VSEPR Chart | Valence Shell Electron Pair Repulsion Theory

A pretty good read on the VESPR current bonding-model explanation from Sigma-Aldrich with notes on the CCCBDB - Computational Chemistry Comparison and Benchmark DataBase  database at NIST. I may need to add portions of these to the Graph Database model for each element on the Periodic Table.

Here is Miles' "Bonding" model:  http://milesmathis.com/ionic.pdf

NIST's CCCBDB database
Calculated Electron Affininty for O (Oxygen atom)
Experimental Electron Affinity is 1.46198 ± 0.00043 eV
https://cccbdb.nist.gov/elecaff2x.asp?casno=17778802
https://cccbdb.nist.gov/glossary.asp --list of acronyms in their Database
------
https://www.sigmaaldrich.com/RO/en/technical-documents/technical-article/chemistry-and-synthesis/organic-reaction-toolbox/vsepr-chart-valence-shell-electron-pair-repulsion-theory

VSEPR Chart | Valence Shell Electron Pair Repulsion Theory

WHAT IS VSEPR THEORY?
The valence shell electron pair repulsion (VSEPR) theory is a model used to predict 3-D molecular geometry based on the number of valence shell electron bond pairs among the atoms in a molecule or ion. This model assumes that electron pairs will arrange themselves to minimize repulsion effects from one another. In other words, the electron pairs are as far apart as possible.

VSEPR SHAPES
The VSEPR model is useful for predicting and visualizing molecular structures. The structures are: linear, trigonal planar, angled, tetrahedral, trigonal pyramidal, trigonal bipyramidal, disphenoidal (seesaw), t-shaped, octahedral, square pyramidal, square planar, and pentagonal bipyramidal.

The VSEPR structures take the names of 3-D geometric shapes, as in the example trigonal bipyramidal. Under the VSEPR model, a trigonal bipyramidal molecule such as phosphorus pentachloride or PCl5, with a central phosphorus atom and five valence shell electron pairs, looks like two (bi) connected triangular-base pyramids, where each atom is the vertex or corner of a triangular face.

vsepr-chart-table
VSEPR Chart Downloads
Downloadable PDF
Printable Image SVG (in-browser)
Molecular Shape Image Downloads
Linear (CN 2)
Trigonal pyramidal
Linear (CN 3) Trigonal bipyramidal
Linear (CN 4) Bisphenoidal (seesaw)
Linear (CN 5) T-shaped
Trigonal planar Octahedral
Angled (CN 3) Square pyramidal
Angled (CN 4) Square planar
Tetrahedral Pentagonal bipyramidal
These resources are provided free of charge. We'd be delighted if you find them useful enough to warrant a bookmark or social media share. Or link to them from your website. Just remember to give us credit.

USING THE VSEPR CHART TO DETERMINE SHAPE AND BOND ANGLE
To use a VSEPR table, first determine the coordination number or number of electron pairs.

Count the valence electrons of the central atom.
Add an electron for each bonding atom.
Subtract an electron if the central atom has a positive charge; and add an electron for a central atom with negative charge.
Halve your count to get the total electron pairs.
Finally, look up your molecule on the chart by coordination number and number of atoms.

Alternatively, you can count the lone electron pairs, which are also indicated on the chart.

Example: PCl5
Once you know PCl5 has five electron pairs, you can identify it on a VSEPR chart as a molecule with a trigonal bipyramidal molecular geometry. Its bond angles are 90 ° and 120 °, where the equatorial-equatorial bonds are 120 ° apart from one another, and all other angles are 90 °.

More VSEPR Examples
Some other examples shown on the VSEPR chart are sulfur hexafluoride, SF6, whose six electron pairs give it octahedral geometry with 90 ° angles, and CO2, which has two electron pairs and linear geometry.

WHAT DOES VSEPR STAND FOR?
VSEPR is an acronym that stands for valence shell electron pair repulsion. The model was proposed by Nevil Sidgwick and Herbert Powell in 1940. Ronald Gillespie and Ronald Nyholm then developed the model into their theory published in 1957; they are considered the developers of the VSEPR theory. The approach was commonly referred to as VSEPR from 1963 to the present.

WHAT ARE THE VSEPR THEORY POSTULATES?
Gillespie summarizes the VSEPR theory rules as:

• Nonbinding domains are larger than single bond domains; they are more spread out and occupy more space in the valence shell than single bond domains. This is understandable because lone pairs are under the influence of only one positive core rather than two.
  
  The size of a single bond domain in the valence shell of a central atom decreases with increasing electronegativity of the ligand.
  
  Although it is often convenient to think of double and triple bonds as composed of a σ or two π bonds or two or three bent single bonds, respectively, it is simpler in the electron pair domain model to consider a double bond as a two electron pair domain and a triple bond as a three electron pair domain in which the individual electron pairs are not distinguished. These bond domains increase in size from a single to a double to a triple bond.1

VSEPR is often explained to beginners as eight simpler postulates:

• Molecular shape can be determined by the number of electron pairs present.
  Electron pairs tend to repel one another.
  Electron pairs arrange themselves to minimize the repulsion between them.
  The valence or outermost electron shell is assumed to be spherical.
  Multiple bonds are accounted as single electron pairs, and bonded electron pairs as a single pair.
  Lone pair electrons have the maximum repulsion, and bond pair electrons the minimum.
  All electron pairs assume positions of least repulsion.
  Repulsive interaction of electron pairs is greatest between lone pairs and least between bond pairs: bond pair – bond pair < lone pair – bond pair < lone pair – lone pair.
  

MOLECULAR GEOMETRY DEFINITION
Molecular geometry is a method to determine the shape of a molecule based on the repulsion occurring between bond electron pairs in the outermost (or valence) electron shell. It’s useful to study molecular geometry to get information beyond that provided in a Lewis structure. Many physical and chemical properties are affected by the shape of molecules.

VSEPR is a molecular geometry model that helps predict the general shape of a molecule but doesn’t provide information about the length or type of bonds. VSEPR theory is not effective in molecules where the central atom is a transition metal and thus has a high atomic mass that offsets or weakens the pull of bonded valence electrons.

ELECTRON GEOMETRY VS MOLECULAR GEOMETRY
The VSEPR model is one way to determine molecular geometry. A more advanced way of determining the shape of a compound is electron geometry. Both approaches depend on information about electrons, but the electron geometry model accounts for all electrons. The two models can predict different shapes for the same molecule.

You can also differentiate the two by thinking of electron geometry as a way of looking at the electrons that surround an atom and molecular geometry as a way of looking at the arrangement of atoms around a central atom.

MOLECULAR MODELING WITH VSEPR
We offer chemistry resources for beginners as well as professionals. Our Cochranes orbitals for Unit molecular models show the effect of unpaired electrons on molecular structure, and our Cochranes orbitals for Unit molecular models allow you to construct various types of molecules. We also sell several types of MolyMod Molyorbital sets and molecular model sets by MolyMod.

References
1.
Gillespie RJ. 1992. Electron densities and the VSEPR model of molecular geometry. Can. J. Chem.. 70(3):742-750. https://doi.org/10.1139/v92-099

Reading this paper from the PaddlePaddle GEM-PGL library creators. They are running ML models on SMILES bonding. One thing that stood out in the paper is determing the "Bond Strength" and "Angles" with the molecular bonds. It seems this was the big gap to fill in the models.

Paper:
https://github.com/PaddlePaddle/PGL/blob/main/examples/kddcup2021/PCQM4M/SuperHelix_PCQM4M.pdf
--------------

LITEGEM: LITE GEOMETRY ENHANCED MOLECULAR
REPRESENTATION LEARNING FOR QUANTUM PROPERTY
PREDICTION

SOLUTION TO KDD CUP 2021 PCQM4M-LSC CHALLENGE
Shanzhuo Zhang∗, Lihang Liu∗
Sheng Gao∗
, Donglong He∗
PaddleHelix Team, Baidu Inc.
{zhangshanzhuo, liulihang}@baidu.com
{gaosheng06, hedonglong}@baidu.com
Weibin Li∗
, Zhengjie Huang∗
Weiyue Su, Wenjin Wang
PGL Team, Baidu Inc.
{liweibin02, huangzhengjie}@baidu.com
{suweiyue, wangwenjin02}@baidu.com


ABSTRACT

In this report, we (SuperHelix team) present our solution to KDD Cup 2021-PCQM4M-LSC, a large-scale quantum chemistry dataset on predicting HOMO-LUMO gap of molecules. Our solution, Lite Geometry Enhanced Molecular representation learning (LiteGEM) achieves a mean absolute error (MAE) of 0.1204 on the test set with the help of deep graph neural networks and various self-supervised learning tasks. The code of the framework can be found in https://github.com/PaddlePaddle/PaddleHelix/tree/dev/competition/kddcup2021-PCQM4M-LSC/ .

1 Introduction

Molecular property prediction has been widely considered as one of the most critical tasks in computational drug and materials discovery, since many methods rely on predicted molecular properties to evaluate, select and generate molecules . With the development of deep neural networks (DNNs), molecular representation learning exhibits a great advantage over feature engineering-based methods, which has attracted increasing research attention to tackle the molecular property prediction problem.

Inspired by our previous work (Fang et al. [2021]), we propose using Lite Geometry Enhanced Molecular representation learning (LiteGEM) for the quantum property prediction: HOMO-LUMO gap (Nakata and Shimazaki [2017]). We add the word “Lite” due to the fact that our original GEM model requires 3D geometry information as input feature, which is absent in this dataset. However, the self-supervised learning strategies proposed in GEM can still be incorporated into LiteGEM and boost the performance.

The report is organized as follows. We briefly introduce graph neural nets and message passing in Section 2. In
Section 3, we present the main architecture of our model LiteGEM and its components. Experiment details such as choices of hyper-parameters and performance of ensemble can be found in Section 4.

2 Preliminaries

A molecule consists of atoms, and the neighboring atoms are connected by the chemical bonds, which can be naturally represented by a graph G = (V, E), where V is a node set and E is an edge set. An atom in the molecule is regarded as a node v ∈ V and a chemical bond in the molecule is regarded as an edge (u, v) ∈ E connecting atoms u and v.

Graph neural networks (GNNs) can be seen as message passing neural networks (Gilmer et al. [2017a]), which are useful for predicting molecular properties. Following the definitions of the previous GNNs (Xu et al. [2019]), the features of a node v are represented by xv and the features of an edge (u, v) are represented by euv. Taking node features, edge features and the graph structure as inputs, a GNN learns the representation vectors of the nodes and the entire graph, where the representation vector of a node v is denoted by hv and the representation vector of the entire graph is denoted by hG. A GNN iteratively updates a node’s representation vector by aggregating the messages from the node’s neighbors. Given a node v, its representation vector h(k)
v at the k-th iteration is formalized by
a
(k)
v = AGGREGATE(k)
(...snipped out formula)

where N (v) is the set of neighbors of node v, AGGREGATE(k)
is the aggregation function for aggregating messages
from a node’s neighborhood, and COMBINE(k)
is the update function for updating the node representation. We initialize
h
(0)
v by the feature vector of node v, i.e., h
(0)
v = xv.
READOUT function is introduced to integrate the nodes’ representation vectors at the final iteration to gain the graph’s representation vector hG, which is formalized as
hG = READOUT(h(K)v|v ∈ V), (2)
where K is the number of iterations. In most cases, READOUT is a permutation invariant pooling function, such as summation and maximization. The graph’s representation vector hG can then be used for downstream task predictions.

....

3.2.1 Geometry-level: Bond Length & Bond Angle Prediction

According to the Hohenberg–Kohn theorems in DFT, the ground-state electron density determines all ground-state properties of a molecule, i.e., the ground-state conformation determines the HOMO-LUMO gap of a molecule. On the other hand, as proved in our previous work ChemRL-GEM (Fang et al. [2021]), by incorporating accurate 3D conformation in the QM9 dataset, we are able to achieve significant improvement in predicting various quantum properties of molecules. However, unlike ChemRL-GEM, due to the inference time limitation of this task, it is nearly impossible to generate accurate 3D conformation using DFT during inference.
To solve such dilemma, we adopt geometry-level self-supervised learning tasks of GEM only in the training procedure.

More concretely, we utilize the bond length and bond angle prediction task of GEM as both the pre-training and
auxiliary task for training LiteGEM. The bond lengths and the bond angles are the most important molecular geometrical parameters. The bond length is the distance between two joint atoms in a molecule, reflecting the bond strength between the atoms, while the bond angle is the angle connecting two consecutive bonds, including three atoms, describing the local spatial structure of a molecule.

....

3.4 Feature Selection

We extend the standard node and edge features provided by the OGB-PCQM4M dataset (Liu et al. [2021], Chen et al. [2019], Kearnes et al. [2016]). To be specific, we incorporate the atom type, chirality, degree, formal charge, number of hydrogen atoms connected, radical electron, hybridization, ring size, van der waals radius, valence of out shell, partial charge as atom features, and aromatic, if in ring, bond type, bond stereo type and conjugated as bond features.