New invariom names of double bonded oxygen for invariom refinementsinvariom refinements

Another improvement concerning the classification of terminal double bonded oxygen atoms also relates to the database in a wider sense. So far multipole parameters for the invariom O2c had always been transferred from formaldehyde and O2n from aminoxide. As expected next-nearest neighbours (NNN) were found to influence the ESP from the invariom charges considerably (see Section 4.3.1) for carbonyl oxygen atoms. Thus, a more differentiated classification of those atoms improved the transferability of invariom point charges. Hence, atoms that are two bonds away are now included in the invariom name for charges so that there are 19 new invariom names instead of the one before.

If this finer classification would also improve the performance of scattering factors was examined for those molecules of the database test set which contained an invariom ’O2c’

or ’O2n’ by the old nomenclature. The results are shown in Table 2.1.

Table 2.1:Comparison of R1 for invariom refinements with scattering factors assigned by different invariom names for double bonded oxygen atoms.

structure structural formula R1 for new model compound

code O2x O2x[NNN]

luckno O O

O

HO OH

OH 1.48 % 1.48 % acetic acid

nactyr

O O

O HN

OH

H₂O

1.04 % 1.02 % acetic acid

npp

N OH N

N O

O 1.88 % 1.87 % 2-nitrobenzeneamine

eg3095

NH OOC

HO

OH O

OH

2.56 % 2.55 % benzoic acid

For three of the four test structures scattering factors transferred based on the elongated invariom names improved the fit to the XRD data. Therefore, this change in invariom names was also implemented inInvariomTool.

compounds

3.1 Introduction

Structure determination from single crystal XRD has become very fast in the past decades, allowing in easy cases data collection, structure solution, refinement and preparation for publication in Acta Crystallographica E within one day.^[111] Reasons for this are improve-ments in instrumentation and software. Data collection has been accelerated by intro-duction of area detectors, which nowadays can be operated in continuous readout mode.

Software has automated most steps and is easier to use, so that small-molecule structures are solved within seconds. Thus, the number of crystal structures published each year has increased exponentially, as shown by the statistics of the Cambridge Structural Database (CSD),^[77,112] where most of the published structures are deposited. Since

"the number of experienced crystallographers dedicated to single-crystal studies has certainly not increased in proportion to the number of reported studies",^[111]

automated validation of completeness, quality and correctness is required before crystal structures are submitted to the CSD. Incorrect structures cause problems especially for research that relies on them.

Numerous studies employ information from the CSD.^[113] Most of them are of statistical nature,^[113] derive properties,^[114] investigate methods^[115] or are simply based on selected structures obtained from the database.^[116] According to the Cambridge Crystallographic Data Center (CCDC) homepage 17 publications were based upon information from the CSD in the first nine month of 2016.

Completeness and quality of a structure are usually ensured by automatic structure vali-dation through CheckCIF.^[111] Validating the correctness of a structure is more challenging.

Erroneous structures in which hydrogen atoms are either missing or misplaced, and ob-viously incorrectly assigned atom-types can be identified by specific indicators.^[111] The information for those indicators is deduced from the structure models. Investigating reflec-tion data and comparing reflecreflec-tion files from different compounds can yield indicareflec-tions to possible fraud.^[117–120] Such comparisons of different data sets can reveal that two probably isomorphous structures have reflection data deviating only by a scale factor, implying a linear correlation if both data are plotted against each other. In such cases only one of the structures can be correct, but automatic methods so far can not tell which of the compared structures is the correct one, if in principle both are chemically possible. Coordination compounds have a rather high flexibility concerning the geometry of ligands. It requires a chemists knowledge and experience to discern which of two structures with different metal centers is more likely. Likelihood, however, is a weak argument when judging other scientists

work. Therefore, a method delivering proof of the correct metal in otherwise isomorphous models is required.

This chapter of the thesis applies a method for identifying the metal atom that com-pares the ability to fit XRD data for models that include atomic asphericity. Since valence density strongly affects low order data,^[30] a better description of bonding ED is also a model improvement for data collected to standard resolution (0.83 Å). The same region of data is affected by the difference between metals that have a similar number of electrons.

Additionally, a better model improves crystallographic phases and hence, the general distin-guishing power between different models. The method applied in this thesis to distinguish metals in crystal structures of coordination compounds by aspherical scattering factors was developed and validated for complexes, which had been synthesized in-house with different metals as central atom.^[99]

In this project eleven pairs of crystal structures^[121–142] from the CSD with pairwise the same cell and compound geometry but different metals as central atom were investigated.

The crystal structure pairs were identified by Jim Simpson and Matthias Weil. In some cases the reflection data sets were not the same, but isomorphism was still questioned due to similar cell and molecule geometries.

Alternatively to the procedure described and applied here, it would of course be better to apply other chemical analyses to identify the correct metal atom, if the compound is at hand.

But for this project only the deposited crystal structure data was available. Thankfully not only the models but also reflection data were deposited, otherwise the method could not be applied to identify the correct structure. A synthetic approach of trying to synthesize and recrystallize each of the possible complexes and redetermination of each crystal structure would show which of the structures can be reproduced. However, doubts concerning the reason for non-reproducible structures would have to be resolved and discovery of new polymorphs or co-crystals would also hinder the conclusion of such an synthetic approach.

3.1.1 Isomorphism

For those compounds where the XRD data sets were not basically equal, real isomor-phism was theoretically possible. Before discussing this subject, the terms isomorisomor-phism and isostructuralism shall be specified. The definition of isostructural crystals as given by the IUCr Online Dictionary of Crystallography is:

"Two crystals are said to be isostructural if they have the same structure, but not necessarily the same cell dimensions nor the same chemical composition, and with a ’comparable’ variability in the atomic coordinates to that of the cell dimensions and chemical composition."^[143]

The definition for isomorphous crystals by the same reference is:

"Two crystals are said to be isomorphous if (a) both have the same space group and unit-cell dimensions and (b) the types and the positions of atoms in both are the same except for a replacement of one or more atoms in one structure with different types of atoms in the other (diadochy), such as heavy atoms, or the presence of one or more additional atoms in one of them (isomorphous addition)."^[143]

From those definitions isomorphism is the more precise description for the pairs of structures investigated, since they have the same cell, space group and just differ by one atom type.

Hence, the isomorphism discussed here concerns diadochy.

Due to more degrees of freedom in the ligand, true isomorphism occurs less frequently when the ligand size increases. In complexes of3d-metals different coordination geometries are favored for different numbers of electrons and thus different electronic configurations.

Bond distances can change due to e.g. Jahn-Teller (JT) splitting.^[144] Therefore, unit cells and molecular geometry are usually not the same, so that true isomorphism occurs rarely. If isomorphous structures are found for two complexes, their chemical properties are commonly compared further, leading to a topic of interest for chemical rather than crystallographic journals.

Several cases¹ for isomorphous structures of octahedrally coordinated manganese(II) and cobalt(II) are known.^[145–147] In an example of isomorphous octahedral nickel(II) and cobalt(II) complexes^[148] a striking geometrical difference between the compounds is that the largest angle between two oxygens atoms is once 152.08(6)^◦ and once 157.17(5)^◦, while the bond distances differ by around 0.08 Å. The changes between bond lengths of nickel and cobalt are smaller (0.03 Å),^[145] but bond angles differ by up to three degrees. Between the cobalt and manganese complexes the bond lengths differ by values between 0.05 and 0.3 Å.^[146] Cell parameters were observed to change by 0.03 Å^[145] to 0.3 Å^[146] and even 0.4 Å.^[147] These numbers indicate the considerable extent of structural changes related to exchange of metal atoms in isomorphous structures.

Additional chemical analytics should confirm the different metal atoms, if differences between ’isomorpous’ complexes in crystal structure models are less or not even significant, especially if the coordination geometry is uncommon for one of the metals. Otherwise the isomorphism is suspicious even if XRD data are collected in different experiments, since the frequently applied IAM can not identify the correct metal atom.