Computing Quantum Chemical Results without Doing

Quantum Chemistry: A Machine Learning Shortcut

Mojtaba Haghighatlari1, Johannes Hachmann

1,2,3

1Department of Chemical and Biological Engineering,

2Computational and Data-Enabled Science and Engineering Graduate Program,

University at Buffalo, SUNY, Buffalo, New York 14260, USA

3New York State Center of Excellence in Materials Informatics,

Buffalo, New York 14203, USA

ABSTRACT

Computational quantum chemistry is a valuable tool that allows us to characterize and assess

compounds, materials, and reactions. It is based on the notion that quantum mechanics

rigorously maps a given chemical structure to its properties. Unfortunately, this mapping is

expensive and does not lend itself to generalizations that correspond to chemical intuition,

empirical rules, and well-known trends. Our work addresses the question if machine learning

– given a suitable training set – can recover this mapping in simplified form, which would

offer a shortcut to quantum chemical results without the need to perform quantum chemical

calculations. Such a shortcut would open the door to rapid high-throughput screening

capabilities, which would enable an unprecedented exploration of chemical space.

We will present our latest chemical data mining approaches that allow us to extract an

understanding of hidden structure-property relationships in large-scale data sets and to

quantify these in model equations. We will discuss the utility of these innovative machine

learning, statistical learning, and informatics techniques as well as the error bars and

predictive performance of the resulting models. In this context, we will introduce CheML,

our machine learning and informatics software suite for the chemical and materials sciences.

Oral presentation preferred.

Coupled-Cluster Interpretation of the Photoelectron Spectra of Ag3-

and Au3-

Nicholas P. Bauman,1 Jared A. Hansen,1 Piotr Piecuch,1,2 and Masahiro Ehara3,4

1Department of Chemistry, Michigan State University, East Lansing, MI 48824, U.S.A.

2Department of Physics and Astronomy, Michigan State University, East Lansing, MI 48824, U.S.A.

3Institute for Molecular Science and Research Center for Computational Science, Okazaki 444-8585, Japan

4Elements Strategy Initiative for Catalysts and Batteries (ESICB), Kyoto University, Kyoto 615-8245, Japan

We undertake a thorough ab initio study of the photoelectron spectra of silver and gold trimer anions,1,2 Ag3

- and Au3-, by examining the ground and excited states of the corresponding neutral

particles, Ag3 and Au3, employing the scalar relativistic ionized equation-of-motion coupled-cluster (IP-EOMCC) approximations. We examine the effects of basis set, number of correlated electrons, level of applied theory including up to 3-hole-2-particle terms, and geometry relaxation. The IP-EOMCC methods allow one to obtain the ground and excited states of the

(N − 1)-electron open-shell system by applying the linear ionizing operator, ��岫�−1岻, to the ground state of the corresponding N-electron closed-shell core obtained with a single-reference CC approach. In this way, one can properly account for spin symmetry that the conventional, particle-conserving, open-shell CC/EOMCC approaches have difficulties with, while determining the electronic spectrum of the (N − 1)-electron system obtained by removing an electron from the N-electron closed-shell species. To further improve the accuracy, we adopt a simple IP-EOMCC-based extrapolation scheme that captures the important higher-order many-electron correlation contributions to the energy in an approximate, but computationally efficient manner, in addition to the basis set effects and the correlation effects associated with semi-core electrons. Our IP-EOMCC calculations provide an accurate and complete assignment of peaks and shoulders in the experimental photoelectron spectra of Ag3

- and Au3- for the first time.

1 H. Handschuh, C.-Y. Cha, P.S. Bechthold, G. Ganteför, and W. Eberhardt, J. Chem. Phys. 102, 6406 (1995). 2 H. Handschuh, G. Ganteför, P.S. Bechthold, and W. Eberhardt, J. Chem. Phys. 100, 7093 (1994). 3 N.P. Bauman, J.A. Hansen, M. Ehara, and P. Piecuch, J. Chem. Phys. 141, 101102 (2014). 4 N.P. Bauman, J.A. Hansen, and P. Piecuch, in preperation.

Structure of the One-particle reduced densitymatrix from Generalized Pauli exclusion

principle

Romit Chakraborty and David A. Mazziotti

June 19, 2015

The Pauli exclusion principle requires that the occupations of the orbitals lie be-tween zero and one. These Pauli conditions hold for one-electron reduced densitymatrices (1-RDMs) from both open and closed quantum systems. More than 40 yearsago, it was recognized that there are additional conditions on the 1-RDM for closedquantum systems. In this review we discuss the structure of the 1-RDM from thegeneralized Pauli exclusion principle in many-electron atoms and molecules and theviolation of the generalized Pauli principle as a sufficient condition for the opennessof a many-electron quantum system.

**I would like to give a contributed talk**

1

Molecular Dynamic simulations of CO-induced surface reconstruction

and island formation on Pt, Pd, and Pt/Pd (557) surfaces Joseph R. Michalka and J. Daniel Gezelter

1

University of Notre Dame 1251 Nieuwland Science Hall, Notre Dame, Indiana, 46556, United States,

(574)-631-7595, gezelter@nd.edu

Platinum and Palladium surfaces and nanoparticles are of prime importance in many

catalytic processes. The catalytic activity of these species is strongly influenced by their

displayed facets and availability of under-coordinated metal binding sites. The presence

of adsorbates, here carbon monoxide (CO), can lead to significant surface restructuring.

Using molecular dynamics simulations, we model CO-induced surface restructuring on

high-index Pt, Pd, and Pt/Pd surfaces. Whereas CO adsorbed to Pt (557) leads to

increased surface mobility of Pt and a doubling of the step-edges, CO on Pd (557)

maintains the high-index facet. CO on a Pd (557) system covered with one layer of Pt

leads to domains of Pt and an exposure of the underlying Pd. Metallic self and cross

interactions are described using the Embedded Atom Method, while metal-CO

interactions were parameterized from experimental data and theoretical (DFT)

calculations.

If a slot is available, I wish to give a talk.

The Molecular Origin of Color tuning of Deep‐water Lake Baikal Cottoid Fish 

Visual Pigments

 

Hoi Ling Luk1, Fabio Montisci2, Federico Melaccio2, Nihar Bhattacharyya3, James 

Morrow3, Belinda S. W. Chang3, Francesca Fanelli4, Massimo Olivucci1,2 

 1Department of Chemistry, Bowling Green State University, Bowling Green OH 43403 2Università di Siena, Dipartimento di Chimica, via A. De Gasperi 2, I‐53100, Siena, Italy  3Department of Ecology and Evolutionary Biology and Department of Cell and Systems 

Biology, University of Toronto, 25 Harbord St., Toronto, ON M5S 3G5, Canada. 4Dulbecco Telethon Institute, Department of Life Sciences, University of Modena and 

Reggio Emilia, I‐41125 Modena, Italy

 

Lake Baikal is located in Eastern Siberia and is the deepest and most ancient lakes in 

the world.  Its deepness provides the colonization of all depth habitats by a unique 

fauna  that  includes  a  remarkable  flock  of  largely  endemic  teleost  fish  of  the  sub‐

order Cottoidei.   These  cottoid  fishes  shows  a  gradual  blueshift  in  the  absorption 

maxima  (λmax) of  their visual pigments  in  relation  to  the depth of  their habitat and 

such  spectral  shifts  do  not  arise  from  alternative  chromophores  as  all  visual 

pigments in the Baikal cottoid fishes contain 11‐cis‐retinal chromophore. Hence, this 

differing λmax could have arisen by single amino acid substitutions from an ancestor 

which  phylogenetics  study  indicates  that  the  ancestral  species  to  the  Baikal  flock 

had a rod pigment with a λmax around 505 nm.  In order  to understand  the origin of 

such  color  tuning,  four  pairs  of  quantum  mechanics/  molecular  mechanics 

(QM/MM)  Baikal  cottoid  fish  rhodopsin  models  (with  A1  and  A2  chromophore) 

which  successfully  reproduce  the  experimental  absorption  maxima  are  prepared 

using  comparative modeling approach. Analysis of  the electrostatic  effect of  every 

amino  acid  in  the  retinal  chromophore  cavity  of  4  Å  are  performed  which 

reasonably  support  the  hypothesis  of  the  observed  mutations  inside  the  retinal 

chromophore cavity in the models are involved in the mechanism used by biological 

evolution to tune the color of the absorption maxima of the lake Baikal cottoid fish 

rhodopsins. Relevant data and results will be presented as a contributed talk.   

Reduced Density Matrix Theory  in Quantum Chemistry and Physics 

David A. Mazziotti 

 

Abstract 

 

Energies and properties of many‐electron molecules can be expressed as a linear functional of 

the two‐electron reduced density matrix (2‐RDM).  The lecture will discuss recent advances and 

applications of 2‐RDM theory including an application to transition‐metal chemistry. 

 

Non-adiabatic current densities, transitions and power absorbed by a molecule in a time-

dependent electromagnetic field

Anirban Mandal and Katharine L. C. Hunt

Department of Chemistry, Michigan State University, East Lansing, MI 48824

The energy of a molecule subject to a time-dependent perturbation separates completely into

adiabatic and non-adiabatic terms, where the adiabatic term reflects the adjustment of the ground

state to the perturbation, while the non-adiabatic term accounts for the transition energy.1 For a

molecule perturbed by a time-dependent transverse electromagnetic field, in this work we show

that the average power absorbed by the molecule is equal to the time rate of change of the non-

adiabatic term in the energy.2 The non-adiabatic term is given by the transition probability to an

excited state k, multiplied by the transition energy from the ground state to k, and then summed

over the excited states. The average power absorbed by the molecule is derived from the integral

over space of the scalar product of the applied electric field and the non-adiabatic current density

induced in the molecule by the field. No net power is absorbed due to the action of the applied

electric field on the adiabatic current density. The work done on the molecule by the applied

field is the time integral of the power absorbed. The result established here shows that work done

on the molecule by the applied field changes the populations of the molecular states. Our results

are based on a perturbed Hamiltonian in the Coulomb gauge. In any arbitrary gauge, the

complete Hamiltonian, including the energy of the electromagnetic field as well as the molecular

energy, is gauge invariant.3

1 A. Mandal and K. L. C. Hunt, J. Chem. Phys. 137, 164109 (2012).

2 A. Mandal and K. L. C. Hunt, J. Chem. Phys. (Submitted).

3 A. Mandal and K. L. C. Hunt, Manuscript in preparation.

This abstract is intended for oral presentation

MWTCC 2015 

N2 vs. O2 Adsorption on Open Iron Sites of Fe2(dobdc): An

Electronic Structure Theory Study              

Pragya Verma,1 Rémi Maurice,

1,2 Laura Gagliardi,

1,* and Donald G. Truhlar

1,*

1Department of Chemistry, Supercomputing Institute, Nanoporous Materials Genome

Center, and Chemical Theory Center, University of Minnesota, Minneapolis, Minnesota

55455-0431, USA

2SUBATECH, UMR CNRS 6457, IN2P3/EMN Nantes/Université de Nantes, 4 rue

Alfred Kastler, BP20722, 44307 Nantes Cedex 3, France

*e-mails: gagliard@umn.edu and truhlar@umn.edu

The presence of open metal sites and high porosity makes the metal–organic framework

Fe2(dobdc) suitable for separating gaseous mixtures. For example O2 can be adsorbed

more strongly than N2 on Fe2(dobdc), and it can, in principle, be used to separate O2

from air [Bloch et al. J. Am. Chem. Soc. 133, 14814 (2011)]. In this work, we investigate

the differential adsorption of N2 and O2 on Fe2(dobdc) with quantum mechanical

methods applied to a finite-size cluster. The cluster is chosen such that it is large enough

to allow an accurate description of the most important contributions to the binding

enthalpies and small enough that one can perform high-level quantum mechanical

calculations. We use state-of-the-art exchange–correlation functionals as well as wave

function approaches to determine the ground state of the Fe–N2 and Fe–O2 interacting

systems. Our calculations reveal that the ground state structure of the Fe–O2 system has

the dioxygen unit in a triplet spin state ferromagnetically coupled to the high-spin state

(quintet state) of the central iron atom of the cluster. Charge Model 5 (CM5) and LoProp

charge calculations have been performed to determine the partial atomic charges on the

guest molecules (N2 and O2) and the central iron atom of the complex, and these

calculations show that the charge transfer from the open iron(II) site is more important

for O2 than for N2. Furthermore, O2 was calculated to bind more strongly than N2, which

is in agreement with experimental results.

 

New software for visualization of hyper-spherical coordinates and for 3D-printing of potential energy surfaces: Application to ozone molecule

Alexander Teplukhin and Dmitri Babikov

Chemistry Department, Marquette University, Milwaukee, WI 53201-1881, USA

Quantum molecular dynamics in triatomic systems takes special place in the fundamental chemical theory and in the chemistry education as well, because it allows introducing and studying spectroscopy and chemical reactions using the smallest possible number of atoms. Also, many important gas-phase molecules are triatomic, which makes this topic directly relevant to the real life problems in environmental chemistry, atmospheric chemistry and astrochemistry.

At first sight, there should be nothing special about the choice of vibrational coordinates: it could be valence coordinates (two bond lengths and an angle between bonds) or normal-mode coordinates (represent symmetric stretching, asymmetric stretching and bending). However, neither the valence nor the normal-mode coordinates are employed for accurate numerical studies of triatomics, because the first set results in the extremely complicated Hamiltonian operator, while the second set is an approximation which breaks down at higher levels of vibrational excitation and/or for anharmonic potentials. Jacobi coordinates could be a better choice, but they are arrangement specific: a set of Jacobi coordinates for A + BC arrangement is inappropriate for AB + C arrangement.

This is where the adiabatically-adjusting principal-axes hyper-spherical (APH) coordinates come to scene [1,2]. The Hamiltonian operator in these coordinates is simple, the symmetry is incorporated rigorously, and all three atom arrangements are treated on equal footing. However, one should admit that the APH coordinates are much less popular, compared to the simpler but more limited Jacobi coordinates.

One reason for this is that the formalism of adiabatic adjustment is mathematically involved [1,2] which creates a barrier to understanding, particularly by students at the beginning of their computational research projects. In order to simplify introduction to the APH coordinates we created an interactive desktop application APHDemo [3] that allows seeing a triatomic system on the screen, dragging atoms by mouse from one arrangement to another and watching how the APH coordinates adjust continuously, for example, from the reagent channel to the product channel, through the reaction intermediate. The Jacobi coordinates can also be made visible for comparison, which allows understanding better their drawbacks, and emphasizing advantages of the APH coordinates.

The major area of application of this program is, probably, in the educational process. We created several animations that illustrate typical examples of vibrational dynamics and can be used in the classroom presentation of the APH coordinates. This tool may also be rather handy to those who plan employing the APH coordinates in their research, particularly to graduate students and postdocs.

Another tricky aspect related to APH coordinates is understanding of potential energy surface in these coordinates. In our recent paper [4] we proposed to combine the isoenergy approach with 3D printing technology to create a plastic model of the PES which can be taken into hands and inspected in detail from any perspective (you can examine one for ozone during a poster session). Both APHDemo application and Matlab script to generate an STL file for 3D printer are freely available. References

1. R. T Pack, Chem. Phys. Lett., 108, 333 (1984). 2. R. T Pack and G. A. Parker, J. Chem. Phys., 87, 3888 (1987). 3. A. Teplukhin and D. Babikov, Chem. Phys. Lett., 614, 99 (2014). 4. A. Teplukhin and D. Babikov, J. Chem. Educ., 92(2), 305 (2015).

Bulk and Water-Vapor Interfacial Aqueous

Electrons are Spectroscopically Indistinguishable

Marc P. Coons, Zhi-Qiang You, and John M. Herbert∗

Department of Chemistry and Biochemistry, The Ohio State University, Columbus, Ohio

43210

E-mail: coons.24@osu.edu

Abstract

The vertical detachment energies (VDEs) of the aqueous electron in bulk water

and at the water-vapor interface have been computed using long-range-corrected den-

sity functional theory and mixed quantum/classical molecular dynamics simulations

based on one-electron pseudopotentials. Quantitative agreement with experimental

measurements of VDEs for the bulk species suggests that the chosen methodologies

provide reasonable models for the excess electron. The VDEs of the interfacial elec-

tron are found to be very similar to those in the bulk which is in contrast to the recently

reported value as measured from liquid-jet photoelectron spectroscopy. Our findings

suggest that a new interpretation of the liquid-jet experiments is required to reconcile

experimental and theoretical data. Furthermore our results may have an important im-

pact on the understanding of processes ocurring at aqueous/biological interfaces such

as dissociative electron attachment.

I originally selected my presentation type to be oral, but I would like to

change that. Please consider this for the poster session.

∗To whom correspondence should be addressed

1

Computational Study of Cold Ions Trapped in a Double-Well Potential

Dmytro Shyshlov and Dmitri Babikov

Chemistry Department, Marquette University, Milwaukee, WI 53201

Rigorous computational treatment is developed to study quantum dynamics of cold ions in

a double-well trap. Numerically accurate approach is adopted, that makes no assumption of weak

coupling between the wells, or harmonic approximation for energy spectrum of the double-well

system. The goal is to reproduce, from first principles, the process of efficient energy swaps

between the wells observed in the experiments at NIST [Nature 471, 196 (2011)] and Innsbruck

[Nature 471, 200 (2011)]. The model parameters and the initial conditions are carefully chosen to

reproduce experimental observables. Accurate energies and wave functions of the system are

obtained, and the evolution of motional wave packets on the accurate potential energy surface

(Fig. 1) is studied, which provides new insight. The experimental results are reproduced in detail,

by this model. Explanation of the energy transfer is given in terms of wave packet dynamics in an

asymmetric potential energy well. Surprisingly, it originates in the well-known classical principle:

angle of incidence equals to the angle of reflection, and has nothing to do with quantum tunneling

effect.

Figure 1. Potential energy surface of the system of two Be+ ions in the double-well potential.

CONFERENCE: 2015 Midwest Theoretical Chemistry Conference

DATES: Friday , June 26 – Sunday, June 28, 2015

PLACE: University of Michigan – Ann Arbor , Michigan

---------------------

TITLE : Is the Faster than Expected Diffusion of Small Neutral Solutes in Ionic Liquids Linked to their Polar/Apolar Nature?

PRESENTING AUTHOR : Juan C. Araque

LIST OF ALL AUTHORS WITH AFFILIATION : Juan C. Araque, University of Iowa; Sharad K. Yadav, University of Iowa; Michael Shadeck, Pennsylvania State University; Mark Maroncelli, Pennsylvania State University; and Claudio J. Margulis, University of Iowa.

ABSTRACT:

In a recent article, 1 we described the mechanisms by which diffusion of small neutral solutes in ionic liquids deviates from Stokes-Einstein behavior. This deviation is significantly more pronounced than in the case of conventional solvents.2 We found that small neutral solutes probe ionic liquid regions that are “soft” (of low friction and less polar) and “stiff” (of high electrostriction). The existence of regions of low and high friction can be traced to the charge and apolar structural nature of ionic liquids. Heterogeneous solute diffusional regimes (cage/jump) are thus imposed by the nanoscale solvent structure. Whereas solute displacements are anomalously short in stiff regions (cage regime), they are anomalously large in soft regions (jump regime). Overall, a clear link emerges between ionic liquid structure in the form of polar/apolar duality and deviations from Stokes-Einstein hydrodynamics.

REFERENCES:

1. Araque, J.C., et al., How Is Diffusion of Neutral and Charged Tracers Related to the Structure and

Dynamics of a Room-Temperature Ionic Liquid? Large Deviations from Stokes–Einstein Behavior

Explained. J. Phys. Chem. B, 2015. 119(23): p. 7015-7029.

2. Kaintz, A., et al., Solute Diffusion in Ionic Liquids, NMR Measurements and Comparisons to

Conventional Solvents. J. Phys. Chem. B, 2013. 117(39): p. 11697-11708.

COMMENTS: I would like to give a CONTRIBUTED TALK .

Minimum Energy Conical Intersections Through Graphical

Processing Unit Acceleration of Two Step Methods

B. Scott Fales1

1Department of Chemistry, Michigan State University

12 June 2015

Multireference methods are often used to describe regions of strong nonadiabatic coupling, suchas near a minimum energy conical intersection (MECI). The complete active space self consistentfield (CASSCF) method has long been a standard tool for describing strongly coupled multiref-erence systems, though vertical excitation energies calculated using state averaged CASSCF arenot size intensive and wavefunction convergence is often poor. In pursuit of computationally effi-cient CASSCF alternatives, we have investigated the improved virtual orbital complete active spaceconfiguration interaction (IVO-CASCI)[1][2][3] and the configuration interaction singles natural or-bitals (CISNO-CASCI)[4] methods. Both IVO-CASCI and CISNO-CASCI provide accurate verticalexcitation energies and topographically correct potential energy surfaces (PES) in the MECI regionwhen compared with CASSCF. These methods have been implemented using graphical processingunits (GPUs), an emerging technology which has proven useful for accelerating electronic structuremethods. We demonstrate that multireference CASCI methods can be applied to nanoparticlesby calculating the first several singlet states of systems including a buckeyball (C60), the retinalprecursor β-carotene (C40H56), and a silicon nanoparticle (Si72H68) using active spaces as large as16 electrons in 16 orbitals, using a single NVIDIA K40 GPU and a single core of an Intel Xeon2.40 GHz processor. To facilitate geometry and MECI optimizations of systems approaching thenanoscale, we couple our GPU based methodologies with a numerical optimizer that parallelizesacross multiple nodes, affording us a hierarchical parallelization scheme that scales approximatelylinearly with the number of nodes and GPUs.

I am interested in providing a contributed talk at this years conference, if possible.

References

[1] William J. Hunt and William A. Goddard III. “Excited States of H2 Using Improved VirtualOrbitals”. In: Chem. Phys. Lett. 3 (1969), pp. 414–418.

[2] S. Huzinaga and C. Arnau. “Virtual Orbitals in Hartree-Fock Theory. II”. In: J. Chem. Phys.

54 (1971), pp. 1948–1951.

[3] Davin M. Potts et al. “The improved virtual orbital-complete active space configuration interac-tion method, a ”packageable” efficient ab initio many-body method for describing electronicallyexcited states”. In: J. Chem. Phys. 114 (2001), pp. 2592–2600.

1

[4] Yinan Shu, Edward G. Hohenstein, and Benjamin G. Levine. “Configuration interaction singlesnatural orbitals: An orbital basis for an efficient and size intensive multireference descriptionof electronic excited states”. In: J. Chem. Phys. 142 (2015), p. 024102.

2

Will ice float on water in hybrid density functional

theory world?

Alex P. Gaiduk,† Francois Gygi,‡ and Giulia Galli†

†Institute for Molecular Engineering, The University of Chicago, Chicago, IL 60637

‡Department of Computer Science, University of California, Davis, CA 95616

First-principles simulations provide a way to study the properties of wa-ter and aqueous solutions without the need for empirical input. General-ized gradient approximations, commonly used as an engine of first-principlesmolecular dynamics, predict the equilibrium density of water to be lower (0.7–0.9 g/ml) than in the experiment. Hybrid functionals improve the structureand hydrogen bonding in water and could potentially yield better density;however, determining the bulk properties of water with hybrid functionalshas been out of reach due to computational complexity. This work presentsthe first robust determination of density and compressibility of water and iceusing the hybrid functional PBE0. We show that the fraction of Hartree–Fock exchange in PBE0 lowers the density of both water and ice, leading tobetter agreement with the experiment for ice but worse agreement for liquidwater. Inclusion of dispersion interactions on computed molecular-dynamicstrajectories led to a substantial improvement of the PBE0 results for thedensity of liquid water, which, however, resulted to be slightly lower thanthat of ice.

Presentation type: Oral

Active Space Decomposition   

Toru Shiozaki 

 

Department of Chemistry, Northwestern University, Evanston, IL 602908, USA.  

In  this  talk  I  will  present  an  approach  to  compute  diabatic  couplings  for  electron  and 

energy transfer processes in covalently linked chromophore pairs from an orbital‐optimized 

active space decomposition (ASD) method [1]. Our method is based on multi‐configuration 

wave functions and  is systematically  improvable. Applications to triplet transfer processes 

in the so‐called Closs systems will be discussed using Marcus theory. 

 

 

  [1] I. Kim, S. M. Parker, and T. Shiozaki, http://arxiv.org/abs/1505.02346 (2015). 

 

Trivalent Metal Loading via Atomic Layer Deposition

Joshua Borycz,b In Soo Kim,

a Ana Platero-Prats,

c Samat Tussupbayev,

b Timothy C.

Wang,d Omar K. Farha,

d,e Joseph T. Hupp,

a,d Laura Gagliardi,

b Karena Chapman,

c

Christopher J. Cramer,*b Alex B. F. Martinson*a

aMaterials Science Division, Argonne National Laboratory, 9700 S. Cass Ave., Argonne, Illinois 60439,

USA bDepartment of Chemistry, Supercomputing Institute, and Chemical Theory Center, University of

Minnesota, Minneapolis, Minnesota 55455, United States cX-ray Science Division, Argonne National Laboratory, 9700 S. Cass Ave., Argonne, Illinois 60439, USA

dDepartment of Chemistry, Northwestern University, 2145 Sheridan Road, Evanston, Illinois 60208, USA

eDepartment of Chemistry, Faculty of Science, King Abdulaziz University, Jeddah, Saudi Arabia

Abstract

Post-synthetic modification of metal organic frameworks (MOFs) can be used to alter

their catalytic function. NU-1000 is a highly stable MOF consisting of Zr6-nodes, which

are reactive with precursors such as trimethylaluminum (AlMe3), trimethylindium

(InMe3), and dimethylaluminum isopropoxide (DMAI) by atomic layer deposition

(ALD). MOFs decorated with Al and In may then be used to catalyze industrially or

environmentally important reactions such as ethane to ethanol conversion. In this work

we used density functional theory to determine the mechanism of addition of the AlMe3,

InMe3, and DMAI precursors to NU-1000 to discern the structure and possible catalytic

utility of these modified MOFs. To confirm our computational results we compared to

experimental X-ray pair distribution function (PDF) data. There is reasonably good

agreement between the proposed computational structures and transition states and

experimental data. Experimental results indicate that a maximum of eight metals add to

each Zr6-node. The computational mechanism predicts that two metals will add to each

face of the Zr6-nodes by reacting with the –OH, -OH2, and µ-OH groups. This leads to a

final structure with eight metals per node with distorted tetrahedral geometries.

MTCC Abstract

Title: LICHEM: A QM/MM Interface for Polarizable Force Fields

Abstract:

We introduce the open source LICHEM software package for QM/MM simulations multipolar and

polarizable force fields. This initial implementation can interface with Gaussian, PSI4, NWChem,

TINKER, and TINKER-HP to enable QM/MM simulations using either the AMOEBA or point-charge

force fields. LICHEM employs wrappers to unmodified QM and MM software packages to perform

geometry optimizations, single-point energies, reaction pathways, and Monte Carlo simulations. We

tested our initial implementation on human DNA polymerase λ and (H2O)n (n=2,3,21) clusters. The

calculations confirm that LICHEM accurately performs high-quality QM/MM simulations and

highlight the importance of polarization.

Authors: Eric G. Kratz and G. Andres Cisneros

Presentation type: Oral presentation preferred if there are still slots available

NOVEL APPROACHES FOR HIGH PERFORMANCE COMPUTATIONAL CHEMISTRY 

 

MARK S. GORDON 

 

IOWA STATE UNIVERSITY 

 

In to enable accurate calculations on large molecules and molecular systems, one 

needs to develop novel methods and algorithms to avoid the typically high order 

scaling of correlated electronic structure methods. One type of method that can 

lower the scaling factor is a fragmentation method, and several of these will be 

discussed. A second approach is to develop algorithms that can make efficient use of 

novel computer architectures. One such approach will also be discussed. 

 

 

 

CH Stretch Vibrations as Probes of Local Environment

Daniel P. Tabor, 1 Timothy S. Zwier,

2 and Edwin L. Sibert III

1

1University of Wisconsin,

2Purdue University

The CH stretch region is an interesting candidate as a probe of structure and local environment.

The functional groups are ubiquitous and their vibration spectra exhibit a surprising sensitivity to

molecular structure. In this talk we review our theoretical model Hamiltonian [J. Chem. Phys.

138 064308 (2013)] for describing vibrational spectra associated with the CH stretch of CH2

groups and then describe an extension of it to molecules containing methyl and methoxy groups.

Results are compared to gas phase, conformer specific infrared spectroscopy of a set of

molecules that highlight the role of symmetry and environment. The curvilinear local-mode

Hamiltonian predicts most of the major spectral features considered in this study and provides

insights into mode mixing. We conclude by considering how the CH stretch spectrum of

cyclohexane is substantially modified when it forms a complex with a series of alkali metals and

what these spectra tell about the structure of the complex.

Completely Renormalized Coupled-Cluster Calculations for Bond Breaking Using Unrestricted Hartree-Fock References

Jared A. Hansen,1 Piotr Piecuch,1,2 and Jun Shen1

1Department of Chemistry, Michigan State University, East Lansing, MI 48824, U.S.A.

2Department of Physics and Astronomy, Michigan State University, East Lansing, MI 48824, U.S.A.

The popular coupled-cluster CCSD(T) approach works well for molecules near their equilibrium geometries, but it fails to describe potential energy surfaces (PESs) along bond breaking coordinates and biradicals if the restricted Hartree-Fock (RHF) determinant is used as a reference. One approach to address this concern is to use the unrestricted Hartree-Fock (UHF) determinant as the reference wave function in CCSD(T) computations. While this eliminates the unphysical characteristics of the RHF-based CCSD(T) PESs in the bond breaking regions, it introduces other issues, such as spin-contamination and non-analytic behavior of the resulting surfaces. Another way in which the failures of RHF-based CCSD(T) calculations can be addressed is by turning to the completely renormalized coupled-cluster (CR-CC) methods, such as CR-CCSD(T) and its more recent, rigorously size extensive, counterpart termed CR-CC(2,3). These approaches and their higher-order extensions provide an accurate description of biradical and single bond-breaking situations using RHF references, but the question remains as to whether or not the UHF-based CR-CC approaches, especially CR-CC(2,3), can further improve their RHF-based counterparts, particularly for single bond breaking into open-shell fragments on singlet PESs. To address this question, we present the UHF-based CR-CCSD(T) and CR-CC(2,3) results for bond breaking in the HF, F2, H2O2, and C2H6 molecules, comparing them with the exact, full configuration interaction, and full CCSDT data and the results of the RHF-based CR-CCSD(T) and CR-CC(2,3) calculations. We show that, unlike the CCSD(T) approach, which is very sensitive to the type of the reference determinant employed in the calculations, the CR-CC approximations provide a robust description regardless of the reference type (RHF or UHF), with the spin-adapted RHF-based CR-CC(2,3) results being most accurate for the bond breaking cases examined in this work.

Accurate Prediction of Lattice Energies of Molecular Crystals with Extended

Symmetry-Adapted Perturbation Theory

Ka Un Lao∗and John M. Herbert

Department of Chemistry and Biochemistry, The Ohio State University, Columbus, OH 43210

The efficient fragment-based method (XSAPT) developed by our group with chemical accuracyfor non-covalent interactions in molecular clusters has been extended to molecular crystals underperiodic boundary conditions (PBC) by incorporating long-range electrostatic interactions usingEwald-summation. The PBC-XSAPT method affords lattice energies within experimental uncer-tainty for a variety of molecular crystals, such as solid NH3, CO2, H2O, and C6H6. Crystallineoxalyl dihydrazide with five experimentally known polymorphs is a challenging system for theoret-ical crystal modeling since the polymorph energy ordering is governed by subtle balances betweendifferent kinds of interactions. Dispersion-corrected density functional approximations (DFT-D)fails to reproduce experimental observations. Nevertheless, our PBC-XSAPT method agrees withthe available experimental ordering. The “embarrassingly parallelizable” property of XSAPT makeit efficient enough to be applied to molecular crystals containing numerous monomer units. Further-more, the PBC-XSAPT interaction energy can be decomposed into physically meaningful energycomponents and such understanding is very useful for rational design of molecular crystals.

∗ I would like to give a contributed talk

Nuclear Quantum Effects are Important in Calculating the Electronic Spectrum of 9-Methylguanine

Yu Kay Lawa and Ali A. Hassanalib

Department of Science, Indiana University East, Richmond, INa and Condensed Matter Physics Section, The Abdus Salaam International Center for Theoretical Physics, Trieste, Italyb

Conformations of 9-methylguanine were sampled using the Cornell et al force field, ab initio Born-Oppenheimer molecular dynamics (BOMD)

simulation with classical nuclei and 9-methylguanine using a path-integral treatment of nuclear quantum effects (NQE). These conformations were

then used to calculate the heterogeneously-broadened absorption spectrum of 9-methylguanine using TD-DFT using the Franck-Condon approximation. We show that incorporation of nuclear quantum effects in

conformational sampling significantly broadens the distribution of bond lengths in 9-methylguanine, leading to significant broadening to the red-end of the calculated absorption spectrum, leading to much improved

correspondence with the experimental spectrum for this molecule. This suggests significant conformational freedom can be attributed to the

presence of quantum effects on the nuclei, something that is often neglected in conformational sampling. The role of the inclusion of water as part of the model in TD-DFT calculations, at the macroscopic level, was

largely found to modulate the transition dipole moments of the guanine base rather than the excitation energies; perturbations due to the

presence of water varied between different conformers, leading to a cancellation of their overall effect on the bulk absorption spectrum. 

Conjugation in Aromatic S-Nitrosothiols: Interaction of a π-

bond and –SNO Moiety

Matthew Flister and Qadir K. Timerghazin*

Department of Chemistry, Marquette University

Milwaukee, WI 53233

The development of new S-nitrosothiols (RSNOs) is key to the advancement of a

growing field of study in the biochemistry of nitric oxide (NO) due to the ubiquitous

nature of RSNOs in living organisms. These thiol derivatives can undergo homolytic

dissociation of the S–N bond releasing NO and possibly nitroxyl HNO, small molecules

with significant impact in many biological processes. While progress has been made in

biological applications of NO, few synthetic or endogenous RSNOs have emerged with

controlled and tunable release of NO. Moreover, many stable synthetic RSNOs involve

bulky substituent groups inducing stability of the –SNO group but preventing the

controlled release of NO. Study of substituent effects on the –SNO group should serve as

a reliable means for systematic development of new RSNOs. We have previously studied

PhSNO as a simple scaffold to use in substituent study. However, due to the non-planar

geometry of PhSNO and limited conjugation between the aromatic ring and –SNO group,

it is still unclear as to the exact effect of conjugation between a π-bond or π-system and

an RSNO. In this contribution, we will detail a thorough investigation of the simplest

form of RSNO in proximity to a π-bond or π-system, vinyl-SNO (VinSNO). We will

discuss the complex electronic structure of VinSNO using results from natural bond

orbital (NBO) calculations with special emphasis on the donor/acceptor interactions

between the π-bond and –SNO group of VinSNO. Furthermore, we will provide

comparison between VinSNO and PhSNO toward a broader understanding of the whole

class of aromatic RSNOs.

Reduced-scaling electronic structure theory approaches for simulating

responsive organic materials

Keith A. Werling and Daniel S. Lambrecht

University of Pittsburgh, Department of Chemistry, 219 Parkman Ave, Pittsburgh PA, 15206

We present a hierarchy of hybrid approaches for embedded many-body expansions combined

with local approximations to enable expedited irst principles calculations of organic materials

properties. We demonstrate calculations of deformation in response to external electric and

mechanical stimuli within this framework, as is required to assess applications as responsive

materials (e.g. piezoelectric sensors, shape-shifting electromechanical materials). Treating

these crystalline or semi-crystalline molecular materials under the inluence of external

perturbations requires a balanced description of subtle intermolecular forces. We show that

the presented approaches provide CCSD(T)-quality results while appealing with reduced

scaling with system size as low as O(N) as well as embarrassing parallelism. Diferent variants

for the many-body embedding, ranging from point charges to quantum mechanical embedding,

are analyzed with respect to the quality of the results and computational eiciency. We then

present applications of our approaches to inding improved hydrogen-bonded organic

piezoelectric materials [1-2].

[1] K. A. Werling, G. R. Hutchison, and D. S. Lambrecht, “Piezoelectric Efects of Applied

Electric Fields on Hydrogen-Bond Interactions: First-Principles Electronic Structure

Investigation of Weak Electrostatic Interactions”, J. Phys. Chem. Lett. 4, 1365-1370 (2013).

[2] K. A. Werling, M. Griin, G. R. Hutchison, and D. S. Lambrecht, "Piezoelectric Hydrogen

Bonding: Computational Screening for a Design Rationale", J. Phys. Chem. A 118, 7404-7410

(2014).

An approach to pure-sampling quantum Monte Carlo.

Egor Ospadov and Stuart M. Rothsteina,

Departments of Chemistry and Physics,

Brock University,

St. Catharines, ON L2S 3A1 CANADA

Diffusion quantum Monte Carlo, the most widely-used quantum Monte Carlo algorithm, samples from

ΨΦ0. This so-called “mixed distribution” is the product of an inputted importance sampling function (Ψ)

and the unknown “exact” ground-state wave function (Φ0), which is exact, save for the mismatch of its

nodal hypersurface with that of the truly exact wave function. The importance sampling function Ψ

substantially biases physical properties represented by operators that commute with the position operator,

such as the dipole moment. The objective of pure-sampling quantum Monte Carlo is to remove Ψ from the

sampled distribution, to sample from the so-called “pure distribution”, |Φ0|2, and thus to calculate physical

properties that are independent of the importance sampling function being employed in the calculation.

We describe a pure-sampling quantum Monte Carlo algorithm that achieves this objective [EO and SMR, J.

Chem. Phys. 142, 024114 (2015)] Our algorithm is implemented by systematically increasing an

algorithmic parameter until the calculations converge to statistically equivalent values. Thereby one

unambiguously determines values for the ground-state energy, static electrical response properties and

other one-electron expectation values. These quantities are free from biases that plague other approaches in

the literature: importance sampling bias, population control bias, time-step bias, extrapolation-model bias,

and the finite-field approximation. Applications of the algorithm to a variety of molecules are described,

with some emphasis on technical challenges poised by large molecules.

______________________________________________________ a srothste@brocku.ca

SMR should like to give a contributed talk.

Performance and Energy Efficiency of Quantum Chemistry Algorithms on Modern Computer Architectures

Kristopher Keiperta. Gaurav Mitra

b, Vaibhav Sunriyal

a, Sarom S. Leang

a, Masha Sosonkina

c,

Alistair Rendellb, and Mark S. Gordon

a

aDepartment of Chemistry and Ames Laboratory, Iowa State University, Ames, Iowa 50011-3111,United States

bResearch School of Computer Science, Australian National University, Acton ACT 0200,Australia

cDepartment of Modeling and Simulation, Old Dominion University, Norfolk, VA 23529 United States

The energy costs associated with cooling and powering a high performance supercomputer system usually surpass

the initial equipment costs over 3-4 years of use. Poor energy efficiency of modern computer hardware is the most

significant barrier to reaching exascale computing. Development of significantly more efficient hardware designs and

software algorithms is necessary to continue advancement of computationally feasible chemical system sizes which can be

studied with accurate ab initio quantum chemistry algorithms . Most mobile consumer devices use RISC-based ARM CPUs

that are designed for low power consumption. Recent improvements in mobile computing performance have enabled use of

ARM CPUs for high performance scientific applications. Modern CISC-based architectures such as Haswell x86 have been

designed with a focus on energy efficiency as well. The performance and energy-to-solution for a variety of quantum

chemistry algorithms in the GAMESS quantum chemistry application have been measured for ARM32, ARM64, and

Haswell x86 architectures. The Haswell x86 system is shown to have 2-4x better performance than the ARM systems for all

benchmark cases, while having worse energy efficiency than ARM64 for some parallel algorithms. The ARM32 device is

more energy efficient than ARM32 and ARM64 for all benchmark cases.

Note: I would like to present this orally.

Midwest Theoretical Chemistry Conference Abstract Title: Using Range-Separated Hybrid Density Functional Theory for Rational Design of Organic Optoelectronic Devices Authors: Heidi Hendrickson Zilong Zheng, Francis Devine, Shaohui Zheng, Eitan Geva, Barry Dunietz Abstract:

A fundamental understanding of charge separation in organic materials is necessary for the rational design of optoelectronic applications. Computational approaches provide fundamental insights into related processes, which in turn guide the synthesis of novel optoelectronic materials. Conventional density functional theory (DFT) methods have been known to fail in accurately characterizing frontier orbital gaps and charge transfer states in molecular systems. We address these shortcomings by implementing an optimally-tuned range-separated hybrid (OT-RSH) functional approach within DFT and time-dependent (TD) DFT. Our work particularly addresses the effect of the surrounding environment on fundamental molecular properties such as ionization potential (IP), electron affinity (EA), and charge transfer excitation energies. Specifically, we have investigated the spurious agreement between thin film IP/EA measurements and gas-phase frontier orbital energies calculated with widely-used density functionals. We show that both gas-phase and environmentally-corrected RSH-DFT frontier orbital energies properly correspond to gas-phase and thin-film experimental measurements, respectively, for a set of organic semiconducting molecules.

We also benchmark the RSH functionals in describing charge transfer excitation by using a model ethene dimer and silsesquioxane molecules currently investigated as candidates for building blocks in photovoltaic applications. In order to account for complex environmental effects on charge transfer energies in silsesquioxane molecules, a protocol combining charge-constrained DFT, a polarizable continuum solvent model, and RSH TDDFT was tested and validated against experimental measurements. The protocol provides a way to understand charge transfer within complex environments for molecules used in photovoltaic applications. Current work is focused on extending this method to various silsesquioxane systems.

Evolving Fluorophores for Organic Light Emitting Diodes

Yinan Shu and Benjamin G. Levine

Department of Chemistry, Michigan State University, East Lansing, MI 48824

Organic light-emitting diodes (OLEDs) are the basis for low-cost, high-resolution flat panel displays.

Recently, Adachi and co-workers developed highly efficient OLEDs which employ thermally activated

delayed fluorescence (TADF) to enhance the luminescence quantum yield. TADF is most efficient when

the gap between S1 and T1 is small and the S1-S0 transition dipole moment is large, but these two properties

are difficult to achieve simultaneously. By taking advantage of parallel graphics processing unit (GPU)

computing and genetic algorithms, we are able to pre-screen a large set of possible candidates and evolve

these molecules towards the desired properties. The fitness of each candidate for TADF is estimated using

GPU-accelerated density functional and time-dependent density functional calculations. The set of optimal

candidates identified in our study includes some molecules known to exhibit TADF and others that have

not been reported in the literature.

Dear Organnizer:

I would like to apply for oral presentation.