Examples

A collection of example usages of the OGOLEM framework.

Homogeneous Lennard-Jones clusters are an important, well-established real-life benchmark for global optimization. With this simple input for the first "challenging" case of 38 atoms, OGOLEM finds the true fcc global minimum-energy structure in about 7261 steps (mean value of 10 runs, the fastest needed 1781 steps, the slowest 19082 steps). This improved input, finds it in about 342 steps (again a mean value of 10 runs, the fastest needing 176 steps, the slowest 677 steps).
More coming soon...

OGOLEM papers

Using global structure optimization to model adhesion and aggregation of complex molecules on surfaces

A. Freibert, J. M. Dieterich and B. Hartke
Exploring self-organization of molecular tether molecules on a gold surface by global structure optimization
J. Comput. Chem. 40 (2019) 1978
DOI: 10.1002/jcc.25853
As an extension of our first global optimization of clusters on surfaces, we show here that the same strategy can be applied to successfully model the surface-adhesion of large, complex molecules onto surfaces, yielding qualitatively improved understanding of how tether molecules act as anchors for surface functionalizations.

Global structure optimization of clusters on surfaces

C. Witt, J. M. Dieterich and B. Hartke
Cluster structures influenced by interaction with a surface
Phys. Chem. Chem. Phys.20 (2018) 15661
DOI: 10.1039/C8CP02694G
By incorporating a rigid surface as an "environment", ogolem can also be used to globally optimize clusters on surfaces. In this proof-of-principle paper, we demonstrate this for the benchmark case of homogeneous and heterogeneous Lennard-Jones clusters on Lennard-Jones surfaces. By tuning the cluster-surface interaction, a wide range of cluster structure changes becomes accessible.

Malleable parallelism with minimal effort for maximal throughput and maximal hardware load

F. Spenke, K. Balzer, S. Frick, B. Hartke and J. M. Dieterich
Malleable parallelism with minimal effort for maximal throughput and maximal hardware load
Comput. Theor. Chem., 1151 (2019) 72
DOI: 10.1016/j.comptc.2019.02.002
Our RMI-based parallel implementation can be used to easily construct parallel jobs that adapt their hardware resource use (e.g., nodes, CPU-cores) to what is available. This happens during runtime, without pre-planning or code-instrumentation, and externally controlled by the queueing system. Here we demonstrate this feature of malleability/adaptiveness, with real-life calculations on parallel high-performance computing installations of various sizes, including a transregional HPC center. During normal queue operation in the presence of dozens of standard, fixed-size jobs, this allows us to achieve 100% machine load, even in the challenging situation of a complete machine drain before the start of a huge job that uses all 17,856 CPU cores.

Hybdridizing Evolutionary Algorithms with Local Heat Pulses

J. M. Dieterich and B. Hartke
Improved Cluster Structure Optimization: Hybdridizing Evolutionary Algorithms with Local Heat Pulses
Inorganics, 5 (2017) 64
DOI: 10.3390/inorganics5040064 (pdf)

We show a successful attempt at hybridizing evolutionary algorithms with other global optimization techniques (here: local heat pulses, LHP). Our combined EA+LHP protocol exhibits superior performance to our normal EA implementation for a range of hard cluster structure optimization problems.

Globally-Optimized Local Pseudopotentials

B. G. del Rio, J. M. Dieterich, and E. A. Carter
Globally-Optimized Local Pseudopotentials for (Orbital-Free) Density Functional Theory Simulations of Liquids and Solids
J. Chem. Theory Comput., 13 (2017) 3684
DOI: 10.1021/acs.jctc.7b00565
We extend the parameter-optimization capabilities of the OGOLEM framework to optimize local pseudopotentials as used on (orbital-free) density functional theory simulations to match properties as obtained from Kohn-Sham density functional theory simulations with either non-local pseudopotentials or projector augmented wave approaches. We show for a range of examples (solids and liquid metals) that our so-obtained goLPSs provide superior accuracy compared to bulk-local pseudopotentials for the targeted properties with good transferability. We also demonstrate that OGOLEM is capable to use the local channel of a non-local pseudopontial as an initial guess, making our algorithm a fully-fledged local pseudopotential parametrization.

An Error-safe, Portable, and Efficient Evolutionary Algorithms Implementation with High Scalability

J. M. Dieterich and B. Hartke
An Error-safe, Portable, and Efficient Evolutionary Algorithms Implementation with High Scalability
J. Chem. Theory Comput., 12 (2016) 5226
DOI: 10.1021/acs.jctc.6b00716
Development of a RMI-based parallelization solution for evolutionary algorithms. Key features include error-safety against hardware or transmission problems at runtime, portability across a wide range of platforms and excellent scalability tested up to 6144 CPU cores.

Observable-targeting structure optimization

J. M. Dieterich and B. Hartke
Observable-targeting global cluster structure optimization
Phys. Chem. Chem. Phys., 17 (2015) 11958
DOI: 10.1039/C5CP01910A (pdf)
Describes a novel strategy for cluster structure optimization using an (experimental) observable instead of the electronic energy to construct the fitness function. Proof of concept implementation for neutral water clusters targeting an experimental IR spectrum as the observable. Initial application to small and medium sized water clusters.

Graph-based shortcut to low-energy minima

J. M. Dieterich and B. Hartke
A graph-based short-cut to low-energy structures
J. Comput. Chem., 35 (2014) 1618
DOI: 10.1002/jcc.23669
Method development in directed mutation strategies. Development of a graph-based identification of non-optimal building blocks and directed movement into potentially better ones. Specialized move class implementation for water clusters. Example global optimizations of water clusters with 24 and 61 water molecules.

Size-resolved water cluster study

U. Buck, C. C. Pradzynski, T. Zeuch, J. M. Dieterich, and B.Hartke
A size-resolved perspective of large water clusters
Phys. Chem. Chem. Phys., 16 (2014) 6859
DOI: 10.1039/c3cp55185g (pdf)
Global optimization of water clusters with an efficient implementation of the TTM3-F force field for 24, 25, and 32 water molecules. IR spectra generated based on V. Buch's empirical model, direct comparison to and interpretation of experimental results.

Relative anion binding for interpenetrated coordination cages

M. Frank, J. M. Dieterich, S. Freye, R. A. Mata, and G. H. Clever
Relative anion binding affinity in a series of interpenetrated coordination cages
Dalton Trans., 42 (2013) 15906
DOI: 10.1039/C3DT51709H
Extension of the previously developed push-and-pull model to more complex interpenetrated cage structures. Comparison to experimental data.

Push-and-pull model for allosteric anion binding

J. M. Dieterich, G. H. Clever, and R. A. Mata
A push-and-pull model for allosteric anion binding in cage complexes
Phys. Chem. Chem. Phys., 14 (2012) 12746
DOI: 10.1039/C2CP41793F
Development of a push-and-pull model to describe allosteric anion binding in cage complexes. Global parametrization of the model parameters against ab initio reference data, comparison of derived, non-empirical cage dimensions with experiment.

DFT-based global optimization of small water clusters with sodium

R. M. Forck, J. M. Dieterich, C. C. Pradzynski, A. L. Huchting, T. Zeuch, and R. A. Mata
Small sodium-doped water clusters: A combined experimental and theoretical study.
Phys. Chem. Chem. Phys., 14 (2012) 9054
DOI: 10.1039/C2CP41066D
Global optimization of small water clusters containing a solvated sodium atom on the density function level of theory. Analysis on the coupled-cluster level of theory and comparison to experimental results.

Assessing the difficulty of benchmark functions

J. M. Dieterich and B. Hartke
Empirical review of standard benchmark functions using evolutionary global optimization
Appl. Math., 3 (2012) 1552
DOI: 10.4236/am.2012.330215 (pdf)
Review of the difficulty and the empirical scaling with problem size of standard benchmark functions in the context of the OGOLEM framework. Reviewed functions are Ackley, Rastrigin, Schwefel, Schaffer's F6 and F7, Lunacek, and randomly distributed Gaussians with up to 1000D problem size.

Global structure optimization of alkaline earth clusters

J. M. Dieterich, S. Gerke, and R. A. Mata
A first-principles based potential for the description of alkaline-earth metals
J. Atom. Mol. Opt. Phys., 2012 (2012) ID 648386
DOI: 10.1155/2012/648386 (pdf)
Global optimization of mixed alkaline-earth clusters using a Gupta force field model globally parametrized against highest-level ab initio data. Analysis of structural effects.

Highly mixed Lennard-Jones clusters

J. M. Dieterich and B. Hartke
Composition-induced structural transitions in mixed LJ clusters: global reparametrization and optimization
J. Comput. Chem., 32 (2011) 1377
DOI: 10.1002/jcc.21721
Global parameter fit of an extended LJ model to CCSD(T)/AV5Z data with relativistic corrections for rare gas dimers. The resulting, new potential is used for a global optimization of homogenous, binary, ternary, quaternary, and quinary clusters of rare gases. Composition induced structural transitions can be observed.

Kanamycin A aggregation

J. M. Dieterich, U. Gerstel, J.-M. Schröder, and B. Hartke
Aggregation of Kanamycin A: dimer formation with physiological cations
J. Mol. Mod., 17 (2011) 3195
DOI: 10.1007/s00894-011-0983-x
Dimers of Kanamycin A, with a NAMD/GAFF force field, with one optional sodium/potassium cation; post-optimized with MOPAC/PM3 and with MOLPRO/DF-LMP2/cc-pVDZ; prediction of IR, Raman and NMR spectra.

Molecular design of optimally switchable molecules

N. O. Carstensen, J. M. Dieterich, and B. Hartke
Design of optimally switchable molecules by genetic algorithms
Phys. Chem. Chem. Phys., 13 (2011) 2903
DOI: 10.1039/C0CP01065K
Global optimization of S₀/S₁ wave lengths for azobenzene cis/trans switching by varying subsituents on a fixed back bone. Target wavelengths of off-the-shelf laser pointers can be met when allowing enough substitution flexibility. Restricting substituents to a subsets allows for simpler molecules and acceptable match with the target wave lengths. Extensive surface hopping dynamics and absorption spectrum analysis of the result molecules.

Initial framework introduction

J. M. Dieterich and B. Hartke
OGOLEM: Global Cluster Structure Optimization for Arbitrary Mixtures of Flexible Molecules - A Multiscaling, Object-Oriented Approach
Mol. Phys., 108 (2010) 270
DOI: 10.1080/00268970903446756
The original introduction of OGOLEM. Presenting LJ clusters using standard LJ parameters and mixing rules of binary, ternary, quaternary and quinary composition. Presenting initial results on kanamycin A aggregation with up to four kanamycin units and up to four sodium/potassium ions using a GAFF force field.