Visualization of Molecular Properties at the Qantum Mechanical Level using Blender Thomas Haschka∗

Manuel Dauchez†

Eric Henon‡

Non-affiliated individual

Universite´ de Reims Champagne Ardenne.

Universite´ de Reims Champagne Ardenne

A BSTRACT Molecular visualization is an active and ever expanding subject of research. Most of the visualization methods are however applied on large biological molecules. The representation of smaller molecules on a quantum mechanical level is to date both cumbersome and rare. We present herein a workflow that allows arbitrary volumetric datasets obtained from quantum mechanical calculations to be visualized. Using the open source Blender software and in house developed scripts we show how a researcher might create high resolution publication ready images from quantum mechanical data, such as electron density, electrostatic potential or spin density. This approach currently beats standard molecular visualization tools such as VMD and PyMOL as it allows to visualize higher resolution volumetric datasets and hence, more complex molecules and molecular properties. Further we boast superior image qualitity due to the availability of high quality volumetric absorbtion, scattering and emission shaders in Blender. Finally we show how a today’s state of the art cloud platforms provide the computational power required to create high quality visual representions. All the scripts presented herein are freely available on github https://github.com/haschka/quantumblendervis Index Terms: I.3.7 [Computer Graphics]: Three-Dimensional Graphics and Realism—Animation I.3.m [Computer Graphics]: Three-Dimensional Graphics and Realism—Miscellaneous I.6.8 [Simulation and Modeling]: Types of Simulation—Visual; 1 I NTRODUCTION Visual representations of molecules date back at least to the mid 19th century. Famous chemists such as Archibald Scott Couper [1] and Friedrich August Kekul´e von Stradonitz [14] have already used sketches to represent molecular structures. Advances in computer technology and especially in computer graphics ultimately leaded, to machine assisted, and machine generated graphics that highlight the structures, properties and functions of molecules and molecular interactions. As such these representations have vastly replaced mechanical models that were used beforehand. Most advances in visualization are made in the world of biological macromolecules [16] where several different representations, especially for secondary protein structures, have been developed. On the scale of quantum chemistry however molecular visualization is largely reduced to classical models. Ball and stick representations enhanced by isosurfaces highlighting molecular properties such as the highest occupied molecular orbitals (HOMO) and lowest unoccupied molecular orbitals (LUMO) are common. Most quantum mechanical properties are represented as scalar or vector fields. A two dimensional surface as such is insufficient to correctly outline the contents and quantum molecular visualization necessitates the treatment of volumetric data. ∗ e-mail:

[email protected]

† e-mail:[email protected] ‡ e-mail:[email protected]

1st International Workshop on Virtual and Augmented Reality for Molecular Science (VARMS) 2015 23 March, Arles, France 978-1-4673-6926-8/15/$31.00 ©2015 IEEE

Volumetric rendering using computer algorithms dates back at least to 1984 [13]. It is nevertheless cumbersome both because of the datasizes involved and the computational effort to be made. Special purpose tools targeting quantum chemistry [11] [17] have been developed, and volumetric renderings started to appear in literature [2]. Both VMD [8] and PyMol [19], arguably the most important free molecular visualization tools, are limited and cumbersome when large volumes and thus fine grained volumetric datasets are used. We present herein a workflow that allows us to overcome these limitations. We show how we create detailed images from quantum calculations of large molecules involving several hundreds of atoms. Our method has proved to be robust to volumetric datasets of over 10003 voxels large. The volumetric shaders implemented in the Blender [22] software allow us to render visualizations of quantum mechanical properties of large molecules at an unprecedented quality. This naturally comes at the cost of computational power which led us to use state of the art cloud computing platforms such as the one offered by Qarnot computing. The workflow described herein shall not be regarded as a fully featured software tool. Even though the collection of scripts is fully functional, we see this presentation as a mere guideline for those interested in writing their own scripts and tools in order to create volumetric representations of quantum mechanical results using open source software. We hope to inspire the quantum chemistry community in showing them what can be done using freely available tools on the internet in combination with cloud computing solutions. A schematic representation of the workflow presented herein is shown in figure 1. The artistic and aesthetic value of molecular visualization is an other point that we would like to highlight. Sophisticated molecular visualization is today a key in communicating and teaching science [4] and is further frequently featured in the artistic sections of international congresses. Use of computational resources for teaching represents an ideal bridge between accumulated research findings and educational contents. The representation at the human scale of a chemical object, an abstract object, from a simple sketch to augmented reality, is an essential tool to stimulate the senses, learning and understanding of fundamental mechanisms. However, for teachers, making effective use of numeric data requires skills and more preparation than printed material. The proposed workflow lowers the barriers which may inhibit the use of quantum mechanical data in the lecture halls of today’s universities. 2

E XPOSITION

During our research around the biological Thrombospondin molecule [7] we performed quantum chemical calculations on parts of this protein that were up to 218 atoms large. This protein features several calcium ions that polarize its surface. One of our goals was to visually investigate how the electron cloud of the molecule is deformed and displaced by a charge approaching it. Using the Gaussian software [5] and its cubegen utility we created large volumetric datasets featuring the electron density of this molecule both, with and without calcium ions attached to it. These datasets contain 5003 voxels and could not be read using standard tools. In VMD [8]

7