D chemical structure) can offer a valuable risk assessment of potential formulation challenges.

**Keywords:** hydrogen-bond propensity; hydrogen-bond coordination; supramolecular synthon; hydrogen-bond energies; Cambridge Structural Database; molecular electrostatic potential; pyrazoles

#### **1. Introduction**

A key question in crystal engineering is, given a molecular structure, can we predict its crystal structure [1]? At the core of the crystal structure of most organic molecular solids is the supramolecular synthon [2–5], a "structural unit within supermolecules which can be formed and/or assembled by known or conceivable synthetic operations involving intermolecular interactions", introduced by Desiraju in 1995 [2]. A reason for why this idea is so important to crystal engineering is the fact that detailed knowledge and control of intermolecular interactions is as vital to this field as is control of the covalent bond to molecular synthesis [6–18]. Furthermore, the synthon can serve as a valuable starting point for identifying the most likely ways in which molecules will aggregate [19,20]. This means that an important step towards predicting a crystal structure often involves finding the most likely synthons in molecules with competing molecular recognition sites, Scheme 1.

Many groups have proposed practical tools and avenues for a priori prediction of synthons in crystal structures of organic molecular solids. Some of these are based on electrostatics [21–23] and lattice energy calculations [24–26], with the focus purely on thermodynamic (enthalpic) factors [27]. For example, Hunter et al. converted maxima/minima on calculated molecular electrostatic potential surfaces into hydrogen-bond energies to determine synthon preference in multicomponent systems [28,29]. The preferred connectivity patterns of a molecule in the solid state in such cases is determined using Etter guidelines which states that the best hydrogen bond acceptor binds to the best donor [22]. The required ranking of best donor-acceptor pairs can be determined using molecular electrostatic potentials [21,23,30–42].

One way of accounting for kinetic factors that inevitably influence crystallization and assembly is to systematically analyze large swaths of structural information in the Cambridge Structural Database (CSD) [43]. Although an individual crystal structure does not provide information about kinetics of the seed formation and crystal growth, it is

51

**Citation:** Sandhu, B.; McLean, A.; Sinha, A.S.; Desper, J.; Aaker¨оy, C.B. Assessment of Computational Tools for Predicting Supramolecular Synthons. *Chemistry* **2021**, *3*, 612–629. https://doi.org/10.3390/ chemistry3020043

Academic Editors: Catherine Housecroft and Katharina M. Fromm

Received: 25 March 2021 Accepted: 25 April 2021 Published: 3 May 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

reasonable to assume that if a particular structural array is abundant in a database, it may reflect thermodynamic stability of a given crystal packing array, as well as a kinetic preference of its formation. Aakeröy, co-workers and others have used structural informatics tools such as hydrogen-bond propensity [44], and hydrogen-bond coordination [45] developed by the Cambridge Crystallographic Data Centre (CCDC) to understand the synthon predictions in the co-crystal formation [23,46–48]. These tools have considerable potential for validation purposes in both single component and multi-component systems. Yet, despite these efforts, there is still a need for more additional studies that systematically analyze and compare the abilities and reliability of different methods for predicting which synthon(s) is/are most likely to appear in crystal structures of high-value organic chemicals such as pharmaceuticals and agrochemicals.

**Scheme 1.** Schematics of using an in-silico approach for predicting synthons in a multifunctional molecule. Hydrogen-bond donor, **D**, and hydrogen-bond acceptor, **A**.

In order to address this issue, we have employed several known methods for synthon prediction in order to map out the structural landscape of a family of relatively flexible pyrazole containing molecules (**P1**–**P12**) capable of forming specific and competing intermolecular interactions, Scheme 2. We selected a pyrazole backbone because many compounds comprising this chemical functionality are known to possess a wide range of biological activities such as anti-microbial, anti-fungal, anti-tubercular, anti-inflammatory, anti-convulsant, anticancer, anti-viral, and so on [49–57]. The pyrazole-amide functionality is also present in some pharmaceutical related compounds such as Entrectinib, Graniseton, and Epirizol as well as antifungal compounds such as Furametpyr, Penthiopyrad and Tolfempyrad [58]. Due to the presence of multifunctional groups, these molecules are always at risk of synthon polymorphism. Therefore, knowledge gained from a successful use of tools such as molecular electrostatic potentials, hydrogen-bond energies, hydrogen-bond propensity and hydrogen-bond coordination, for predicting synthon appearances could have significant practical applications.

The target molecules, **P1**–**P12** can be divided into two groups: Group 1 (**P1**–**P8**) includes molecules with two hydrogen-bond (HB) donors (pyrazole NH and amide NH) and two HB acceptors (pyrazole N and C=O of amide), leading to four possible/likely synthons; A-D, Scheme 3. It can be assumed that each molecule forms a combination of two synthons to satisfy all hydrogen-bond donors and acceptors leading to two possible synthon combinations; (A + D) or (B + C)).

Group 2 (**P9**–**P12**) comprises molecules with two HB donors (pyrazole NH and amide NH) and three HB acceptors (pyrazole N, carbonyl C=O and pyridine N) groups. Two additional synthons can therefore be postulated in this group, E and F, leaving a total of six possibilities, Scheme 3.

The overall outline for testing different predictive methods against experimental data is summarized in Scheme 4. We used four protocols for determining the most likely synthon in the crystal structures of **P1**–**P12**; molecular electrostatic potentials (MEPs), hydrogenbond energies (HBE), hydrogen-bond propensities (HBP) and hydrogen-bond coordination (HBC). The prediction from each method was then compared to experimental data. The overall goal was to identify which approach is likely to deliver robust and transferable

guidelines for estimating/predicting which hydrogen bonds are most likely to appear in molecular solids when there are numerous potential avenues for assembly.

**Scheme 2.** Target molecules (P1–P12) employed in this study.

**Scheme 3.** Representations of six postulated synthons in a generic pyrazole-amide; A, N-H (pyrazole) ... N (pyrazole); B, N-H (pyrazole) ... C=O (amide); C, N-H (amide) ... N (pyrazole); D, N-H (amide) ... C=O (amide); E, NH (pyrazole) ... N (aromatic) and F, NH (amide) . . . N (aromatic). (Y = H, CH3; X = methyl, ethyl, or benzyl).

**Scheme 4.** The road map for synthon predictions.

The study was undertaken to address the following questions,


#### **2. Materials and Methods**

2-Amino-pyrazole, 2-amino-5-methyl-pyrazole, acetic anhydride, propionic anhydride and benzoyl chloride were purchased from Sigma Aldrich (Milwaukee, WI, USA) and utilized without further purification. Synthetic procedures and characterization of all molecules are provided in the Supporting Information (SI). Melting points were measured using Fisher-Johns melting point apparatus. 1H NMR data were collected on a Varian Unity plus 400 MHz spectrophotometer in DMSO.

Additionally, Cambridge Structural Database (Version 5.38 and Mercury 3.9) based structural informatics tools such as hydrogen-bond propensity (HBP), and hydrogen-bond coordination (HBC) were used [59–62] to predict the synthons in the crystal structures of **P1**–**P12**.

#### *2.1. Molecular Electrostatic Potentials (MEPs)*

Molecular electrostatic potential surfaces of **P1**–**P12** were generated with DFT B3LYP level of theory using 6-311++G\*\* basis set in vacuum. All calculations were carried out using Spartan'08 software (Wavefunction, Inc., Irvine, CA, USA) [63]. All molecules were geometry optimized with the maxima and minima in the electrostatic potential surface (0.002 e/au isosurface) determined using positive point charge in the vacuum as a probe. The numbers indicate the Coulombic interaction energy (kJ/mol) between the positive point probe and the surface of the molecule at that particular point.

#### *2.2. Hydrogen-Bond Energies (HBE)*

The hydrogen-bond energies (HBE) were calculated using molecular electrostatic potentials (MEPs) combined with Hunter's parameters [28,29]. The hydrogen-bond parameters, α (hydrogen-bond donor) and β (hydrogen-bond acceptor) were determined using maxima and minima on the MEPS, respectively (Equations (1) and (2)), and the free energy of interaction is given by the product, −α β [28]. Only conventional hydrogen-bonds donors (pyrazole N-H, O-H) and acceptors (O=C, pyridine, pyrazole N) were included. In the HBE approach, both single-point interactions and two-point interactions (dimeric synthons) were considered.

$$\alpha = 0.0000162 \, MEP \text{max}^2 + 0.00962 \, MEP\_{\text{max}} \tag{1}$$

$$\beta = 0.000146 \, MEPmin^2 - 0.00930 MEP\_{\text{min}} \tag{2}$$

$$E = -\sum\_{ij} \alpha\_i \beta\_j \tag{3}$$

#### *2.3. Hydrogen-Bond Propensity (HBP)*

The hydrogen-bond propensity (HBP) is the probability of formation of an interaction based on defined functional groups and fitting data [44,46]. Possible values fall in the range of zero to one, where a value closer to zero indicates less likely occurrence and a value closer to one indicates a higher likely occurrence of an interaction in the given molecule.

#### *2.4. Hydrogen-Bond Coordination (HBC)*

The hydrogen-bond coordination (HBC) is the probability of observing a coordination number for any given hydrogen-bond donor/acceptor atom [45]. The coordination number (CN) is defined as the number of intermolecular hydrogen-bonds formed between a donor and an acceptor. =0, =1, =2, =3 denotes the number of times a functional group donates or accepts. The CN with the highest probability corresponds to the most optimal (likely) hydrogen-bond interaction. The numbers that are colored relate to the outcome present in the selected H-bonding network, if this is green it indicates that the outcome is optimal, whereas if it's red that indicates the outcome is sub-optimal.

#### **3. Results**

#### *3.1. Molecular Electrostatics Potentials*

MEPs values for each hydrogen-bond donor (pyrazole NH and amide NH) and hydrogen-bond acceptor (pyrazole N, C=O (amide)) for **P1**–**P8** (and additional pyridine N for **P9**–**P12**) are presented in Figures 1 and 2, respectively. In group 1 (**P1**–**P8**), pyrazole NH is the best donor and C=O (amide) is the best acceptor (as ranked by electrostatic potentials) followed by amide NH and pyrazole N, therefore we can postulate that B + C is the most likely synthon. In group 2 (**P9**–**P12**), pyrazole NH is the best donor and pyridine N is the best acceptor followed by amide NH and C=O (amide), therefore D + E is the most likely synthon.

#### *3.2. Hydrogen-Bond Energies (HBE)*

Hydrogen-bond energies for each combination of synthons is presented in Table 1, See Supporting Information (SI) for additional details. Based on hydrogen-bond energies, in **P1**–**P8**, both A + D and B + C have very similar energies, and thus they can be expected to have equal chances of appearing. Based on a similar analysis, in **P9**–**P12,** A + F, A + D, and C + E, are equally possible.

**Figure 1.** Electrostatic potential values (in kJ/mol); (**a**) **P1,** (**b**) **P2,** (**c**) **P3,** (**d**) **P4,** (**e**) **P5,** (**f**) **P6,** (**g**) **P7** and (**h**) **P8**. Red indicates MEPs on hydrogen-bond acceptors and blue indicates MEPs on hydrogen-bond donors.

**Figure 2.** Electrostatic potential values (in kJ/mol); (**a**) **P9,** (**b**) **P10,** (**c**) **P11** and (**d**) **P12**. Red indicates MEPs on hydrogenbond acceptors and blue indicates MEPs on hydrogen-bond donors.

**Table 1.** Hydrogen-bond energies (in kJ/mol) for each combination synthon for molecules **P1**–**P12**. Green shade indicates the most likely predicted synthons by this method.


*3.3. Hydrogen-Bond Propensities (HBP)*

The propensities calculations consider all possible interactions between two donors (pyrazole NH and amide NH) and two acceptors (pyrazole N and C=O) resulting in four propensity numbers for **P1**–**P8**. In molecules with an additional acceptor, **P9**–**P12,** six different combinations of propensities are obtained. The propensities of individual (see Supporting Information (SI) for details) and combination synthon are presented in Table 2. Based on combination approach in HBP, the most likely synthons to appear in **P1**–**P8** are A + D and B + C and for **P9**–**P12,** synthons A + F and B + F.


**Table 2.** Hydrogen-bond propensities (probability of interaction between a hydrogen-bond acceptor and donor) for combination synthons in molecules **P1**–**P12**. Combination synthon propensities are calculated by multiplying the individual synthon propensities. Green shade indicates the most likely predicted synthons by this method.

#### *3.4. Hydrogen-Bond Coordination (HBC)*

HBC was calculated for each molecule, and the highest donor Coordination Number (CN) was matched with the highest acceptor CN in each molecule to determine the most likely synthon, Figure 3. Synthon (B + C) in **P1**–**P6**, (B + C) and (A + D) in **P7**–**P8,** (B + F) and (D + E) in P9–P12 were predicted to be the most likely synthon.

**Figure 3.** Predicted hydrogen-bond coordination **P1**–**P12** (D: hydrogen-bond donor, A: hydrogen-bond acceptor; green: optimal interactions, red: sub-optimal interactions).

#### *3.5. Experimentally Observed Crystal Structures*

Suitable crystals of **P1, P2, P3, P4, P7, P8, P10 and P11** were obtained by slow solvent evaporation method using methanol solvent. Few other solvents such as ethanol, THF, ethyl acetate etc. were tried to grow crystals for P5, P6, P9 and P12 with no positive hit. We obtained crystallographic data for eight (**P1, P2, P3, P4, P7, P8, P10** and **P11**) out of twelve molecules. Our re-determination of the structure of **P2** was consistent with the reported structure in the CSD (ARAGUV) [64] and the crystal structure of **P8** is also reported in the CSD (PESQEK) [65], Figures 4 and 5.

**Figure 4.** Crystal structures showing synthons (B + C) observed in **P1, P2, P3, P4, P7** and (A + D) observed in **P8**.

**Figure 5.** Crystal structures showing synthon (A + F) observed in **P10** and (C + E) in **P11**.

In group 1, six out of eight crystal structures (**P1**–**P4**, **P7** and **P8**) were obtained. In five of the six, B + C was observed (**P1**–**P4** and **P7**) and synthon (A + D) was found in **P8**. In group 2, when an extra acceptor group was added to the benzyl group, two crystal structures were obtained (**P10** and **P11**)**.** Synthon (A + F) is observed in **P10** and (C + E) was observed in **P11**.

#### **4. Discussion**

#### *4.1. Molecular Conformational Analysis*

Molecular conformational analysis using DFT B3LYP (6-311++G\*\* basis set in vacuum) shows that **P1**–**P12** with amide functionality occur as trans instead of cis isomer as the stable conformation, which was further confirmed based on a CSD search (Scheme 5, See Supporting Information (SI) for details). When a pyrazole group is added to the amide functionality, the most stable conformation is when pyrazole functional groups are cis to the amide NH group. When meta-substituted pyridine is added to this group, the more stable conformation is when pyridine N is trans to the amide NH group. However, the second conformation is only ~4 kJ/mol higher in energy and therefore is observed in **P10**. It is worth noting that the energy optimized conformations are not necessarily completely identical to those that may appear in the solid state, where a variety of close

contacts and packing forces may distort some geometric parameters away from idealized gas phase values. However, these idealized conformations are likely to be most relevant in the solution phase at the point when target molecules begin to recognize and bind to each other during nucleation and growth (Scheme 5, See Supporting Information (SI) for additional details).

**Scheme 5.** Molecular conformation analysis of **P1**–**P12** using DFT B3LYP (6-311++G\*\* basis set in vacuum). Green highlights the more stable conformation. Green highlights the likely conformation whereas red highlight the less likely conformation. ƻa indicates acyclic moiety.

#### *4.2. Prediction Analysis*

Four different methods (MEPs, HBE, HBP, and HBC) were used to predict synthons in group 1 and group 2 molecules.

In group 1 (**P1**–**P8**), Synthon (B + C) was predicted by all four methods, however, synthon (A + D) was predicted by HBE and HBP method. In group 2 (**P9**–**P12**), A total of six synthons (A, B, C, D, E and F) and six different combination possibilities (A + F, A + D, C + E, B + C, D + E and B + F), made it complex to predict such synthons. In group 2, (D + E) was predicted by MEPs and HBC. (B + F) was predicted by HBP and HBC. Synthon (A + F) was predicted by both HBE and HBP. (C + E) and (A + D) were predicted by HBE only (Figure 6).

**Figure 6.** Summary of synthons predicted in **P1**–**P12**.

*4.3. CSD Search-Molecular Geometric Complementarity*

The molecular complementarity approach was used to determine whether synthon A or C is more likely to occur in pyrazole molecules with more than one binding site. The fragments used for CSD and resulting bond angles are listed in Figure 7. Based on this limited dataset, NH(amide) ... .N(pyrazole) interaction is more linear compared to NH(pyrazole....N(pyrazole) indicating that synthon C has a geometric preference over synthon A. Moreover, a search performed with similar pyrazole binding pockets in the CSD gave 12 hits and in every case, synthon C was preferred over A.

**Figure 7.** Pyrazole fragments used to perform CSD molecular complementarity search (**a**) Average bond angle for A (22 structures in CSD); (**b**) average bond angle for B (35 structures in CSD) and (**c**) pyrazole fragment for CSD search.

#### *4.4. Validation Analysis*

The comparison between the predictions from MEPs, HBE, HBP and HBC, and experimentally observed results is given in Table 3. In **P1**–**P8** with two donors and two acceptor sites, B + C was predicted to be the most likely synthon by all four methods which was also observed experimentally in five of the six compounds, **P1**–**P4**, and **P7**. In **P8**, A + D was present which was predicted by HBE and HBP as a possible synthon. By increasing the number of acceptor choices to three in **P9**–**P12** the challenge of getting the correct answer increased due to the enhanced possibility of synthon crossover and synthon polymorphism.

In **P10**, A + F was observed but it was only predicted correctly by HBE and HBP. In **P11**, C + E was observed experimentally, and this was only predicted correctly by HBE.


**Table 3.** Experimental vs. predicted synthons in the crystal structures of **P1**–**P14, P7**–**P8** and **P10-P11.**

#### *4.5. Polymorph Assessment of Experimental Structures*

A hydrogen-bond likelihood analysis was performed to understand the risk of synthon polymorphism in these molecules with multiple binding sites. Experimental structure was imported into the predicted hydrogen-bond coordination table to compare it against the predicted structures. Hypothetical structures in this tool are generated based on the combination of HBP and HBC parameters for each molecule. A correlation of the HB propensity vs. the mean HB coordination for all putative synthons possible in the structure of a given molecule is plotted. The most likely synthons should be found in the lower right corner of the plot. In **P1** (which is also a representative of **P2**, **P3**, **P4**, and **P7**), the experimental structure matched with the most likely synthon predicted by combined HBP an HBC. In **P8, P10** and **P11**, the experimental structures do not contain the most likely synthon predicted by a combination of HBP and HBC, highlighting the possibility of other structures with more optimal hydrogen-bond patterns, which indicates a reasonable risk of polymorphism in these three compounds. In **P8**, B + C was predicted by all four methods to be the most likely hydrogen bond, yet A + D was observed experimentally. Interestingly, this structure leads to a polymeric structure as reported by Daidone et al. [65], Figure 8.

In **P10**, multiple synthons were predicted by different methods, such as D + E, A + F, A + D, C + E and B + F. However A + F is in fact observed experimentally (and accurately predicted by HBE and HBP methods). These results indicate that even though A + F is geometrically constrained, as seen in CSD search, it can still form experimentally. Due to steric hindrance with meta-substituted pyridine, synthon A is observed as singlepoint interaction instead of dimeric synthon, resulting in a herringbone type arrangement, Figure 9.

In **P11**, experimentally observed synthon C + E was predicted correctly only by the HBE method, in addition, it was not deemed to be the most likely interaction by the hydrogen-bond likelihood analysis. This is less surprising as synthon C is the most likely synthon among all possibilities due to its linearity and dimeric chelate effect, leaving E as an alternate option because the pyridine nitrogen atom is a better acceptor than the C=O moiety, Figure 10.

**Figure 8.** Hydrogen-bonding observed in the crystal structure of (**a**) **P7** and (**b**) **P8** along the crystallographic *a*-axis.

**Figure 9.** Synthon and packing observed in the crystal structure of **P10** leading to herringbone-like structure.

**Figure 10.** Synthon and packing observed in the crystal structure of **P11** leading to 1-D chains along the crystallographic *a*-axis.

Additionally, molecules for which crystal structure was known were screened through hydrogen-bond coordination likelihood analysis to understand the risk of synthon polymorphism. This tool suggested that three out of eight molecules (**P8**, **P10**, **P11**), where crystal structure is known, have a chance of synthon crossover and synthon polymorphism,

Figure 11. This tool is particularly useful as it can guide chemists to narrow down list of APIs that are at risk of delivering new structural polymorphs, which can cause havoc in late-stage formulation efforts.

**Figure 11.** Polymorph assessment of **P1, P8, P10** and **P11** molecules showing the experimental structure and more stable hypothetical structure.

This work also highlights that small changes to a molecule can lead to profound changes in the hydrogen bonding, resulting in different packing arrangements in the crystal structure, Figure 12. For example, adding an electron donating methyl group to the pyrazole ring in **P7** changes the synthon from (B + C) in **P7** to (A + D) in **P8**. Comparing the electrostatic potential of **P7** and **P8**, adding a methyl group decreases the charges on both H-bond donors (amide NH and pyrazole NH) whereas increase the charges on the H-bond acceptor (pyrazole N), highlighting these two molecules as a classic example of synthon crossover. The addition of pyridine (para or meta) group changes the ranking of H-bond acceptors in **P10** and **P11**. Therefore, pyrazole NH and amide NH binds to pyridine N in **P11** and **P10** instead of C=O (as was observed in **P7** and **P8**) respectively. The position of a substituent (meta vs. para) as well as addition of methyl group on the pyrazole ring also leads to different crystal packing. For example, changing the pyridine N from meta to para position and adding methyl group on the pyrazole ring changes the synthon from (A + F) in **P10** to (C + E) in **P11**, resulting in varying packing arrangements.

**Figure 12.** Comparison of subtle variations on the molecular fragments leading to different hydrogen bonding interactions in crystal structures of **P7**, **P8**, **P10** and **P11**.

Another key observation is the ability of hydrogen-bond propensity model to incorporate effect of aromaticity on the probability of hydrogen–bond interaction. For example, presence of aromatic groups such as phenyl and pyridine rings in molecules **P8**–**P12** leads to higher values of donor and acceptor aromaticity (0.56) in the logistic regression model compared to molecules **P1**–**P6** (0.27). This further impacts the final probability values where **P1**–**P6** has higher HBP values for synthon B and D compared to **P7** and **P8**, see Table 2 and Table S5 in Supporting Information for more details.

This study didn't include extensive crystallization experiments to grow suitable crystals for **P5**, **P6**, **P9** and **P12** but it is possible that with further crystallization work, crystals of these molecules can be achieved. One of the reasons behind why **P5** and **P6** might be difficult to crystallize compared to **P1**–**P4** is due to the presence of butyramide side chain which increases the number of rotatable bonds and hence number of conformations in the solution state. The possibility of multiple conformations very close in energy to each other can cause difficulties in crystallizing molecules by hindering the nucleation and crystal growth phase. Additionally, further crystallization work can also be done for molecules with risk of synthon polymorphism such as **P8**, **P10** and **P11** to validate the computational studies.

In this study, we examined two energy based, and two informatics-based protocols for predicting hydrogen—bond based synthons in crystal structures of a family of pyrazoles. The overall outcome gave the following ranking in terms pf predictive quality: HBE (100%) > HBP (87.5%) > HBC = MEPs (62.5%), Figure 13. Even though HBE is a step

forward of MEPs method, it worked better than the latter in pyrazole based molecules because of inclusion of energies of dimeric interactions due to a chelating effect. It is important to note that even though pyrazole molecules studied in this work with high predictive success rate have with very limited structural diversity (<250 gmol−<sup>1</sup> molecular weight, <3 rotatable bonds, limited functional groups), such models can be applicable to diverse and more flexible drug-like molecules with similar functional groups for prediction analysis. We encourage scientific community to test these CCDC provided tools on actual drug molecules to further understand their applicability and range of predictive ability of such tools.

**Figure 13.** Validation results of supermolecular synthons observed in the pyrazole analogues using four different prediction models.

#### **5. Conclusions**

One useful consequence of predicting a correct supramolecular synthon is the ability to predict the right crystal structure. Four prediction models based on molecular electrostatic potentials (MEPs), hydrogen-bond energies (HBE), hydrogen-bond propensity (HBP) and hydrogen-bond coordination (HBC) were studied for their ability to predict synthons in small pyrazole based targets, **P1**–**P12**. Molecules were grouped into categories based on number of conventional hydrogen-bond donors and acceptors, group 1 (**P1**–**P8**) and group 2 (**P9**–**P12**). In group 1, both HBE and HBP gave 100% success rate as predicted synthon matched with the experimental results for molecules **P1**–**P4** and **P7**. However, when a strong acceptor group such as pyridine nitrogen was added as in **P9**–**P12**, synthon prediction became complex, so did the experimentally observed synthons. Only two crystal structures (**P10** and **P11**) were obtained experimentally. HBE predicted the synthons correctly for both molecules whereas HBP predicted it correctly for **P10**. Additionally, hydrogen-bond coordination likelihood analysis suggested that **P8**, **P10** and **P11** are at risk of synthon crossover and synthon polymorphism based on the net putative interaction likelihoods. Methodologies used in this study are a valuable tool to determine which synthon is likely to form in the crystal structure of a molecule and if the molecule is at a risk of synthon polymorphism. Therefore, a simple health check on these molecules using structural informatics tools such as MEPs, HBE, HBP and HBC for mapping out the structural landscape of these types of molecules will have significant practical applications in various fields.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/10 .3390/chemistry3020043/s1, Figure S1. (a) *Cis* and *trans* amide functionality (both bonds are acyclic representing using symbol @) used to perform the torsion angle search. (b) Pie chart indicating number of structures with torsions for *cis* (yellow, ~32 structures, 0.5%) and *trans* (red, ~6303 Structures, 99.5%) conformations; Figure S2. (a)Propensity-coordination chart of **P1**-**P6** molecules and (b) coordination of each functional group in all predicted motifs; Figure S3. Propensity-coordination chart of **P7**-**P8** molecules and the coordination of each functional group in all predicted motifs; Figure S4. Propensity-coordination chart of **P9**-**P10** molecules and the coordination of each functional group in all predicted motifs; Figure S5. Propensity-coordination chart of **P11-P12** molecules and the coordination of each functional group in all predicted motifs; Figure S6. 1H NMR of 3-acetamido-1H-pyrazole, **P1**; Figure S7. 1H NMR of 3-acetamido-5methyl-1H-pyrazole, **P2**.; Figure S8. 1H NMR of 3-propamido-1H-pyrazole, **P3**; Figure S9**.** 1H NMR of 3-propamido-5methyl-1H-pyrazole, **P4**; Figure S10. 1H NMR spectrum of 3-butyramido pyrazole, **P5**; Figure S11. NMR spectrum of 3-butyramido 5-methyl pyrazole, **P6**; Figure S12. 1H NMR of 3-benzamido-1H-pyrazole, **P7**; Figure S13. 1H NMR of 3-benzamido-5methyl-1H-pyrazole, **P8**; Figure S14. 1H NMR of N-(pyrazole-2-yl)nicotinamide, **P9**; Figure S15. NMR spectrum of N-(5-methylpyrazol-2-yl)nicotinamide, **P10**; Figure S16. NMR spectrum of N-(pyrazole-2-yl)isonicotinamide, **P11**; Figure S17. NMR spectrum of N-(5-methylpyrazol-2-yl)isonicotinamide, **P12**; Table S1. Energies of each *trans* amide conformation relative to most stable *trans* conformation is shown below in kJ/mol. The conformations with duplicate energies were ignored. Note: methyl-based target molecule conformations are not shown here; Table S2. Hydrogen-bond energies (in kJ/mol) for each individual synthon for molecules **P1**-**P12**. Synthon A and C are dimeric synthons; therefore, energies are presented for pairs of molecules; Table S3. Hydrogen-bond energies (in kJ/mol) for each combination synthon for molecules **P1**-**P12**; Table S4. Functional groups used to determine the hydrogen-bond propensities for the **P1-P12** target molecules. The labels in the figures can be explained as follows: Tn = atom makes n bonds, c = atom is cyclic, = bond is acyclic, and Hn = n bonded hydrogen atoms; Table S5. Hydrogen-bond propensities for each individual synthon possible in molecules **P1-P12**; Table S6. Hydrogen-bond propensities for combination synthons possible in molecules **P1-P12**. Combination synthon propensities are calculated by multiplying the individual synthon propensities; Table S7. Experimental details of crystals obtained in this study.

**Author Contributions:** B.S. and C.B.A. conceived and designed the experiments; B.S. and A.M. performed the experiments; A.S.S. and J.D. performed the single-crystal X-ray crystallography; B.S. and C.B.A. analyzed the data and wrote the paper. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Johnson Cancer Research Center, Kansas State University.

**Data Availability Statement:** Details of chemical synthesis, calculations, and NMR and IR spectra are available online. The crystallographic data for this paper (CCDC 2072631-2072636) can be obtained free of charge via www.ccdc.cam.ac.uk/data\_request/cif, or by emailing data\_request@ccdc.cam.ac.uk, or by contacting The Cambridge Crystallographic Data Centre, 12 Union Road, Cambridge CB2 1EZ, UK; Fax: +44-1223-336033.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


## *Review* **The Crystal Chemistry of Inorganic Hydroborates** †

#### **Radovan Cern ˇ ý \*, Matteo Brighi and Fabrizio Murgia**

Laboratory of Crystallography, DQMP, University of Geneva, 24 quai E. Ansermet, CH-1211 Geneva, Switzerland; Matteo.Brighi@unige.ch (M.B.); Fabrizio.Murgia@unige.ch (F.M.)

**\*** Correspondence: Radovan.Cerny@unige.ch

† Dedicated to Dr. Howard Flack (1943–2017).

Academic Editor: Catherine Housecroft Received: 28 August 2020; Accepted: 23 September 2020; Published: 29 September 2020

**Abstract:** The crystal structures of inorganic hydroborates (salts and coordination compounds with anions containing hydrogen bonded to boron) except for the simplest anion, borohydride BH4 −, are analyzed regarding their structural prototypes found in the inorganic databases such as Pearson's Crystal Data [Villars and Cenzual (2015), Pearson's Crystal Data. Crystal Structure Database for Inorganic Compounds, Release 2019/2020, ASM International, Materials Park, Ohio, USA]. Only the compounds with hydroborate as the only type of anion are reviewed, although including compounds gathering more than one different hydroborate (mixed anion). Carbaborane anions and partly halogenated hydroborates are included. Hydroborates containing anions other than hydroborate or neutral molecules such as NH3 are not discussed. The coordination polyhedra around the cations, including complex cations, and the hydroborate anions are determined and constitute the basis of the structural systematics underlying hydroborates chemistry in various variants of anionic packing. The latter is determined from anion–anion coordination with the help of topology analysis using the program TOPOS [Blatov (2006), IUCr CompComm. Newsl. 7, 4–38]. The Pauling rules for ionic crystals apply only to smaller cations with the observed coordination number within 2–4. For bigger cations, the predictive power of the first Pauling rule is very poor. All non-molecular hydroborate crystal structures can be derived by simple deformation of the close-packed anionic lattices, i.e., cubic close packing (*ccp*) and hexagonal close packing (*hcp*), or body-centered cubic (*bcc*), by filling tetrahedral or octahedral sites. This review on the crystal chemistry of hydroborates is a contribution that should serve as a roadmap for materials engineers to design new materials, synthetic chemists in their search for promising compounds to be prepared, and materials scientists in understanding the properties of novel materials.

**Keywords:** hydroborate; anions packing; crystal structure

#### **1. Introduction**

Hydroborates are anions containing hydrogen bonded to boron. They are also sometimes referred to as "boranes"; this term is, however, used for neutral molecules BxHy according to IUPAC [1]. Inorganic hydroborates are salts or coordination compounds where one of the ligands is the hydroborate. The bonding in boron clusters of boranes and hydroborates was explained by Lipscomb [2] using the concept of a three-electron-two-center bond. The concept was further developed in the Polyhedral Skeletal Electron Pair theory (PSEPT) also known as Wade–Mingos rules based on a molecular orbital treatment of the bonding [3,4]. The naming of the boron and hydroborate clusters, used in this review, follows the Wade-Mingos rules: (i) a number of vertices in the polyhedral boron cluster, i.e., dodecaborate for B12H12<sup>2</sup><sup>−</sup> anion, and (ii) the name that describes the topology of the polyhedral cluster, i.e., *closo*-, *nido*- and *arachno*-borates for all polyhedron vertices occupied by boron, and one or two boron atoms missing, respectively.

Inorganic hydroborates were studied as fuels for military applications [5], reducing agents in organic syntheses [6], weakly coordinating anions in catalysis [7,8], for the delivery of 10B for Boron Neutron Capture Therapy (BNCT) [9], as nanocarriers for the delivery of various chemotherapy drugs [10], magnetic resonance imaging (MRI) agents [11], liquid electrolytes [12], and more recently as solid ionic conductors [13,14]. It is exactly the increase of interest in hydroborates as solid-state electrolytes for application in all-solid-state batteries that motivated us to assemble the structural characteristic of known compounds, analyze the coordination of cations and anions, study the type of anion packing, and find the structural aristotypes. The structural systematics discussed in this review attempt to provide a guide to functional inorganic hydroborates design. The hydroborates containing the simplest anion, borohydride BH4 −, were recently reviewed by us [15] and are not included herein. We will limit our review only to the compounds containing the hydroborate as the only type of anion, although including compounds gathering more than one different hydroborate (mixed anion). We will include in our review also the carbaborane anions where one or more boron atoms are replaced by carbon and partly halogenated hydroborates, i.e., where there is the partial replacement of hydrogen by a halogen. Hydroborates containing anions other than hydroborate or neutral molecules such as NH3 will not be discussed. Only the compounds in which the crystal structure has been experimentally fully characterized are included; the structures predicted by ab initio calculations or any other method of prediction, but not experimentally confirmed, are excluded. According to our analysis, the crystal chemistry of inorganic large hydroborates is much simpler compared to that of containing small BH4 − [15] reducing the anion arrangement, for most of the presented compounds, to the packing of hard spheres. We will also show that the first Pauling rule on preferred cation coordination in these (mostly) ionic compounds is limited to small cations, with the coordination number within 2–4 [16].

#### **2. Anions Packing**

The hydroborates that form the compounds mostly used in various applications are *closo*-dodecaborates B12H12<sup>2</sup><sup>−</sup> and *closo*-decaborates B10H10<sup>2</sup><sup>−</sup> (Figure 1). While the boron cluster in the first has the form of an ideal icosahedron with the symmetry Ih, the latter is the gyroelongated square bipyramid (bi-capped square antiprism) with the symmetry D4d. The non-crystallographic symmetry of both clusters is at the origin of their orientation disorder in the *ht* phases of many hydroborates.

**Figure 1.** The two frequently used hydroborate anions: *closo*-dodecaborate (**left**) and *closo*-decaborate (**right**) and their point group symmetry.

The high symmetry of the two anions is also responsible for the building principle of corresponding crystal structures based on the packing of hard spheres. Inorganic hydroborates usually contain three types of anion packing: two close packings, cubic close packing (*ccp*) and hexagonal close packing (*hcp*), and one less dense packing in a body-centered cubic cell (*bcc*). However, the poly-anion nature of hydroborates complicates the detections of the packing type, which usually deviates from the ideal packing. For instance, the hexagonal layers of *ccp* or *hcp* are not always parallel to simple crystallographic planes. In some works, the anion packing was analyzed with the algorithms developed for molecular

dynamics (MD) simulation [17], which gives the frequency with which each basic packing maps locally the analyzed structure. This is justified when studying with MD the temporal evolution of some microscopic aspect (such as the distribution of energy barriers for diffusion), and necessarily the atomic positions are given as the atoms' distribution, hence anions packing distribution. However, it makes no sense to use such an algorithm for the analysis of a periodic structure, which gives the averaged description of the structure and where the atoms in the unit cell are located on Wyckoff sites even when the structure is disordered. Instead, the easiest method is to analyze the shape of the first and second coordination sphere of anion–anion interaction, i.e., considering the poly-anions centers: in both close packings, the coordination number of the first sphere is 12, and the polyhedron shape is cuboctahedron for *ccp* (Figure 2a) and anti-cuboctahedron for *hcp* (Figure 2b). In *bcc* packing, the first coordination sphere has a cubic shape, since there are eight nearest neighbors, while the second coordination sphere contains six anions forming an octahedron. Together, the first and second coordination spheres form a rhombic dodecahedron in *bcc* (Figure 2c). We will allocate a type of packing to a given hydroborate, based on the anion–anion coordination polyhedron (which may be deformed) and on the number, type, and connectivity of the interstitial sites, which must not change with respect to the ideal packing.

**Figure 2.** Anion–anion coordination for (**a**) monoclinic *rt*-Na2B12H12 with cubic close packing (*ccp*) of anions. The first coordination sphere contains 12 nearest neighbors at the distance 7.01–7.42 Å forming a cuboctahedron; (**b**) monoclinic *rt*-K2B10H10 with hexagonal close packing (*hcp*) of anions. The first coordination sphere forms an anti-cuboctahedron with a bond distance 6.67–7.68 Å; (**c**) orthorhombic *rt*-NaCB9H10 with body-centered cubic (*bcc*) packing of anions. The first coordination shell is shown on the left, while the second shell is shown on the right (cube and octahedron, respectively). (**d**) Wrong determination of coordination polyhedron for the *rt*-Na2B12H12 leading to the false determination of a *bcc* sublattice. The apical atoms in the octahedron are indeed at the distance of 9.51 Å, while all the others fall in the range 7.0–7.4 Å.

#### **3. Controlling Anions Packing**

The type of anion packing is an important parameter controlling hydroborates properties such as cation mobility [18]. While the packing type for monoatomic structures such as metals is simply driven by space-filling efficiency, it becomes more complex for poly-atomic structures such as ionic compounds. The ability of smaller cations to fill available interstitial sites in the packing of larger anions may control the packing type of the latter [16]. In the case of metal hydroborates, the polyanions are only approximately spherical, and their orientation is controlled by directional metal–hydrogen bonding. It means that in addition to the space-filling and cation/anion size ratio, the cation–anion directional bonding is a third parameter controlling the anion packing [19] at a given temperature and pressure.

Cation distribution on interstitial sites and cation–anion directional bonding are at the basis of the anion packing design and engineering: (1) cation mixing where the larger cation fills the interstitial sites left empty by the smaller (mobile) cation [20,21]; (2) anion mixing, which perturbates the directional cation–anion bonding [22,23], and with the same effect, (3) anion modification such as partial halogenation [21]. In all cases, the volume change and its effect on the total crystal energy cannot be excluded [21].

Anions packing varies also when a neutral molecule such as NH3 is introduced into the structure of the hydroborate salt [24]. In that case, when the molecule coordinates to the cation, the mobility of the latter is strongly decreased [25].

From the analysis of the anion packing presented here below, it seems that increasing anion anisotropy, i.e., partial halogenation of B12H12, use of elongated B10H10, or by *carba*-borate and *nido*-borate, increases the probability of having an *hcp* lattice.

#### **4. Preferred Cation Sites**

The first Pauling rule predicts the preferred cation coordination in iono-covalent compounds from the ratio of cation and anion radii [16]. We have analyzed all single metal hydroborates presented in this review according to the first Puling rule. The cation radii are according to Shannon [26]. We have used an ionic radius for a given oxidation state contrary to Pauling's original work, which is based on univalent radii. As only very few estimations of hydroborate radii are available in the literature (for BH4 see [15]), we have used the radii from [27], where the diameter of the anion was calculated as the maximal distance between two terminal hydrogens. Clearly, these radii correspond to the lower limit of the estimation. The calculated cation/anion radius ratios are plotted in Supplementary Materials Figure S1 together with regions corresponding to different cation coordination polyhedra. The agreement of the observed coordination (color of the data point) with the predicted region is relatively good for smaller cations with the observed coordination number within 2–4. For larger cations, the predictive power of the first Pauling rule is very poor, which is in agreement with the recent analysis of Pauling rules validity for oxides [28]. For example, we observe the tetragonal coordination in the region where the octahedral coordination should exist. This discrepancy would be removed if larger anion radii were used, such as estimated from solid crystal structures: 3.18 Å for B10H10 [29] or 3.28 Å for B12H12 [30] (3.46–3.5 Å from hard spheres approximation) or even larger from the calculation of electrostatic potential surface maps [31]. With larger estimation of the anion size, the predicted region of existence would be shifted to smaller coordination numbers.

#### **5. Anions and Cations Dynamics**

The complex anions in the hydroborates can participate in the fast reorientation (rotational) motion. This important dynamical feature contributes to the entropic part of the free energy balance determining thermodynamic stability. Therefore, the information on the anion reorientation dynamics is important for understanding the fundamental properties of hydroborate salts such as the thermal stability and symmetry of the crystal structure. When rotating, the anions approach the spherical symmetry, averaging off the local hydrogen–cation interactions and following more precisely the ideal packing of hard spheres, which leads to stabilizing the higher crystal symmetries in *ht* phases. In addition to the localized reorientation motion of the anions, the long-range motion of the cations is observed in the compounds with high cation conductivity. These two types of atomic motion may be related, i.e., the fast cations diffusion is accompanied by the fast reorientation motion of the

anions [32], as was also shown by ab initio molecular dynamics calculations [33]. The anion and cation dynamics in hydroborates are studied by nuclear magnetic resonance (NMR) [34] and quasi-elastic neutron scattering (QENS) [35]. For more details on the experimental and theoretical studies of the ions dynamics in hydroborates, we refer to the recent review [34].

The ions dynamic is practically invisible in the diffraction experiment using Bragg scattering, which corresponds to the space and time average of atomic positions. The anions reorientation motion and cations long-range diffusion can be only indirectly judged from the observed disorder in atomic positions. If both dynamic and static disorders are present, they may be separated by a temperature-dependent diffraction. The static disorder, if present, persists down to low temperatures where the dynamic one disappears. Some details of the local anion–anion correlation can be obtained by modeling the total scattering (Bragg and diffuse scattering), for example using the Pair Distribution Function analysis. Such experiments have not yet been done on inorganic hydroborates except for salts with a BH4 − anion [36,37]. The disorder of atomic positions observed from Bragg scattering needs a correct understanding of the superposition of anion point symmetry (Figure 1) with the point symmetry of Wyckoff positions in the crystal. The high non-crystallographic symmetry of the hydroborate anion can fit only partly to a Wyckoff position, and the anion needs to be oriented in such a way that the symmetry elements of a given site match the corresponding symmetry elements of the anions point group. Only in this case do the atomic positions of boron and hydrogen appear fully ordered. If in such ordered structure the anions reorientation happens between orientations that are equivalent by the site symmetry, Bragg scattering cannot detect it. However, if the reorientation occurs between non-equivalent orientations, the structure appears disordered in atomic positions. If this is the case, the Bragg scattering cannot still say whether the disorder is dynamic (i.e., the reorientation of the anion) or whether it is of static nature (i.e., a long-range disorder of anions orientation) [38].

We can demonstrate the interaction between the anions point symmetry and Wyckoff site symmetry on two examples: Cs2B12H12 crystallizes at *rt* in an ordered anti-CaF2 structure with the crystal symmetry reduced from *Fm*-3*m* to *Fm*-3 due to the icosahedral symmetry of the *closo*-anion localized in one orientation (light gray, Figure 3 left). At 529 K, it undergoes a second-order transition of order–disorder type disordering the *closo*-anion by {110} mirror planes (absent in the point symmetry of the anion) increasing the crystal symmetry to *Fm-*3*m*. Then, the observed disordered anion is a superposition of the anion in two orientations (light and dark gray, Figure 3 right) related by the {110} mirror plane or by the rotation around the anions axis C3 by 45◦ (which is equivalent), as shown in Figure 3. This disorder is static as shown by QENS studies where different reorientation jumps about two molecular axes, C3 by 120◦ and C5 by 72◦, were suggested [39]. Both jumps exist in *lt* ordered and *ht* disordered phases. They operate independently at a lower temperature, but combine to two-axial jumps at higher temperatures, as shown by 1H and 11B NMR [40].

Li2B12H12 crystallizes at *rt* in an ordered *anti*-CaF2 structure with its symmetry reduced from *Fm*-3*m* to *Pa*-3 resulting in two *closo*-anions in the unit cell in different orientations but related by the *a*-glide plane (light and dark gray, Figure 3 left). Li2B12H12 undergoes at 628 K a first-order phase transition into the *s.g. Fm*-3*m* where the disordered anion may be explained as a superposition of the two orientations from the *rt*-phase. As the reorientation dynamics in Li2B12H12 has not yet been studied, we cannot conclude whether the disorder in the *ht*-Li2B12H12 is of dynamic or static nature. The phase transition in Li2B12H12 may be compared with that in C60, which crystallizes in an ordered *Pa*-3 structure below 250 K and transforms with first-order order–disorder phase transition connected with the rotation of C60 molecules into an *Fm*-3*m* structure (see Figure S2) [41].

The understanding of cations disorder is easier as there is no interaction between the cations symmetry (single atom) and the symmetry of a Wyckoff site. The observed disorder is always a position disorder, but its nature (dynamic or static) cannot be concluded from Bragg intensities. For example, it is easily identified by measuring the electrical conductivity based on the given cation. As local correlations between moving cations are suggested by molecular dynamic studies [33], the total scattering studies on ionic conductors are of high interest.

*rt*-Li2B12H12 and *rt*-Cs2B12H12 *ht*-Li2B12H12 and *ht*-Cs2B12H12

**Figure 3.** Two orientations, dark and light gray, of the *closo*-anion in the *rt* ordered (s.g. *Pa*-3) and in the *ht* disordered Li2B12H12 (s.g. *Fm*-3) viewed along the C3 rotation axis of the *closo*-anion. The *rt*-Cs2B12H12 (s.g. *Fm*-3) contains one anion in the orientation identical to the light gray anion in *rt*-Li2B12H12, while the *ht*-Cs2B12H12 (s.g. *Fm*-3*m*) has an identical orientation disorder (rotation around C3 by 45◦) as *ht*-Li2B12H12.

#### **6. Anions Chemistry**

According to the aforementioned PSEPT, hydroborates can be thought of as formally made from simple electron donors, i.e., B-H groups, but also C-H in the case of carbaboranes, providing electron density to build up a cluster. The main building block of hydroborates, the B-H unit, accounts for four valence electrons. Excluding the electron pair involved in the boron–hydrogen bond, the remaining two electrons are available to link the boron atoms in non-classical connectivity, where on average two electrons are shared between three boron atoms (three-centers two-electrons). This multicenter interaction is typical of boron-based compounds, due to the intrinsic boron electron deficiency and relatively low electronegativity. As a direct consequence, the possibility of arranging several B-H groups is limited (with fewer exceptions in some open-cage hydroborates [42]), according to 4*n*'s Wade–Mingos rules, to a deltahedron, i.e., a polyhedron with all faces as equilateral triangles. From the point of view of molecular orbitals (MO), each B-H group participates in the formation of three MOs for the cluster bonding, where two are tangent to the cage surface, giving rise to σ superposition of *p* orbitals, while the remaining is oriented inside the cluster and accounts for a further MO. Therefore, *n* B-H building blocks lead to *n*+1 total MOs in the framework. Such non-classical electron distribution implies electron resonance all along the σ-bonding in the cage, which, in analogy with carbon chemistry, is often referred to as "superaromaticity". This resonance energy (that accounts for the superior stability of hydroborates compared to other compounds) has been determined for several *closo*-boranes and carba*closo*-boranes up to 12 vertices [43,44]. It is worth noting that the greater the number of B-H units, the higher the stability due to the resonance effect, with a special stabilization shown by 6- and 12-vertex polyhedra. However, in the mono and bi-substituted carba*closo* series, the difference in energy decreases progressively [44], which is likely because the greater electronegativity of the carbon atom(s) withdraws electronic density and limits the resonance along the cage [45]. The progressive removal of one or two boron vertices from a *closo* cage leads to stable open-cage structures, namely *nido*- and *arachno*-boranes. Indeed, retrieving one or two vertices does not change the number of MOs involved in the skeletal binding, since the missing electrons are provided by the addition of hydrogen atoms. However, the loss of quasi-spherical geometry and the consequent reduction of electronic delocalization, besides the more hydridic character of apical hydrogens and their ability to withdraw electrical density from the cage [46], results in the lower chemical stability of the *nido*- and *arachno*-frameworks, with respect to their parent *closo* compounds [47].

#### **7. Classification of Inorganic Hydroborates**

Contrary to metal borohydrides [15], the known hydroborates containing larger poly-anions are mostly limited to alkali-metals, alkali-earth, and 3*d* transition metals. No double-cation hydroborate with a large electronegativity difference between the cations is known. Contrary to the borohydrides,

many hydroborate structures with larger hydroborate clusters can be derived not only from a close packed anion lattice (*ccp* or *hcp*) by filling the interstitial tetrahedral (T) and octahedral (O) sites with cations but also from the *bcc* anion packing. The O and T sites are deformed and interpenetrate in *bcc* packing, which makes their occupation by cations less favorable, and neighboring sites cannot be simultaneously occupied. Only very few ionic compounds are based on *bcc* anion packing, contrary to monoatomic structures such as metals. Well-known examples are the *ht* phases of AgI and Ag chalcogenides (S-Te) with Ag2S being a structural prototype for the *ht* phase of Na2B12H12 (see Figure S2). A common feature of all three *ht* phases, AgI, Ag2S, and Na2B12H12, is being a superionic conductor with *bcc* anions packing. It is interesting to note that while not a single metal borohydride is based on a *bcc* packing of anions, this anion packing is very often observed among the here-discussed hydroborates with poly-anions, especially at higher temperatures. This is probably a consequence of the weaker cation–hydrogen interaction compared to borohydride, since in hydroborates, the hydridic nature of hydrogen is depleted by the strong electronic density withdrawing toward the boron cage, which is due to the high delocalization among the boron–boron σ bonds in the cage [31,48].

The hydroborates discussed in this review are listed in the following tables: Table 1 contains single cation hydroborates, Table 2 contains double cation hydroborates, and Table 3 contains the hydroborates with inorganic poly-cation. The structural prototypes given in the tables can be found in different crystallographic databases of inorganic compounds such as the Inorganic Crystal Structure Database [49] or Pearson's Crystal Data [50], and in the following, we will not be providing the references for each structural prototype. The occurrence of structural prototypes among the hydroborates is shown in Figure 4, and it is compared with the occurrence among all inorganic compounds as extracted from the Linus Pauling File (LPF) [51].

While the inorganic compounds are dominated by four structure prototypes, NaCl, perovskite, sphalerite, and CsCl, the hydroborates prefer CaF2, Ni2In, perovskite (including Ag3SI), NaCl, and wurtzite (anti-types are included). As already discussed, the prototypes based on anionic *bcc* packing (AgI, Ag2S, and perovskites) are very common among the hydroborate salts. Other information on the chemistry, synthesis, and application of inorganic hydroborates can be found in recent reviews [15,27,52–58].

The content of the tables is organized in the following way: *First column*: Hydroborates are ordered according to the order of cations in the periodic table. The name describing the topology of the polyhedral cluster is given, too. *Second column*: The symbol of the phase is given as a prefix of the chemical formula only if defined in the original publication. *Third column*: The space group symbol as used in the original publication. If different from the standard setting, the latter is given in parentheses. *Fourth column*: The structural prototype among inorganic compounds. The choice of the prototype is not always in the sense of an *aristotype*, i.e., the most symmetrical structure in the Bärnighausen tree of the given structure (see for example [59]), but rather as a closest structure from which the hydroborate may be derived by replacing anions (i.e., oxides, halides) with hydroborates. The naming of the structural prototypes is according to LPF, i.e., name, Pearson symbol, and space group number [51]. The space group of the prototype or its mineral name is given only if more than one polymorph of the prototype exists. *Fifth column*: Cation coordination. The denticity of the hydroborate coordination, i.e., the number of hydrogen atoms of one hydroborate group coordinating to a given cation, is listed only if the crystal structure was determined by neutron diffraction, single crystal X-ray diffraction, or ab initio solid-state calculations. *Sixth column*: Anion coordination. *Seventh column*: Type of anions packing. If different from three basic packings, the symbol of the anion's net topology is given. The topological analysis was performed with the program TOPOS [60]. The type of molecules packing is given for molecular compounds. The nomenclature used to represent the different nets has been chosen according to the symbolism used in the TOPOS Topological Database (TTD) [61]. *Eighth column*: The given references relate to the most reliable crystal structure determination. If available, one reference for the X-ray and one for the neutron diffraction experiment are given as well as one for the validation of the structure by ab initio solid-state calculations.



*Chemistry* **2020** , *2*


**Table 1.** *Cont.*


*Chemistry* **2020** , *2*


**Table 1.** *Cont.*


**Table 1.** *Cont.*



monoclinic, *o* =

orthorhombic, *h* =

= room temperature, *ht* = high temperature, *hp* = high pressure, *m* =

hexagonal close packed, *bcc* =

(dodecadeltahedron).

 Color coding: alkali metals: brown; alkali earth metals: orange; rare earths: magenta; transition elements: blue; actinides: dark green; other metals: light green.

body-centered

 cubic, *aa* = simple sequence of hexagonal layers. 1:

hexagonal (refers to different polymorphs). *ccp* = cubic close packed, *hcp* =

gyroelongated

 square bipyramid; 2: triangular

orthobicupola;

 3: snub disphenoid




#### *Chemistry* **2020** , *2*

**Figure 4.** Occurrence of selected structural prototypes among binary inorganic compounds extracted from the Linus Pauling File [51], and among the hydroborates (excluding borohydrides with BH4 − anion). Both graphs include ternary solid solutions. The fractions are renormalized, to sum up to one for each graph.

#### **8. Conclusions**

Contrary to borohydrides, where all types of inorganic solid compounds can be found ranging from salts based on anions packing up to 2D and 3D coordination polymers, the higher hydroborates usually form salts where the large hydroborate anions pack, in most cases, according to one of the three common packings. The exception are hydroborates of Be that form molecular compounds, and transition metals hydroborates, which have still a packed anion sublattice, but the resulting metal–anion interaction is covalent, classifying these compounds as coordination polymers. The same is valid also for double cation hydroborates, while the situation is different for poly-cations, where the compact anion sublattice is broken as a consequence of the directional anion–cation bonds. The exception is the nearly spherical (if orientationally disordered) NH4 <sup>+</sup> cation.

Contrary to other inorganic compounds, the hydroborate salts crystallize very often with the *bcc* anion packing, favoring 3D pathways based on tetrahedral sites for high cation mobility. The anion packing can be best controlled by anion mixing and to a certain extend by anion modification. The mixing of cations and inclusion of neutral molecules in the crystal has a negative effect on cations mobility. The knowledge of anions packing and the ability to control it play an important role in designing the materials with desired properties such as high cation mobility at *rt* in solid-state electrolytes. We hope that this review clarifies the existing relation between different hydroborates poly-anions and the resulting crystal structures, sustaining the material science community in designing novel hydroborate-based compounds.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2624-8549/2/4/53/s1, Figure S1: Cation coordination in single metal hydroborates as a function of cation/anion radius ratio C/A. Figure S2: Analogy between temperature polymorphs of Li2B12H12 and C60, and between Na2B12H12 and Ag2S.

**Author Contributions:** Conceptualization, R.C.; writing—original draft preparation, R. ˇ C., M.B. and F.M.; ˇ writing—review and editing, R.C., M.B. and F.M.; All authors have read and agreed to the published version of ˇ the manuscript.

**Funding:** This research received no external funding and it was funded by the University of Geneva.

**Acknowledgments:** We acknowledge all colleagues who shared with us their unpublished results, i.e., the references marked "in preparation".

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References and Note**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Stereosopecificity in [Co(sep)][Co(edta)]Cl2***·***2H2O**

**Peter Osvath, Allen Oliver and A. Graham Lappin \***

Department of Chemistry and Biochemistry, University of Notre Dame, Notre Dame, IN 46556-5670, USA; peter.osvath@gmail.com (P.O.); aoliver2@nd.edu (A.O.)

**\*** Correspondence: alappin@nd.edu

**Abstract:** The X-ray structure of racemic [Co(sep)][Co(edta)]Cl2·2H2O is reported and reveals heterochiral stereospecificity in the interactions of [Co(sep)]3+ with [Co(edta)]−. Hydrogen-bonding along the molecular *C*2-axes of both complexes accounts for the stereospecificity. The structure of Λ- [Co(en)3]Δ-[Co(edta)]2Cl·10H2O has been re-determined. Previous structural data for this compound were collected at room temperature and the model did not sufficiently describe the disorder in the structure. The cryogenic temperature used in the present study allows the disorder to be conformationally locked and modeled more reliably. A clearer inspection of other, structurally interesting, interactions is possible. Again, hydrogen-bonding along the molecular *C*2-axis of [Co(en)3] 3+ and the equatorial carboxylates of [Co(edta)]− is the important interaction. The unique nature of the equatorial carboxylates and molecular *C*2-axis in [Co(edta)]−, straddled by two pseudo*-C*3-faces where the arrangement of the carboxylate groups conveys the same helicity, is highlighted. Implications of these structures in understanding stereoselectivity in ion-pairing and electron transfer reactions are discussed.

**Keywords:** stereoselectivity; hydrogen-bonding; complex ion

**Citation:** Osvath, P.; Oliver, A.; Lappin, A.G. Stereosopecificity in [Co(sep)][Co(edta)]Cl2·2H2O. *Chemistry* **2021**, *3*, 228–237. https://doi.org/10.3390/ chemistry3010017

Received: 11 December 2020 Accepted: 2 February 2021 Published: 6 February 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

#### **1. Introduction**

Structural studies have played a critical role in the determination of absolute configuration [1], an essential component of investigations of the chiral discriminations involving transition metal complexes. For over forty years, there have been attempts to understand the underlying principles that govern the diastereoselectivity that results from hydrogen-bonded interactions between complex cations such as [Co(en)3] 3+ (en = ethane-1,2-diamine) and complex anions such as [Co(edta)]<sup>−</sup> (edta4− = 2,2- ,2"3,2"'-(ethane-1,2-diyldinitrilo)tetraacetate(−4)) [2–5]. Indeed, the first reported resolution of [Co(edta)]<sup>−</sup> involved formation of a diastereoselective salt with Λ-[Co(en)3] 3+ as resolving agent [6].

Yoneda and co-workers employed the structurally related cations Δ-[Co(en)3] 3+, Δ- [Co(sep)]3+, and Δ(*λ*,*λ*,*λ*)-[Co((*RR*)-chxn)3] 3+ (sep = (1,3,6,8,10,13,16,19-octaazabicyclo [6.6.6]icosane), chxn = *trans*-1,2-cyclohexanediamine) to infer orientation effects in a series of careful chromatographic experiments, and suggested the dominant interactions in solution align major symmetry axes in the participating complexes that are homochiral [2,3]. Thus, Δ-[Co(sep)]3+, where the P(*C*3)-axis is sterically inhibited, interacts predominantly through the M(*C*2)-axis and with an anion interacting through its *C*3-axis, will have a M(*C*2)M(*C*3) or ΔΛ-preference, while with an anion interacting through its *C*2-axis, the preference will be M(*C*2)M(*C*2) or ΔΔ, Figure 1.

The complex Δ-[Co(en)3] 3+, like Δ-[Co(sep)]3+, is found to interact through the M(*C*2) axis while, conversely, the sterically more restricted Δ(*λ*,*λ*,*λ*)-[Co((*RR*)-chxn)3] 3+ has a preference for interactions through the P(*C*3)-axis. It is through such arguments that [Co(ox)3] <sup>3</sup>−, [Co(gly)(ox)2] <sup>2</sup>−, and [Co(edta)]<sup>−</sup> (ox2<sup>−</sup> = oxalate(−2), gly<sup>−</sup> = glycinate(−1)) have been identified as using their *C*<sup>3</sup> or pseudo-*C*<sup>3</sup> carboxylate faces in hydrogen bonding with the cations, see Figure 2.

**Figure 1.** View down the molecular *C*3- and *C*2-axes of Δ(*λ*,*λ*,*λ*)-[Co(en)3] 3+ showing the different helicities, P and M respectively, established by the arrangement of the chelating ligands. In Δ(*λ*,*λ*,*λ*)- [Co(sep)]3+, the molecular *C*3-axis is capped.

**Figure 2.** View down the molecular *C*2-axis of ΔΛΔ-[Co(edta)]<sup>−</sup> showing how it is flanked by the carboxylate pseudo-*C*3-faces.

Quantitative studies of ion pairing by conductivity, notably by Tatehata and coworkers, are consistent with this simple explanation [4], although the discrimination is at the limit of detection. Thus, the ion pairing constant in solutions containing Λ-[Co(en)3] 3+, Δ-[Co(edta)]−, is 125(5) M<sup>−</sup>1, whereas for the pair Λ-[Co(en)3] 3+, Λ-[Co(edta)]−, it is 119(5) M−<sup>1</sup> (25 ◦C, 0.01 M ionic strength (KI)) consistent with a ΛΔ-preference. Limited solution structural data from NMR relaxation measurements in the presence of [Cr(en)3] 3+, are also consistent and reveal [Co(edta)]− to use the pseudo-*C*<sup>3</sup> carboxylate face in discriminations. However, in [Co(edta)]−, it must be noted that the chirality resulting from the arrangement of the carboxylate groups of the equivalent pseudo-*C*<sup>3</sup> faces straddle the molecular *C*<sup>2</sup> axis, see Figure 2. Consequently, unlike in the case of the tris-bidentate chelates, the helicity conveyed by the ligands along the principal *C*2-axis and the pseudo-*C*<sup>3</sup> faces is the same [5].

Our interest derives from the useful correlation between the chiral recognition in these ion pairs and the chiral induction in the electron transfer reactions of [Co(edta)]− and other complexes with the pseudo-*C*<sup>3</sup> face with [Co(en)3] 2+ and derivatives [7–13]. A conclusion might be that the ion pairs serve as reasonable analogues for the precursor complex for the electron transfer process, despite the difference in the charge on the cation. The [Co(edta)]− system is particularly apt, as the complex has two very distinct sides; the carbon CH2-backbone of the ligand that is not capable of forming hydrogen-bonds, and the molecular *C*2-carboxylate axis surrounded by the two-pseudo-*C*<sup>3</sup> faces, all, in the parlance

of Yoneda, with the same handedness. Dipole considerations dictate that it is this latter side that should interact with hydrogen bonding cations.

Information derived from crystal structures on the interactions of [Co(edta)]− with [Co(en)3] 3+ and its derivatives or their analogues is very limited. In an earlier study, the structure of Λ-[Co(en)3]Δ-[Co(edta)]2Cl·10H2O was reported [13]. In the present paper, this structure has been re-determined at cryogenic temperature to provide a better model for a chelate ring conformational disorder and improve an understanding of the hydrogen bonding between the complexes. Indeed, a characteristic of the chiral induction in electron transfer with [Co(edta)]− and diastereomeric derivatives of [Co(en)3] 3+ is a dependence on chelate ring conformation. The structure of racemic [Co(sep)][Co(edta)]Cl2·2H2O has been determined for the first time. The relevance of these static structures in understanding electron transfer is discussed.

#### **2. Materials and Methods**

The compounds [Λ-(+)D-[Co(en)3]Cl3, Na[Δ-(+)546Co(edta)]·4H2O, K[Co(edta)]·2H2O, and [Co(sep)]Cl3 [13,14] were prepared by literature methods. Crystals of Λ-[Co(en)3]Δ- [Co(edta)]2Cl·10H2O were obtained by diffusion of propan-2-ol into a solution prepared by the addition of 0.051 g (0.12 mmol) of Na[Δ-(+)546Co(edta)]·4H2O, to a solution of 0.038 g (0.10 mmol) of [Λ-(+)D-[Co(en)3]Cl3, in 3 mL water as previously described [13]. An arbitrary sphere of data was collected on a violet rod-like crystal, having approximate dimensions of 0.267 × 0.062 × 0.040 mm, on a Bruker APEX-II diffractometer using a combination of ω- and ϕ-scans of 0.5◦ [15].

A sample of red-pink block microcrystals of [Co(sep)][Co(edta)]Cl2·2H2O (calc. (found): C 33.1 (33.5), H 5.81 (6.03); N 17.5 (17.2); Cl 8.87 (8.72) was obtained by slow evaporation of an aqueous solution containing 0.281 g (0.635 mM) of Na[Co(edta)]·4H2O and 0.10 g (0.212 mM) of [Co(sep)]Cl3·H2O. An arbitrary sphere of data was collected on a red-pink block-like crystal, having approximate dimensions of 0.031 × 0.018 × 0.010 mm, on a Bruker PHOTON-2 CMOS diffractometer using a combination of ω- and ϕ-scans of 0.5◦ [15].

For both structures, data were corrected for absorption and polarization effects, and analyzed for space group determination [16]. The structures were solved by dual-space methods and expanded routinely [17]. Models were refined by full-matrix least-squares analysis of F<sup>2</sup> against all reflections [18]. All non-hydrogen atoms were refined with anisotropic atomic displacement parameters.

Atomic displacement parameters for hydrogen atoms in Λ-[Co(en)3]Δ-[Co(edta)]2Cl· 10H2O were modeled as a mixture of refined and constrained geometries. Hydrogen atoms on the Co complexes were modeled with atoms riding on the coordinates of the atom to which they are bonded with atomic displacement parameters tied to that of the atom to which they are bonded (*U*iso(H) = 1.2 *U*eq (C/N)). Water hydrogen atoms were included in positions located from a difference Fourier map. Most water hydrogen atoms were refined freely; several that did not model well were modeled with atomic displacement parameters tied to that of the oxygen to which they are bonded (*U*iso(H) = 1.5 *U*eq(O)).

In [Co(sep)][Co(edta)]Cl2·2H2O, one water was found to be disordered over two positions during refinement. In the final structure, this atom was modeled with two, halfoccupancy oxygen atoms, and concomitant hydrogen atoms, at sites suggested as the loci of the original extended displacement parameters. Residual electron density (2.24 e<sup>−</sup>/Å3) is located near (0.87 Å) Co4 of one of the [Co(edta)]− anions. It is unclear what this residual density might be, and is likely due to Fourier ripple or a very small amount of otherwise unresolvable molecular disorder. Hydrogen atoms for [Co(sep)][Co(edta)]Cl2·2H2O were treated as a mixture of freely refined and geometrically constrained atoms. Hydrogen atoms bonded to carbon and nitrogen were treated as riding models with *U*iso(H) = 1.2 *U*eq(C). Water hydrogen atoms were modeled at locations initially located from a difference Fourier map and subsequently tied to the coordinates of the oxygen to which they are

bonded. Atomic displacement parameters for water hydrogen atoms in this model were restrained to *U*iso(H) = 1.5 *U*eq(O).

The disorder in Λ-[Co(en)3]Δ-[Co(edta)]2Cl·10H2O was resolved more thoroughly with a cryogenic measurement of the data for the complex. The ethane-1,2-diamine was observed to be a twist disorder of the ethylene backbone, across the crystallographic twofold axis that bisects the ethylene chain. Only carbon atom C3 is the unique atom in the model. The two sites were modeled from density observed in a Fourier difference map and the site occupancy ratios summed to unity yielding an approximately 0.77:0.23 ratio. The major component was refined with anisotropic displacement parameters and the minor component with an isotropic atom. Hydrogen atoms about the disorder (on nitrogen and the disordered carbon) were modeled using routine methodology (occupancies tied to the disorder component, riding atom positions, and displacement parameters).

#### **3. Results**

### *3.1. [Co(sep)][Co(edta)]Cl2*·*2H2O*

The structure of [Co(sep)][Co(edta)]Cl2·2H2O (Supplementary Materials) consists of two crystallographically independent, yet chemically identical, [Co(sep)]3+ cations, two [Co(edta)]− anions, four chlorine anions, and four water molecules of crystallization in the asymmetric unit of the primitive, centrosymmetric, triclinic space group P-1. Crystal data are summarized for [Co(sep)][Co(edta)]Cl2·2H2O [19].

The cations have a cobalt atom encapsulated in an octahedral fashion by a sepulchrate ligand. The cobalt is coordinated by the amine nitrogen atoms, that retain their hydrogen atoms. The "apical" nitrogen atoms of the sepulchrate are non-coordinating. The conformation of the complex is Δ(*λ*,*λ*,*λ*) or Λ(*δ*,*δ*,*δ*) (*lel*3). In the anion, the cobalt is chelated by edta4<sup>−</sup> in a six-coordinate coordination geometry with geometry ΔΛΔ or ΛΔΛ, abbreviated Δ and Λ, respectively.

The complex ions are arranged in a hetero-chiral pairwise fashion Δ-[Co(sep)]Λ- [Co(edta)] or Λ-[Co(sep)]Δ-[Co(edta)]. The Co-Co distances are 5.162(1) Å and 5.170(1) Å for the two independent pairs. The chlorine atoms form hydrogen bonds with two neighboring sepulchrate amide nitrogen atoms, Table 1, occupying two of the molecular *C*2-axes of the complex cation, but do not participate in further H-bonding. Each chloride atom is hydrogen bonded by two N-H atoms from a single sepulchrate. The third molecular *C*2-axis of the [Co(sep)]3+ is occupied in stereospecific fashion by [Co(edta)]<sup>−</sup> with a pair of N-H hydrogen bonds to the equatorially coordinated *G*-ring oxygens of its neighboring [Co(edta)]<sup>−</sup> anion (graph set notation *R*<sup>2</sup> <sup>2</sup>(8)), see Figure 3. Thus, they also do not propagate the hydrogen bonded network.

**Figure 3.** Hydrogen bonding between Δ-[Co(sep)]3+ and Λ-[Co(edta)]−. The view is down the molecular *C*3-axis of [Co(sep)]3+ and shows the interaction with Cl<sup>−</sup> on the molecular *C*2-axes. Bond distances for complexes (Å): Co-N (Δ-[Co(sep)]3+) 1.973(3), 1.974(3) 1.979(3) 1.987(3) 1.987(3) 1.993(3); Co-N (Λ-[Co(edta)]−) 1.921(3), 1.930(3); Co-O (G-ring) 1.898(3), 1.911(3); Co-O (R-ring) 1.879(3), 1.898(3).


**Table 1.** Hydrogen bonds between metal-ion complexes [Å and ◦].

<sup>a</sup> Coordinated oxygen, <sup>b</sup> terminal oxygen.

The hydrogen bonded network is extended through the structure with water molecules linking non-coordinated acetate oxygen atoms of [Co(edta)]−. The arrangement of molecules results in a 2D sheet of H-bonded molecules parallel to the b/c plane. Each [Co(edta)]− accepts four hydrogen bonds and is the "corner" of a 4-connected square. Located in the center of each square is a [Co(sep)]3+ from an adjacent sheet, related by inversion symmetry. Due to the orientation of the ligands, these sheets are bi-layers, with hydrophobic regions between layers, see Figure 4.

### *3.2. (*Λ*-[Co(en)3]*Δ*-[Co(edta)]2Cl*·*10(H2O)*

The complex, Λ-[Co(en)3]Δ-[Co(edta)]2Cl·10(H2O) crystallizes as violet rod-like crystals from an aqueous solution. There are, collectively, two molecules of the cation, four molecules of [Co(edta)]− anion, two chloride anions, and 20 water molecules of crystallization in the unit cell of the primitive, acentric, orthorhombic space group P21212. The correct enantiomorph of the space group and absolute stereochemistry of the complex were determined both by comparison with the known configuration of the complex and by comparison of intensities of Friedel pairs of reflections. Friedel pair analysis (Flack *x* parameter = 0.006(6) [20] and Hooft *y* parameter = 0.001(4) [21] support the assignment. Crystal data are summarized for Λ-[Co(en)3]Δ-[Co(edta)]2Cl·10(H2O) [22].

**Figure 4.** Packing diagram for [Co(sep)][Co(edta)]Cl2·2H2O looking down the crystallographic c-axis. Note the hydrophobic channels parallel to the b/c plane alternating with the hydrogen-bonded channels containing the Cl− ions.

The [Co(en)3] 3+ cation is located on the crystallographic two-fold axis at [0, 0.5, z] and the chloride anion on the two-fold axis at [1, 0, z], while the [Co(edta)]− anion is located in a general position. Thus, the charge carrying species in the asymmetric unit consist of one half [Co(en)3] 3+ cation, one half chloride anion, and one [Co(edta)]<sup>−</sup> anion. One ethylene diamine ligand is disordered over two sites (the ligand in question bisects the two-fold axis) and was routinely modeled with partial occupancy atoms (0.77:0.23).

The chloride ions are coordinated by eight water molecules, part of an extensive hydrogen bonding network forming channels, parallel to the c-axis. The amine nitrogen atoms of the [Co(en)3] 3+ cations also participate, with alternating [Co(en)3] 3+ and Cl<sup>−</sup> in the ac-plane. The [Co(edta)]− anions alternate in orientation in a parallel ac-plane, completing a layered structure along the b-axis. The layers are held together by hydrogen bonding between the [Co(en)3] 3+ and [Co(edta)]−. The two [Co(edta)]<sup>−</sup> anions interact with [Co(en)3] 3+ in different fashion. One interaction, with a Co-Co distance of 5.844(1) Å lies roughly along a molecular *C*2-axis of [Co(en)3] 3+. There is hydrogen-bonding between one N-H proton along the *C*2-axis, and a second nitrogen on the C3-face of [Co(en)3] 3+ with the coordinated and un-coordinated oxygen atoms of the carboxylate group of one of the G-rings on [Co(edta)]<sup>−</sup> (graph set notation *R*<sup>2</sup> <sup>2</sup>(8)). The interaction with the *C*2-axis nitrogen involves the disordered ethane-1,2-diamine ring on [Co(en)3] 3+, and examination of the linearity of the hydrogen bonds for the two conformations reveals that the *λ* conformation is preferred, Table 1, and that the 5.844 Å interaction favors Λ(*λ*,*λ*,*λ*) (*ob*3), see Figure 5.

The other interaction with a Co-Co distance of 7.509(1) Å shows a hydrogen bonding interaction between an uncoordinated carbonyl oxygen of an out-of-plane R-ring of [Co(edta)]− with an N-H proton from the disordered 1,2-diaminoethane ring on [Co(en)3] 3+. Again, examination of the linearity of the hydrogen bonds for the two conformations reveals that the *δ* conformation is preferred, and that the 7.509 Å interaction favors Λ(*δ*,*λ*,*λ*) (*lelob*2).

**Figure 5.** Hydrogen bonding between Λ-[Co(en)3] 3+ and Δ-[Co(edta)]<sup>−</sup> with a Co-Co distance of 5.844 (1) Å. The view is roughly down the molecular *C*3-axis of Λ-[Co(en)3] 3+ and shows the Λ(*λ*,*λ*,*λ*) (*ob*3) conformation. Bond distances for complexes (Å): Co-N (Λ-[Co(en)3] 3+) 1.955(2), 1.965(2), 1.960(3); Co-N (Δ-[Co(edta)]−) 1.921(3), 1.927(2); Co-O (G-ring) 1.914(2), 1.921(3); Co-O (R-ring) 1.888(2), 1.877(2).

#### **4. Discussion**

As expected, bond distances and angles within the molecular ions in both structures are comparable with related studies. However, the focus of this communication is the interactions between the metal-ion complexes.

It has been a generally accepted concept that the chiral discriminations of tris-bidentate chelate complexes such as [Co(en)3] 3+ are the result of different orientations of hydrogenbonding interactions since the *C*<sup>3</sup> axis has the opposite helicity to the *C*<sup>2</sup> axis [2–4]. Thus, Δ-[Co(en)3] 3+ is P(*C*3)M(*C*2) using the nomenclature for P or positive helicity referring to a right-handed screw and M to the left-handed screw. Ion pairing discrimination studies with metal-complex carboxylate liganded anions carried out by chromatography and conductivity measurements have highlighted the importance of the hydrogen-bonded match of the *C*<sup>3</sup> and *C*2-axes of [Co(en)3] 3+ with a pseudo-*C*<sup>3</sup> carboxylate face where three carboxylate groups form a facial motif that is not subtended by a chelate ring. The helicity of the pseudo-*C*<sup>3</sup> carboxylate face in the anion projects to the axis with the same helicity in the cation. Thus the pseudo-*C*<sup>3</sup> carboxylate face of Δ-[Co(edta)]<sup>−</sup> interacts preferentially with the N-H hydrogens on the *C*3-axes of Δ-[Co(chxn)3] 3+ where the *C*2-axes are sterically encumbered, but with the N-H hydrogens on *C*2-axes of Λ-[Co(sep)]3+ where the *C*3-axes are encumbered.

However, the hexa-coordinated complex ion, [Co(edta)]−, differs in symmetry from a tris-bidentate chelate. The two pseudo-*C*<sup>3</sup> carboxylate faces flank the *C*<sup>2</sup> axis, and the ligand arrangement is such that all three present the same overall helicity. What is notable in both of the structures reported here is that the interactions involving the closest Co-Co distances involve the *C*<sup>2</sup> axis or the in-plane *G*-rings of the anion, strongly suggesting a more important role for the helicity conveyed along the *C*2-axis of [Co(edta)]− in determining the discriminations. This observation is also consistent with the solution NMR structure of the ion pair, {[Cr(en)3] 3+[Co(edta)]−}, where by symmetry, the paramagnetic cation straddles the *C*2-axis of [Co(edta)]− [5].

There is a structural comparison with Δ-[Ni(en)3]Δ-[Ni(edta)]·4H2O that is also relevant [23]. The cation and anion share two interactions. The closest Ni-Ni distance is 5.40 Å and reveals a direct hydrogen bond formed between the non-coordinated *G*-ring oxygen of [Ni(edta)]2<sup>−</sup> and an N-H on the *C*3-axis of [Ni(en)3] 2+ with a second interaction involving the coordinated O of the other *G*-ring, a bridging water molecule, and a second N-H on the *C*3-axis of [Ni(en)3] 2+ (Graph set notation *R*<sup>2</sup> <sup>2</sup>(12)). There is no direct pseudo-*C*<sup>3</sup> carboxylate

interaction involving the three N-H groups of the *C*3-axis of [Ni(en)3] 2+. Instead, it is the in-plane *G*-ring carboxylates that again play a dominant role. A longer Ni-Ni distance at 6.14 Å involves a non-coordinated *R*-ring oxygen of [Ni(edta)]2<sup>−</sup> with two N-H protons on the *C*3-axis of [Ni(en)3] 2+ (Graph set notation *R*<sup>1</sup> <sup>2</sup>(6)).

While the distinction between the ligand arrangement conveying the same helicity through the *C*2-axis and pseudo-*C*<sup>3</sup> faces in [Co(edta)]− with the same helicity may seem semantic, it should be noted that the dipole moment of [Co(edta)]− projects along the *C*2-axis. The idea that discriminations by [Co(edta)]− can be projected through the *C*2 axis and not only by the pseudo-*C*3-faces has implications in the interpretation of the extensive data on related outer-sphere stereoselective electron transfer. The reductions of [Co(edta)]− with both [Co(en)3] 2+ and [Co(sep)]2+ have been shown to occur by an outer-sphere mechanism [7,9].

Computational work on the effects of distance and orientation in outer-sphere electrontransfer reactions between metal ion complexes has focused predominantly on the [Fe(OH2)6] 3+/2+ and [Ru(OH2)6] 3+/2+ self-exchange reactions [24–26]. Increasingly, sophisticated calculations [27–31] have not markedly changed the conclusions first reached that face-to-face interactions along the *C*3-axes represent the closest approach of the two metal centers, at distances roughly 5–6 Å, and the most favorable configuration for overlap of the donor and acceptor orbitals resulting in electron transfer. Other orientations over a range of distances provide less favorable pathways with super-exchange mechanisms involving the ligands more likely at distances in excess of 6 Å.

Unlike the self-exchange reactions, the outer-sphere oxidations of [Co(en)3] 2+ and [Co(sep)]2+ by [Co(edta)]<sup>−</sup> involve complexes with opposite charges, and the additional electrostatic attraction can provide a more intimate interaction, generally at hydrogenbonded distances. An attractive model for a *C*3-*C*<sup>3</sup> interaction is provided by the structure of [Cr(en)3] 3+[Cr(ox)3] 3+ where the metal-metal distance is 4.98 Å [32], shorter than the closest approach distances found in the present structural studies with [Co(en)3] 3+ (5.84 Å) and [Co(sep)]3+ (5.17 Å). Although the electron transfer precursor is dynamic and comparisons with static structures are fraught with problems, were distance the only factor, then for [Co(edta)]− projecting discrimination through the pseudo-*C*3-face, one might well expect the reaction stereoselectivity to reflect a dominant *C*3-*C*<sup>3</sup> interaction and hence a homochiral, ΔΔ or ΛΛ, preference. That is not what is observed, providing a further piece of evidence that the discrimination by [Co(edta)]− is more complex.

Further, a detailed analysis of structural, charge, and dipolar effects on the stereoselective electron transfer data also revealed [33] that there is a distinction between [Co(edta)]− and the oxalate containing reagents that possess the pseudo-*C*<sup>3</sup> motif such as *C*1-*cis*(N)-[Co(gly)2(ox)]<sup>−</sup> (gly<sup>−</sup> = glycinate(−1)). Stereoselectivity and chiral discriminations involving [Co(edta)]− should be considered in a class of their own.

It is noted that the chelate ring conformations of [Co(en)3] 3+ (*ob*<sup>3</sup> or *lelob*2) and [Co(sep)]3+ (*lel*3) differ in the two structures presented. Both cations employ the *C*2 axis in the closest interaction, and it is not unreasonable to expect that [Co(en)3] 3+ (*lel*3) would interact in similar fashion to [Co(sep)3] 3+ (*lel*3) since they differ mainly along the *C*3-axis. In stereoselective electron-transfer studies where [Co(edta)]− is used as an oxidant for [Co(*R*,*S*-pn)3] 2+, (*R*,*S*-pn = *R*,*S*-propane-1,2-diamine), [Co(*RR*,*SS*-bn)3] 2+ (*RR*,*SS*-bn = *RR*,*SS*-butane-2,3-diamine), and [Co(*RR*,*SS*-chxn)3] 2+, the individual stereoselectivities for the conformation isomers can be determined and show a trend from homo-chiral to hetero-chiral with increasing *ob*-character, previously ascribed to differences in hydrogenbonding [9,11]. The weighted average stereoselectivities are [Co(*RR*,*SS*-chxn)3] 2+ (8% homo-), [Co(*RR*,*SS*-bn)3] 2+ (3% homo-), [Co(*R*,*S*-pn)3] 2+ (4% hetero-), and can be compared with those of the conformationally labile [Co(en)3] 2+ (10% hetero-), and [Co(sep)]2+ (17% hetero-), reflecting the clear trend in discrimination in the tris-bidentate derivatives of [Co(en)3] 2+. However, this does not provide insight into the discriminating nature of the oxidant [Co(edta)]−. The most apt comparison is with the oxidant Λ-[Co((*R*)-tacntp)] ((*R*)-tacntp3<sup>−</sup> = 1,4,7-tri-aza-cyclo-nonane-1,4,7-tris [2- (*R*)-2- -propionate](-3)), which has

3-fold symmetry with a strong dipole, but no *C*2-axis [34]. While the selectivity in the oxidation of [Co(en)3] 2+ (11% hetero-) is comparable in sense and magnitude with the value for reaction with [Co(edta)]−, the weighted average selectivity with [Co(*RR*,*SS*-chxn)3] 2+ (31% homo-) reflects a much stronger *C*3-*C*<sup>3</sup> preference.

#### **5. Conclusions**

In conclusion, the structures presented highlight an important role for hydrogen bonding involving the unique *C*2-axis of [Co(edta)]− in chiral discriminations with [Co(en)3] 3+ and derivatives. Stereoselectivity and chiral discriminations involving [Co(edta)]− should be considered in a class of their own. This has implications in the interpretation of data for related stereoselective electron transfer reactions and suggest that generalizations should be avoided.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/2624-854 9/3/1/17/s1, Figure S1: Labelling diagram for [Co(sep)][Co(edta)]Cl2·2H2O; Figure S2: Packing diagram for [Co(sep)][Co(edta)]Cl2·2H2O; Figure S3: View along c-axis for [Co(sep)][Co(edta)]Cl2·2H2O; Figure S4: Labelling diagram for Λ-[Co(en)3]Δ-[Co(edta)]2Cl·10H2O. X-Ray crystallographic files of [Co(sep)][Co(edta)]Cl2·2H2O and Λ-[Co(en)3]Δ-[Co(edta)]2Cl·10H2O, CCDC 2049761, 2049762.

**Author Contributions:** Conceptualization, P.O. and A.G.L.; methodology, P.O., A.O., and A.G.L.; validation, P.O., A.O., and A.G.L.; formal analysis, A.O. and A.G.L.; investigation, P.O., A.O., and A.G.L.; resources, P.O., A.O., and A.G.L.; data curation, A.O., and A.G.L.; writing—original draft preparation, A.G.L.; writing—review and editing, A.G.L.; visualization, P.O., A.O., and A.G.L.; supervision, A.G.L.; project administration, A.G.L.; funding acquisition, A.O., and A.G.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded under National Science Foundation (Grant No. CHE84-06113). The ALS is supported by the U.S. Dept. of Energy, Office of Energy Sciences, under contract DE-AC02-05CH11231.

**Acknowledgments:** Samples for synchrotron crystallographic analysis were submitted through the SCrALS (Service Crystallography at Advanced Light Source) program. Crystallographic data were collected at Beamline 12.2.1 at the Advanced Light Source (ALS), Lawrence Berkeley National Laboratory.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


*Article*

## **The Tyranny of Arm-Wrestling Methyls on Iron(II) Spin State in Pseudo-Octahedral [Fe(didentate)3] Complexes** †

#### **Neel Deorukhkar 1, Timothée Lathion 1, Laure Guénée 2, Céline Besnard <sup>2</sup> and Claude Piguet 1,\***


Received: 11 March 2020; Accepted: 31 March 2020; Published: 2 April 2020

**Abstract:** The connection of a sterically constrained 3-methyl-pyrazine ring to a *N*-methylbenzimidazole unit to give the unsymmetrical α,α'-diimine ligand **L5** has been programmed for the design of pseudo-octahedral spin-crossover [Fe(**L5**)3] <sup>2</sup><sup>+</sup> units, the transition temperature (*T*1/2) of which occurs in between those reported for related facial tris-didentate iron chromophores fitted with 3-methyl-pyridine-benzimidazole in a LaFe helicate (*T*1/<sup>2</sup> ~ 50 K) and with 5-methyl-pyrazine-benzimidazole **L2** ligands (*T*1/<sup>2</sup> ~350 K). A thorough crystallographic analysis of [Fe(**L5**)3](ClO4)2 (I), [Ni(**L5**)3](ClO4)2 (**II**), [Ni(**L5**)3](BF4)2·H2O (**III**), [Zn(**L5**)3](ClO4)2 (**IV**), [Ni(**L5**)3](BF4)2·1.75CH3CN (**V**), and [Zn(**L5**)3](BF4)2·1.5CH3CN (**VI**) shows the selective formation of pure facial [M(**L5**)3] <sup>2</sup><sup>+</sup> cations in the solvated crystals of the tetrafluoroborate salts and alternative meridional isomers in the perchlorate salts. Except for a slightly larger intra-strand interannular twist between the aromatic heterocycles in **L5**, the metric parameters measured in [Zn(**L5**)3] <sup>2</sup><sup>+</sup> are comparable to those reported for [Zn(**L2**)3] <sup>2</sup>+, where **L2** is the related unconstrained ligand. This similitude is reinforced by comparable ligand-field strengths (Δoct) and nephelauxetic effects (as measured by the Racah parameters *B* and *C*) extracted from the electronic absorption spectra recorded for [Ni(**L5**)3] <sup>2</sup><sup>+</sup> and [Ni(**L2**)3] <sup>2</sup>+. In this context, the strictly high-spin behavior observed for [Fe(**L5**)3] 2+ within the 5–300 K range contrasts with the close to room-temperature spin-crossover behavior of [Fe(**L2**)3] <sup>2</sup><sup>+</sup> (*T*1/<sup>2</sup> = 349(5) K in acetonitrile). This can be unambiguously assigned to an intraligand arm wrestling match operating in bound **L5**, which prevents the contraction of the coordination sphere required for accommodating low-spin FeII. Since the analogous 3-methyl-pyridine ring in [Fe(**L3**)3] <sup>2</sup><sup>+</sup> derivatives are sometimes compatible with spin-crossover properties, the consequences of repulsive intra-strand methyl–methyl interactions are found to be amplified in [Fe(**L5**)3] <sup>2</sup><sup>+</sup> because of the much lower basicity of the 3-methyl-pyrazine ring and the resulting weaker thermodynamic compensation. The decrease of the stability constants by five orders of magnitude observed in going from [M(**L2**)3] <sup>2</sup><sup>+</sup> to [M(**L5**)3] <sup>2</sup><sup>+</sup> (M = NiII and ZnII) is diagnostic for the operation of this effect, which had been not foreseen by the authors.

**Keywords:** pyrazine-benzimidazole; spin crossover; iron(II); ligand field; nephelauxetic

#### **1. Introduction**

In line with the formulation of the ligand field theory [1,2], or as it was originally called by Bethe, crystal-field theory [3], it was realized that an open-shell metal with at least two valence electrons in a specific chemical environment could exist with either high-spin or low-spin configuration [4].

Following van Vleck's approach to magnetism [5], Pauling perceptively recognized that it would be feasible to obtain systems in which two spin states could be present simultaneously, while their ratio should depend on the energy difference between them [6,7]. The discovery of thermal spin-state equilibria operating in FeIII dithiocarbamate by Cambi et al. [8–10] at the same period, indeed confirmed these predictions. Since then, a myriad of metal coordination complexes and polymeric materials have been shown to display spin transitions, often referred to as spin-crossover (SCO) materials. These have been studied in detail and extensively reviewed over the last two decades [11–20]. Due to the 'on–off' switching of the magnetic properties accompanying the spin transition from the low-spin diamagnetic configuration (1A1 label in octahedral symmetry) to the high-spin paramagnetic form (5T2 label in octahedral symmetry) for d6 transition metals in pseudo-octahedral geometry (Scheme 1a), the 'magic' [FeIIN6] chromophores, where N is a heterocyclic nitrogen donor atom, have been intensively investigated [11–20]. Various external stimulations such as changes in temperature [21,22], pressure [23,24], magnetic field [25] or light-irradiation [26,27] can be used for inducing the SCO processes, which makes these microscopic magneto-optical switches very attractive for their introduction into responsive macroscopic materials [12,13,16,28–31]. The most common and accepted approach for rationalizing the design of spin-crossover pseudo-octahedral FeII complexes relies on the energetic balance Δ*E*HL = *E*<sup>0</sup> hs <sup>−</sup> *<sup>E</sup>*<sup>0</sup> ls = 2(Δoct <sup>−</sup> *<sup>P</sup>*) between the ligand-field stabilization energy as measured by Δoct = 10 Dq and the spin pairing energy modeled with the Racah parameters *B* and *<sup>C</sup>* with *<sup>P</sup>* <sup>=</sup> <sup>2</sup>*<sup>B</sup>* <sup>+</sup> <sup>4</sup>*<sup>C</sup>* <sup>≈</sup> <sup>19</sup>*<sup>B</sup>* (Scheme 1a) [31]. When <sup>Δ</sup>oct >> *<sup>P</sup>*, <sup>Δ</sup>*E*HL = *<sup>E</sup>*<sup>0</sup> hs <sup>−</sup> *<sup>E</sup>*<sup>0</sup> ls = <sup>2</sup>(Δoct <sup>−</sup> *<sup>P</sup>*) >> <sup>0</sup> and the pseudo-octahedral FeII complex adopts a low spin configuration with a diamagnetic 1A1 electronic ground state, as shown in the right part of the Tanabe–Sugano diagram built for the electronic d<sup>6</sup> configuration (Δoct/*B* > 20 in Scheme 1b). The reverse situation occurs when *P* >> Δoct, which leads to Δ*E*HL = *E*<sup>0</sup> hs <sup>−</sup> *<sup>E</sup>*<sup>0</sup> ls <sup>=</sup> <sup>2</sup>(Δoct <sup>−</sup> *<sup>P</sup>*) << 0 and the existence of the paramagnetic high-spin 5T2 ground state (left part of Tanabe–Sugano diagram with Δoct/*B* < 10 in Scheme 1b). Finally, for intermediate values 0 ≤ Δ*E*HL <sup>=</sup> <sup>2</sup>(Δoct <sup>−</sup> *<sup>P</sup>*) ≤ *mRT* (*m* ≤ 10), the two spin states coexist and are thermally populated at accessible temperatures. However, the latter statement is misleading and physically unsound since both Δoct and *P* change during the spin transition as a result of the population of the antibonding orbitals in the high-spin form. For pseudo-octahedral spin-crossover [FeIIN6] complexes, the Fe–N bond lengths extend by approximately 10% upon the low-spin to high-spin transition and Δoct consequently decreases according to a 1/*r<sup>n</sup>* dependence with *n* = 5–6 (Equation (1)) [31].

$$\frac{\Delta\_{\rm{oct}}^{\rm{HS}}}{\Delta\_{\rm{oct}}^{\rm{LS}}} = \left(\frac{r\_{\rm{LS}}}{r\_{\rm{HS}}}\right)^{n} \tag{1}$$

Taking typical Fe–N bond distances of *r*LS = 2.0 Å and *r*HS = 2.2 Å [11–20] leads to ΔLS oct/ΔHS oct ≈ 1.75 accompanying the spin transition, whereas *P* changes very little (*P*HS : *P*LS - 19*B*), except for a faint reduction of the nephelauxetic effect with larger bond lengths [32].

Taking into account the 10% bond length expansion accompanying the spin transition, the simplistic zero-point energy differences between the two states summarized in Scheme 1a (i.e., Δ*E*HL = *E*<sup>0</sup> hs <sup>−</sup>*E*<sup>0</sup> ls = 2(Δoct − *P*)) should be replaced with Equation (2), which is transformed into Equations (3) and (4) upon introducing ΔLS oct/ΔHS oct ≈ 1.75.

$$
\Delta E\_{\rm HL} = E\_{\rm hs}^{0} - E\_{\rm ls}^{0} = \frac{12}{5} \Delta\_{\rm oct}^{\rm LS} - \frac{2}{5} \Delta\_{\rm oct}^{\rm HS} - 38B \tag{2}
$$

$$
\Delta E\_{\rm HL} = E\_{\rm hs}^{0} - E\_{\rm ls}^{0} = 3.8 \Delta\_{\rm oct}^{\rm HS} - 38B \tag{3}
$$

$$
\Delta E\_{\rm HL} = E\_{\rm hs}^{0} - E\_{\rm ls}^{0} = 2.17 \Delta\_{\rm oct}^{\rm LS} - 38B \tag{4}
$$

**Scheme 1.** (**a**) The thermodynamic origin of FeII spin state equilibrium in octahedral symmetry (*P* is the electron spin pairing energy). (**b**) Tanabe–Sugano diagram for a d<sup>6</sup> metal ion calculated using the electrostatic matrices in the strong-field basis and using the free ion Racah parameters *B* = 917 cm−<sup>1</sup> and *C* = 4.41*B* = 4040 cm−<sup>1</sup> [31]. The red areas illustrate the limiting domain of co-existence of low-spin and high-spin complexes for which 0 ≤ *E*0 hs <sup>−</sup> *<sup>E</sup>*<sup>0</sup> ls <sup>≤</sup> 2000 cm−<sup>1</sup> with a Racah parameter fixed at 75% of the free ion values, whereas the dashed green domain corresponds to non-accessible ligand-field strengths (see text) [31].

Solving Equations (3) and (4) for Δ*E*HL = 0 provides ΔHS oct/*<sup>B</sup>* = 38/3.8 = <sup>10</sup> and <sup>Δ</sup>LS oct/*B* = 38/2.17 = 17.5 as the lower limit of the existence of thermally accessible spin state equilibria, while the higher limits can be estimated for Δ*E*HL = 2000 cm−1, which gives ΔHS oct/*B* = 10 + [2000/(3.8*B*)] and ΔLS oct/*B* = 17.5 + [2000/(2.17*B*)] (red areas in Scheme 1b). More sophisticated calculations using Racah parameters *B* and *C*, reduced by 70–80% of their free ion values, predict narrow ranges of ligand field strengths 11,000 <sup>≤</sup> <sup>Δ</sup>HS oct <sup>≤</sup> 12,500 cm−<sup>1</sup> and 19,000 <sup>≤</sup> <sup>Δ</sup>LS oct <sup>≤</sup> 22,000 cm<sup>−</sup>1, for which the phenomenon of a thermal spin transition can be expected in FeII coordination complexes [31]. Following this theoretical approach, the toolkit of coordination chemists for programming and tuning the thermodynamic spin transition parameters in molecular [FeIIN6] complexes logically relied on the manipulation of Δoct and *B* via (i) some controlled distortions of the coordination geometry from a perfect octahedron

by using chelating ligands with fixed bite angles [33] and (ii) specific programming of metal–ligand bonding interactions via ligand design [15,19,34]. Benefiting from the huge amount of experimental data collected during the last decades for [Fe(N∩N)3] <sup>2</sup><sup>+</sup> complexes, where N∩N is an α,α'-diimine chelate ligand possessing two N-heterocyclic donor atoms, it was shown that the connection of a six-membered heterocycle to a five-membered heterocycle in N∩N provides favorable ligand-field strengths around FeII for promoting spin-state equilibria (Equation (5)) with transition temperatures *T*1/<sup>2</sup> = Δ*H*SCO/Δ*S*SCO (i.e., the temperature at which Δ*G*SCO = 0 and *x*hs = *x*ls = 0.5) within the 30–500 K range [14,15,33].

$$\mathrm{Fe^{II}\_{\mathrm{low}-\mathrm{spin}}} \stackrel{\mathrm{K\ominus}}{\rightleftharpoons} \mathrm{Fe^{II}\_{\mathrm{high}-\mathrm{spin}}} K\_{\mathrm{SCO}} = \mathrm{x\_{\mathrm{ls}}/\mathrm{x\_{\mathrm{ls}}}} = e^{-\left(\Lambda \mathrm{\bar{c}} \mathrm{s} \mathrm{co}/RT\right)} = e^{\left(\Lambda \mathrm{\bar{s}} \mathrm{c} \mathrm{o}/RT - \Lambda \mathrm{\bar{h}} \mathrm{c} \mathrm{o}/RT\right)}\tag{5}$$

The didentate ligands **L1** and **L2** match the latter criteria and the associated pseudo-octahedral complexes [Fe(**L***k*)3] <sup>2</sup><sup>+</sup> indeed exhibit spin-crossover behaviors in acetonitrile solutions, the transition temperatures of which reveal the stronger Fe–N bonds induced by the strong-accepting pyrazine units in [Fe(**L2**)3] <sup>2</sup><sup>+</sup> (*T*1/<sup>2</sup> ~350 K) compared with pyridine units in [Fe(**L1**)3] <sup>2</sup><sup>+</sup> (*T*1/<sup>2</sup> ~310 K, Scheme 2) in the absence of sterical constraints [35].


*D*7KH*IDF*>)H**/**@ FKURPRSKRUHZKHQHPEHGGHGLQWRDQRQFRQVWUDLQHG/D)H KHOLFDWHGLVSOD\HG*7* a .

**Scheme 2.** Chemical structures of the didentate ligands **L1**–**L5**, together with the electronic properties of the associated pseudo-octahedral [Ni(**L***k*)3] <sup>2</sup><sup>+</sup> complexes and thermodynamic spin-crossover (SCO) properties of the pseudo-octahedral [Fe(**L***k*)3] <sup>2</sup><sup>+</sup> complexes in CD3CN solutions [35,36].

Moving the methyl group bound to the pyridine ring from the 5-position in **L1** to the 3-position in **L3** and to the 6-position in **L4** (Scheme 2) is well-known to stepwise decrease the ligand-field strengths in the resulting [Fe(**L***k*)3] <sup>2</sup><sup>+</sup> complexes because the operation of additional sterical constraints, produced by intra-strand interactions in [Fe(**L3**)3] <sup>2</sup><sup>+</sup> [37] and by inter-strand interactions in [Fe(**L4**)3] <sup>2</sup>+, extends the Fe–N bond lengths (see Equation (1)) [38–40]. The associated trend Δoct(**L1**) ≈ Δoct(**L2**) > Δoct(**L3**) > Δoct(**L4**) observed for the isostructural [Ni(**L***k*)3] <sup>2</sup><sup>+</sup> complexes (Scheme 2), for which the determination of ligand field Δoct and Racah *B* parameters are not complicated by any SCO behavior, are in line with the observation of pure high-spin configurations for the [Fe(**L3**)3] <sup>2</sup><sup>+</sup> and [Fe(**L4**)3] <sup>2</sup><sup>+</sup> complexes in solution (Scheme 2) [35]. Whereas the connection of methyl groups adjacent to the donor nitrogen atom in the bound 6-methyl-pyridine groups in [Fe(**L4**)3] <sup>2</sup><sup>+</sup> produces such large inter-strand interactions that the

contraction accompanying the high-spin to low-spin transition cannot be envisioned [41], the situation with the remote 3-methyl substituted pyridine units in [Fe(**L3**)3] <sup>2</sup><sup>+</sup> is less clear and a sophisticated triple-stranded heterometallic LaFe helicate containing the facial [Fe(**L3**)3] <sup>2</sup><sup>+</sup> chromophore has been shown to display partial SCO behavior at low temperature (*T*1/<sup>2</sup> ~50 K) [36]. Taking into account that (i) the replacement of a pyridine with a pyrazine ring in going from [Fe(**L1**)3] <sup>2</sup><sup>+</sup> and [Fe(**L2**)3] 2+ stabilizes the low-spin state by *T*1/<sup>2</sup> = 40 K (Scheme 2) and (ii) moving the methyl group from the 5-position in [Fe(**L1**)3] <sup>2</sup><sup>+</sup> to the 3-position in *fac*-[Fe(**L3**)3] <sup>2</sup><sup>+</sup> (as found in the related LaFe helicate) produces an opposite trend with the stabilization of the high-spin form by *T*1/<sup>2</sup> ≈ 50–310 = −260 K [36], we thus ingenuously explored the possibility of combining both aspects in the didentate ligand **L5** where the methyl group is now connected to the 3-position of a pyrazine ring (Scheme 3a) with the hope of pushing the transition temperature toward cryoscopic temperatures for [Fe(**L5**)3] <sup>2</sup><sup>+</sup> around *T*1/<sup>2</sup> ~ *T*1/<sup>2</sup> ([Fe(**L2)**3] <sup>2</sup>+) <sup>−</sup> 260 = 350 <sup>−</sup> 260 = 90 K (3-methyl-pyrazine). This effort is justified by our long-term quest for designing a pseudo-octahedral spin-crossover [Fe(**L***k*)3] <sup>2</sup><sup>+</sup> unit that can modulate the luminescence of adjacent emissive lanthanides in (supra)molecular assemblies via energy transfers within a temperature domain (77–150 K) accessible to optical reading and addressing [36]. Finally, since minor structural variations may induce large changes in ligand-field strength, the systematic exploration of unpredictable intermolecular packing interactions [42,43] operating in crystalline samples of [Fe(**L5**)3]X2 complexes (X- = monoanionic counter-ions) may contribute to the lucky search for some 'ideal' FeII complexes, which additionally exhibit hysteretic behavior and bistability [17,44,45].

**Scheme 3.** (**a**) Chemical structure and synthesis of the didentate ligand **L5** shown in its anti-conformation and (**b**) associated 1H NMR spectrum with numbering scheme (CDCl3, 298 K).

#### **2. Experimental**

Chemicals were purchased from Sigma-Aldrich (Gmbh, Buchs) and Acros and used without further purification unless otherwise stated. Dichloromethane, 1,2-dichloroethane, tert-butylmethyl ether, and *N*,*N*-dimethylformamide were dried through an alumina cartridge. Silica-gel plates (Merck, 60 F254) were used for thin-layer chromatography, SiliaFlash® silica gel P60 (0.04–0.063 mm,) and Acros silica gel 60 (0.035–0.07 mm) were used for preparative column chromatography.

**Preparation of** *N***-methyl-2-nitroaniline (4)**. 1-chloro-2-nitrobenzene (**3**, 31.75 g, 201.5 mmol, 1.0 eq) and methylamine (198 mL, 40% weight in H2O, 2295.6 mmol, 11.4 eq) were introduced into a Carius tube equipped with a magnetic stirrer and heated at 120 ◦C for 48 h. Excess of methylamine was rotatory evaporated and the residual brown oil was partitioned between CH2Cl2 (300 mL) and half sat. aq. NH4Cl (300 mL). The organic layer was separated and the aq. phase was further extracted with CH2Cl2 (3 × 150 mL). The combined organic extracts were dried over anhydrous Na2SO4, filtered, and the solvent evaporated to dryness. The resulting red oil was purified by column chromatography (Silica, CH2Cl2) to give 28.99 g of *N*-methyl-2-nitroaniline (**4**, 190.5 mmol, yield 94%) as a deep red orange oil, which slowly crystallized within hours. 1H NMR (CDCl3, 400 MHz, 298 K) δ/ppm: 8.13 (1H, dd, <sup>3</sup>*J* = 8.8 Hz, <sup>4</sup>*J* = 1.6 Hz), 8.00 (1H, bs), 7.43 (1H, ddd, <sup>3</sup>*J* = 8.8 Hz, <sup>3</sup>*J* = 7.2 Hz, <sup>4</sup>*J* = 1.6 Hz), 6.81 (1H, dd, <sup>3</sup>*J* = 8.6 Hz, <sup>4</sup>*J* = 1.0 Hz), 6.62 (1H, ddd, <sup>3</sup>*J* = 8.4 Hz, <sup>3</sup>*J* = 7.2 Hz, <sup>4</sup>*J* = 1.2 Hz), 2.99 (3H, s).

**Preparation of** *N***,3-dimethyl-***N***-(2-nitrophenyl) pyrazine-2-carboxamide** (**2**, left pathway in Scheme 3a). A suspension of 3-methyl-pyrazine-2-carboxylic acid (**1**, 5 g, 35.5 mmol, 1 eq) and di-isopropyl-ethylamine (7.2 mL, 5.5 g, 45 mmol, 1.26 eq) in CH2Cl2 (4 mL) was added dropwise into a two-necked flask containing isobutyl chloroformate (5.4 mL, 5.408 g, 45 mmol, 1.26 eq) dissolved in CH2Cl2 (1 mL). The mixture was stirred at −15 ◦C for 120 min, after which a solution of *N*-methyl-2-nitroaniline (**4**, 5.5 g, 0.0352 mol, 1 eq) in CH2Cl2 was added. After stirring for 15 h at room temperature, the solution was partitioned between CH2Cl2 (100 mL) and half-sat. aq. NH4Cl (250 mL). The organic layer was separated and the aqueous phase was further extracted with CH2Cl2 (3 × 100 mL). The organic fractions were dried over anhydrous Na2SO4, concentrated under vacuum, and purified by column chromatography (Silica, CH2Cl2/MeOH 99.2:0.8) to yield *N*,3-dimethyl-*N*-(2-nitrophenyl) pyrazine-2-carboxamide (**2**, 8.22 mmol, yield 22%). 1H NMR (CDCl3, 400 MHz, 298 K) δ/ppm: mixture of two rotamers A (72.5%) and B (27.5 %): 2.67 (3H, s, A), 2.72 (3H, s, B), 3.58 (3H, s, A), 3.33 (3H, s, B), 7.37–7.80 (3H, m, A and B), 8.29 (1H, d, <sup>3</sup>*J* = 2.5 Hz, A), 8.60 (1H, d, <sup>3</sup>*J* = 2.5 Hz, B), 8.0 (1H, dd, <sup>3</sup>*J* = 2.5 Hz, <sup>5</sup>*J* = 0.6 Hz, A), 8.50 (1H, dd, <sup>3</sup>*J* = 2.5 Hz, <sup>5</sup>*J* = 0.6 Hz, B), 7.86 (1H, dd, <sup>3</sup>*J* = 8.5 Hz, <sup>4</sup>*J* = 1.5 Hz, A), 8.13 (1H, dd, <sup>3</sup>*J* = 8.5 Hz, <sup>4</sup>*J* = 1.5 Hz, B). ESI-MS (soft-positive mode; MeOH+CHCl3+HCOOH): *m*/*z* =273.0 ([2 + H]+), 295.1 ([2 + Na]+).

**Preparation of** *N***,3-dimethyl-***N***-(2-nitrophenyl) pyrazine-2-carboxamide (2**, right pathway in Scheme 3a). A suspension of 3-methyl-pyrazine-2-carboxylic acid (**1**, 0.2 g, 1.44 mmol, 1 eq) and di-isopropyl-ethylamine (0.37 mL, 0.28 g, 2.16 mmol, 1.5 eq) in 1,2-dichloroethane (4 mL) was added dropwise into a two-necked flask containing trimethylacetyl chloride (0.194 mL, 0.19 g, 1.58 mmol, 1.1 eq) in 1,2-dichloroethane (1 mL). The mixture was stirred at −20 ◦C for 60 min, after which a solution of *N*-methyl-2-nitroaniline (4, 0.329 g, 2.16 mmol, 1.5 eq) in 1,2-dichloroethane (5 mL) was added. After refluxing for 16 h, the solution was concentrated under vacuum and partitioned between CH2Cl2 (80 mL) and half-saturated aqueous solution of NH4Cl (200 mL). The organic layer was separated and the aqueous phase was further extracted using CH2Cl2 (3 × 80 mL). The combined organic fractions were concentrated under vacuum after drying with anhydrous Na2SO4. Ultimate purification using column chromatography (Silica, CH2Cl2/MeOH 99.2:0.8) yielded 0.106 g of *N*, 3-dimethyl-*N*-(2-nitrophenyl) pyrazine-2-carboxamide (**2**, 0.387 mmol, yield 27%).

**Preparation of 1-methyl-2-(3-methylpyrazin-2-yl)-1***H***-benzo[***d***]imidazole (L5)**. *N*,3-dimethyl-*N*-(2-nitrophenyl)pyrazine-2-carboxamide (**2**, 2.56 g, 7.58 mmol, 1 eq) was dissolved in EtOH:DMF (20 mL:25 mL). Sodium dithionite (8.0 g, 40 mmol, 5.2 eq) was added to the mixture and the temperature of the system was raised to 80 ◦C when 20 mL of water was added. After refluxing for 36 h, the mixture was neutralized using aqueous ammonia and the solvents were removed in vacuo. The concentrate was dissolved in CH2Cl2 (50 mL) and washed with water (3 × 200 mL). The aqueous layers were collectively further extracted with CH2Cl2 (3 × 50 mL). The combined organic fraction was then concentrated under vacuum and purified by column chromatography (silica, CH2Cl2/MeOH 98:2) to yield **L5** (7.67 mmol, yield 60%). The compound was crystallized as needles by slow diffusion of n-hexane into a concentrated CH2Cl2 solution of **L5**. 1H NMR (CDCl3, 400 MHz, 298 K) δ/ppm: 2.91 (3H, s), 3.96 (3H, s), 7.35–7.44 (2H, m), 7.49 (1H, d, <sup>3</sup>*J* = 7 Hz), 7.88 (1H, d, <sup>3</sup>*J* = 7 Hz), 8.57 (1H-pz, dd, <sup>3</sup>*J* = 2.4 Hz, <sup>5</sup>*J* = 0.5 Hz), 8.60 (1H-pz, dd, <sup>3</sup>*J* = 2.4 Hz). 13C NMR (CDCl3, 101 MHz, 298K) δ/ppm: 155.68 (Cq), 149.12 (Cq), 144.54 (Cq), 143.84 (CHpz), 142.45 (Cq), 140.86 (CHpz), 136.15 (Cq), 123.85 (CH), 122.90 (CH), 120.45 (CH), 110.01 (CH), 31.83 (CH3), 23.26 (CH3). ESI-MS (soft-positive mode; MeOH+CHCl3+HCOOH): *m*/*z* 225.1([**L5** + H]<sup>+</sup>). Elemental analysis calculated for C13H12N4 (%): C 69.62, H 5.39, N 24.98; Found (%): C 69.46, H 5.09, N 25.20.

**Preparation of mononuclear FeII, ZnII, and NiII complexes with 1-methyl-2-(3-methylpyrazin-2-yl)-1***H***-benzo[***d***]imidazole (L5)**. In a typical synthesis, 0.3 mmol (3 eq) of the ligand **L5** dissolved in acetonitrile (2 mL) was added to 0.1 mmol (1 eq) of Fe(ClO4)2·6H2O or Fe(CF3SO3)2 or Ni(BF4)2·6H2O or Zn(CF3SO3)2 in acetonitrile (2 mL). The resulting mixture was stirred under an inert atmosphere for 3 h, then evaporated to dryness under vacuum to yield microcrystalline powders of the respective complexes. These powders were dissolved in acetonitrile and allowed to crystallize by evaporation or by slow diffusion of tert-butyl methyl ether to give 64–78% of primary [Fe(**L5**)3](CF3SO3)2·1.5H2O, [Ni(**L5**)3](BF4)2·1.5H2O·1.5CH3CN and [Zn(**L5**)3](BF4)2·4H2O complexes (Table S1). Single crystals suitable for characterization by X-ray diffraction could be obtained by slow evaporation of acetonitrile solution containing 10 eq of (*<sup>n</sup>*Bu)4NClO4 or (*<sup>n</sup>*Bu)4NBF4 to give [Fe(**L5**)3](ClO4)2 (I), [Ni(**L5**)3](ClO4)2 (II), [Ni(**L5**)3](BF4)2·H2O (III), [Zn(**L5**)3](ClO4)2 (IV), [Ni(**L5**)3](BF4)2·1.75CH3CN (V) and [Zn(**L5**)3](BF4)2·1.5CH3CN (VI).

**Caution!** Dry perchlorates may explode and should be handled in small quantities and with the necessary precautions [46,47].

#### *2.1. Spectroscopic and Analytical Measurements*

1H and 13C NMR spectra were recorded at 298 K on a Bruker Avance 400 MHz spectrometer. Chemical shifts are given in ppm with respect to tetramethylsilane. Spectrophotometric titrations were performed with a J&M diode array spectrometer (Tidas series) connected to an external computer. In a typical experiment, 25 cm<sup>3</sup> of ligand in acetonitrile (2 <sup>×</sup> <sup>10</sup>−<sup>4</sup> M) was titrated at 298 K with a solution of Fe(CF3SO3)2 or Ni(BF4)2·6H2O or Zn(CF3SO3)2 (2 <sup>×</sup> <sup>10</sup>−<sup>3</sup> M) in acetonitrile under an inert atmosphere. After each addition of 33 μL, the absorbance was recorded using Hellma optrodes (optical path length 0.1 cm) immersed in the thermostated titration vessel and connected to the spectrometer. Mathematical treatment of the spectrophotometric titrations was performed with factor analysis [48–50] and with ReactLabTM Equilibria (previously Specfit/32) [51–53]. Pneumatically-assisted electrospray (ESI-MS) mass spectra were recorded from 10−<sup>4</sup> M (ligands) and 10−<sup>3</sup> M (complexes) solutions on an Applied Biosystems API 150EX LC/MS System equipped with a Turbo Ionspray source. Elemental analyses were performed by K. L. Buchwalder from the Microchemical Laboratory of the University of Geneva. Elemental analysis was not conducted for perchlorate salts for security reasons, while crystals of the tetrafluoroborate salts lost their solvent upon separation from the mother liquor and were not further characterized. Electronic spectra in the UV–Vis region were recorded at 293 K from solutions in CH3CN with a Perkin-Elmer Lambda 1050 using quartz cells of a 0.1 or 1.0 mm path length. Solid-state absorption spectra were recorded with a Perkin-Elmer Lambda 900 using capillaries. Solid-state magnetic data were recorded on a MPMS 3 or MPMS 5 QUANTUM DESIGN magnetometers using magnetic fields of 1000–5000 Oe at 1 K/min rates within the 5–300 K range. The magnetic susceptibilities were corrected for the magnetic response of the sample holder and for the diamagnetism of the compounds by using the approximation <sup>χ</sup><sup>D</sup> = <sup>−</sup> MW <sup>2</sup> · <sup>10</sup>−<sup>6</sup> cm3·mol−<sup>1</sup> [54].

#### *2.2. X-Ray Crystallography*

Summary of crystal data, intensity measurements, and structure refinements for compounds **L5**, [Fe(**L5**)3](ClO4)2 (**I**), [Ni(**L5**)3](ClO4)2 (**II**), [Ni(**L5**)3](BF4)2·H2O (**III**), and [Zn(**L5**)3](ClO4)2 (**IV**) is presented in Tables S2–S4. Pertinent bond lengths, bond angles, and interplanar angles are collected in Tables S5–S14 together with ORTEP views and pertinent numbering schemes gathered in Figures S1–S5. The crystals were mounted on MiTeGen kapton cryoloops with protection oil. X-ray data collection was performed with an Agilent SuperNova Dual diffractometer equipped with a CCD Atlas detector (Cu[Kα] radiation). The structures were solved by using direct methods [55,56] or dual-space methods [57]. Full-matrix least-square refinements on *F*<sup>2</sup> were performed with SHELX2014 [58]. CCDC 1988655-1988659 contained the supplementary crystallographic data. These data can be obtained free of charge from the Cambridge Crystallographic Data Centre via www.ccdc.cam.ac.uk/. Single crystals of [Ni(**L5**)3](BF4)2·1.75CH3CN (**V**) and [Zn(**L5**)3](BF4)2·1.5CH3CN (**VI**) could also be obtained as inversion twins. The two complexes were isostructural and crystallized in the trigonal system (*P*3*c*1 space group) with five independent complexes in the asymmetric unit, all located on three-fold rotation axes (the metal content of the asymmetric unit is 5/3) and *Z* = 10 (Table S15). Although there is no doubt that the three ligands adopt facial arrangements around the metal to give exclusively *fac*-[M(**L5**)3] 2+ cations (Figures S6–S7), we were only able to locate unambiguously three BF4 - counter-anions in the asymmetric unit. Despite numerous efforts, we were not able to obtain a satisfying model for the last third of a BF4 − counter-anion and gave up to further discuss these structures and to deposit the cif files.

#### **3. Results and Discussion**

Synthesis, characterization, and solid-state structures were obtained for the didentate ligand **L5** and its pseudo-octahedral complexes [M(**L5**)3]X2 (M = Fe, Ni, Zn and X = BF4, ClO4). Compared with pyridine-carboxylic acids, which are easily activated via their transformation into acyl chloride with the help of thionyl chloride or oxalyl chloride [59], the electron-rich pyrazine analogue **1** produced only negligible yield (<1%) of the target amide product **2** under these standard conditions [60]. The 3-methylpyrazine-2-carboxylic acid **1** was thus activated as its anhydride through reaction with either isobutyl chloroformate (left path in Scheme 3a) or pivaloyl chloride (right path in Scheme 3a). Subsequent nucleophilic attack with *N*-methyl-2-nitroaniline **4** yielded the ortho-nitroamide compound **2** in moderate yield. A subsequent reductive cyclisation reaction provided ligand **L5**, which was characterized by its 1H-NMR spectrum (Scheme 3b). The lack of NOE effect observed between the methyl groups in positions 5 and 8 indicates an anti-conformation for the α,α'-diimine chelate unit, which was confirmed by (i) the crystal structure of **L5** (Figure 1a) and (ii) gas-phase calculations predicting a global energy minimum for the planar anti-conformation (interplanar angle between the two aromatic rings α = 180◦, Figure 1b) [61]. Given that the same anti-conformations are (i) found in the solid state (Figure S8) and (ii) predicted in the gas phase for the ligands **L2** [60] and **L5**, their computed EHMO frontiers orbitals are comparable (Figure S9a) and lead to akin electronic absorption spectra dominated by intense π\*←π covering the near UV range (Figure S9b).

**Figure 1.** (**a**) Molecular structure of ligand **L5** in its crystal structure highlighting the dihedral angle α = 144.8◦ (α = 180◦ for a planar anti-conformation and α = 0◦ for a planar syn-conformation). (**b**) Gas-phase energy computed for **L5** at the MM2 level as a function of the interplanar angle [61].

Interestingly, the gas-phase energy of **L5** displayed two additional local energy minima for α = ±16◦ (Figure 1b), which shifted from α = 0◦ previously reported for the second local minimum in **L2** (Figure S10) [60]. The larger interplanar angle of 16◦ in the optimized syn-conformation of **L5** was the result of the sterical crowding between the close methyl groups connected to the adjacent aromatic rings (positions 5 and 8 in the numbering of Scheme 3b). Taking the latter conformation as a limiting structural model when **L5** is bound to a metal cation provides *d*N···<sup>N</sup> = 2.83 Å between the two nitrogen donor atoms of the α,α'-diimine chelate. According to Phan et al. [33], the latter separation matches the 2.78 ≤ *d*N···<sup>N</sup> ≤ 2.93 Å range for which a diimine ligand might be used to achieve spin-crossover behavior in tris-homoleptic FeII complexes.

Stoichiometric mixing of **L5** (3 eq.) with Fe(CF3SO3)2, Ni(BF4)2·6H2O or Zn(BF4)2·6H2O (1 eq,) in acetonitrile gave fair yields of microcrystalline primary precipitates of [Fe(**L5**)3](CF3SO3)2·1.5H2O, [Ni(**L5**)3](BF4)2·1.5H2O·1.5CH3CN, and [Zn(**L5**)3](BF4)2·4H2O complexes (Table S1). A series of isostructural complexes [Fe(**L5**)3](ClO4)2 (**I**), [Ni(**L5**)3](ClO4)2 (**II**), and [Zn(**L5**)3](ClO4)2 (**IV**) could be obtained as single crystals by recrystallization in acetonitrile containing 10 eq of (*<sup>n</sup>*Bu)4NClO4.

Monocrystals suitable for x-ray diffractions were also obtained for [Ni(**L5**)3](BF4)2·H2O (**III**), [Ni(**L5**)3](BF4)2·1.75CH3CN (**V**), and [Zn(**L5**)3](BF4)2·1.5CH3CN (**VI**) using the same method except for the replacement of (*<sup>n</sup>*Bu)4NClO4 with (*<sup>n</sup>*Bu)4NBF4 (Tables S3 and S4). The crystal structures of the perchlorate salts systematically displayed the formation of *mer*-[M(**L5**)3] <sup>2</sup><sup>+</sup> cations (Figure 2), in which the [MN6] chromophores adopted a geometry close to the perfect octahedron as ascertained by SHAPE's scores close to zero (Table 1) [62–66].

**Figure 2.** Perspective views of the molecular structures of the [M(**L5**)3] <sup>2</sup><sup>+</sup> cations in the crystal structures of [Fe(**L5**)3](ClO4)2 (**I**), [Ni(**L5**)3](ClO4)2 (**II**), and [Zn(**L5**)3](ClO4)2 (**IV**). The nitrogen atoms of the pyrazine rings bound to the metals are displayed with blue spheres in order to highlight their meridional arrangements around the central cation. Color codes: C = grey, N = blue. Hydrogen atoms and ionic perchlorate counter-anions are omitted for clarity. For [Fe(**L5**)3](ClO4)2 (**I**), the carbon atoms of the methyl groups are shown as green spheres using the Corey–Pauling–Koltun (CPK) model to highlight the intra-strand steric hindrance.

On the contrary, the crystal structures of the tetrafluoroborate salts showed the existence of *fac*-[M(**L5**)3] <sup>2</sup>+, in which the three didentate ligands adopted the same orientation along the pseudo-threefold axis passing through the metal (Figure 3). Having previously established that the energy gap between the facial (*C*3-symmetry) and meridional (*C*1-symmetry) geometries in [Zn(**L1**)3] 2+ and [Zn(**L2**)3] <sup>2</sup><sup>+</sup> roughly followed a pure statistical (i.e., entropic) trend and does not overcome thermal energy at room temperature [60], we concluded that packing forces, specific to the use of perchlorate or tetrafluoroborate counter anions, are more than enough for the quantitative and selective crystallization of pure meridional, respectively facial isomers. For the [Ni(**L5**)3] <sup>2</sup><sup>+</sup> and [Zn(**L5**)3] <sup>2</sup><sup>+</sup> chromophores, the M–N bonds are systematically shorter for the more basic benzimidazole nitrogen donor (Table 1, entry 4; p*K*a(bzim) = 5.68) than with its pyrazine counterpart (Table 1, entry 5; p*K*a(pyrazine) = 0.65) [67], a trend in complete agreement with that reported for the analogous complex [Zn(**L2**)3] <sup>2</sup><sup>+</sup> (Table 1, column 8) [60]. Moreover, the shift of the methyl group bound to the pyrazine ring from the 5-position in **L2** to the 3-position in **L5** has globally no geometric influence on the [ZnN6] coordination sphere, thus leading to Zn–N bond distances surrounding the standard value of Zn–N = 0.74 + 1.46 = 2.20 Å deduced from the effective ionic radii [68]. In other words, the close methyl groups found in the bound didentate ligand **L5** do not induce major intramolecular steric constraints in [Zn(**L5**)3] <sup>2</sup><sup>+</sup> and only a slight increase of the interannular intraligand angles can be detected in going from [Zn(**L2**)3] <sup>2</sup><sup>+</sup> (α =

21(13)◦) to [Zn(**L5**)3] <sup>2</sup><sup>+</sup> (α = 38(4)◦, entry 5 in Table 1). The molecular structures of [Ni(**L5**)3] <sup>2</sup><sup>+</sup> were very similar to those observed for the ZnII analogues (Figures 2 and 3), except for the slightly shorter Ni–N bond distances, a trend in line with the contraction of Shannon's effective ionic radii predicted to be 0.74 Å for six-coordinate Zn2<sup>+</sup> and 0.69 Å for six-coordinate Ni2<sup>+</sup> [68]. The detection of long Fe–N bond distances in [Fe(**L5**)3] <sup>2</sup><sup>+</sup> (*d*(Fe-Nbz) = 2.14(1) Å and *d*(Fe-Npz) = 2.24(3) Å, entries 3–4 in Table 1) is more remarkable since it suggests that the FeII metal center adopts a pure high spin configuration at 180 K as previously found for the analogous 3-methylpyridine-benzimidazole ligand in [Fe(**L3**)3] 2+ (*d*(Fe-Nbz) = 2.14(2) Å and *d*(Fe-Npy) = 2.26(5) Å) [35].

Spin-state, magnetic, and electronic properties of the pseudo-octahedral complexes [M(**L5**)3]X2 (M = Fe, Ni and X = BF4, ClO4) were obtained in the solid state. Molar magnetic susceptibilities (χM), corrected for diamagnetism of solid state samples of [Fe(**L5**)3](ClO4)2 (I) were recorded at variable temperatures in a constant magnetic field of 5000 Oe. The χM*T* versus *T* plot shows a smooth and regular increase of the χM*T* product in the 50–300 K range (red trace in Figure 4), which can be fitted to Equation (6) using a high Curie constant *<sup>C</sup>* <sup>=</sup> 3.5708(9) cm3·K·mol−<sup>1</sup> and a non-negligible temperature independent paramagnetism TIP = 848(5) <sup>×</sup> 10−<sup>6</sup> cm3·mol−1, a behavior diagnostic for a high-spin Fe(II) complex [69].

$$
\chi\_{\rm M}T = \mathbb{C} + T \cdot \text{TIP} \tag{6}
$$

**Figure 3.** Perspective views of the molecular structures of the [M(**L5**)3] <sup>2</sup><sup>+</sup> cations in the crystal structures of [Ni(**L5**)3](BF4)2·H2O (**III**, top left), [Ni(**L5**)3](BF4)2·1.75CH3CN (**V**, top right) and [Zn(**L5**)3](BF4)2·1.5CH3CN (**VI**). The nitrogen atoms of the pyrazine rings bound to the metals are displayed with blue spheres in order to highlight their facial arrangements around the central cation. Color codes: C = grey, N = blue. Hydrogen atoms and ionic tetrafluoroborate counter-anions are omitted for clarity.


**Table 1.** Structural data for hexa-coordinate metallic centers (FeII, NiII or ZnII) found in the crystal structures of [Fe(**L5**)3](ClO4)2 (**I**), [Ni(**L5**)3](ClO4)2 [Ni(**L5**)3](BF4)2·H2O (**III**), [Zn(**L5**)3](ClO4)2 (**IV**), [Ni(**L5**)3](BF4)2·1.75CH3CN ( **V**), and [Zn(**L5**)3](BF4)2·1.5CH3CN (**VI**) at 180 K collecting average bond lengths,interannularintraligandangles(α),interchelateangles(β),chelatebiteangles(γ),andSHAPE's

 the asymmetric unit.

**a**

(**II**),

The additional abrupt decrease in the magnetic susceptibility occurring at low temperature (*T* < 40 K) can be assigned to zero-field splitting (ZFS) of high-spin Fe(II), which was modeled with Equation (7), where *D* and *E* are the axial and rhombic ZFS parameters, respectively [70–73].

$$E\_n^0 = D \cdot \left( S\_\sharp^2 - \frac{S(S+1)}{3} \right) + E \cdot \left( S\_\sharp^2 - S\_\circ^2 \right) \tag{7}$$

The pseudo-threefold axis characterizing the [FeN6] chromophore in [Fe(**L5**)3](ClO4)2 implies that *E* can be neglected (*E* ~ 0). Consequently, the electron–electron interaction splits the *S* = 2 manifold at zero magnetic field into three energy levels located at *E*<sup>0</sup> <sup>1</sup> <sup>=</sup> <sup>−</sup>2*<sup>D</sup>* (*m*<sup>s</sup> <sup>=</sup> 0), *<sup>E</sup>*<sup>0</sup> <sup>2</sup> = −*D* (*m*<sup>s</sup> = ±1), and *E*0 <sup>3</sup> <sup>=</sup> <sup>2</sup>*<sup>D</sup>* (*m*<sup>s</sup> <sup>=</sup> <sup>±</sup>2). Application of the van Vleck Equation (8), where *<sup>k</sup>*<sup>B</sup> <sup>=</sup> 0.695039 cm−1·K−<sup>1</sup> is the Boltzmann constant and *N*<sup>A</sup> is Avogadro number, *E*<sup>1</sup> *<sup>n</sup>* are first-order spin-only Zeeman effects given in Equation (9), where <sup>μ</sup><sup>B</sup> = <sup>−</sup>4.6686 <sup>×</sup> 10−<sup>5</sup> cm−1·G−<sup>1</sup> is the Bohr magneton, and *<sup>E</sup>*<sup>2</sup> *<sup>n</sup>* stands for the second-order Zeeman effects, leads to a satisfying fit (dotted black trace in Figure 4 with agreement factor AF <sup>=</sup> 3.73 <sup>×</sup> <sup>10</sup><sup>−</sup>3) with Land<sup>é</sup> factor *<sup>g</sup>* <sup>=</sup> 2.20(1), *<sup>D</sup>* <sup>=</sup> 0.52(1) cm−<sup>1</sup> and TIP = <sup>−</sup>*N*A*E*<sup>2</sup> *<sup>n</sup>* = 319(4) × <sup>10</sup>−<sup>6</sup> cm3·mol<sup>−</sup>1.

$$\chi\_{\rm M}T = T \cdot N\_{\rm A} \cdot \frac{\sum\_{n} \left[ \left( \frac{{\left( {E\_n^1} \right)^2}{k\_{\rm B} T} - 2E\_n^2 \right) \cdot \exp \left( - \frac{E\_n^0}{k\_{\rm B} T} \right) \right]}{\sum\_{n} \left[ \exp \left( - \frac{E\_n^0}{k\_{\rm B} T} \right) \right]} \tag{8}$$

$$E\_n^1 = -\mathcal{g} \cdot \mu\_{\rm B} \cdot m\_{\rm S} \tag{9}$$

$$\begin{array}{r} \text{4.0} \\ \text{50.5} \\ \text{50.5} \\ \text{20.0} \\ \text{3.0} \\ \text{1} \\ \text{2.5} \\ \text{2.0} \\ \text{0} \end{array} \qquad \begin{array}{r} \\ \text{50.5} \\ \\ \text{1} \\ \text{2.5} \\ \text{2.0} \\ \text{30.0} \\ \text{100.1} \end{array} \qquad \begin{array}{r} \\ \text{1} \\ \text{2.5} \\ \text{30.0} \\ \text{150.0} \\ \text{250.0} \end{array} \qquad \begin{array}{r} \\ \text{1} \\ \text{250.0} \\ \text{30.0} \\ \text{30.0} \\ \text{1} \\ \text{250.0} \end{array}$$

**Figure 4.** χ*MT* versus *T* plot of the molar magnetic susceptibility (χ*M*) between 5–300 K, corrected for diamagnetism recorded for the complex [Fe(**L5**)3](ClO4)2 (**I**, red trace) at 5000 Oe. The dotted black trace shows the best fit (agreement factor AF <sup>=</sup> 3.73 <sup>×</sup> <sup>10</sup><sup>−</sup>3) obtained using Equation (8) with *<sup>g</sup>* <sup>=</sup> 2.20(1), *<sup>D</sup>* <sup>=</sup> 0.52(1) cm−<sup>1</sup> and TIP = <sup>−</sup>*N*A*E*<sup>2</sup> *<sup>n</sup>* <sup>=</sup> 319(4) <sup>×</sup> <sup>10</sup>−<sup>6</sup> cm3·mol<sup>−</sup>1.

The latter magnetic data closely matched those reported for the analogous [Fe(**L3**)3](CF3SO3)2 complex (*g* = 2.20(2), *D* = 0.85(1) cm−<sup>1</sup> [36]), and demonstrate that our novel [Fe(**L5**)3](ClO4)2 complex, in which the 3-methyl-pyridine group of **L3** is replaced with a 3-methyl-pyrazine group in **L5**, is also purely high-spin within the 5–300 K range with no trace of SCO behavior. A careful inspection of the experimental curve around 80 K (Figure 4) showed a very minor deviation from the theoretical model, which could be tentatively assigned to traces of trapped low-spin form as previously reported for *fac*-[Fe(**L3**)3] <sup>2</sup><sup>+</sup> when it is incorporated into a LaFe triple-stranded helicate [36].

The electronic absorption spectrum recorded for [Fe(**L5**)3](ClO4)2 (**I**) in the solid state shows the expected Jahn–Teller split FeII( 5E←5T2) ligand-field transition (Figure 5a) [31]. A deconvolution using two Gaussian functions gives νmax <sup>=</sup> 8881 cm−<sup>1</sup> and 11,887 cm<sup>−</sup>1, thus leading to a barycenter at 10,384 cm<sup>−</sup>1, which provides a direct estimation of ΔHS oct(FeII) <sup>=</sup> 10 Dq, a value only 700 cm−<sup>1</sup> below the minimum of ΔHS oct <sup>≈</sup> 11,000 cm−<sup>1</sup> suggested to be the lower limit for inducing spin state equilibria [31]. However, the latter criterion strongly depends on the choice of the Racah parameter *B*, which is not easily extracted from the single intrashell d–d transition observed in the electronic spectra of high-spin FeII complexes. For this reason, we have recorded the electronic absorption spectrum of the analogous NiII complex [Ni(**L5**)3](BF4)2·H2O (**III**), for which the combination of two spin-allowed d–d transitions Ni(3T2←3A2) and Ni(3T1←3A2) and one spin-forbidden transition Ni(1E←3A2) (Figure 5b) allows a complete characterization of the electronic parameters <sup>Δ</sup>oct(NiII), *<sup>B</sup>*(NiII), and *<sup>C</sup>*(NiII) with the help of Equations (10)–(13) [74,75].

$$E\{^3\mathbf{A}\_{2\mathbf{g}}\} = 0\tag{10}$$

$$E\left(^{1}\mathrm{E\_{g}}\right) = 8B + 2\mathrm{C} - \frac{6B^{2}}{\Delta\_{\mathrm{oct}}}\tag{11}$$

$$E\left(^{3}\mathbf{T}\_{2\mathbf{g}}\right) = \Delta\_{\text{tot}}\tag{12}$$

$$E\left(^{3}\mathrm{T}\_{1\mathrm{g}}\right) = 1.5\Delta\_{\mathrm{oct}} + 7.5B - 0.5\sqrt{225B^2 + \Delta\_{\mathrm{oct}}^2 - 18\Delta\_{\mathrm{oct}}B} \tag{13}$$

**Figure 5.** Electronic absorption spectra recorded at 298 K for (**a**) [Fe(**L5**)3](ClO4)2 (**I**) in the solid state and (**b**) [Ni(**L5**)3](BF4)2·H2O (**III**) in the solid state (black trace, left axis) and in acetonitrile solution (0.1 M, red trace, right axis).

A Gaussian deconvolution of the visible part of the absorption spectrum into three peaks yielded two broad bands, diagnostic for the spin-allowed, but parity-forbidden, transitions at 10,672 cm−<sup>1</sup> ( 3T2←3A2) and 16,763 cm−<sup>1</sup> ( 3T1←3A2; Table S16), together with a third weaker band at 12,584 cm−1, which can be ascribed to the spin-forbidden 1E←3A2 component (Figure 5b). Subsequent non-linear least-squares fits of the energies of these transitions with Equations (10)–(13) provides a first rough set of ligand field strength Δoct = 10,672 cm−<sup>1</sup> and Racah parameters *B* = 760 cm−<sup>1</sup> and *C* = 3413 cm−<sup>1</sup> (Table S16). However, the mixing of the spin-allowed 3T2←3A2 transition with the spin-forbidden 1E←3A2 transition via spin-orbit coupling for apparent ligand field strengths around 11,000–12,000 cm−1, as found for [Ni(**L5**)3](BF4)2·H2O, requires further refinements [76]. A detailed analysis of a series of NiII complexes led *Hancock* and coworkers to propose three empirical Equations (14)–(16) to obtain more reliable ligand field strengths Δoct and Racah parameters *B* and *C* in cm−<sup>1</sup> units (ε<sup>1</sup> and

<sup>ε</sup><sup>2</sup> are the extinction coefficients at the observed frequencies of the 1E←3A2 transition and 3T2←3A2 transition, respectively) [76]. The analysis of the experimental absorption spectra using this model gives the corrected parameters gathered in Table 2 for the [Ni(**L5**)3] <sup>2</sup><sup>+</sup> and [Ni(**L2**)3] <sup>2</sup><sup>+</sup> chromophores.

$$
\Delta\_{\rm oct} = 10 \text{Dq} = 10630 + 1370(\varepsilon\_1/\varepsilon\_2) \tag{14}
$$

$$B = 1120 - 0.022 \cdot \Delta\_{\text{oct}} \tag{15}$$

$$C = 15B - 9975 \tag{16}$$

**Table 2.** Ligand-field strengths (oct) and Racah parameters (*B*, *C*) computed with Equations (14)–(16) for [Ni(**L5**)3](BF4)2·H2O (**III**) in the solid state and in 0.1 M acetonitrile solution at 298 K.


*<sup>a</sup>* Nephelauxetic parameter β = *B*/*B*<sup>o</sup> using *B*<sup>o</sup> = 1042 cm−<sup>1</sup> for free Ni2<sup>+</sup> ion [77].

The refined Δoct, *B* and *C* parameters computed for [Ni(**L5**)3] <sup>2</sup><sup>+</sup> (Table 2) almost exactly matched those previously reported for the unconstrained [Ni(**L2**)3] <sup>2</sup><sup>+</sup> complex, for which the related [Fe(**L2**)3] 2+ complex displayed spin-crossover behavior above room temperature (*T*1/<sup>2</sup> ~400 K in the solid state, *T*1/<sup>2</sup> ~ 350 K in acetonitrile solution [35]). Moreover Δoct ([Ni(**L5**)3] <sup>2</sup>+) = 11,630 cm−<sup>1</sup> is compatible with the ligand field range 11,200 <sup>≤</sup> <sup>Δ</sup>oct (NiII) <sup>≤</sup> 12,400 cm−<sup>1</sup> established by Busch and coworkers [78] as a reliable and useful benchmark for predicting and rationalizing the spin-crossover of the related FeII complexes [40]. The absence of SCO behavior depicted by [Fe(**L5**)3](ClO4)2 is thus difficult to assign to some inadequate electronic properties of the [FeN6] chromophore, but more probably to the impossibility of the coordination sphere to shrink for adopting short-enough Fe–N bonds compatible with low-spin FeII. This pure sterical limitation can be tentatively assigned to the intraligand sterical constraints programmed to occur between the methyl groups bound to the pyrazine and benzimidazole rings in each coordinated syn-**L5** ligand in [Fe(**L5**)3] <sup>2</sup>+. However, packing forces operating in the solid state may be as important, or even much larger than intramolecular constraints and a definitive assessment requires the extension of our analysis to isolated complexes in solution, where intermolecular interactions are significantly reduced.

Stabilities and electronic properties of the pseudo-octahedral complexes [M(**L5**)3]X2 (M = Fe, Ni, Zn and X = BF4, CF3SO3) were obtained in acetonitrile solutions. Following the procedure previously detailed for analogous [Zn(**L***k*)3] <sup>2</sup><sup>+</sup> [60] and [Fe(**L***k*)3] <sup>2</sup><sup>+</sup> [35] with the didentate ligands **L1** and **L2**, spectrophotometric titrations of submillimolar concentrations of **L5** with M(CF3SO3)2 (M = Ni, Zn) in dry acetonitrile (Figure 6a,b and Figure S11a,b) showed the successive formation of two absorbing complexes [M(**L5**)*n*] <sup>2</sup><sup>+</sup> (*n* = 2, 1; equilibria (17)–(18)) as ascertained by their independent eigenvectors found in the factor analyses (Figure 6c and Figures S11c–S12c) [48–50] and their satisfying re-constructed absorption spectra (Figure 6d and Figure S11d) [51–53].

$$\left[\text{M}^{2+} + \text{L5} \; \left[\text{M}(\text{L5})\right]^{2+} \beta\_{1,1}^{\text{M,L5}}\right] \tag{17}$$

$$\text{M}^{2+} + 2 \text{ L5 } \left[ \text{M}(\text{L5})\_2 \right]^{2+} \beta\_{1,2}^{\text{M,L5}} \tag{18}$$

$$\text{M}^{2+} + 3 \text{ L5 } \left[ \text{M}(\text{L5})\_3 \right]^{2+} \beta\_{1,3}^{\text{M,L5}} \tag{19}$$

The 4000 cm−1 red-shift of the ligand-centered π\*←π transition observed upon complexation to MII (Figure 6a and Figure S11a) is diagnostic for the anti→syn conformational change of the <sup>α</sup>,α'diimine unit accompanying the coordination of **L5** to MII [79,80]. Non-linear least-square fits [51–53] of the spectrophotometric data to equilibria (17)–(18) provide the macroscopic cumulative formation constants gathered in Table 3 (entries 3–4) together with speciation curves [81] showing a maximum formation of ca. 50% of the ligand speciation under the form of [M(**L5**)2] <sup>2</sup><sup>+</sup> at submillimolar concentrations (Figure 6e and Figure S11e). Attempts to consider the formation of an additional [M(**L5**)3] <sup>2</sup><sup>+</sup> complex according to equilibrium (19) only failed in our hands, which suggests that the βM,**L**<sup>5</sup> 1,3 constant is too low for providing significant quantities of the latter complex at this concentration.

**Figure 6.** (**a**) Variation of absorption spectra and (**b**) corresponding variation of observed molar extinctions at different wavelengths recorded for the spectrophotometric titration of **L5** with Zn(CF3SO3)2 (total ligand concentration: 2.0 <sup>×</sup> <sup>10</sup>−<sup>4</sup> mol.dm−<sup>3</sup> in acetonitrile, 298 K). (**c**) Evolving factor analysis using four absorbing eigenvectors [48–50], (**d**) re-constructed individual electronic absorption spectra [51–53] and (**e**) associated computed speciation [81].

**Table 3.** Cumulative stability constants (log <sup>β</sup>M,**L***<sup>k</sup>* 1,*n* , Equations (17)–(19), intrinsic affinities (Δ*G*M,**L***<sup>k</sup>* = <sup>−</sup>*RT* ln *f* M,**L***<sup>k</sup>* ) and global interligand interactions (Δ*E***L***k*,**L***<sup>k</sup>* <sup>=</sup> <sup>−</sup>*RT* ln *u***L***k*,**L***<sup>k</sup>* ) estimated in acetonitrile at 298 K.


*<sup>a</sup>* Computed with Equation (22).

According to the site-binding model [82,83], the first stability constant <sup>β</sup>M,**L***<sup>k</sup>* 1,1 reflects the simple intermolecular affinity *f* M,**L***<sup>k</sup>* between the didentate ligand **L***k* and the entering MII cation (including the change in solvation, Equation (20)) modulated by a pure entropic contribution ω1,1 = 24 [60] produced by the change in rotational entropies accompanying the transformation of the reactants into products, a parameter often referred to as the statistical factor [84,85].

$$
\beta\_{1,1}^{\text{M,Lk}} = 24^{\text{M,Lk}} \tag{20}
$$

Applying Equation (20) to the stability constants <sup>β</sup>M,**L***<sup>k</sup>* 1,1 collected in Table 3 (entry 3) leads to intrinsic free energy affinities <sup>Δ</sup>*G*M,**L***<sup>k</sup>* <sup>=</sup> <sup>−</sup>*RT* ln *f* M,**L***<sup>k</sup>* <sup>=</sup> <sup>−</sup>*RT* ln <sup>β</sup>M,**L***<sup>k</sup>* 1,1 /24 (entry 6), which are roughly reduced by a factor two in going from the 5-methyl-pyrazine ligand **L2** to the 3-methyl-pyrazine analogue **L5**, as illustrated for [Zn(**L2**)]2<sup>+</sup> (Δ*G*Zn,**L2** = <sup>−</sup>31.1(3) kJ·mol−1) and [Zn(**L5**)]2<sup>+</sup> (Δ*G*Zn,**L5** = <sup>−</sup>17.1(7) kJ·mol<sup>−</sup>1).

The fixation of two ligands to give [M(**L***k*)2] <sup>2</sup><sup>+</sup> obeying equilibrium (18) requires twice the intermolecular metal-ligand affinity, a statistical factor of ω1,2 = 120, which takes into account all the possible geometric isomers [60] and the operation of allosteric cooperativity factors *u***L***k*,**L***<sup>k</sup>* measuring the extra energy cost (*u***L***k*,**L***<sup>k</sup>* < 1), respectively, energy benefit (*u***L***k*,**L***<sup>k</sup>* > 1) produced by the binding of two ligands to the same metal (Equation (21)) [82,83,86].

$$
\beta\_{1,2}^{\text{M,Lk}} = 120 \left( f^{\text{M,Lk}} \right)^2 \mu^{\text{Lk,Lk}} \tag{21}
$$

Applying Equation (21) to the stability constants <sup>β</sup>M,**L***<sup>k</sup>* 1,2 (entry 4 in Table 3) with the help of the intrinsic affinities *<sup>f</sup>* M,**L***<sup>k</sup>* <sup>=</sup> exp −Δ*G*M,**L***k*/*RT* deduced from Δ*G*M,**L***<sup>k</sup>* (entry 6 in Table 3) provides interligand free energies interactions <sup>Δ</sup>*E***L***k*,**L***<sup>k</sup>* <sup>=</sup> <sup>−</sup>*RT* ln *u***L***k*,**L***<sup>k</sup>* <sup>=</sup> <sup>−</sup>*RT* ln <sup>β</sup>M,**L***<sup>k</sup>* 1,2 /120 <sup>−</sup> <sup>2</sup>Δ*G*M,**L***<sup>k</sup>* (entry 7 in Table 3) close to zero (non-cooperativity) for [M(**L2**)2] <sup>2</sup><sup>+</sup> complexes, but negative (positive cooperativity) for [M(**L5**)2] <sup>2</sup>+. However, the neglect of the expected [M(**L5**)3] <sup>2</sup><sup>+</sup> for modeling the spectrophotometric titrations results in a slight overestimation of the second cumulative constant <sup>β</sup>M,**L***<sup>k</sup>* 1,2 , fully compatible with the apparent, but probably not pertinent, minor positive cooperativity observed for the successive binding of **L5** ligands to M2<sup>+</sup> centers. Finally, the introduction of the estimated intrinsic affinities and interligand interactions into Equation (22) allows some predictions concerning the inaccessible third cumulative stability constants [35].

$$
\beta\_{1,3}^{\text{M,Lk}} = 64 \left( f^{\text{M,Lk}} \right)^3 \left( \mu^{\text{Lk,Lk}} \right)^3 \tag{22}
$$

*Chemistry* **2020**, *2*

The resulting values of log βM,**L5** 1,3 ≈ 12.0 (Table 3, entry 5) correspond to a reduction by five orders of magnitude with respect to log βM,**L2** 1,3 ≈ 17.0 experimentally found with the less constrained ligand **L2**. There is no doubt that the hindered planar arrangement of the two aromatic heterocycles in **L5** has a deep impact on the strength of the M–N bonds because of the misalignment of the nitrogen lone pairs with metal d-orbitals in pseudo-octahedral geometry. Nevertheless, at 1 M concentration of **L5** in acetonitrile, the ligand speciation curve computed [81] by using βM,**L5** 1,*<sup>n</sup>* (*n* = 1–3) gathered in Table 3 shows that [M(**L5**)3] <sup>2</sup><sup>+</sup> (M = Ni, Zn) corresponds to more than 90% of the distribution at the stoichiometric M:**L5** = 1/3 ratio (Figure S12). At a total ligand concentration of 0.1 M in acetonitrile, [Ni(**L5**)3] <sup>2</sup><sup>+</sup> stands for 84% of the ligand speciation and its absorption spectrum closely matches that recorded for related solid state samples (Figure 5). Repeating the detailed analysis described in the previous section (Equations (10)–(16)) provides ligand-field strengths (Δoct) and Racah parameters (*B*, *C*) similar to those found in the solid state (Table 2), especially for the crucial ratio Δoct/*B* = 13.3, which is identical for [Ni(**L2**)3] <sup>2</sup><sup>+</sup> and [Ni(**L5**)3] <sup>2</sup>+, both in solution and in the solid state. In other words, the moving of the methyl group connected to the pyrazine ring from the 5-position in **L2** to the 3-position in **L5** indeed strongly reduces the affinity of the didentate ligand for Ni(II) in solution, but only has a weak effect on the ligand field strengths and on the Ni–N (or Zn–N) bond length. Interestingly, the variable-temperature 1H NMR spectra recorded for diamagnetic [Zn(**L5**)3] <sup>2</sup><sup>+</sup> (>80% for a total concentration of 0.1 M in CD3CN in the 233–333 K range, Figure S13) showed a single set of signals compatible with the exclusive formation of an averaged *C*3-symmetrical species with no contribution from either blocked facial and meridional isomers or from partial decomplexation to give [Zn(**L5**)2] <sup>2</sup><sup>+</sup> + **L5** (equilibrium (19)). These observations are in contrast with the detection at low temperature in CD3CN of two well-resolved spectra characteristics of a slow exchange operating between *fac*-[Zn(**L2**)3] <sup>2</sup><sup>+</sup> and *mer*-[Zn(**L2**)3] <sup>2</sup><sup>+</sup> [60], and suggests that the weaker stability constants are accompanied by faster ligand exchange processes around Zn2<sup>+</sup> in [Zn(**L5**)3] <sup>2</sup>+. This decrease in affinity reaches its paroxysm for the coordination of Fe2<sup>+</sup> since the spectrophotometric titration of **L5** with Fe(CF3SO3)2 conducted at submillimolar concentration displays only a minor drift of the absorption spectra with no pronounced end point (Figure 7).

**Figure 7.** (**a**) Variation of absorption spectra and (**b**) corresponding variation of observed molar extinctions at different wavelengths recorded for the spectrophotometric titration of **L5** with Fe(CF3SO3)2 (total ligand concentration: 2.0·10−<sup>4</sup> mol.dm−<sup>3</sup> in acetonitrile, 298 K.

Attempts to model these limited variations within the frame of equilibria (17)–(18) only failed. The amount of [Fe(**L5**)*n*] <sup>2</sup><sup>+</sup> in solution is strongly limited by (very) low cumulative stability constants, a situation produced by the impossibility for FeII to adopt a compact low-spin configuration in the sterically constrained complex [35]. In the absence of a significant amount of [Fe(**L5**)3] <sup>2</sup><sup>+</sup> complexes in solution, no spin state equilibria could be investigated for this system.

#### **4. Conclusions**

Having established that the shift of the methyl group connected to the pyridine ring in going from the didentate ligand **L1** (5-position) to **L3** (3-position) was accompanied by a drift of the transition temperature in the associated spin-crossover complexes from *T*1/<sup>2</sup> ~310 K in [Fe(**L1**)3] <sup>2</sup><sup>+</sup> [34] to *T*1/<sup>2</sup> ~50 K in a pure facial version of the [Fe(**L3**)3]2+ chromophore [36], we attempted in this work to transpose this trend for pyrazine analogues **L2** and **L5** with the preparation of the missing member of the series [Fe(**L5**)3] <sup>2</sup><sup>+</sup> (3-methyl-pyrazine), for which an ideal *T*1/<sup>2</sup> ~ 90 K could be naively predicted since *T*1/<sup>2</sup> ~ 350 K in [Fe(**L2**)3]2+ (5-methyl-pyrazine). At first sight, this approach appeared to be promising since the molecular structures of pseudo-octahedral [M(**L2**)3] <sup>2</sup><sup>+</sup> and [M(**L5**)3] <sup>2</sup><sup>+</sup> (M = NiII and ZnII) were comparable, except for the expected larger interannular twist between the connected aromatic rings produced by the close methyl groups in bound **L5**. The diagnostic ratio Δ/*B* = 13.3 was identical for both NiII complexes, thus pointing to electronic properties also compatible with the induction of SCO behavior in the analogous FeII complexes. Surprisingly and disappointingly, [Fe(**L5**)3] <sup>2</sup><sup>+</sup> exists as a pure high spin complex within the 5–300 K range with no trace of spin state equilibrium. A thorough analysis of its thermodynamic formation in solution highlights a huge decrease in affinity of the ligand **L5**, compared with **L2**, for its binding to M2<sup>+</sup> cations despite the presence of the same nitrogen donor atoms. Compared with the pyridine analogues [M(**L3**)3] <sup>2</sup>+, which possess similar intra-strand sterical constraints (3-methyl substituents), the much weaker σ-donating N-pyrazine donor atoms were unable to compensate for the additional interstrand constraints required for chelating **L5** around M2+. This limiting factor, which can be compared to a sort of arm wrestling match, is amplified with small cations and low-spin FeII cannot be complexed to **L5**.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2624-8549/2/2/15/s1. Tables S1–S16 collecting elemental analyses, crystallographic, and photophysical data. Figures S1–S13 showing the crystal structures, 1H NMR spectra, spectrophotometric titrations, and theoretical calculations.

**Author Contributions:** Conceptualization, T.L. and C.P.; Methodology and practical chemical and spectroscopic studies, N.D. and T.L.; Crystallography, L.G. and C.B.; Writing draft report, N.D.; Writing of manuscript and editing, C.P., N.D., T.L., and C.B.; Project administration and funding acquisition C.P. All authors have read and agreed to the published version of the manuscript.

**Funding:** Financial support from the Swiss National Science Foundation is gratefully acknowledged (grant number: 200020\_178758).

**Acknowledgments:** The authors thank Kerry Buchwalder for performing the elemental analysis and for efficient technical support.

**Conflicts of Interest:** The authors declare no conflicts of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Solid Phase Nitrosylation of Enantiomeric Cobalt(II) Complexes**

**Mads Sondrup Møller, Morten Czochara Liljedahl, Vickie McKee and Christine J. McKenzie \***

Department of Physics, Chemistry and Pharmacy, University of Southern Denmark, Campusvej 55, 5230 Odense M, Denmark; madssm@sdu.dk (M.S.M.); mlilj18@student.sdu.dk (M.C.L.); mckee@sdu.dk (V.M.) **\*** Correspondence: mckenzie@sdu.dk

**Abstract:** Accompanied by a change in color from red to black, the enantiomorphic phases of the cobalt complexes of a chiral salen ligand (L2−, Co(L)·CS2, and Co(L) (L = L*S,S* or L*R,R*)) chemisorb NO (g) at atmospheric pressure and rt over hours for the CS2 solvated phase, and within seconds for the desolvated phase. NO is installed as an axial nitrosyl ligand. Aligned but unconnected voids in the CS2 desorbed Co(L*R,R*)·CS2 structure indicate conduits for the directional desorption of CS2 and reversible sorption of NO, which occur without loss of crystallinity. Vibrational circular dichroism (VCD) spectra have been recorded for both hands of LH2, Zn(L), Co(L)·CS2, Co(L), Co(NO)(L), and Co(NO)(L)·CS2, revealing significant differences between the solution-state and solid-state spectra. Chiral induction enables the detection of the νNO band in both condensed states, and surprisingly also the achiral lattice solvent (CS2 (νCS at 1514 cm<sup>−</sup>1)) in the solid-state VCD. Solution-state spectra of the paramagnetic Co(II) complex shows a nearly 10-fold enhancement and more extensive inversion of polarity of the vibrations of dominant VCD bands compared to the spectra of the diamagnetic compounds. This enhancement is less pronounced when there are fewer polarity inversions in the solid state VCD spectra.

**Keywords:** chirality; vibrational circular dichroism; solid–gas reaction; chemisorption; nitrosyl

#### **1. Introduction**

NO is biologically important, but also a highly toxic gas, and materials for its sorptive removal from exhausts are of great interest. A number of diverse materials, from metal oxides, zeolites, and metal-organic frameworks have shown the ability to selectively sorb NO, in some cases reversibly [1–6]. Chemisorptive processes can result in the transformation of the sorbed NO into less toxic compounds and precedence for this was demonstrated recently using the crystalline solid-state of a dicobalt(II) complex, which co-chemisorbs NO and O2. Accompanied by metal oxidation, their conversion to a coordinated nitrite and nitrate counter anion ensues through a series of unidentified in-crystal reactions [7].

We were interested in investigating the chemisorptive reactivity of NO by an isolated mononuclear Co(II) site inside the crystalline lattice of a molecular solid in efforts to understand the mechanism of the host–guest, solid–gas chemistry of the aforementioned molecular dicobalt(II) complexes [7]. In particular, we wanted to investigate whether the in-crystal conversion of NO and O2 to NO2 − and NO3 - is dependent on the presence of two closely-located cobalt ions working cooperatively to activate these guest substrates. Mononuclear Co(salen) has been shown to react with NO in both solution- [8] and solidstates [8,9], and offers the opportunity to explore this possibility. Unfortunately, however, no details of the structural or spectroscopic changes in the solids have been reported for the solid–gas reaction. It is unknown whether or not well-defined pores to allow transport through the solid are requisite, whether or not the process is true chemisorption (where bonds are formed), whether the presumed in-solid NO coordination to Co(II) results in its oxidation, and whether the reaction is reversible. We have, therefore, reinvestigated the NO gas–solid reaction for the salen system, however, the parent scaffold has been replaced with *N,N*- -bis(3,5-di-t-butylsalicylidene)-1,2-cyclohexanediamine (L2<sup>−</sup>), the derivative used

**Citation:** Møller, M.S.; Liljedahl, M.C.; McKee, V.; McKenzie, C.J. Solid Phase Nitrosylation of Enantiomeric Cobalt(II) Complexes. *Chemistry* **2021**, *3*, 585–597. https://doi.org/10.3390/ chemistry3020041

Academic Editors: Catherine Housecroft and Katharina M. Fromm

Received: 31 March 2021 Accepted: 26 April 2021 Published: 28 April 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

in the manganese(III) chloride complex that is also known as 'Jacobsen's catalyst' [10], and is used for catalyzing asymmetric epoxidation reactions [10–12]. A survey of the Cambridge structural database [13] of the crystal structures of metal complexes of L2<sup>−</sup> suggests that the peripheral bulky tert-butyl groups of L2<sup>−</sup> ensure that mononuclear Co sites are isolated from each other by preventing the formation of crystal phases containing dimeric (M(salen))2 [14–18], which form when a phenolato oxygen atom of each salen on adjacent complexes bridge between the two metal ions. Another advantage is that, unlike some phases of the parent Co(salen) and its derivatives, Co(L) does not bind O2.

Another aspect of this work is the fact that L2<sup>−</sup> provides chirality by virtue of the aliphatic backbone carbon atoms to give L*S,S* and L*R,R* (Scheme 1). With respect to the use of L2<sup>−</sup> for constructing enantiopure complexes to catalyze asymmetric reactions, we wished also to learn whether a chiral ligand would induce chirality into an achiral axial co-ligand, since this is a proxy for a bound substrate.

**Scheme 1.** *S*,*S* and *R*,*R* conformations of CoII(L*S*,*S*) and CoII(L*R*,*R*).

#### **2. Materials and Methods**

**Caution** NO is a highly toxic gas. CS2 vapors are also toxic.

#### *2.1. Instrumentation*

Mass spectra are recorded with electrospray ionization (ESI) on a Bruker micrOTOF-Q II spectrometer (nanospray, capillary temperature = 180 ◦C, spray voltage = 3.7 kV). UV-vis spectra were recorded on an Agilent 8453 spectrophotometer in 1 cm quartz cuvettes. VCD and IR spectra were recorded on a CHIRALIR-2XTM spectrometer equipped with a single PEM, a resolution of 4 cm−1, optimized at 1400 cm−1, and in a single block with 50000 scans. All VCD spectra are baseline corrected by half differentiating from the other enantiomer, i.e., subtracting the VCD spectrum of one enantiomer from that of the other enantiomer and dividing the intensity of the resulting spectrum by two, and vice versa. All solution state VCD spectra were recorded in CDCl3 using an ICLSL-4 liquid cell with BaF2 windows and a path length of 75 μm. Solutions were prepared by dissolving 30 mg sample in 0.3 mL CDCl3. All solid phase VCD spectra were recorded as mulls applied between two circular BaF2 windows (Ø25 mm × 4 mm), and the sample holder was rotated at a constant speed throughout the recording to reduce artifacts. Mulls were prepared by grinding, with an agate mortar and pestle, 20 mg sample with 50 μL Nujol oil. Powder X-ray diffraction data were collected on a Synergy, Dualflex, AtlasS2 diffractometer using CuKα radiation (λ = 1.54184 Å) and the CrysAlis PRO 1.171.40.67a suite [19]. Powdered samples were adhered to the mounting loop using Fomblin®Y, and diffractograms were recorded with a detector distance of 120 mm, using Gandolfi scans with a single kappa setting, and an exposure time of 200 s.

#### *2.2. Single Crystal X-ray Diffraction*

The crystals used for Single Crystal X-ray Diffraction (SCXRD) were taken directly from the mother liquor and mounted using Fomblin®Y to adhere the crystal to the mounting loop. X-ray crystal diffraction data were collected at 100(2) K on a Synergy, Dualflex, AtlasS2 diffractometer using CuKα radiation (λ = 1.54184 Å) and the CrysAlis PRO 1.171.40.67a suite [19], and corrected for Lorentz-polarization effects and absorption. Using SHELXLE [20], the structure was solved by dual space methods (SHELXT [21]) and refined on F2 using all the reflections (SHELXL-2018 [22]). All the non-hydrogen atoms were refined using anisotropic atomic displacement parameters, and hydrogen atoms were inserted at calculated positions using a riding model. Parameters for data collection and refinement are summarized in Table 1. The chirality in these complexes is provided by the cyclohexane ring in the ligand backbone, however, the remainder of the structures, including the heavier cobalt and (for Co(L*R,R*)·CS2 and Co(NO)(L*R,R*)·CS2) sulfur atoms, are arranged almost centrosymmetrically. As a result, the |E2 − 1| statistics, cumulative intensity distribution plots, and Wilson plots (SI Figures S6–S8) appear to favor a centrosymmetric structure (as does the initial estimate of the Flack x parameter in SHELXT for Co(NO)(L*R,R*)·CS2), and Platon ADDSYM in checkCIF suggests a (pseudo) center of symmetry may be present. Nevertheless, all three structures were successfully refined in the chiral space group *P*21, showing the expected chair conformation of the cyclohexane rings, all with *R,R* chirality. When carbon disulfide is desorbed from Co(L*R,R*)·CS2 to produce Co(L*R,R*), the crystal quality is reduced slightly, resulting in weak high angle data. The structure of Co(L*R,R*), therefore, contains more uncertainties.

#### *2.3. Computational Details*

The VCD and IR calculations considered in this work were performed using (unrestricted) density functional theory (DFT) calculations performed using Jaguar [23] through the Maestro graphical interface. [24] All structures have been geometry optimized in isolation using B3LYP/LACVP\*\* [25,26], and vibrational frequencies have also been calculated at this level of theory. All calculations are solution gas phase calculations. For L*R,R*H2, different conformations were calculated using the OPLS2005 [27] forcefield, and the VCD and IR spectra were based on a Boltzmann average of these conformations. For Zn(L*R,R*), the VCD and IR spectra are based on only the most stable conformation.

#### *2.4. Synthesis*

The*R,R*and *S,S* enantiomers of *N,N*- -bis(3,5-di-t-butylsalicylidene)-1,2-cyclohexanediamine, L*R,R*H2 and L*S,S*H2, were prepared according to literature methods with a reported ee. of >95% [28]. The *R,R* and *S,S* enantiomers of *N,N*- -bis(3,5-di-tert-butylsalicylidene)-1,2 cyclohexanediaminozinc(II) (Zn(L*R,R*) and Zn(L*S,S*)) [29] and*N,N*- -bis(3,5-di-tert-butylsalicylidene)- 1,2-cyclohexanediaminocobalt(II) (Co(L*RR*) and Co(L*SS*)) [30] were also synthesized according to literature procedure. All other chemicals were purchased from Sigma-Aldrich and were used without further purification.

## 2.4.1. Co(LR,R)·CS2 and Co(LS,S)·CS2

A degassed solution of Co(L*RR*) or Co(L*SS*) (0.250 g) in CS2:*n*-hexane (16 mL, 1:1 *v*/*v*) was placed in a vial which was then placed inside a large glass jar with an air tight lid, and the solution was allowed to slowly evaporate at 5 ◦C in the fridge, yielding ruby red block crystals suitable for single crystal X-ray diffraction. Yield 0.227 g, 80.6%.

IR (Nujol) cm<sup>−</sup>1: 1609 (m, C=C), 1591 (m, C=N), 1524 (vs, C=S), 1254 (m, C–O).

ESI-MS (pos. mode, MeCN): found (calcd) *m*/*z* = 603.217 (603.34, [Co(LR,R)]+, C36H52CoN2O2 100%).

## 2.4.2. Co(LR,R) and Co(LS,S) (Desolvated from Co(LR,R)·CS2 and Co(LS,S)·CS2)

A sample of Co(L*R,R*)·CS2 or Co(L*S,S*)·CS2 (0.25 g, size 0.05–1 mm3) was placed inside a 10 mL round bottom flask, and the flask was attached to a rotary evaporator. The crystals were heated to 95 ◦C in vacuo (~10−<sup>2</sup> mbar) and rotated for one hour, resulting in a color change from translucent dark red to opaque red/orange. Co(L*R,R*) and Co(L*S,S*) were isolated in a quantitative yield.

IR (Nujol) cm<sup>−</sup>1: 1608 (s, C=C), 1596 (vs, C=N), 1524 (m, C=N), 1255 (m, C–O). ESI-MS (pos. mode, MeCN): found (calcd) *m*/*z* = 603.217 (603.34, [Co(L*R,R*)]+, C36H52CoN2O2 100%).

UV-vis (DCM) λmax/nm (ε/M−<sup>1</sup> cm<sup>−</sup>1): 247 (36,012), 428 (11,446).

#### 2.4.3. Co(NO)(LR,R) and Co(NO)(LS,S)

Crystals of Co(L*R,R*) or Co(L*S,S*) (0.25 g, size 0.05–1 mm3) were placed in a Schlenk tube (30 mL), and the Schlenk line was evacuated and filled with N2 (3 cycles). NO (1.2 bar) was then admitted into the system, resulting in a color change from opaque red/orange to opaque black within seconds, and the closed tube was then allowed to stand for one hour. Before opening to air, the system was evacuated and flushed with N2 (3 cycles). Co(NO)(L*R,R*) and Co(NO)(L*S,S*) were isolated in a quantitative yield. Recrystallization of Co(NO)(L*R,R*) (0.25 g) from CS2:*n*-hexane (8 mL, 1:1 *v*/*v*) via the same procedure as described above for Co(L*R,R*)·CS2 yielded black needle crystals of Co(NO)(L*R,R*)·CS2 suitable for single crystal X-ray diffraction.

IR (Nujol) cm−1: 1659 (w, N=O) 1638 (vs, C=N), 1608 (s, C=C,), 1524 (w, C=N), 1255 (w, C–O).

ESI-MS (pos. mode, MeCN): found (calcd) *m*/*z* = 603.317 (603.34, [Co(L*R,R*)]+, C36H52CoN2O2, 100%), 634.320 (634.34, [Co(NO)(L*R,R*)H]+, C36H53CoN3O3, 4.98%)

UV-vis (DCM) λmax/nm (ε/M−<sup>1</sup> cm<sup>−</sup>1): 267 (18,040), 366 (5391).

#### **3. Results and Discussion**

#### *3.1. Reaction of Solid-State Co(II) Complexes with NO*

Two new enantiomorphic phases of Co(L) (L = L*S,S*, L*R,R*) have been prepared and structurally characterized. Co(L)·CS2, was obtained by recrystallization of Co(L) from CS2/*n*-hexane. A strong band in the IR spectrum at 1524 cm−<sup>1</sup> is associated with the νCS of co-crystallized CS2. This band is very close to that for free CS2 at 1520 cm<sup>−</sup>1, indicating little interaction with the cobalt atom. Co(L)·CS2 undergoes CS2 loss on heating at 95 ◦C for 1 h at 10−<sup>2</sup> mbar to reproduce Co(L), however it is now a new unreported phase that is different to the starting phase. Both Co(L)·CS2 and Co(L) react in the solid state with NO gas (1 atm, rt, unground crystals ranging in size from 0.05 to 1 mm3) (Scheme 2).

**Scheme 2.** Preparation of Co(NO)(L) from Co(L) by a solid-state gas reaction with NO (1 atm).

This reaction is accompanied by a color change from red to black (Figure 1, SI film), with the process occurring over several hours for the CS2 solvate and in seconds to minutes for the desolvated phase, depending on sample size. Co(NO)(L) is formed inside the lattice. The IR spectrum of the product shows a low intensity νNO band at 1657 cm<sup>−</sup>1.

The NO can be removed stoichiometrically from Co(NO)(L) by heating to 195 ◦C. This process has been cycled three times without significant decomposition (Figure 2). Powder X-ray diffraction (PXRD) shows that crystallinity is retained after CS2 desorption, NO chemisorption, and subsequent desorption without significant change to the pattern (Figure 3). The pattern for the recrystallized sample of the nitrosyl complex is also similar. These suggest that that the packing is similar throughout these sequential gas (CS2/NO)– solid sorption/chemisorption and desorption processes.

**Figure 1.** Crystals of (**a**) Co(L)·CS2; (**b**) Co(L), obtained by the desorption of CS2 from Co(L)·CS2; and (**c**) Co(NO)(L) after the NO gas–solid reaction with Co(L).

**Figure 2.** Thermogravimetric analysis showing reversible and ostensibly stoichiometric binding of NO to Co(L) to give Co(NO)(L) over three cycles. The red lines are the heating profile, while the black lines are the weight loss profile.

**Figure 3.** Series of PXRD patterns for an enantiomorphic phase of the CS2 solvate, before and after desorption of CS2, followed by the product of NO chemisorption and subsequent NO desorption. Top pattern is of recrystallized Co(NO)(L) as a CS2 solvate. Note: Black—measured; red—calculated from single crystal structural data.

#### *3.2. Structures of Enantiomorphic Phases*

Single crystal X-ray structures were obtained for Co(L*R,R*)·CS2, Co(L*R,R*) and Co(NO)(L*R,R*)·CS2. Notably, the data for Co(L*R,R*) were obtained from a crystal of Co(L*R,R*)·CS2 after desorption of the CS2, which occurs in a single crystal-to-single crystal (SCSC) transformation for many of the individual crystals, i.e., many of the crystals do not break but retain their morphology in the process. This is not the case for the subsequent NO sorption, where the crystals break into smaller pieces and no crystal of adequate quality for a structure determination by SCXRD was found. PXRD establishes, however, that crystallinity is retained. Recrystallization of the solid Co(NO)(L*R,R*) (from CS2/*n*-hexane) was necessary for obtaining a single crystal X-ray structure of the cobalt nitrosyl. Details of the data collections are provided in Table 1. Solid state iron porphyrin complexes have been shown to be capable of NO gas sorption, in this case the process occurs in an SCSC transformation [31,32].

**Table 1.** Selected crystallographic data for Co(L*R,R*)·CS2, Co(L*R,R*), and Co(NO)(L*R,R*)·CS2. All data obtained using Cu Kα radiation.


The structures of Co(L*R,R*)·CS2 and Co(NO)(L*R,R*)·CS2 are shown in Figure 4. That of Co(L*R,R*) is similar (SI Figure S1). The cobalt ion in Co(L*R,R*)·CS2 is close to square planar, with the cobalt atom lying 0.023 Å above the O–N–N–O plane of L2−. The sum of angles around the Co(II) is 360.11◦, with cis angles of between 86.73◦ and 94.02◦. The average Co–Osalen and Co–Nsalen bond distances are 1.844(2) Å and 1.861(2) Å, respectively. The cyclohexane backbones show the expected chair conformation of the *R,R* enantiomer. The metric parameters for the unsolvated phase of Co(L*R,R*) are similar (Table 2).

The bond lengths and angles for the nitrosyl complex are comparable to those found for its homologue Co(NO)(salen) [34], with the geometry around the cobalt atom being close to square pyramidal. The angles deviate an average of 4.9◦ from 90◦, and the cobalt atom lies even further (0.237 Å) above the O–N–N–O plane of L2<sup>−</sup> compared to the Co(II) complexes, with no significant interaction to the cobalt ion in the other axial position. The distances and the associated angles are very similar to the cobalt(II) complexes, however, the flat chelating salen ligand can be expected to impose constraints. Analysis of the O/N–Co distances in other cobalt complexes of L2<sup>−</sup> (SI Figure S2) show that it is typical for

square pyramidal Co(III) complexes to display similar values to the Co(II) square planar complexes. Nitrosyl is a well-known non-innocent ligand, and has been assigned formally to the extremes of NO+, NO· and NO−. The N–O distance in free radical NO is 1.15 Å, and in Co(NO)(L*R,R*)·CS2 it is less than 0.03 Å greater. With various structural ambiguities, we describe the nitrosyl complex using the Enemark-Feltham notation [35], i.e., {Co(NO)}8.

**Figure 4.** Single crystal X-ray structures of (**a**) Co(L*R,R*)·CS2, and (**b**) Co(NO)(L*R,R*)·CS2. Atom colors: red—oxygen; blue nitrogen; white—carbon; yellow—sulfur; dark blue—cobalt. Atomic displacement ellipsoids are drawn at 50% probability and hydrogen atoms are omitted for clarity.

**Table 2.** Selected mean bond distances and angles.


The packing in the three structurally characterized compounds show layers of the complexes in a herringbone arrangement, where the tert-butyl groups para to the phenolato O atom reside in columns parallel to the *a* axis (Figure 5, SI Figures S3 and S4). In the solvates, the co-crystallized CS2 molecules are also located in these columns. Desorption of CS2 from Co(L*R,R*)·(CS2), to give Co(L*R,R*), results in a relatively large unit cell change with approximate doubling of the *a* axis from 13.55410(10) Å to 26.2740(4) Å, and an increase in the *β*-angle from 92.1290(10)◦ to 113.379(2)◦. The packing in Co(L*R,R*) is shown in Figure 5. The voids occur where the CS2 occupied the lattice of Co(L*R,R*)·(CS2). They are not connected and have an approximate size of 216.34 Å3 (3.2% of unit cell volume). The kinetic diameters of CS2 and NO are approximately 3.6 Å and 3.17 Å, respectively [36], and, on the basis of the structures and the approximate void diameter of 7.4 Å, it is tempting to posit that the exit of CS2 will be synchronously accompanied by rotation about the C–C bond attaching the tert-butyl groups to the aryl ring. The chemisorption of NO occurs with small shifts in the PXRD pattern (Figure 3), suggesting this unit cell is changed slightly but with related packing. The recrystallized Co(NO)(L*R,R*)·CS2 shows a marginally (1.7%) expanded unit cell compared to that of Co(L*R,R*)·CS2. The two most intense diffraction peaks in the PXRD of Co(NO)(L*R,R*) are moved towards a slightly higher 2θ angle upon chemisorption of NO compared to the parent phase Co(L*R,R*), which corresponds to a change in the interplanar distance (d) from 13.8962 Å and 12.0268 Å in Co(L*R,R*) to 13.4325 Å and 11.8659 Å in Co(NO)(L*R,R*) (Figure 3). This can be attributed to a slight molecular rearrangement, in order to make room for binding of NO at the cobalt atom.

**Figure 5.** Packing diagram for Co(L*R,R*) with voids in yellow (1.2 Å probe radius and 0.7 Å grid spacing). Hydrogen atoms are omitted for clarity.

#### *3.3. Solution-State Vibrational Circular Dichroism (VCD) Spectra*

VCD allows the characterization of enantiopure compounds through the analysis of IR-active bands. The primary quantity associated with IR absorbance is the dipole strength, however, VCD measures the differential absorbance (ΔA(ν)= Aleft(ν) − Aright(ν)) that is proportional to the rotational strength; a quantity which depends both on the electric and magnetic dipole transition moments. Thus, the VCD intensities are not directly correlated with the intensities in the IR spectrum. VCD has predominantly been used to analyze solution and pure liquid phases of organic molecules. In this work, VCD spectra for both hands of LH2 and the Zn(II), Co(II), and Co(III)NO complexes of L2<sup>−</sup> have been recorded. The IR spectra (purple) and the corresponding VCD spectra for the L*S,S*- and L*R,R*-based systems (colored blue and red respectively) are shown in Figure 6. The IR spectra show that the band due to the salen νC=N, the most intense in the IR spectra, shifts from 1632 cm−<sup>1</sup> in LH2 to a lower frequency (1598 cm−1) in coordination with Zn2+ in accordance with the decrease in the C=N bond order as a consequence of co-ordinate bond formation by the azomethine nitrogen lone pair. The band due to the νC=N, is the most intense in the VCD spectrum of the Zn(II) complexes. Despite being the most intense in the IR spectrum for L*S,S*H2 and L*R,R*H2, this is now a relatively moderate signal in the VCD spectra of L*S,S*H2 and L*R,R*H2, where, instead, several unassignable C–C and C–H vibrations are the most intense. Computational analysis to aid in the assignment of bands in VCD spectra is common in the study of enantiopure organic compounds, and we have performed such an analysis with reasonable success for L*R,R* and Zn(L*R,R*) (SI Figure S5). The introduction of transition metal ions makes calculation a formidable task, perhaps even impossible for the paramagnetic complexes. In this work, we take a fingerprint approach to the characterization. The fact that we have characterized both enantiomers and achieved consistently opposite band polarities demonstrates the reliability of the results.

**Figure 6.** Solution (CDCl3) IR (top, purple line, obtained using equimolar concentrations of both hands) and VCD (bottom) spectra of (**a**) L*S,S*H2 and L*R,R*H2, (**b**) Zn(L*S,S*) and Zn(L*R,R*), (**c**) Co(L*S,S*) and Co(L*R,R*), and (**d**) Co(NO)(L*S,S*) and Co(NO)(L*R,R*). The L*S,S*- and L*R,R*-based systems are colored blue and red, respectively.

An increase in VCD signal strength is noticeable in the comparison of the VCD spectra of the Zn(II) complexes and those for LH2. This arises because coordination locks the ligand into fewer conformations [37]. The VCD spectra for Co(L*S,S*) and Co(L*R,R*) are very different to those of the structurally analogous Zn(II) complexes. Apart from many different positions and relative intensities, the spectrum is close to monosignate (all bands for one hand showing one polarity). A change in the absorbance of left and right circularized IR radiation causes the polarity inversion of some bands. In addition, the signals are approximately nine times as intense as those for all the other compounds. This is associated with the diamagnetism and paramagnetism of low-lying excited electronic states in the Co(II) complexes. Electronic transitions are strongly magnetic-dipole dependent and hence, when present, can induce enhanced VCD by the coupling of the electronic magnetic-dipole transition moments with the smaller vibrational magnetic-dipole moments responsible for normal VCD intensity [37].

Azide is a linear, non-chiral ligand with a track record as a proxy vibrational spectroscopic and structural marker for O2 binding sites. Accordingly, it binds to the iron center at the distal position of the heme group. Bormett et al. [38] have reported the VCD spectra of ferric hemoglobin azide, and show that the azide ligand somehow borrows magnetic dipole intensity from the chiral environment inherent to this protein to produce a VCD signal. The effect we see in this work is analogous, and we have used VCD to assess the chiral induction of the signals for achiral NO in the complexes of L2−. This co-ligand can be regarded as a proxy substrate in a reaction catalyzed by complexes of enantiopure L2<sup>−</sup>, which, particularly with respect to asymmetric catalytic synthesis, is easily accessible in both hands. Although close to the dominating salen imine νCN (1616 cm−1) band in Co(NO)(L), the νNO (1657 cm<sup>−</sup>1) can be distinguished in the IR spectra as the second most intense band (Figure 6d, top). Notably, a signal for νNO can also be seen at this position in the VCD spectrum, but its intensity is now very low by comparison to many other bands in the spectrum. This indicates that, while chiral induction occurs, it is not particularly strong. In this context, it is interesting to note that the enantiomeric excess in catalytic reactions with complexes of L*R,R* and L*S,S* are sometimes disappointing [39–43].

#### *3.4. Solid State VCD Spectra*

While the use of Nujol mull, KBr discs, or even neat powders in sample preparation for recording solid-state IR spectra is commonplace, irreproducibility due to a high sensitivity to inhomogeneity in particle size, especially because the technique is a transmission method, makes classic solid-state sample preparation for VCD spectroscopy less straightforward. This is probably why solid state VCD spectra are rarely reported [37]. In the context of the

molecular solid-state chemistry described here, it was interesting for us to develop this technique. We found that, with consistent grinding and Nujol, it was possible to obtain reproducible spectra. The solid-state IR and VCD spectra of Co(L)·CS2 and Co(L) are shown in Figure 7a,b. In contrast to the IR spectra, the VCD spectra of the two phases are easily distinguishable. Obviously, this is information that was lost in the solution state. The differences between the solution-state and solid-state spectra are greater in the VCD spectra compared to the IR spectra: There is far less tendency towards a monosignate spectra for both phases in the solid-state spectra. Fascinatingly, an nCS signal related to co-crystallized CS2 is visible in VCD spectrum at 1514 cm−<sup>1</sup> (Figure 7a). This indicates supramolecular chiral induction of this achiral molecule. The signal disappears on CS2 desorption to form Co(L) (Figure 7b). The intensity of the IR band at 1524 cm−<sup>1</sup> also decreases. This band comprises an overlap of the nCS with a band that does not appear in the VCD spectrum.

**Figure 7.** Solid-state IR spectra of a mixture of equal amounts of both hands of the complexes (purple line) and solid-state VCD spectra of (**a**) Co(L)·CS2, (**b**) Co(L), and (**c**) Co(L)(NO) (blue line complexes of L*S,S*; red line complexes of L*R,R*).

The solid-state IR and VCD spectra of Co(NO)(L) are shown in Figure 7c. It is difficult to assign the bands in the spectra, however, the band at 1638 cm−1, assigned to the imine νCN, is significantly more intense than all the others. It also shows an increase of 22 cm−<sup>1</sup> compared to the band assigned to this vibration in the solution state spectra of the same complex. This suggests a change in the electronics of the complex due to different supramolecular interactions in the solution-state versus the solid-state. The higher wavenumber for νCN suggests that this bond is stronger in the solid state. Indirectly, this suggests that the NO is a stronger donor to the Co atom in the solid-state. The wellresolved, but low intensity, band at 1659 cm−<sup>1</sup> is assigned to νNO. There is significant polarity inversion between comparable signals in the paramagnetic Co(II) compounds, the diamagnetic {Co(NO)}8 phase, and the {Co(NO)}8 systems in the solution and solid states. Curiously, an unassigned band at 1344 cm−1, which is most intense in the solution-state spectrum of Co(NO)(L), appears to be absent in the solid-state VCD spectrum.

#### **4. Conclusions**

We have demonstrated chemisorption of NO into an ostensibly non-porous crystalline material, but postulate that transient conduits for CS2 desorption and NO chemisorption form because the tert-butyl groups can rotate around the bond between the aryl and the tert-butyl carbon atoms. The gas–solid reactions involved in the transitions between Co(L)·CS2, Co(L), and Co(L)(NO) occur in crystal to crystal transformations, and in one case an SCSC transformation. The NO binds reversibly, and the materials do not allow O2 to react with the bound NO. This supports the hypothesis that the in-crystal reactivity seen for the crystalline dicobalt(II) complexes described in the introduction, where sorbed NO and O2 were transformed to a nitrite ligand and a nitrate counter anion [7], is dependent on two cobalt(II) ions in close proximity. Accessibility and the specific coordination sphere and geometry, however, also clearly play important roles.

The crystalline solid state offers advantages beyond the solution state, through the provision of tailored cavities that might amplify chirality compared to the solution state. This study shows that in-crystal chemistry for the first time, with the induction of chirality onto both a chemisorbed achiral guest and a physisorbed co-crystallized achiral guest. Crystal phases themselves can occur in chiral space groups without being constructed from enantiopure molecules. Both situations offer feasibility for absolute asymmetric synthesis (AAS) [44] inside solid states. The rare use of VCD spectroscopy to characterize the complexes in this work illustrates its untapped potential in the study of enantiomers of metal coordination complexes in both solution- and solid-states.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/10 .3390/chemistry3020041/s1. CCDC 2073657–2073659 contain the supplementary crystallographic data for this paper. These data can be obtained free of charge from The Cambridge Crystallographic Data Centre via www.ccdc.cam.ac.uk/structures.

**Author Contributions:** M.S.M. synthesized the compounds, performed calculations, and prepared the first manuscript draft; M.S.M. and M.C.L. performed measurements, processed and analyzed the data, and designed the figures; V.M. assisted with X-ray crystallography; C.J.M. conceived and supervised the work. All authors reviewed and edited the manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** The work was supported by the Danish Council for Independent Research (Grant 9041- 00170B) and the Carlsberg Foundation grant CF15-0675 for the X-ray diffractometer.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data presented in this study are available in supplementary material.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


*Article*

## **Bis(diphenylphosphino)methane Dioxide Complexes of Lanthanide Trichlorides: Synthesis, Structures and Spectroscopy** †

#### **Robert D. Bannister, William Levason and Gillian Reid \***

School of Chemistry, University of Southampton, Southampton SO17 1BJ, UK;

r.d.bannister@soton.ac.uk (R.D.B.); wxl@soton.ac.uk (W.L.)

**\*** Correspondence: gr@soton.ac.uk

† Dedicated to Dr. Howard Flack (1943–2017).

Received: 18 October 2020; Accepted: 16 November 2020; Published: 19 November 2020

**Abstract:** Bis(diphenylphosphino)methane dioxide (dppmO2) forms eight-coordinate cations [M(dppmO2)4]Cl3 (M = La, Ce, Pr, Nd, Sm, Eu, Gd) on reaction in a 4:1 molar ratio with the appropriate LnCl3 in ethanol. Similar reaction in a 3:1 ratio produced seven-coordinate [M(dppmO2)3Cl]Cl2 (M = Sm, Eu, Gd, Tb, Dy, Ho, Er, Tm, Yb), whilst LuCl3 alone produced six-coordinate [Lu(dppmO2)2Cl2]Cl. The complexes have been characterised by IR, 1H and 31P{1H}-NMR spectroscopy. X-ray structures show that [M(dppmO2)4]Cl3 (M = Ce, Sm, Gd) contain square antiprismatic cations, whilst [M(dppmO2)3Cl]Cl2 (M = Yb, Dy, Lu) have distorted pentagonal bipyramidal structures with apical Cl. The [Lu(dppmO2)2Cl2]Cl has a *cis* octahedral cation. The structure of [Yb(dppmO2)3(H2O)]Cl3·dppmO2 is also reported. The change in coordination numbers and geometry along the series is driven by the decreasing lanthanide cation radii, but the chloride counter anions also play a role.

**Keywords:** lanthanide trichloride complexes; diphosphine dioxide; coordination complexes; X-ray structures

#### **1. Introduction**

Early work viewed the chemistry of the lanthanides (Ln) (Ln = La–Lu, Pm unless otherwise indicated) in oxidation state III as very similar and often only two or three elements were examined, and the results were assumed to apply to all. More recent work has shown this to be a very unreliable approach and detailed studies of all fourteen elements (excluding only the radioactive Pm) are required to establish properties and trends [1,2]. Sometimes yttrium is also included since it is similar in size to holmium. The main changes along the series are due to the lanthanide contraction, the reduction in the radius of the M3<sup>+</sup> ions between La (1.22 Å) and Lu (0.85 Å), and at some point a reduction in coordination number may be driven by steric effects, especially with bulky ligands. However, the decrease in radius also results in an increase in the charge/radius ratio along the series and this can lead to significant electronic effects on the ligand preferences. This interplay of steric and electronic effects means that changes in coordination number or ligand donor set can occur at different points along the series with different ligands. The effects are very nicely demonstrated in a recent article, which examined the changes which occurred in the series of lanthanide nitrates with complexes of 2,2 -bipyridyl, 2,4,6-tri-α-pyridyl-1,3,5-triazine and 2,2 ; 6 ,2"-terpyridine [2]. Tertiary phosphine oxides have proved popular ligands to explore lanthanide chemistry and the area has been the subject of a comprehensive review [3], and several detailed studies of trends along the series La-Lu have been reported [4–7]. We reported bis(diphenylphosphino)methane dioxide (dppmO2) formed square-antiprismatic cations [La(dppmO2)4] <sup>3</sup><sup>+</sup> with Cl, I or [PF6] counter ions, but lutetium gave only octahedral

[Lu(dppmO2)2X2] <sup>+</sup> (X = Cl, I) and [Lu(dppmO2)2Cl(H2O)]2<sup>+</sup> [8]. Other dppmO2 complexes reported include several types with Ln(NO3)3 [4], [Dy(dppmO2)4][CF3SO3]3 [9], [Eu(dppmO2)4][ClO4]3 [10], [La(dppmO2)4][CF3SO3]3 and [Lu(dppmO2)3(H2O)][CF3SO3]3 [11]. Here, we report a systematic study of the systems LnCl3-dppmO2 for all fourteen accessible lanthanides.

#### **2. Materials and Methods**

Infrared spectra were recorded as Nujol mulls between CsI plates using a Perkin-Elmer Spectrum 100 spectrometer over the range 4000–200 cm−1. 1H and 31P{1H}-NMR spectra were recorded using a Bruker AV–II 400 spectrometer and are referenced to the protio resonance of the solvent and 85% H3PO4, respectively. Microanalyses were undertaken by London Metropolitan University or Medac. Hydrated lanthanide trichlorides and anhydrous LnCl3 (Ln–Nd, Pr, Gd, Ho) were from Sigma-Aldrich and used as received. The Ph2PCH2PPh2 (Sigma-Aldrich) in anhydrous CH2Cl2 was converted to Ph2P(O)CH2P(O)Ph2 by air oxidation catalysed by SnI4 [12].

X-Ray Experimental. Details of the crystallographic data collection and refinement parameters are given in Table 1. Many attempts were made to grow crystals for X-ray examination from a variety of solvents including EtOH and CH2Cl2, either by slow evaporation or layering with hexane or pentane. The crystal quality was often rather poor, and all of the structures have disordered co-solvent, either water or ethanol. No attempt was made to locate the protons on the co-solvent. Several showed disorder in one or more of the phenyl rings. Good-quality crystals used for single crystal X-ray analysis were grown from [Lu(dppmO2)4}Cl2]Cl (CH2Cl2/hexane), [Ce(dppmO2)4]Cl3, [Sm(dppmO2)4]Cl3, [Gd(dppmO2)4]Cl3 (EtOH), [Yb(dppmO2)3Cl]Cl2, [Yb(dppmO2)3(H2O)]Cl3·dppmO2 (EtOH), [Lu(dppmO2)3Cl]Cl2 (CH2Cl2).

Data collections used a Rigaku AFC12 goniometer equipped with a HyPix-600HE detector mounted at the window of an FR-E+ SuperBright molybdenum (λ = 0.71073 Å) rotating anode generator with VHF Varimax optics (70 μm focus) with the crystal held at 100 K (N2 cryostream). Structure solution and refinements were performed with either SHELX(S/L)97 or SHELX(S/L)2013 [13,14]. The crystallographic data in cif format have been deposited as CCDC 2033611-2033618.

All samples were dried in high vacuum at room temperature for several hours, but this treatment does not remove lattice water or alcohol. Heating the samples in vacuo is likely to cause some decomposition of the complexes [7] and was not applied.

[La(dppmO2)4]Cl3·4H2O and [Lu(dppmO2)2Cl2]Cl·H2O were made as described [8]. The individual new complexes were isolated as described below, with yields of 50–80%.

[Ce(dppmO2)4]Cl3·6H2O—CeCl3·7H2O (0.025 g, 0.067 mmol) and dppmO2 (0.112 g, 0.268 mmol) afforded colourless crystals of [Ce(dppmO2)4]Cl3·4H2O, by concentrating the ethanolic solution and layering with *n*-hexane (1 mL). Required for C100H100CeCl3O12P8 (2020.1): C, 59.46; H, 4.99%. Found: C, 59.50; H, 4.50%. 1H-NMR (CD2Cl2): δ = 1.52 (s, H2O) 3.60 (vbr, [8H], PCH*2*P), 7.10 (s, [32H], Ph), 7.35 (m, [16H], Ph), 7.70 (m, [32H], Ph). 31P{1H}-NMR (CD2Cl2): δ = 48.6 (s). IR (Nujol mull)/cm<sup>−</sup>1: 3500 br, 1630 (H2O), 1158, 1099s (P=O).

[Pr(dppmO2)4]Cl3·6H2O—To a solution of PrCl3·6H2O (0.025 g, 0.070 mmol) in ethanol (5 mL) was added a solution of dppmO2 (0.117 g, 0.281 mmol) in ethanol (10 mL). A white powdered solid formed on slow evaporation of the ethanol. Required for C100H100Cl3O14P8Pr (2020.9): C, 59.43; H, 4.99%. Found: C, 59.06; H, 4.62% 1H-NMR (CD2Cl2): δ = 4.63 (m, [8H], PCH*2*P), 7.19 (s, [32H], Ph), 7.44 (m, [16H], Ph), 8.19 (m, [32H], Ph). 31P{1H}-NMR (CD2Cl2): δ = 64.0 (s). IR (Nujol mull)/cm−1: 3500 br, 1630 (H2O), 1161, 1102 (P=O).


**Table 1.** X-ray crystallographic data a.

#### *Chemistry* **2020** , *2*

[Nd(dppmO2)4]Cl3·4H2O—To a solution of NdCl3·6H2O (0.025 g, 0.070 mmol) in ethanol (5 mL) was added a solution of dppmO2 (0.116 g, 0.279 mmol) in ethanol (10 mL). A white powdered solid formed on slow evaporation of the ethanol. Required for C100H96Cl3NdO12P8 (1988.2): C, 60.41; H, 4.87%. Found: C, 60.41; H, 4.62%. 1H-NMR (CD2Cl2): δ = 1.52 (s, H2O) 3.66 (m, [8H], PCH*2*P), 7.14 (s, [32H], Ph), 7.35 (m, [16H], Ph), 7.76 (m, [32H], Ph). 31P{1H}-NMR (CD2Cl2): δ = 62.9 (s). IR (Nujol mull)/cm<sup>−</sup>1: 3500 br, 1630 (H2O), 1159 s, 1101 s (P=O).

[Sm(dppmO2)4]Cl3·4H2O—To a solution of SmCl3·6H2O (0.025 g, 0.069 mmol) in ethanol (5 mL) was added a solution of dppmO2 (0.114 g, 0.274 mmol) in ethanol (10 mL). Colourless crystals were formed via slow evaporation of the ethanol. Required for C100H96Cl3O12P8Sm (1994.3): C, 60.22; H, 4.85%. Found: C, 60.05; H, 4.50%. 1H-NMR (CD2Cl2): δ = 2.10 (s, H2O), 5.08 (br, [8H], PCH*2*P), 7.20 (s, [32H], Ph), 7.39 (m, [16H], Ph), 7.83 (m, [32H], Ph). 31P{1H}-NMR (CD2Cl2): δ = 35.6. IR (Nujol mull)/cm<sup>−</sup>1: 3500 br, 1630 (H2O), 1162 s, 1101 (P=O).

[Eu(dppmO2)4]Cl3·4H2O—To a solution of EuCl3·6H2O (0.025 g, 0.068 mmol) in ethanol (5 mL) was added a solution of dppmO2 (0.114 g, 0.274 mmol) in ethanol (10 mL) and the solution was stirred for 20 min. The solution was then concentrated, and colourless crystals were formed through layering with *n*-hexane (1 mL). Required for C100H96Cl3EuO12P8 (1995.9): C, 60.41; H, 4.87%. Found: C, 60.73; H, 4.71%. 1H-NMR (CD2Cl2): δ = 2.15 (s, H2O) 3.12 (br, [8H] PC*H2*P), 7.18 (s, [32H], Ph), 7.38 (m, [16H], Ph), 7.83 (m, [32H], Ph). 31P{1H}-NMR (CD2C2): <sup>δ</sup> = 25.0 (br, "free" dppmO2), <sup>−</sup>13.4. IR (Nujol mull)/cm<sup>−</sup>1: 3500 br, 1630 (H2O), 1159, 1099 (P=O).

[Gd(dppmO2)4]Cl3·4H2O—To a solution of GdCl3·6H2O (0.025 g, 0.067 mmol) in ethanol (5 mL) was added a solution of dppmO2 (0.112 g, 0.269 mmol) in ethanol (10 mL). Colourless crystals were formed through slow evaporation of the solvent. Required for C100H96Cl3GdO12P8 (2001.2): C, 60.02; H, 4.83%. Found: C, 60.05; H, 4.86%. 1H-NMR (CD2Cl2): δ = no resonance. 31P{1H}-NMR (CD2Cl2): δ = no resonance. IR (Nujol mull)/cm<sup>−</sup>1: 3500 br, 1630 (H2O), 1160, 1099 (P=O).

[Sm(dppmO2)3Cl]Cl2—To a solution of SmCl3·6H2O (0.025 g, 0.069 mmol) in ethanol (5 mL) was added a solution of dppmO2 (0.086 g, 0.206 mmol) in ethanol (10 mL). The solvent was removed in vacuo and the resulting white solid was washed with cold ethanol. Colourless crystals were obtained via slow evaporation of an ethanolic solution of the product. Required for C75H66Cl3O6P6Sm (1505.9): C, 59.80; H, 4.42%. Found: C, 59.62; H, 4.55%. 1H-NMR (CD2Cl2): δ = 3.67 (br m, [6H], PCH2P), 7.15 (br, [24H], Ph), 7.35 (m, [12H], Ph), 8.05 (m, [24H], Ph). 31P{1H}-NMR (CD2Cl2): δ = 38.15 (s). IR (Nujol mull)/cm<sup>−</sup>1: 1153 s, 1097 s (P=O).

[Eu(dppmO2)3Cl]Cl2—To a solution of EuCl3·6H2O (0.025 g, 0.068 mmol) in ethanol (5 mL) was added a solution of dppmO2 (0.085 g, 0.205 mmol) in ethanol (10 mL). The solvent was removed in vacuo and the resulting white solid was washed with cold ethanol. Required for C75H66EuCl3O6P6 (1507.49): C, 59.76; H, 4.41%. Found: C, 59.71; H, 4.56%. 1H-NMR (CDCl3): δ = 3.66 (br, [6H], PCH2P), 7.03 (br m, [36H], Ph), 7.87 (br, [24H], Ph). 31P{1H}-NMR (CDCl3): <sup>δ</sup> <sup>=</sup> <sup>−</sup>14.8 (s). IR (Nujol mull)/cm<sup>−</sup>1: 1153 s, 1098 s (P=O).

[Gd(dppmO2)3Cl]Cl2·3H2O—To a solution of GdCl3·6H2O (0.025 g, 0.067 mmol) in ethanol (5 mL) was added a solution of dppmO2 (0.084 g, 0.201 mmol) in ethanol (10 mL). The solvent was removed in vacuo and the resulting white solid was washed with cold ethanol. Required for C75H66Cl3O6P6Gd (166.8): C, 57.49, H, 4.63%; Found: C, 57.17; H, 4.43%. 1H-NMR (CD2Cl2): no resonance. 31P{1H}-NMR (CD2Cl2): no resonance. IR (Nujol mull)/cm<sup>−</sup>1: 3500 br, 1630 (H2O), 1155 s, 1098 s (P=O).

[Tb(dppmO2)3Cl]Cl2·H2O—To a solution of TbCl3·6H2O (0.025 g, 0.067 mmol) in ethanol (5 mL) was added a solution of dppmO2 (0.084 g, 0.201 mmol) in ethanol (10 cm3). The solvent was removed in vacuo and the resulting white solid was washed with cold ethanol. Required for C75H68Cl3O7P6Tb (1532.5): C, 58.78; H, 4.47%. Found: C, 59.41; H, 4.54%. 1H-NMR (CD2Cl2): δ = 1.9 (br H2O), 3.50 (br m, [6H], PCH2P), 5.89 (br, [36H], Ph), 7.46 (br, [24H], Ph). 31P{1H}-NMR (CD2Cl2): <sup>δ</sup> <sup>=</sup> <sup>−</sup>29.2 (s). IR (Nujol mull)/cm<sup>−</sup>1: 3500 br, 1630 (H2O), 1153 s, 1097 s (P=O).

[Dy(dppmO2)3Cl]Cl2·H2O—To a solution of TbCl3·6H2O (0.025 g, 0.066 mmol) in ethanol (5 mL) was added a solution of dppmO2 (0.083 g, 0.199 mmol) in ethanol (10 cm3). The solution was filtered

then concentrated and layered with hexane (1 mL) yielding a white powdered product. Colourless crystals were formed by layering a CH2Cl2 solution of the product with hexane. Required for C75H68Cl3DyO7P6 (1536.0): C, 58.64; H, 4.46%. Found: C, 58.21; H, 4.63%. 1H-NMR (CD2Cl2): δ = 1.9 (vbr H2O), 3.66 (br m, [6H], PCH*2*P), 7.33 (br, [36H], Ph), 8.66 (br, [24H], Ph). 31P{1H}-NMR (CD2Cl2): δ = 18 (vbr, s). IR (Nujol mull)/cm<sup>−</sup>1: 3500 br, 1630 (H2O), 1156 s, 1099 s (P=O).

[Ho(dppmO2)3Cl]Cl2·H2O—To a solution of HoCl3 (0.050 g, 0.124 mmol) in ethanol (5 mL) was added a solution of dppmO2 (0.230 g, 0.55 mmol) in ethanol (10 mL). The solvent was removed in vacuo and the resulting pale pink solid was washed with cold ethanol. Required for C75H68Cl3HoO7P6 (1538.5): C, 58.66; H, 4.55%. Found: C, 59.41; H, 4.52%. 1H-NMR (CD2Cl2): δ = 2.1 (br, H2O), 3.72 (br s, [6H], PCH2P), 6.78 (br, [36H], Ph), 7.68 (br, [24H], Ph)]. 31P{1H}-NMR (CD2Cl2): <sup>δ</sup> <sup>=</sup> <sup>−</sup>13.5 (s). IR (Nujol mull)/cm<sup>−</sup>1: 3500 br, 1630 (H2O), 1154 s, 1097 s (P=O).

[Er(dppmO2)3Cl]Cl2·3H2O—To a solution of ErCl3·6H2O (0.025 g, 0.065 mmol) in ethanol (5 mL) was added a solution of dppmO2 (0.082 g, 0.196 mmol) in ethanol (10 mL). The solvent was removed in vacuo and the resulting white solid was washed with cold ethanol. Required for C75H72Cl3ErO9P6 (1576.8): C, 57.13; H, 4.60%. Found: C, 57.08; H, 4.54%. 1H-NMR (CD2Cl2): δ = δ = 1.2 (br, H2O), 3.25 (br s, [6H], PCH2P), 5.52 (vbr, [12H], Ph), 7.15 (br s, [24H], Ph)], 7.28 (br s, [24H], Ph)]. 31P{1H}-NMR (CD2Cl2): <sup>δ</sup> <sup>=</sup> <sup>−</sup>60.8 (s). IR (Nujol mull)/cm<sup>−</sup>1: 3500 br, 1630 (H2O), 1155 s, 1097 s (P=O).

[Tm(dppmO2)3Cl]Cl2·3H2O—To a solution of TmCl3·6H2O (0.025 g, 0.065 mmol) in ethanol (5 mL) was added a solution of dppmO2 (0.081 g, 0.195 mmol) in ethanol (10 mL). The solvent was removed in vacuo and the resulting white solid was washed with cold ethanol. Required for C75H72Cl3O9P6Tm (1578.5): C, 57.07; H, 4.60%. Found: C, 56.61; H, 4.45%. 1H-NMR (CD2Cl2): δ = 3.48 (m, [6H], PCH2P), 7.11 (br, [24H], Ph), 7.68 (br, [36H], Ph)]. 31P{1H}-NMR (CD2Cl2): <sup>δ</sup> = <sup>−</sup>54.8 (s). IR (Nujol mull)/cm<sup>−</sup>1: 3500 br, 1630 (H2O), 1156 s, 1096 s (P=O).

[Yb(dppmO2)3Cl]Cl2·H2O—To a solution of YbCl3·6H2O (0.025 g, 0.065 mmol) in ethanol (5 mL) was added a solution of dppmO2 (0.080 g, 0.194 mmol) in ethanol (10 mL). The solvent was removed in vacuo and the resulting white powder was washed with cold ethanol. Required for C75H68Cl3O7P6Yb (1546.58): C, 58.24; H, 4.43%. Found: C, 58.73; H, 4.45%. 1H-NMR (CD2Cl2): δ = 3.50 (m, [6H], PCH*2*P), 6.64 (br, [24H], Ph), 7.15 (br, [36H], Ph). 31P{1H}-NMR (CD2Cl2): δ = +9.2 (s). IR (Nujol mull)/cm−1: 3500 br, 1630 (H2O), 1154 s, 1097 s (P=O).

#### **3. Results**

The reaction of LnCl3·nH2O (Ln = La [8], Ce, Pr, Nd, Sm, Eu or Gd; *n* = 6 or 7) with four mol. equivalents of dppmO2 in ethanol gave good yields of tetrakis-dppmO2 complexes, [Ln(dppmO2)4]Cl3. The IR and 1H-NMR spectra show the the isolated complexes retain significant amounts of lattice water, and sometimes EtOH, which is not removed by prolonged drying of the bulk powders in vacuo. The high molecular weights make the microanalyses rather insensitive to the amount of water, but are generally consistent with a formulation [Ln(dppmO2)4]Cl3·nH2O (*n* = 6: Ce, Pr; *n* = 4: Nd, Sm, Eu, Gd), although the amount of lattice solvent probably varies with the sample and is unlikely to be stoichiometric. The presence of significant amounts of lattice solvent is common in lanthanide phosphine oxide systems [7–10], and although evident in X-ray crystal structures, it is often disordered and difficult to model. Obtaining good quality crystals of the complexes proved difficult, but crystals of the Ce, Sm and Gd salts were obtained from various organic solvents and the compositions are shown in Table 1. The crystals contain different amounts of solvent of crystallisation to the bulk samples as they were grown from different media (and crystals were not dried in vacuo). The IR spectra (Table 2) show that the υ(PO) stretch in dppmO2 at 1187 cm−<sup>1</sup> has been lost and replaced by a new very strong and broad band ~1160 cm−<sup>1</sup> and a second band at ~ 1100 cm<sup>−</sup>1, which are due to the coordinated phosphine oxide groups. The frequencies appear invariant with the lanthanide present, which may be due to small differences being obscurred by the width of the bands. In [LnCl3(OPPh3)3] and [LnCl2(OPPh3)4] <sup>+</sup> the frequency of the υ(PO) stretch increases by ~ 10 cm−<sup>1</sup> between La and Lu [7]. The 31P{1H}-NMR chemical shift of dppmO2 at δ = +25.3 shows a high frequency shift to +33.1

in [La(dppmO2)4]Cl3, whilst the corresponding spectra of the Ce, Pr, Nd and Sm complexes show larger shifts due to the presence of the paramagnetic lanthanide ion (Table 2). In contrast, although the solid [Eu(dppmO2)4]Cl3 complex was isolated without difficulty, the 31P{1H}-NMR spectrum shows a strong feature at δ ~ +25 ("free" dppmO2), along with a second resonance at δ = −13.4, which may be assigned to [Eu(dppmO2)3Cl]Cl2 (see below), indicating substantial dissociation of one dppmO2 in solution; the broad resonance of the free dppmO2 is indicative of exchange on the NMR timescale. [Gd(dppmO2)4]Cl3 was isolated, and its constitution confirmed by its X-ray crystal structure, but no 1H or 31P{1H}-NMR resonances were observed, an effect seen in other gadolinium systems [6,7] and ascribed to fast relaxation by the f7 configuration of the metal. Attempts to isolate [Ln(dppmO2)4]Cl3 complexes for Ln = Dy-Lu were unsuccessful. We note that [Dy(dppmO2)4][CF3SO3]3 [9] was isolated with triflate counter ions, but with chloride only [Dy(dppmO2)3Cl]Cl2 was produced (below). An in situ 31P{1H}-NMR spectrum of CeCl3·7H2O <sup>+</sup> 2 dppmO2 in CH2Cl2 showed a single resonance at <sup>δ</sup> = +48, which is consistent with formation of [Ce(dppmO2)4] <sup>3</sup>+, confirming the preference for formation of the tetrakis complexes early in the series, even when there is a deficit of ligand.


<sup>a</sup> In CD2Cl2 solution 298 K; <sup>b</sup> Nujol mull; <sup>c</sup> Ref. [8].

The X-ray structures of [Ce(dppmO2)4]Cl3 (Figure 1), [Sm(dppmO2)4]Cl3 (Figure 2) and [Gd(dppmO2)4]Cl3 (Figure 3) show distorted square antiprismatic cations, very similar to those in [La(dppmO2)4][PF6]3 [8] and [Nd(dppmO2)4]Cl3 [15]. The average Ln-O distances in this series are: La = 2.514 Å, Ce = 2.486 Å, Nd = 2.465 Å, Sm = 2.429 Å and Gd = 2.420 Å, correlating well with the decreasing Ln3<sup>+</sup> radii (La = 1.216 Å, Ce =1.196 Å, Nd = 1.163 Å, Sm = 1.132 Å, Gd = 1.107 Å). The P = O bond lengths and the O-Ln-O chelate angles do not vary significantly along the series. The Ce-O(P) distances in [Ce(dppmO2)4]Cl3 are markedly longer than those in [Ce(Me3PO)4(H2O)4]Cl3 (2.372(2)-2.423(2) Å) [16], which has a distorted dodecahedral geometry with a CeO8 donor set.

**Figure 1.** The cation in [Ce(dppmO2)4]Cl3. The chloride anions and solvate molecules are omitted. Selected bond lengths (Å): Ce1–O1 = 2.4874(14), Ce1–O2 = 2.4790(14), Ce1–O3 = 2.4967(13), Ce1–O4 = 2.4803(14), P1–O1 = 1.5031(14), P2–O2 = 1.5021(14), P3–O3 = 1.5018(14), P4–O4 = 1.5031(14). Chelate angle O-Ce-O = 73.1◦ (av).

**Figure 2.** The cation in [Sm(dppmO2)4]Cl3. The chloride anions and solvate molecules are omitted. Selected bond lengths (Å): Sm1–O1 = 2.4160(14), Sm1–O2 = 2.4400(15), Sm1–O3 = 2.4358(14), Sm1–O4 = 2.4268(15), P1–O1 = 1.5025(15), P2–O2 = 1.5019(15), P3–O3 = 1.4961(15), P4–O4 = 1.4961(16). Chelate angle O-Sm-O = 72.9◦ (av).

**Figure 3.** The cation in [Gd(dppmO2)4]Cl3. The chloride anions and solvate molecules are omitted. Selected bond lengths (Å): Gd1–O1 = 2.420(2), Gd1–O2 = 2.409(3), Gd1–O3 = 2.415(2), Gd1–O4 = 2.398(2), P1–O1 = 1.504(2), P2–O2 = 1.501(3), P3–O3 = 1.501(3), P4–O4 = 1.501(3). Chelate angle O-Sm-O = 73.1◦ (av).

The reaction of LnCl3·6H2O (Ln =Sm, Eu, Gd, Tb, Dy, Ho, Er, Tm, Yb) with 3 mol. equivalents of dppmO2 in EtOH, followed by concentration of the solution or precipitation with hexane, afforded [Ln(dppmO2)3Cl]Cl2 complexes. Examination of the IR and 1H-NMR spectra indicated these incorporated less water or ethanol lattice solvent molecules than the [Ln(dppmO2)4]Cl3, and this was confirmed by the microanalyses. The Sm and Eu complexes appear largely free of solvent of crystallisation, whilst the Tb, Ho and Yb approximate to [Ln(dppmO2)3Cl]Cl2·H2O, and the Gd, Er and Tm complexes are [Ln(dppmO2)3Cl]Cl2·3H2O; again, this is likely to vary from sample to sample and with the isolation method. The IR spectra (Table 2) show the two υ(PO) bands as in the tetrakis complexes, but the higher energy bands of the tris complexes are ~ 5–10 cm−<sup>1</sup> lower in frequency than in the former. We were unable to identify υ(Ln-Cl) vibrations in the far IR spectra. The 31P{1H}-NMR spectra of the [Ln(dppmO2)3Cl]Cl2 show single resonances to high or low frequency of dppmO2 depending on the f<sup>n</sup> configuration of the Ln ion present (Table 2) and are generally similar to those found in other systems [5–7], although the magnitude of the shifts varies widely with the specific f <sup>n</sup> configuration. The line broadening is also highly variable between complexes of different Ln ions. The addition of dppmO2 to a solution of [Ln(dppmO2)3Cl]Cl2 (Ln = Eu, Gd, Tb, Dy, Ho, Er, Tm, Yb) in CH2Cl2 showed 31P{1H}-NMR resonances assignable to "free" dppmO2 and [Ln(dppmO2)3Cl]Cl2, but no new resonances that could be attributed to the formation of significant amounts of [Ln(dppmO2)4] <sup>3</sup>+. Although the resonances are broad in some cases, the observed chemical shifts are identical to those in pure [Ln(dppmO2)3Cl]Cl2. For [Sm(dppmO2)3Cl]Cl2 δ( 31P{1H}) = 38, the resonance shifts to δ = 35.6 upon addition of dppmO2, attributable to the formation of [Sm(dppmO2)4]Cl3, showing that both tris- and tetrakis-dppmO2 complexes exist in solution for samarium in the presence of the appropriate amount of ligand.

The X-ray structures of [Er(dppmO2)3Cl]Cl2 (Er-O = 2.28 Å av.) [17], [Yb(dppmO2)3Cl]Cl2 (Figure 4; Yb-O = 2.28 Å av.) and [Dy(dppmO2)3Cl]Cl2 (Figure S43) show pentagonal bipyramidal cations with an apical chloride. The Ln-O distances are rather variable (Er-O = 2.244(6)–2.328(6) Å; Yb-O = 2.250(2)–2.269(3) Å), but are shorter than those in the tetrakis-dppmO2 cations, reflecting both the reduced coordination number and the smaller metal ion radii (Er = 1.062, Yb = 1.042 Å). The contraction in ionic radii is also evident in the Ln-Cl distances of 2.598(2) Å (Er) and 2.5829(9) Å (Yb). Crystals

of [Dy(dppmO2)3Cl]Cl2 were also obtained and show the same cation type, but during refinement, several of the phenyl rings exhibited severe disorder and the data are therefore not included here (Figure S43).

**Figure 4.** The X-ray structure of [Yb(dppmO2)3Cl]Cl2. The chloride anions and solvate molecules are omitted. Selected bond lengths (Å) and angles (◦): Yb1–Cl1 = 2.5834(9), Yb1–O1 = 2.298(3), Yb1–O2 = 2.282(3), Yb1–O3 = 2.250(2), Yb1–O4 = 2.248(2), Yb1–O5 = 2.338(2), Yb1–O6 = 2.269(3), P–O = 1.494(3)-1.509(3), Cl1–Yb1–O4 = 173.89(9), O1–Yb1–O2 = 73.79(9), O3–Yb1–O4 = 80.64(9), O5–Yb1–O6 = 73.95(9).

Lutetium was previously reported to form the only bis-dppmO2 complex, [Lu(dppmO2)2Cl2]Cl, in this series [8], and this has now been confirmed by the X-ray crystal structure which shows a *cis*-octahedral geometry (Figure 5). The Lu-O distance of 2.230 Å (av) is shorter than the Ln-O distances in the seven- or eigth-coordinate complexes, and correlates both with the reduced coordination number and the smaller radius of Lu3<sup>+</sup> (1.032 Å). Treatment of a CH2Cl2 solution of [Lu(dppmO2)2Cl2]Cl with dppmO2 caused the 31P{1H}-NMR resonance to shift from +40 to +38.3, which suggests that [Lu(dppmO2)3Cl]Cl2 forms in solution. A few crystals of this product were isolated from a mixture containing excess dppmO2. These showed a pentagonal bipyramidal dication (Figure 6). As expected, the Lu-Cl and Lu-O bond lengths are slightly longer than in the six-coordinate cation, but are shorter than the corresponding bonds in [Yb(dppmO2)3Cl]Cl2, showing that the expected contraction continues along the series. The complex, [Lu(dppmO2)3(H2O)][CF3SO3]3, is known and its X-ray crystal structure showed seven-coordinate lutetium [11]. Although not confirmed by an X-ray structure, yttrium is reported to form a six-coordinate complex, [Y(dppmO2)2Cl2]Cl [18].

A different crystal isolated from the YbCl3-dppmO2 reaction proved, on structure solution, to be [Yb(dppmO2)3(H2O)]Cl3·dppmO2·12H2O (Figure 7), which contains a seven-coordinate Yb centre coordinated to three dppmO2 and a water molecule, with the Lu-coordinated water hydrogen-bonded to an adjacent uncoordinated dppmO2 molecule. The geometry is best described as a very distorted pentagonal bipyramid with the water occupying an equatorial position and is similar to the geometry found in [Lu(dppmO2)3(H2O)][CF3SO3]3 [11]. The Yb-OH2 distance of 2.3263(14) Å is ~ 0.05 Å longer than the Yb-O(P).

A large number of disordered solvate water molecules were also present, which proved very difficult to model, but the geometry of the ytterbium cation is clearly defined.

**Figure 5.** The cation in [Lu(dppmO2)2Cl2]Cl. Selected bond lengths (Å) and angles (◦): Lu1–Cl1 = 2.5581(8), Lu1–Cl2 = 2.5163(7), Lu1–O1 = 2.227(2), Lu1–O2 = 2.227(2), Lu1–O3 = 2.274(2), Lu1–O4 = 2.200(2), P1–O1 = 1.510(2), P2–O2 = 1.506(2), P3–O3 = 1.513(2), P4–O4 = 1.507(2), Cl2–Lu1–Cl1 = 95.97(3), O1–Lu1–Cl1 = 97.27(6), O1–Lu1–Cl2 = 99.68(6), O1–Lu1–O2 = 81.56(8), O1–Lu1–O3 = 85.65(7), O2–Lu1–Cl2 = 91.94(5), O2–Lu1–O3 = 84.89(7), O3–Lu1–Cl1 = 87.22(6), O4–Lu1–Cl1 = 94.66(6), O4–Lu1–Cl2 = 92.55(6), O4–Lu1–O2 = 84.74(8), O4–Lu1–O3 = 81.36(7).

**Figure 6.** The X-ray structure of [Lu(dppmO2)3Cl]Cl2. The chloride anions and solvate molecules are omitted. Selected bond lengths (Å) and angles (◦): Lu1–Cl1 = 2.5604(7), Lu1–O1 = 2.341(5), Lu1–O2 = 2.268(5), Lu1–O3 = 2.268(5), Lu1–O4 = 2.297(5), Lu1–O5 = 2.354(5), Lu1–O6 = 2.227(5), P–O = 1.497(5)-1.510(5), Cl1–Lu1–O6 = 176.35(14), O1–Lu1–O2 = 73.37(17), O3–Lu1–O4 = 73.37(17), O5–Lu1–O6 = 83.45(18).

**Figure 7.** The cation in [Yb(dppmO2)3(H2O)]Cl3·dppmO2·12H2O also showing the hydrogen-bonded dppmO2 molecule. Selected bond lengths (Å): Yb1–O3 = 2.2341(14), Yb1–O2 = 2.2899(13), Yb1–O9 = 2.3263(14), Yb1–O4 = 2.2683(13), Yb1–O6 = 2.2208(13), Yb1–O1 = 2.2328(13), Yb1–O5 = 2.2696(14), Pn–On (*n* = 1–6) = 1.5034(14)–1.5072(14), P7–O7 = 1.4924(15), P8–O8 = 1.4926(15).

#### **4. Discussion**

The chemistry of dppmO2 with lanthanides described in the previous section proves to be very systematic along the series La–Lu. For La–Gd, it was possible to isolate [Ln(dppmO2)4]Cl3. Although it could be isolated in the solid state, the solution 31P-NMR spectroscopic data indicate that [Eu(dppmO2)4]Cl3 was largely dissociated in CH2Cl2 solution into [Eu(dppmO2)3Cl]<sup>2</sup><sup>+</sup> and dppmO2; the isolation of the tetrakis-dppmO2 complex no doubt resulting from it being the least soluble species in an exchanging mixture in solution, although present in very minor amounts. The case of [Gd(dppmO2)4]Cl3 is likely to be similar, although the fast relaxation of the f7 ion precluded 31P-NMR study. For the elements Sm-Yb, the complexes [Ln(dppmO2)3Cl]Cl2 were readily isolated, but only for samarium was it possible to convert [Ln(dppmO2)3Cl]Cl2 to [Ln(dppmO2)4]Cl3 in CH2Cl2 solution by treatment with more dppmO2. Similarly, at the end of the series, the complex isolated was [Lu(dppmO2)2Cl2]Cl, for which treatment with dppmO2 afforded a new species in solution, identified as [Lu(dppmO2)3Cl]Cl2 by a structure determination from a few crystals obtained in the presence of excess dppmO2, although a bulk sample could not be isolated [8]. The change from eight-coordination in [Ln(dppmO2)4]Cl3 at the beginning of the series, to seven-coordination from Sm onwards, and finally to six-coordination at Lu, parallels the reduction in Ln3<sup>+</sup> radii. Isolation of both the eight- and seven-coordinate complexes was possible only for Sm, Eu and Gd. However, one should note that the chloride counter ions also have some role, in that whilst in the LnCl3/dppmO2 series tetrakis-dppmO2 species did not form beyond Gd, the complex [Dy(dppmO2)4][CF3SO3]3 [9] has been isolated from dmf solution with triflate counter ions. The role that anions and solvents play in lanthanide chemistry is often overlooked [2], but can be critical in determining which complex is isolated from solution. For example, the reaction of LnCl3 with Ph3PO results in isolation of [Ln(Ph3PO)3Cl3] from acetone, but [Ln(Ph3PO)4Cl2]Cl from ethanol [7]. On further examination by 31P-NMR spectroscopy, both species were found to be present in either solvent (in varying amounts), and the form isolated

reflected the least soluble complex in the particular solvent, which then precipitated from the mixture of rapidly interconverting species.

#### **5. Conclusions**

Through this synthetic, structural and spectroscopic study of the coordination of dppmO2 to the lanthanide trichlorides, we have established where the switch from eight-, to seven-, to six-coordination at the Ln(III) centre occurs along the lanthanide series, with X-ray crystallographic authentication for representative examples. The data also reveal subtle, but systematic, variations in the spectroscopic (e.g., ν(PO)) and structural parameters across the series, reflecting the change in ionic radii, the charge:radius ratio and also the influence of the presence of the competitive chloride ions.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2624-8549/2/4/60/s1; Figure S1–1H-NMR spectrum of [Ce(dppmO2)4]Cl3 in CD2Cl2; Figure S2–31P{1H} spectrum of Ce(dppmO2)4]Cl3 in CD2Cl2; Figure S3–Infrared spectrum of [Ce(dppmO2)4]Cl3 (Nujol mull); Figure S4–1H-NMR spectrum of [Pr(dppmO2)4]Cl3 in CD2Cl2; Figure S5–31P{1H}-NMR spectrum of [Pr(dppmO2)4]Cl3 in CD2Cl2; Figure S6–Infrared spectrum of [Pr(dppmO2)4]Cl3 (Nujol mull); Figure S7—1H-NMR spectrum of [Nd(dppmO2)4]Cl3 in CD2Cl2; Figure S8—31P{1H}-NMR spectrum of [Nd(dppmO2)4]Cl3 in CD2Cl2; Figure S9—Infrared spectrum of [Nd(dppmO2)4]Cl3 (Nujol mull); Figure S10—1H-NMR spectrum of [Sm(dppmO2)4]Cl3 in CD2Cl2; Figure S11—31P{1H}-NMR spectrum of [Sm(dppmO2)4]Cl3 in CD2Cl2; Figure S12—Infrared spectrum of [Sm(dppmO2)4]Cl3 (Nujol mull); Figure S13—1H-NMR spectrum of [Eu(dppmO2)4]Cl3 in CD2Cl2 **;** Figure S14—31P{1H}-NMR spectrum of [Eu(dppmO2)4]Cl3 in CD2Cl2; Figure S15—Infrared spectrum of [Eu(dppmO2)4]Cl3 (Nujol mull); Figure S16—Infrared spectrum of [Gd(dppmO2)4]Cl3 (Nujol mull); Figure S17—1H-NMR spectrum of [SmCl(dppmO2)3]Cl2 in CD2Cl2 (\* = EtOH)**;** Figure S18—31P{1H}-NMR spectrum of [SmCl(dppmO2)3]Cl2 in CD2Cl2; Figure S19—Infrared spectrum of [SmCl(dppmO2)3]Cl2 (Nujol mull); Figure S20—1H-NMR spectrum of [EuCl(dppmO2)3]Cl2 in CDCl3; Figure S21—31P{1H}-NMR spectrum of [EuCl(dppmO2)3]Cl2 in CDCl3; Figure S22—31P{1H}-NMR spectrum of [EuCl(dppmO2)3]Cl2 + excess dppmO2 in CDCl3; Figure S23—Infrared spectrum of [EuCl(dppmO2)3]Cl2 (Nujol mull); Figure S24—Infrared spectrum of [GdCl(dppmO2)3]Cl2 (Nujol mull); Figure S25—1H-NMR spectrum of [TbCl(dppmO2)3]Cl2 in CD2Cl2 (\* = EtOH)**;** Figure S26—31P{1H}-NMR spectrum of [TbCl(dppmO2)3]Cl2 in CD2Cl2; Figure S27—Infrared spectrum of [TbCl(dppmO2)3]Cl2 (Nujol mull); Figure S28—1H-NMR spectrum of [DyCl(dppmO2)3]Cl2 in CD2Cl2; Figure S29—31P{1H}-NMR spectrum of [DyCl(dppmO2)3]Cl2 in CD2Cl2; Figure S30—Infrared spectrum of [DyCl(dppmO2)3]Cl2 (Nujol mull); Figure S31—1H-NMR spectrum of [HoCl(dppmO2)3]Cl2 in CD2Cl2 (\* = EtOH); Figure S32—31P{1H}-NMR spectrum of [HoCl(dppmO2)3]Cl2 in CD2Cl2; Figure S33—Infrared spectrum of [HoCl(dppmO2)3]Cl2 (Nujol mull); Figure S34—1H-NMR spectrum of [ErCl(dppmO2)3]Cl2 in CD2Cl2 (\* = EtOH); Figure S35—31P{1H}-NMR spectrum of [ErCl(dppmO2)3]Cl2 in CD2Cl2; Figure S36—Infrared spectrum of [ErCl(dppmO2)3]Cl2 (Nujol mull); Figure S37—1H-NMR spectrum of [TmCl(dppmO2)3]Cl2 in CD2Cl2 (\* = EtOH); Figure S38—31P{1H}-NMR spectrum of [TmCl(dppmO2)3]Cl2 in CD2Cl2; Figure S39—Infrared spectrum of [TmCl(dppmO2)3]Cl2 (Nujol mull); Figure S40—1H-NMR spectrum of [YbCl(dppmO2)3]Cl2 in CD2Cl2 (\* = EtOH); Figure S41—31P{1H}-NMR spectrum of [YbCl(dppmO2)3]Cl2 in CD2Cl2; Figure S42—Infrared spectrum of [YbCl(dppmO2)3]Cl2 (Nujol mull); Figure S43—The cation in [DyCl(dppmO2)3]Cl2. The chloride anions and solvate molecules are omitted.

**Author Contributions:** Conceptualization, R.D.B., W.L. and G.R.; formal analysis, R.D.B.; investigation, R.D.B.; data curation, R.D.B.; writing—original draft preparation, R.D.B. and W.L.; writing—review and editing, R.D.B., W.L. and G.R.; supervision, W.L. and G.R. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Chloropentaphenyldisiloxane—Model Study on Intermolecular Interactions in the Crystal Structure of a Monofunctionalized Disiloxane †**

**Jonathan O. Bauer \* and Tobias Götz**

Institut für Anorganische Chemie, Fakultät für Chemie und Pharmazie, Universität Regensburg, Universitätsstraße 31, D-93053 Regensburg, Germany; Tobias.Goetz@chemie.uni-regensburg.de

**\*** Correspondence: Jonathan.Bauer@chemie.uni-regensburg.de

† Dedicated to Dr. Howard Flack (1943–2017).

**Abstract:** Small functional siloxane units have gained great interest as molecular model systems for mimicking more complex silicate structures both in nature and in materials chemistry. The crystal structure of chloropentaphenyldisiloxane, which was synthesized for the first time, was elucidated by single-crystal X-ray diffraction analysis. The molecular crystal packing was studied in detail using state-of-the-art Hirshfeld surface analysis together with a two-dimensional fingerprint mapping of the intermolecular interactions. It was found that the phenyl C–H bonds act as donors for both weak C–H···π and C–H···Cl hydrogen bond interactions. The influence of intramolecular Si–O–Si bond parameters on the acceptor capability of functional groups in intermolecular hydrogen bond interactions is discussed.

**Keywords:** disiloxanes; intermolecular interactions; Hirshfeld surface analysis; molecular models; silicon; X-ray crystallography

#### **1. Introduction**

Siloxanes are known to be quite resistant towards thermal and chemical decomposition [1]. Their structural motif, the Si–O–Si bond, therefore, not only forms the basis for silicate minerals in nature, which are built from both geological [2] and biosilicification processes [3], but also the backbone for technologically important organic–inorganic hybrid polymers (silicones) [4] and for new synthetic silicate materials [5,6]. Studies on small and defined molecular siloxane models can provide very useful information on structure and reactivity of more complex siloxane-based materials and surfaces [7–17]. Silica-based biomimetic model systems [18] have also gained much interest in order to understand natural coral shapes [19] and shell formation of unicellular organisms such as diatoms [20]. We recently reported on monofunctionalized disiloxane units that served as simplified molecular model systems for investigating the reactivity and chemoselectivity in targeted further transformations [15].

The identification of weak intermolecular interactions in molecular crystals is an interesting undertaking with the aim of gaining knowledge about structure-forming forces and making it usable for the targeted formation of functional crystalline networks [21,22]. Siloxanes are of particular interest, since a large number of three-dimensional architectures can be formed through Si–O bond formation [23]. The assembly of several siloxane units to form complex framework structures therefore requires a more detailed study of intermolecular interactions.

As part of our studies on weak intermolecular interactions in molecular crystals [24–26], we were now interested in taking a closer look at monofunctionalized disiloxanes with regard to their crystal packing and intermolecular contacts. Disiloxanes as the smallest units of oligo- and polysiloxanes are a fascinating class of substances, ideal for model studies [27]. Unymmetrically substituted crystalline disiloxanes with only one heterofunction

**Citation:** Bauer, J.O.; Götz, T. Chloropentaphenyldisiloxane— Model Study on Intermolecular Interactions in the Crystal Structure of a Monofunctionalized Disiloxane . *Chemistry* **2021**, *3*, 444–453. https:// doi.org/10.3390/chemistry3020033

Academic Editor: Catherine Housecroft

Received: 14 March 2021 Accepted: 25 March 2021 Published: 29 March 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

are extremely rare [15,28–33], but they shed light on the influence of single substituents on the Si–O–Si unit [15]. Chlorosilanes in general are important precursors for the synthesis of other functional silanes [34] and used as silylation reagents for surface modifications [35]. We therefore chose a chlorodisiloxane (**2**) as an appropriate model system, which is only equipped with aryl groups as additional substituents in order to examine the role of C–H···π and C–H···Cl–Si interactions in the crystalline state more closely by using stateof-the-art analytical methods, Hirshfeld surface analysis [36] along with two-dimensional (2D) fingerprint plots [37]. As already successfully applied in previous work [15,38–40], we took advantage of the good crystallization properties that result when compounds are equipped with triphenylsiloxy groups.

#### **2. Experimental Details**

#### *2.1. General Remarks*

All experiments were performed under an inert atmosphere of purified nitrogen by using standard Schlenk techniques. Glassware was heated at 140 ◦C prior to use. Dichloromethane, pentane, tetrahydrofuran, and toluene were dried and degassed with an MBraun SP800 solvent purification system. *n*-Butyllithium (2.5 M solution in hexane, Merck KGaA, Darmstadt, Germany), dichlorodiphenylsilane (98%, Merck KGaA, Darmstadt, Germany), and triphenylsilanol (98%, Merck KGaA, Darmstadt, Germany) were used without further purification. [D6]-Benzene used for NMR spectroscopy was dried over Na/K amalgam. NMR spectra were recorded on a Bruker Avance 300 (300.13 MHz, Bruker Corporation, Billerica, MA, USA) and a Bruker Avance III HD 400 (400.13 MHz) spectrometer at 25 ◦C. Chemical shifts (*δ*) are reported in parts per million (ppm). 1H and 13C{1H} NMR spectra are referenced to tetramethylsilane (SiMe4, *δ* = 0.0 ppm) as external standard, with the deuterium signal of the solvent serving as internal lock and the residual solvent signal as an additional reference. The 29Si NMR spectrum is referenced to SiMe4 (*δ* = 0.0 ppm) as the external standard. For the assignment of the multiplicities, the following abbreviations were used: s = singlet, m = multiplet. High resolution mass spectrometry was carried out on a Jeol AccuTOF GCX spectrometer. Elemental analysis was performed on a Vario MICRO cube apparatus. The IR spectrum was recorded on a Bruker ALPHA FT-IR spectrometer equipped with a diamond ATR unit. For the intensities of the bands, the following abbreviations were used: s = strong, m = medium, w = weak.

#### *2.2. Synthesis of Ph2SiCl(OSiPh3)* (**2**)

*n*-Butyllithium (22.0 mL of a 2.5 M solution in hexane, 55.0 mmol, 1.1 equiv.) was added dropwise to a solution of triphenylsilanol (13.82 g, 50.0 mmol, 1.0 equiv.) in tetrahydrofuran (200 mL) at 0 ◦C. The clear colorless solution was then allowed to slowly warm to room temperature and stirred for 1 h. Then, the reaction mixture was again cooled to 0 ◦C, dichlorodiphenylsilane (**1**) (10.5 mL, 50.0 mmol, 1.0 equiv.) was added and the mixture was allowed to warm to room temperature. The reaction mixture was refluxed for 5 h. After cooling down to room temperature, all volatiles were removed in vacuo and the product was extracted in dichloromethane (100 mL). Again, all volatiles were removed in vacuo and the crude solid material was recrystallized from hot toluene (30 mL). The crystals were isolated via filtration and washed with pentane to obtain compound **2** as a white crystalline solid (15.09 g, 30.6 mmol, 61%). 1H NMR (400.13 MHz, C6D6): *δ* = 7.00–7.15 (m, 15H, *HPh*), 7.71–7.79 (m, 10H, *HPh*). 13C{1H} NMR (75.44 MHz, C6D6): *δ* = 128.3 (s, *CPh*), 130.5 (s, *CPh*), 131.1 (s, *CPh*), 134.2 (s, *CPh*), 134.7 (s, *CPh*), 135.0 (s, *CPh*), 135.7 (s, *CPh*). 29Si NMR (79.49 MHz, C6D6): *δ* = −19.6 (m, *Si*ClPh2), −15.7 (m, *Si*Ph3). HRMS (EI+): C30H25ClOSi2 calcd. *m*/*z* for [M+] 492.1127; found 492.1119. CHN analysis: C30H25ClOSi2 calcd. C 73.07%, H 5.11%; found C 73.11%, H 4.97%. FT-IR (cm−1): 3070 (w), 3024 (w), 1590 (w), 1486 (w), 1427 (m), 1116 (s, Si–O–Si), 1096 (s), 1026 (m), 997 (m), 711 (s), 696 (s), 540 (s), 507 (s), 491 (s), 475 (s).

#### *2.3. X-Ray Crystallography*

Single-crystal X-ray diffraction analysis of chloropentaphenyldisiloxane (**2**) was performed on a GV50 diffractometer equipped with a TitanS2 CCD detector at 123(2) K using graphite-monochromated Cu-K*β* radiation (*λ* = 1.39222 Å). Data collection and reduction was performed using the CrysAlisPro software system, version 1.171.40.14a [41]. The crystal structure was solved with SHELXT 2018/2 [42,43] and a full-matrix least-squares refinement based on *F*<sup>2</sup> was carried out with SHELXL-2018/3 [43–45] using Olex2 [46] and the SHELX program package as implemented in WinGX [47]. A multi-scan absorption correction using spherical harmonics as implemented in SCALE3 ABSPACK was employed [41]. The non-hydrogen atoms were refined using anisotropic displacement parameters. The hydrogen atoms were located on the difference Fourier map and refined independently. The Hirshfeld surface was mapped over *d*norm ranging from −0.0425 to 1.3719 a.u. *d*<sup>i</sup> and *d*<sup>e</sup> in the 2D fingerprint diagrams are the distances from the surface to the nearest atom *interior* and *exterior* to the surface, respectively, and are each given in the range of 0.4 to 3.0 Å. Details on crystal data and structure refinement are summarized in Table 1 (see also Supplementary Materials). The Hirshfeld surface and 2D fingerprint plots including Figures 1–3 and Appendix A Figure A1 were created using CrystalExplorer 17.5 [48].


**Table 1.** Crystal data and structure refinement of chloropentaphenyldisiloxane (**2**).

**Figure 1.** Molecular structure of chloropentaphenyldisiloxane (**2**) (displacement ellipsoids set at the 50% probability level). Selected bond lengths (Å) and angles (◦): Si1–C1 1.8649(12), Si1–C7 1.8636(13), Si1–C13 1.8595(13), Si2–C19 1.8543(14), Si2–C25 1.8516(13), Si2–Cl 2.0700(5), Si1–O 1.6305(10), Si2– O 1.6012(10), Si1–O–Si2 165.08(8), O–Si2–C19 109.26(6), O–Si2–C25 111.83(6), O–Si2–Cl 105.62(4). Shortest intramolecular H···H contact: H2···H8 2.423 Å.

**Figure 2.** 2D fingerprint plots of chloropentaphenyldisiloxane (**2**) showing (**a**) all contributions of intermolecular contacts, (**b**) C···H/H···C (37.9%), and (**c**) Cl···H/H···Cl (8.9%) contacts.

**Figure 3.** Hirshfeld surface analysis of chloropentaphenyldisiloxane (**2**) highlighting C–H···Cl and C–H···π hydrogen bonds (displacement ellipsoids set at the 50% probability level). Distances (Å) and angle (◦) of the C9–H9···Cl contact: C9–H9 0.917, H9···Cl 2.913, C9···Cl 3.669, C9–H9···Cl 140.61. Distances (Å) and angle (◦) of the C15–H15···C22 contact: C15–H15 0.973, H15···C22 2.822, C15···C22 3.748, C15–H15···C22 159.46. Symmetry transformations used to generate equivalent atoms: (i) –1 + x, y, z; (ii) 1–x, 1–y, 1–z; (iii) 1 + x, y, z.

#### **3. Results and Discussion**

According to a synthetic protocol recently published by us [15], chloropentaphenyldisiloxane (**2**) was easily obtained in 61% isolated yield for the first time after reacting dichlorodiphenylsilane (**1**) with lithium triphenylsiloxide (Scheme 1). The use of metallated siloxide reagents for the stepwise and controlled building of organopolysiloxane polymers was impressively shown by Muzavarov and Rebrov [49]. Recrystallization from hot toluene afforded single-crystals of disiloxane **2** suitable for single-crystal X-ray diffraction analysis (Table 1, Figure 1, see also Supplementary Materials). The asymmetric unit of the monoclinic crystal system, space group *P*21/*n*, contains one molecule of compound **2**. The intramolecular bond parameters of the Si–O–Si backbone show a significantly shortened Si2–O bond [1.6012(10) Å], i.e., the bond that contains the silicon atom attached to the chlorine substituent, in comparison to the Si1–O bond [1.6305(10) Å] (Figure 1). It has already been noticed earlier that the Si–O–Si bond angle in chloro-substituted disiloxanes is remarkably larger than in the respective methoxy- and aminodisiloxanes [15]. In compound **2**, the Si1–O–Si2 bond angle of 165.08(8)◦ is even larger than in the previously described [15] chlorodisiloxane MesPhSiCl(OSiPh3) and, together with the short Si2–O bond, may be indicative for a pronounced negative hyperconjugation of the type LP(O)→σ\*(Si–R) [50,51]. However, neither the Si2–Cl bond [2.0700(5) Å] nor the Si–C bonds show any appreciable elongation when compared to other aryl-substituted chlorosilanes [52–54]. A thorough analysis of the Si–O–Si bonding parameters is not only important for organosiloxanes, but also in the crystal chemistry of minerals and has contributed significantly to a deeper understanding of mineral properties [55].

**Scheme 1.** Synthesis of chloropentaphenyldisiloxane (**2**).

Figure 2 shows the 2D fingerprint diagrams of intermolecular interactions in the crystal structure of disiloxane **2**, all contributions (plot **a**) and subdivided into the individual contributions between atoms inside and outside the Hirshfeld surfaces (plots **b** and **c**). As expected, isotropic H···H contacts (52.9%) make the largest percentage contribution to the intermolecular interactions. The point on the Hirshfeld surface where *d*<sup>i</sup> = *d*<sup>e</sup> ≈ 1.2 Å belongs to the shortest intermolecular H···H contact, i.e., H5···H14 (2.455 Å), which is not unusually short for H···H contacts between phenyl groups [25,37,38,56] and almost as long as the shortest intramolecular H···H contact (H2···H8 2.423 Å) (Figures 1 and 2, plot **a**). Two types of short C–H···π (i.e., H···C) contacts can be found in the crystal structure of compound **2** (Figure 2, plot **b**). The closest H···C contact amounts to 2.822 Å (*d*<sup>i</sup> ≈ 1.65 Å, *d*<sup>e</sup> ≈ 1.15 Å), is represented by the spikes, and contains a C–H bond directed towards a single carbon atom (C15–H15···C22) (Figure 3). The other of these shortest C–H···π contacts, located at *d*<sup>i</sup> ≈ 1.8 Å and *d*<sup>e</sup> ≈ 1.1 Å, within the only faintly indicated but typical wing at the lower right of the C···H/H···C contact plot points almost directly to the center of a phenyl ring and can be identified as the C21–H21···π(Ph) interaction with the π-bonded acceptor group containing the carbon atoms C13 to C18 (shortest contact: H21···C18 2.901 Å) (Table 2 and Figure 3).


**Table 2.** H21···C distances of the almost centered C21–H21···π(Ph) contact in compound **2**.

The designation of a C–H···π contact as a hydrogen bond [57,58] applies at least to the most acidic C–H donors such as alkynyl C≡C–H groups [59,60]. C–H···π(Ph) interactions, even with weak C(sp2)–H or even C(sp3)–H donors, generally still have important structuredetermining and directing abilities, although they are borderline cases at the weak end of the hydrogen bond classification [26,59]. It was impressively shown by Nishio et al. [61] that C–H···π interactions can play a crucial role in molecular recognition, for the formation of inclusion compounds, and in controlling specificities in organic reactions. Furthermore, due to their weak but still orienting character, they should also play an important role in the dynamic formation of supramolecular structures of biopolymers during the processes in living cells. Recently, the importance of anisotropic C–H···π interactions in the crystal structure formation of arylmethoxysilanes has also been pointed out [26].

There are no intermolecular C–H···O contacts to be found, which, on the one hand, can be explained by the difficult steric accessibility of the effectively shielded siloxane oxygen atom as a consequence of the large Si–O–Si angle of 165.08(8)◦. On the other hand, this might also have an electronic reason, as recently pointed out by theoretical investigations on the hydrogen bond interaction energy as a function of the Si–O–Si angle [51]. In this picture, the decreased accessibility of oxygen lone electron pairs due to increased negative hyperconjugation may be the reason for the low basicity of the Si–O–Si linkage in compound **2**. This could be interesting with regard to a siloxane—functional group cooperation and lead to the design of precisely defined functional units in which intramolecular Si–O–Sispecific bond parameters can influence the acceptor capabilities of functional groups or vice versa.

The fingerprint plot for the Cl···H/H···Cl contacts shows distinct spikes that closely resemble that of typical hydrogen bonding pattern (Figure 2, plot **c**) [26,37]. In the meanwhile, the existence of C–H···Cl hydrogen bonds has been well documented and evidenced [62–65]. The H···Cl contact in disiloxane **2** is represented by the spike where *d*<sup>i</sup> ≈ 1.7 Å and *d*<sup>e</sup> ≈ 1.1 Å (actually found in the crystal structure: 2.913 Å) and belongs to the C9–H9···Cl–Si2 hydrogen bond (C9···Cl 3.669 Å, C9–H9···Cl 140.61◦) (Figure 3). It is in the range of the sum of the van der Waals radii for hydrogen (1.2 Å) and chlorine (1.75 Å) [66] and is quite the same as found for H···Cl contacts in chloroform at around 2.95 Å [37]. Since the H9···Cl distance is also in the typical range for chloro-substituted hydrocarbons [63], we therefore anticipate an essentially anisotropic contribution of the C– H···Cl–Si hydrogen bond with a directional influence on the crystal packing. The essential directing structure-forming interactions that were identified from the Hirshfeld surface analysis are also clearly reflected in the crystal packing of disiloxane **2** (Figure A1).

For comparison: In MesPhSiCl(OSiPh3) [15], the C···H/H···C and Cl···H/H···Cl contacts with 29.8% and 6.4%, respectively, contribute less to the intermolecular interactions. Although, the directionality of these contacts seems to be less pronounced in MesPhSiCl(OSiPh3), the mesityl CH3 groups can also participate in intermolecular interactions.

#### **4. Conclusions**

Monofunctional disiloxanes are scarce, but helpful model systems in order to provide information on substituent effects on the Si–O–Si structural motif and on the packing in the molecular crystalline state. The present investigation on intermolecular interactions in the crystal structure of a chlorodisiloxane (**2**) was carried out using Hirshfeld surface analysis and 2D fingerprint plots. Two major types of anisotropic short C–H···π contacts and a C– H···Cl–Si hydrogen bond-like interaction were identified to have the strongest directional influence on the packing within the molecular crystal. Although the siloxane unit does not appear to have a pronounced effect on the chlorine substituent in this molecule, it seems worthwhile to address the influence of the siloxane motif on the acceptor capabilities of functional groups directly connected to the Si–O–Si unit in future investigations. The information on intermolecular interactions provided herein may be of particular interest with regard to the design of supramolecular functional polysiloxane architectures.

**Supplementary Materials:** CCDC-2068445 (compound **2**) contains the supplementary crystallographic data for this paper. These data can be obtained free of charge from the Cambridge Crystallographic Data Centre, CCDC, 12 Union Road, Cambridge CB21EZ, UK (Fax: + 44-1223-336-033; E-mail: deposit@ccdc.cam.ac.uk; https://www.ccdc.cam.ac.uk/structures/).

**Author Contributions:** Conceptualization, J.O.B.; methodology, J.O.B. and T.G.; validation, J.O.B. and T.G.; formal analysis, J.O.B.; investigation, J.O.B. and T.G.; resources, J.O.B.; data curation, J.O.B. and T.G.; writing—original draft preparation, J.O.B.; writing—review and editing, J.O.B. and T.G.; visualization, J.O.B.; supervision, J.O.B.; project administration, J.O.B.; funding acquisition, J.O.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the University of Regensburg, the Elite Network of Bavaria (ENB), and the Bavarian State Ministry of Science and the Arts (StMWK), grant number N-LW-NW-2016-366.

**Data Availability Statement:** The data presented in this study are available in Supplementary Materials.

**Acknowledgments:** The authors would like to thank Manfred Scheer and Jörg Heilmann for continuous support and providing laboratory facilities.

**Conflicts of Interest:** The authors declare no conflict of interest.

**Appendix A**

**Figure A1.** Crystal packing of chloropentaphenyldisiloxane (**2**) along the a axis. Hydrogen atoms are omitted for clarity.

#### **References**


*Article*

## **Synthesis and Crystallographic Characterization of X-Substituted 2,4-Dinitrophenyl-4 -phenylbenzenesulfonates** †

#### **Brock A. Stenfors 1, Richard J. Staples 2, Shannon M. Biros <sup>1</sup> and Felix N. Ngassa 1,\***


Received: 9 May 2020; Accepted: 21 May 2020; Published: 15 June 2020

**Abstract:** Treatment of 2,4-dinitrophenol with sulfonyl chlorides in the presence of pyridine results in the formation of undesired pyridinium salts. In non-aqueous environments, the formation of the insoluble pyridinium salt greatly affects the formation of the desired product. A facile method of producing the desired sulfonate involves the use of an aqueous base with a water-miscible solvent. Herein, we present the optimization of methods for the formation of sulfonates and its application in the production of desired x-substituted 2,4-dinitrophenyl-4 -phenylbenzenesulfonates. This strategy is environmentally benign and supports a wide range of starting materials. Additionally, the intermolecular interactions of these sulfonate compounds were investigated using single-crystal x-ray diffraction data.

**Keywords:** arylsulfonates; pyridinium salt formation; single-phase solvent system; sulfonate synthesis; sulfonyl chlorides; X-ray crystallography

#### **1. Introduction**

A reliable method for producing arylsulfonates involves the nucleophilic substitution reaction of alcohols and sulfonyl halides. This reaction is highly efficient in creating the sulfonate ester, a synthetically important electrophile in organic chemistry due to its high chemical reactivity [1]. This property has been utilized in the detection and fluorescent imaging of biologically important compounds. In particular, recent research has shown that 2,4-dinitrobenzenesulfonate-functionilized carbon dots (g-CD-DNBS) are significant regarding the detection and fluorescence imaging of biothiols (Scheme 1) [2].

**Scheme 1.** The fluorescence response from the treatment of g-CD-DNBS with biothiols.

The 2,4-dinitrobenzenesulfonyl (DNBS) group resembles the x-substituted 2,4-dinitrophenyl-4 -phenylbenzenesulfonate compounds. Similarly, this resemblance is found in the fluorescent probe O-hNRSel. O-hNRSel was developed for the detection and imaging of selenocysteine (Sec) in living cells (Scheme 2) [3]. Sec is a selenium-containing amino acid that resides in proteins of organisms and viruses. Selenium is linked to several health benefits, including the prevention of cancer and cardiovascular diseases [4]. The role of Sec in human health underlines the importance of Sec fluorescence probes.

**Scheme 2.** The formation of the probe O-hNRSel from the treatment of O-hNR with Sec in the presence of triethylamine (Et3N) and dichloromethane (CH2Cl2).

A facile synthesis of sulfonates is necessary to produce these biologically significant compounds. A review of the current literature suggests the use of sulfonic acids or sulfonyl halides for synthesizing sulfonates [5,6]. The compounds synthesized hereafter were done so through the sulfonylation of 2,4-dinitrophenol with sulfonyl chloride in the presence of a base. Basic conditions allow for the neutralization of hydrochloric acid. The presence of hydrochloric acid lowers the pH of the reaction, resulting in poor deprotonation of the intermediary structure. Preliminary experiments for the synthesis of 2,4-dinitrophenyl-4-methylbenzenesulfonate involved the treatment of 4-methylbenzenesulfonyl chloride (**1**) with 2,4-dinitrophenol (**2**) in the presence of pyridine and dichloromethane. This reaction was done under N2 atmosphere to avoid hydrolysis of 4-methylbenzenesulfonyl chloride. The production of the insoluble pyridinium salt, the pyridine adduct of 2,4-dinitrophenyl 4-methylbenzenesulfonate (**3**), was evidence of an unsuccessful reaction. Compound **3** is insoluble in dichloromethane and pulled out of the solution as a white precipitate. The proposed mechanism for the formation of the pyridinium salt (**3**) in the presence of pyridine is found in Scheme 3. The desired sulfonate (**4**) was not formed as a result of compound **3** being insoluble in solution. A review of the literature supports the theory that a pyridinium salt was formed [7].

**Scheme 3.** The proposed mechanism for the treatment of 4-methylbenzenesulfonyl chloride (**1**) with 2,4-dinitrophenol (**2**) in the presence of pyridine and dichloromethane to form the pyridinium adduct of 2,4-dinitrophenyl 4-methylbenzenesulfonate (**3**). Shown below this is the desired rearrangement of compound **3** to give the sulfonate product (**4**).

A new synthetic method was proposed and applied to the synthesis of three x-substituted 2,4-dinitrophenyl-4 -phenylbenzenesulfonates. The resulting products were characterized by crystallographic and spectroscopic means. Crystallographic characterization offers insight into the structural features and inter- and intramolecular interactions of molecules. Herein, we report the facile synthesis and characterization of x-substituted 2,4-dinitrophenyl-4 -phenylbenzenesulfonates.

#### **2. Experimental**

The reagents used in the synthesis of x-substituted 2,4-dinitrophenyl-4 -phenylbenzenesulfonates were obtained from commercial sources and used without further purification. Thin-layer chromatography (TLC) was used to track reaction progress and obtain Rf values for the reactions.

*Preparation of 2,4-Dinitrophenyl 4-methylbenzenesulfonate in the presence of N,N-diisopropylethylamine* (**4**). 2,4-dinitrophenol (2.025 g, 11.0 mmol) was dissolved in a flask containing 11 mL of chilled dichloromethane. 4-Methylbenzenesulfonyl chloride (2.090 g, 11.0 mmol) was added dropwise to the stirred solution. This was followed by the dropwise addition of N,N-diisopropylethylamine (3.8 mL, 21.8 mmol). The stirred solution was left at room temperate for 24 h under N2 atmosphere. After 24 h, the mixture was diluted with 15 mL of dichloromethane and transferred to a separatory funnel. The organic layer was washed with water (3 × 10 mL). The aqueous layers were combined and back-extracted with 10 mL of dichloromethane. All organic layers were combined, washed with brine (10 mL), and dried over anhydrous sodium sulfate. Evaporation of solvent yielded a solid, yellow residue as a crude product. Purification via column chromatography with dichloromethane as the solvent yielded a yellow powder (1.712 g, 46%). M.p. 196–202 ◦C. Rf = 0.74 (CH2Cl2). 1H-NMR (400 MHz, DMSO-*d*6) δ 8.79 (d, *J* = 2.8 Hz, 1H), 8.54 (dd, *J* = 9.1, 2.8 Hz, 1H), 7.78–7.71 (m, 2H), 7.54 (d, *J* = 9.1 Hz, 1H), 7.49 (m, 2H), 2.42 (s, 3H). HRMS (ESI): cald. For C13H10N2NaO7S [M + Na]<sup>+</sup> 361.0100; found 361.0110.

*Preparation of pyridinium adduct of 2,4-Dinitrophenyl 4-methylbenzenesulfonate in the presence of pyridine* (**3**). 2,4-dinitrophenol (2.016 g, 11.0 mmol) was dissolved in a flask containing 11 mL of chilled dichloromethane. 4-Methylbenzenesulfonyl chloride (2.088 g, 10.9 mmol) was added dropwise to the stirred solution. This was followed by the dropwise addition of pyridine (1.80 mL, 21.8 mmol). The stirred solution was left at room temperate for 24 h under N2 atmosphere. After reaction completion, an insoluble precipitate was isolated via vacuum filtration giving a fine, white powder. Recrystallization in 2:1 acetone/water, followed by trituration with hexanes afforded the product as small, translucent crystalline sheets (1.875 g, 41% yield). M.p. 240–248 ◦C. Rf = 0.70 (80:20 ACN/H2O w/1 mL NH4OH). 1H-NMR (400 MHz, DMSO-*d*6) δ 9.40–9.33 (m, 2H), 9.05 (d, *J* = 2.5 Hz, 1H), 8.96–8.82 (m, 2H), 8.50–8.35 (m, 3H), 7.43–7.36 (m, 2H), 7.06 (d, *J* = 7.8 Hz, 2H), 2.24 (s, 3H).

*Preparation of 2,4-Dinitrophenyl 4-methylbenzenesulfonate in the presence of triethylamine* (**4**). 2,4-dinitrophenol (1.999 g, 10.9 mmol) was dissolved in a flask containing 11 mL of chilled dichloromethane. 4-Methylbenzenesulfonyl chloride (2.076 g, 10.9 mmol) was added dropwise to the stirred solution. This was followed by the dropwise addition of triethylamine (3.0 mL, 21.9 mmol). The stirred solution was left at room temperate for 24 h under N2 atmosphere. After 24 h, the mixture was diluted with 15 mL of dichloromethane and transferred to a separatory funnel. The organic layer was washed with water (3 × 10 mL). The aqueous layers were combined and back-extracted with 10 mL of dichloromethane. All organic layers were combined, washed with brine (10 mL), and dried over anhydrous sodium sulfate. The resulting solution was evaporated to afford a solid, yellow residue. The crude product was recrystallized in 2:1 DCM/ethyl acetate to afford large, pale-yellow, translucent crystals (1.768, 50%).

*Preparation of 2,4-Dinitrophenyl 4-methylbenzenesulfonate in the presence of aqueous sodium hydroxide* (**4**). 2,4-dinitrophenol (1.021 g, 5.25 mmol) was dissolved in a flask containing 10 mL of THF. 4-Methylbenzenesulfonyl chloride (1.052 g, 5.25 mmol) was then added to the flask, followed by the dropwise addition of 1 M NaOH (10 mL, 10.5 mmol). The solution was stirred at room temperate for 6 h. After reaction completion, a yellow precipitate was isolated via vacuum filtration to afford a yellow powder. The crude product was recrystallized in ethanol to afford large, pale-yellow, translucent crystals (1.204 g, 66%).

*Preparation of 2,4-Dinitrophenyl 2,4,6-trimethylbenzenesulfonate* (Table 1, *entry 1*). 2,4-dinitrophenol (0.842 g, 4.57 mmol) and 2,4,6-trimethylbenzenesulfonyl chloride (1.002 g, 4.58 mmol) were added to a flask

containing 10 mL of tetrahydrofuran. Of aqueous potassium carbonate 0.915 M (10 mL, 9.15 mmol) was added dropwise to the stirred solution. The solution was then stirred at room temperate for 6 h. After reaction completion, a yellow precipitate was isolated via vacuum filtration to afford a yellow powder. The crude product was recrystallized in ethanol to afford large, pale-yellow, translucent crystals (1.470 g, 88%). M.p. 128–132 ◦C. Rf = 0.86 (CH2Cl2). 1H-NMR (400 MHz, Chloroform-*d*) δ 8.75 (d, *J* = 2.8 Hz, 1H), 8.44 (dd, *J* = 9.0, 2.8 Hz, 1H), 7.58 (d, *J* = 9.0 Hz, 1H), 7.03 (s, 2H), 2.57 (s, 6H), 2.35 (s, 3H). HRMS (ESI): cald. For C15H14N2NaO7S [M + Na]<sup>+</sup> 389.0400; found 389.0410.

**Table 1.** Solvent and counter-ion effects on the synthesis of 2,4-dinitrophenyl sulfonate derivatives a.


<sup>a</sup> Reaction conditions: 2,4-dinitrophenol (1.0 eq.) was dissolved in 10 mL of solvent. Sulfonyl chloride (1.0 eq) was added dropwise to the solution, followed by the dropwise addition of base (2.0 eq). All reactions were run at room temperature. <sup>b</sup> Reaction performed under N2 atmosphere. <sup>c</sup> Only pyridinium salt isolated.

*Preparation of 2,4-Dinitrophenyl-4 -phenylbenzenesulfonate* (**1b**). 2,4-dinitrophenol (0.729 g, 3.96 mmol) and biphenyl-4-sulfonyl chloride (1.005 g, 3.96 mmol) were added to a flask containing 10 mL of tetrahydrofuran. Of aqueous potassium carbonate 1.6 M (5.0 mL, 7.92 mmol) was added dropwise to the stirred solution. The stirred solution was left at room temperate for 8 h. After reaction completion, a yellow precipitate was isolated via vacuum filtration to afford a yellow powder. The crude product was recrystallized in a small amount of dichloromethane to afford pale-yellow translucent crystals (1.136 g, 72% yield). M.p. 137–140 ◦C, Rf = 0.69 (CH2Cl2). 1H-NMR (400 MHz, Chloroform-*d*) δ 8.78 (d, *J* = 2.7 Hz, 1H), 8.50 (dd, *J* = 9.0, 2.7 Hz, 1H), 8.01–7.96 (m, 2H), 7.82–7.75 (m, 3H), 7.64–7.60 (m, 2H), 7.53–7.43 (m, 3H). HRMS (ESI): cald. For C18H12N2NaO7S [M + Na]<sup>+</sup> 423.0300; found 423.3700.

*Preparation of 2,4-Dinitrophenyl-4 -(4-methylphenyl)-benzenesulfonate* (**2b**). 2,4-dinitrophenol (0.692 g, 3.75 mmol) and 4 -methylbiphenyl-4-sulfonyl chloride (1.001 g, 3.75 mmol) were added to a flask containing 10 mL of tetrahydrofuran. Of aqueous potassium carbonate 1.6 M (5.0 mL, 7.92 mmol) was added dropwise to the stirred solution. The stirred solution was left at room temperate for 8 h. After reaction completion, a yellow precipitate was isolated via vacuum filtration to afford a yellow powder. The crude product was recrystallized in a small amount of dichloromethane to afford pale-yellow translucent crystals (1.075 g, 69% yield). M.p. 152–155 ◦C, Rf = 0.66 (CH2Cl2). 1H NMR (400 MHz, Chloroform-*d*) δ 8.78 (d, *J* = 2.7 Hz, 1H), 8.50 (dd, *J* = 9.0, 2.7 Hz, 1H), 7.99–7.93 (m, 2H), 7.81–7.74 (m, 3H), 7.56–7.49 (m, 2H), 7.30 (d, *J* = 7.9 Hz, 2H), 2.42 (s, 3H). HRMS (ESI): cald. For C19H14N2NaO7S [M + Na]<sup>+</sup> 437.0400; found 437.3800.

*Preparation of 2,4-Dinitrophenyl-4 -(4-flourophenyl)-benzenesulfonate* (**3b**). 2,4-dinitrophenol (0.680 g, 3.69 mmol) and 4 -flourobiphenyl-4-sulfonyl chloride (1.006 g, 3.69 mmol) were added to a flask containing 10 mL of tetrahydrofuran. Of aqueous potassium carbonate 1.6 M (5.0 mL, 7.92 mmol) was added dropwise to the stirred solution. The stirred solution was left at room temperate for 8 h. After reaction completion, a yellow precipitate was isolated via vacuum filtration to afford a yellow powder. The crude product was recrystallized in a small amount of dichloromethane to afford pale-yellow translucent crystals (0.917 g, 59% yield). M.p. 140–143 ◦C, Rf = 0.64 (CH2Cl2). 1H-NMR

(400 MHz, Chloroform-*d*) δ 8.78 (d, *J* = 2.7 Hz, 1H), 8.51 (dd, *J* = 9.0, 2.7 Hz, 1H), 8.02–7.94 (m, 2H), 7.82–7.71 (m, 3H), 7.65–7.55 (m, 2H), 7.24–7.14 (m, 2H). HRMS (ESI): cald. For C18H11FN2NaO7S [M + Na]<sup>+</sup> 441.0200; found 441.3500.

#### *Spectroscopic and Crystallographic Characterization*

1H-NMR spectra (400 MHz) were recorded on a JEOL ECZ400 spectrometer using a DMSO-*d*<sup>6</sup> or Chloroform-*d* solvent. Chemical shifts are reported in parts per million (ppm, δ) relative to the residual solvent peak, and coupling constants (J) are reported in Hertz (Hz). Results were analyzed and figures were created with the use of MestReNov [8]. The spectra of all compounds synthesized can be found in Figures S1–S8. Spectra were obtained from 16 scans.

The data yielded from the crystallographic characterization of x-substituted 2,4-dinitrophenyl -4 -phenylbenzenesulfonates can be found in Tables S1–S15. Molecular structures of 2,4-dinitrophenyl -4 -phenylbenzenesulfonate (**1b**), 2,4 -dinitrophenyl-4 -(4-methylphenyl)-benzenesulfonate (**2b**), and 2,4 -dinitrophenyl-4 -(4-flourophenyl)-benzenesulfonate (**3b**) can be found in Figures S9–S11. X-ray diffraction was carried out on a Bruker APEXII CCD diffractometer with Mo *K*α radiation. The software used for data collection is as follows: Data collection: *APEX2* [9]; cell refinement: *SAINT* [10]; data reduction: *SAINT* [10]; program used to solve structure: ShelXT [11]; program used to refine structure: *OLEX2* [12,13]; program used to generate figures: Mercury [14–18]; and absorbance correction: *SADABS* [19]. Interactive links for the structures of x-substituted 2,4-dinitrophenyl-4 -phenylbenzenesulfonates are found in the following figures: 2,4-dinitrophenyl -4 -phenylbenzenesulfonate (**1b**), Figure S9; 2,4-dinitrophenyl -4 -(4-methylphenyl)-benzenesulfonate (**2b**), Figure S10; 2,4-dinitrophenyl-4 -(4-flourophenyl)-benzenesulfonate (**3b**), and Figure S11.

#### **3. Results and Discussion**

In our work towards developing a facile synthesis of sulfonates, various reaction conditions were investigated. A variety of solvent and base combinations were used to ascertain their effects on the yield and reaction time of sulfonate derivatives, the results of which can be found in Table 1. Entries are presented in order of decreasing yield. Successful formation of the desired product occurred in all cases, except entry 8 where dichloromethane and pyridine were used. This led to the development of a new synthetic method using an aqueous base and the water-miscible solvent, tetrahydrofuran. This single-phase solvent system defeats the need for a phase transfer catalyst and can support a wider range of starting materials. The use of an aqueous base and tetrahydrofuran resulted in higher yields, less environmental impact, and shorter reaction times. Additionally, the desired sulfonate product was isolated directly from the reaction mixture with good purity. An initial concern with this new synthetic method was the hydrolysis of sulfonyl chloride due to the presence of water. However, results show the rate of hydrolysis has little effect on yield.

A comparison of entries 2, 3, and 5 revealed the electronic and steric effects from biphenyl sulfonyl chloride derivatives and their implications in the synthesis of x-substituted 2,4-dinitrophenyl-4 -phenylbenzenesulfonates. The structure and yield of these sulfonates are shown in Figure 1. The highest yielding x-substituted 2,4-dinitrophenyl-4 -phenylbenzenesulfonate was compound **1b**, which has no substituent. Compounds **2b** and **3b**, which contain electron-donating and electron-withdrawing groups, respectively, were slightly lower in yield. In general, the reaction yield may be affected by electronic factors, steric factors or a combination of both. Although it is not clear to us which factor is responsible for the difference in yields, one can imagine a combination of electronic factors and steric factors to be responsible. However, the difference in yields was not statistically significant enough for us to draw clear conclusions on what exactly accounts for the difference in reaction yields.

**Figure 1.** The structure and yield of the synthesized x-substituted 2,4-dinitrophenyl-4 -phenylbenzenesulfonates. Reaction conditions: 2,4-dinitrophenol (1.0 eq) was dissolved in 10 mL of THF. Sulfonyl chloride (1.0 eq) was added dropwise to the solution, followed by the dropwise addition of 1.6 M aq. K2CO3 (2.0 eq). All reactions were run at room temperature.

Crystal structures of the x-substituted 2,4-dinitrophenyl-4 -phenylbenzenesulfonates were obtained by single crystal X-ray diffraction. Pertinent data such as bond angles, bond lengths, torsion angles, and other crystallographic parameters can be found in the supplementary material. The asymmetric units of 2,4-dinitrophenyl-4 -phenylbenzenesulfonate (**1b**), 2,4-dinitrophenyl-4 -(4-methylphenyl)-benzenesulfonate (**2b**), and 2,4-dinitrophenyl-4 -(4-flourophenyl)-benzenesulfonate (**3b**) can be found in Figure 2.

**Figure 2.** The molecular structure of 2,4-dinitrophenyl-4 -phenylbenzenesulfonate (**1b**; **top**, **middle**), 2,4-dinitrophenyl-4 -(4-methylphenyl)-benzenesulfonate (**2b**; **bottom**, **right**), and 2,4-dinitrophenyl-4 -(4-flourophenyl)-benzenesulfonate (**3b**; **bottom**, **left**) with atom labeling scheme. Displacement of ellipsoids is shown at the 50% probability level. Hydrogen atoms have been omitted for clarity.

The x-substituted 2,4-dinitrophenyl-4 -phenylbenzenesulfonates exhibited a two-fold screw axis (−x, 1/2 + y, 1/2 − z), and glide plane geometry (x, 1/2 − y, 1/2 + z) with an inversion center (−x, −y, −z). Screw axis and glide plane geometries are indicative of efficient packing. All three sulfonate structures exhibited a monoclinic system (*P*21/*c* space group). The central sulfur atom, S1, is a slightly distorted tetrahedron according to the τ<sup>4</sup> descriptor for four-fold coordination [20]. The aryl groups of the sulfonates were oriented gauche about the S1–O1 bond with the following C7–S1–O1–C1 torsion angles: 131.6 (1)◦, **1b**; −94.0 (1)◦, **2b**; −92.7 (1)◦, and **3b** (Table 2). The S1 = O2 and S1 = O3 bond lengths were in good agreement with known values. The S1–C7 and S1–O1 bond lengths were 1.751 (2) Å and 1.626 (1) Å for compound **1b**; 1.745 (2) Å and 1.619 (1) Å for compound **2b**; and 1.743 (2) Å and 1.623 (1) Å for compound **3b**, respectively. The S1–O1–C1 bond angles for compounds **1b**, **2b**, and **3b** were 120.5 (1)◦, 120.4 (1)◦, and 119.2 (1)◦, respectively. Molecules were liked by π–π interactions, C–H···O

hydrogen bonds, C–H···π interactions, and, in the case of compound **3b**, C–H···F hydrogen bonds (Figure 3). Hydrogen atoms bonded to carbon atoms were placed in calculated positions and refined as riding: C–H = 0.95–1.00 Å with a fixed *U*iso(H) = 1.2*U*eq(C) for all C–H groups.

**Table 2.** Bond lengths of S1–C7, S1–O1, S1 = O2, and S1 = O3 (Å), bond angles of S1–O1–C1 (◦) and torsion angles of C7–S1–O1–C1 (◦) for 2,4-dinitrophenyl-4 -phenylbenzenesulfonate (**1b**), 2,4-dinitrophenyl-4 -(4-methylphenyl)-benzenesulfonate (**2b**), and 2,4-dinitrophenyl-4 -(4 flourophenyl)-benzenesulfonate (**3b**).


**Figure 3.** A depiction of the inter- and intramolecular hydrogen bonds present in the crystal structures of 2,4-dinitrophenyl-4 -phenylbenzenesulfonate (**1b**; **right**), 2,4-dinitrophenyl-4 -(4-methylphenyl)-benzenesulfonate (**2b**; **middle**), and 2,4-dinitrophenyl -4 -(4-flourophenyl)-benzenesulfonate (**3b**; **left**) shown as capped sticks with standard CPK colors. Hydrogen bonds and contacts are depicted with cyan dashed lines.

The extent of hydrogen bonding varied significantly in the sulfonate derivatives. In the case of compound 2b, only two C–H··· O hydrogen bonds were observed (Table 3). A greater number of hydrogen bonds in compounds **1b** and **3b** lend credence to a lattice dependent on hydrogen bond contacts for efficient packing. The F1···H17 bond in compound **3b** had a bond length of 2.450 Å, the shortest and therefore strongest hydrogen bond of the sulfonate compounds.


**Table 3.** Length of hydrogen-bond contacts (Å) and corresponding symmetry codes for 2,4 -dinitrophenyl-4 -phenylbenzenesulfonate (**1b**), 2,4-dinitrophenyl-4 -(4-methylphenyl)-benzenesulfonate (**2b**), and 2,4-dinitrophenyl-4 -(4-flourophenyl)-benzenesulfonate (**3b**). Atoms labels follow the atom numbering scheme in Figure 2.

#### **4. Conclusions**

The proposed synthetic method is useful in producing a broad range of sulfonates in good yields with the elimination of chlorinated solvents and organic bases, compared to previous methods. Additionally, the new synthetic method significantly decreased reaction time and work up. It was determined that the reaction favored a single-phase solvent system involving an aqueous base and water-miscible solvent. This facile method produced sulfonate products in high purity. The resulting products were characterized by crystallographic and spectroscopic methods. Crystallographic characterization revealed sulfonate structures exhibiting a monoclinic system (*P*21/*c* space group) with a two-fold screw axis (−x, 1/2 + y, 1/2 − z) and glide plane geometry (x, 1/2 − y, 1/2 + z) and inversion center (−x, −y, −z). Values obtained for the S = O bond lengths closely resembled known values. Four-fold coordination about the S1 atom reviled a slightly distorted tetrahedron geometry. The crystallographic results supported the successful formation of pure sulfonate products. The resulting products will be used to investigate the reaction conditions for the regioselective cleavage of C–O and S–O bonds in the amination of arylsulfonates.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2624-8549/2/2/36/s1, Figures S1–S8 showing 1H-NMR spectra, Figures S9–S11 depicting asymmetric units of x-substituted 2,4-dinitrophenyl-4 -phenylbenzenesulfonates with interactive links, and tables S1–S15 containing the crystallographic data for these compounds.

**Author Contributions:** Conceptualization, B.A.S. and F.N.N.; Methodology and Practical Chemical and Spectroscopic Studies, B.A.S. and F.N.N.; Crystallography, B.A.S., R.J.S., S.M.B., and F.N.N.; Writing Draft Report, B.A.S. and F.N.N.; Writing of Manuscript and Editing, B.A.S. and F.N.N.; Project Administration and Funding Acquisition, F.N.N. All authors have read and agreed to the published version of the manuscript.

**Funding:** Funding for this research was provided by the following: National Science Foundation (grant No. MRI CHE-1725699); Grand Valley State University Chemistry Department's Weldon Fund.

**Acknowledgments:** The authors thank Pfizer, Inc. for the donation of a Varian INOVA 400 FT NMR spectrometer. The CCD-based X-ray diffractometers at Michigan State University were upgraded and/or replaced by departmental funds.

**Conflicts of Interest:** No potential conflict of interest was declared by the authors.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

#### *Article*
