The deMon-Nano User's Guide

The deMon-Nano User's Guide
The deMon-Nano User’s Guide
Installation Guide and Reference Manual
The deMon-Nano User’s Guide
Installation Guide and Reference Manual
Version deMon-Nano 2009 Experiment
September 2009
Reference: The deMon User’s Guide, Version deMon-Nano Experiment 2009
Authors: T. Heine, M. Rapacioli, S. Patchkovskii, J. Frenzel,
A.M. K¨oster, P. Calaminici, H. A. Duarte, S. Escalante, R. Flores-Moreno,
A. Goursot, J.U. Reveles, D.R. Salahub, A. Vela
Title Picture: A [email protected] multiwalled nanocable
The procedures and applications presented in this guide have been included for their
instructional value. They have been tested with care but are not guaranteed for any
particular purpose. The authors do not offer any warranties or representations, nor do
they accept any liabilities with respect to the programs or applications.
The software described in this guide is furnished under a license and may be used or
copied only in accordance with the terms of such license.
c
Copyright 2003-2009
by the authors
All rights reserved. No part of this publication may be reproduced, stored in a retrieval
system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without prior written permission of one of the authors.
Contents
i
Contents
1 Getting Acquainted
1
1.1 The Game of the Name . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1
1.2 How to Avoid Reading this Manual . . . . . . . . . . . . . . . . . . . . .
2
1.3 How to Read this Manual . . . . . . . . . . . . . . . . . . . . . . . . . .
2
1.4 How to Use deMon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3
2 Getting Started
7
2.1 Before you Begin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7
2.2 How to Install deMon . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7
2.2.1
Contents of the Distribution Package . . . . . . . . . . . . . . . .
7
2.2.2
Installation Prereqisites
. . . . . . . . . . . . . . . . . . . . . . .
8
2.2.3
Express Installation . . . . . . . . . . . . . . . . . . . . . . . . . .
9
2.2.4
Advanced Installation
. . . . . . . . . . . . . . . . . . . . . . . .
9
2.3 Porting deMon to a new platform . . . . . . . . . . . . . . . . . . . . . .
11
2.4 Tuning deMon for a specific host . . . . . . . . . . . . . . . . . . . . . .
12
2.5 What to do when things go wrong . . . . . . . . . . . . . . . . . . . . . .
13
2.6 How to Run deMon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
16
3 Carrying On
19
3.1 Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
19
3.2 Files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
19
3.3 Input Syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
20
4 Keywords
23
4.1 Geometry Input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
23
4.1.1
Keyword GEOMETRY . . . . . . . . . . . . . . . . . . . . . . .
23
4.1.2
Keyword CONSTANTS . . . . . . . . . . . . . . . . . . . . . . .
32
4.1.3
Keyword VARIABLES . . . . . . . . . . . . . . . . . . . . . . .
33
4.1.4
Keyword SYMMETRY . . . . . . . . . . . . . . . . . . . . . . .
33
4.2 Basis Set Input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
35
4.2.1
Keyword BASIS . . . . . . . . . . . . . . . . . . . . . . . . . . .
35
4.2.2
Keyword AUXIS . . . . . . . . . . . . . . . . . . . . . . . . . . .
36
4.2.3
Keyword ECPS . . . . . . . . . . . . . . . . . . . . . . . . . . .
38
ii
Contents
4.3 Electronic State Control . . . . . . . . . . . . . . . . . . . . . . . . . . .
40
4.3.1
Keyword MULTIPLICITY . . . . . . . . . . . . . . . . . . . . .
40
4.3.2
Keyword CHARGE . . . . . . . . . . . . . . . . . . . . . . . . .
40
4.3.3
Keyword MOEXCHANGE . . . . . . . . . . . . . . . . . . . . .
40
4.3.4
Keyword FIXMOS . . . . . . . . . . . . . . . . . . . . . . . . . .
41
4.3.5
Keyword SMEAR . . . . . . . . . . . . . . . . . . . . . . . . . .
41
4.3.6
Keyword CONFIGURE . . . . . . . . . . . . . . . . . . . . . . .
42
4.4 SCF Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
46
4.4.1
Keyword SCFTYPE . . . . . . . . . . . . . . . . . . . . . . . . .
46
4.4.2
Keyword ORBITALS . . . . . . . . . . . . . . . . . . . . . . . .
47
4.4.3
Keyword ERIS . . . . . . . . . . . . . . . . . . . . . . . . . . . .
48
4.4.4
Keyword GUESS . . . . . . . . . . . . . . . . . . . . . . . . . .
50
4.4.5
Keyword MIXING . . . . . . . . . . . . . . . . . . . . . . . . . .
51
4.4.6
Keyword SHIFT . . . . . . . . . . . . . . . . . . . . . . . . . . .
52
4.4.7
Keyword DIIS . . . . . . . . . . . . . . . . . . . . . . . . . . . .
53
4.4.8
Keyword BROYDEN . . . . . . . . . . . . . . . . . . . . . . . .
54
4.4.9
Keyword VXCTYPE . . . . . . . . . . . . . . . . . . . . . . . .
54
4.4.10 Keyword GRID . . . . . . . . . . . . . . . . . . . . . . . . . . .
56
4.5 Molecular mechanics and QM/MM Control . . . . . . . . . . . . . . . . .
59
4.5.1
Keyword QMMM . . . . . . . . . . . . . . . . . . . . . . . . . .
59
4.5.2
Keyword FORCEFIELD . . . . . . . . . . . . . . . . . . . . . .
60
4.5.3
Keyword MMOPTIONS . . . . . . . . . . . . . . . . . . . . . .
61
4.5.4
Keyword VDWAALS . . . . . . . . . . . . . . . . . . . . . . . .
62
4.5.5
Keyword MADELUNG . . . . . . . . . . . . . . . . . . . . . . .
62
4.5.6
Keyword PERIOIC . . . . . . . . . . . . . . . . . . . . . . . . .
63
4.6 DFTB Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
65
4.6.1
Keyword DFTB . . . . . . . . . . . . . . . . . . . . . . . . . . .
65
4.7 Optimization Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
69
4.7.1
Keyword OPTIMIZATION . . . . . . . . . . . . . . . . . . . . .
69
4.8 MD Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
70
4.8.1
Keyword MDYNAMICS . . . . . . . . . . . . . . . . . . . . . .
70
4.8.2
Keyword MDSTEPS . . . . . . . . . . . . . . . . . . . . . . . .
71
4.8.3
Keyword MDTEMP . . . . . . . . . . . . . . . . . . . . . . . . .
72
Contents
iii
4.8.4
Keyword TIMESTEP . . . . . . . . . . . . . . . . . . . . . . . .
72
4.8.5
Keyword MDBATH . . . . . . . . . . . . . . . . . . . . . . . . .
73
4.8.6
Keyword CONSERVE . . . . . . . . . . . . . . . . . . . . . . . .
74
4.8.7
Keyword MDCONSTRAINTS . . . . . . . . . . . . . . . . . . .
75
4.8.8
Keyword MDPRESSURE . . . . . . . . . . . . . . . . . . . . . .
76
4.8.9
Keyword HEATPIPE . . . . . . . . . . . . . . . . . . . . . . . .
77
4.8.10 Keyword CARPAR . . . . . . . . . . . . . . . . . . . . . . . . .
80
4.9 Property Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
82
4.9.1
Keyword POPULATION . . . . . . . . . . . . . . . . . . . . . .
82
4.9.2
Keyword DIPOLE . . . . . . . . . . . . . . . . . . . . . . . . . .
82
4.9.3
Keyword MAGOUT . . . . . . . . . . . . . . . . . . . . . . . . .
82
4.9.4
Keyword POLARIZABILITY . . . . . . . . . . . . . . . . . . .
83
4.9.5
Keyword FREQUENCY . . . . . . . . . . . . . . . . . . . . . .
84
4.9.6
Keyword THERMO . . . . . . . . . . . . . . . . . . . . . . . . .
85
4.9.7
Keyword FNMC . . . . . . . . . . . . . . . . . . . . . . . . . . .
85
4.9.8
Keyword HARDNESS . . . . . . . . . . . . . . . . . . . . . . .
85
4.10 Solvent Effects
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
87
4.10.1 Keyword SOLVENT . . . . . . . . . . . . . . . . . . . . . . . . .
87
4.10.2 Keyword PCM-CARD . . . . . . . . . . . . . . . . . . . . . .
91
4.11 Visualization and Topology . . . . . . . . . . . . . . . . . . . . . . . . .
93
4.11.1 Keyword VISUALIZATION . . . . . . . . . . . . . . . . . . . .
93
4.11.2 Keyword PLOT . . . . . . . . . . . . . . . . . . . . . . . . . . .
94
4.11.3 Keyword CPSEARCH . . . . . . . . . . . . . . . . . . . . . . . .
96
4.11.4 Keyword ISOSURFACE . . . . . . . . . . . . . . . . . . . . . . .
97
4.11.5 Keyword GEOSURFACE . . . . . . . . . . . . . . . . . . . . . . 100
4.11.6 Keyword BOX . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
4.11.7 Keyword POINTS . . . . . . . . . . . . . . . . . . . . . . . . . . 103
4.12 Miscellaneous Keywords . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
4.12.1 Keyword MAXMEM . . . . . . . . . . . . . . . . . . . . . . . . 105
4.12.2 Keyword TITLE . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
4.12.3 Keyword PRINT
. . . . . . . . . . . . . . . . . . . . . . . . . . 105
4.12.4 Keyword EMBED . . . . . . . . . . . . . . . . . . . . . . . . . . 107
4.12.5 Keyword CHOLESKY . . . . . . . . . . . . . . . . . . . . . . . . 108
iv
Contents
4.12.6 Keyword MATDIA . . . . . . . . . . . . . . . . . . . . . . . . . 109
4.12.7 Keyword WEIGHTING . . . . . . . . . . . . . . . . . . . . . . . 109
4.12.8 Keyword QUADRATURE . . . . . . . . . . . . . . . . . . . . . 110
4.12.9 Keyword ECPINTEGRATION . . . . . . . . . . . . . . . . . . . 110
4.12.10 Keyword GENAO . . . . . . . . . . . . . . . . . . . . . . . . . . 110
5 Examples
112
5.1 Example ps18 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
5.2 Examples amk5, amk6, amk7, and amk8 . . . . . . . . . . . . . . . . . . 112
5.3 Examples amk29, amk30, and amk31 . . . . . . . . . . . . . . . . . . . . 112
5.4 Example ps86 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
5.5 Example th6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
5.6 Example amk17 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
5.7 Example ps84 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
5.8 Example amk19 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
5.9 Example amk20 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
5.10 Example ps15 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
5.11 Example ps6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
5.12 Example amk24 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
5.13 Example amk26 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
5.14 Example amk33 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
5.15 Example amk35 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
5.16 Example ps88 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
5.17 Example ps90 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
6 Using QM/MM in deMon
113
7 Using MD in deMon
114
8 Vu
115
8.1 What is Vu? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
8.2 Vu and deMon
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
8.3 Running Vu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
9 MAG
118
Contents
v
10 Quick Keyword Reference
119
11 Troubleshooting
120
A Automatic Generation of Auxiliary Functions
121
B Format of a platform description file
123
B.1 Philosophy of deMon configuration files . . . . . . . . . . . . . . . . . . . 123
B.2 Structure of a platform configuration files . . . . . . . . . . . . . . . . . . 123
C Working with the deMon Test Suite
129
D Global Counters, Limits and Pointers
130
References
133
vi
List of Tables
List of Tables
1
Platform-specific installation notes . . . . . . . . . . . . . . . . . . . . .
7
2
Parameter Settings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
20
3
Logical and Physical Filenames . . . . . . . . . . . . . . . . . . . . . . .
21
4
Special Input Symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . .
21
5
Internal Coordinate Substitution . . . . . . . . . . . . . . . . . . . . . .
27
6
Basis Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
37
7
ERI Calculation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
49
8
Exchange and Correlation Functionals . . . . . . . . . . . . . . . . . . .
56
9
Molecular Fields
96
10
Structure of a platform configuration file . . . . . . . . . . . . . . . . . . 124
11
Platform Specific Functions . . . . . . . . . . . . . . . . . . . . . . . . . 128
12
Counters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
13
Limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
14
Pointers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
List of Figures
vii
List of Figures
1
Z-Matrix Input Orientation . . . . . . . . . . . . . . . . . . . . . . . . .
25
2
Dihedral Angle Definition . . . . . . . . . . . . . . . . . . . . . . . . . .
26
3
Mixed Input Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . .
29
4
Set up of heatpipe zone masks . . . . . . . . . . . . . . . . . . . . . . . .
78
5
Representation of the benzene cavity surface. . . . . . . . . . . . . . . . .
91
1 Getting Acquainted
1
1
Getting Acquainted
1.1
The Game of the Name
deMon is a system of programs for density functional theory (DFT) calculations of atoms,
molecules and solids [1–3]. Its first widely available version [4] appeared in 1992. Shortly
after its appearance the original deMon code was substantially modified for commercialization by BIOSYM Technologies. The beta-release of this version appeared 1993. It was
the basis of the deMon-KS1 [5] series of programs developed in Montreal until 1997.
Meanwhile the original deMon version was further developed in Montpellier and Stockholm. These developments were first independent from each other. In 1997 they merged
to the deMon-KS3 [6] series of programs.
Independently from the deMon development the AllChem project started [7] in Hannover in 1995. The aim of this project was to write a well structured DFT code from
scratch. The first AllChem version appeared in 1997. The structured programming of
AllChem proved very useful for the development and testing of new DFT approaches
and algorithms.
In March 2000 the first deMon developers meeting was hold in Ottawa. At this meeting
the deMon and AllChem developers agreed to merge their codes in order to keep a
Tower of Babel from rising. As a result the new code couples the deMon functionality
with the stable and efficient integral and self-consistent field (SCF) part from AllChem.
The new code, which is named deMon 2003 [8], was presented the first time at the third
deMon developers meeting in Geneva.
The new deMon 2003 code is available for just about any computer made today, including
parallel architectures. The versions for the different platforms are essentially the same.
The input file created according to this guide should give, within the numerical accuracy
of the used compile flags, the same output.
The written word carries more legal complications than the spoken, and the need to
distinguish deMon 2003 from similarly spelled products restricts how you may write it.
The best way to refer to the program in the text is to use the unusual capitalization. The
new deMon code has to be cited as:
A.M. K¨oster, R. Flores-Moreno, G. Geudtner, A. Goursot, T. Heine,
J.U. Reveles, A. Vela, S. Patchkovskii, D.R. Salahub, deMon 2004, NRC, Canada
There is another companion to this guide, The deMon Programmer’s Guide, which cur-
2
1 Getting Acquainted
rently under development. This companion is an in-depth guide about the programming
philosophy and style of the new deMon code. Its the place to look if you want to extend or
modify the program. To learn more about other deMon utility programs and interfaces as
well as activities of the deMon user community look at http://www.deMon-software.com.
1.2
How to Avoid Reading this Manual
If you are familiar with an older version of deMon or you rather prefer to learn by
example, look at the directory examples in the deMon tree you have obtained. It contains
sample inputs for the most common applications (see Chapter 5 for the description of
the examples). In Chapter 10 you will find a quick reference guide of the input keywords.
Common problems and their solutions are described in Chapter 11, ”Troubleshooting”.
If you prefer to learn more about the deMon input syntax before you use it, read on.
Almost everything used in the sample inputs is explained in the next two chapters.
1.3
How to Read this Manual
While the example inputs cover the most common standard applications of an quantum
chemistry program like deMon, there is much more to learn about input possibilities.
Eventually, you will want to control specific parts of a calculation or you will need more
output information. You will then have to look in this manual for the necessary information. You can read the information containing section without the preceding ones.
However, all later chapters assume that you have read Chapters 2 and 3.
For example, suppose you want to print molecular orbitals (MOs). Looking up the table
of contents will lead you to Section 4.12.3, which describes the PRINT Keyword. The
concept of keywords and options, as well as their notation in this guide, is described
in Chapter 3. It will take you just a moment to construct the input line for printing
molecular orbitals if you have already read chapter 3; it could be quite frustrating if you
haven’t. So it’s best to read the three first chapters now. Here is a brief sketch of what’s
in the remaining chapters:
Chapter 2 gives an overview about how to install and run the program using the makefiles and scripts which come with deMon.
Chapter 3 explains how to customize the program parameters and input/output files.
It also explains the input syntax.
1.4 How to Use deMon
3
Chapter 4 gives a detailed description of all input keywords and their options.
Chapter 5 discusses the input and output of the examples.
Chapter 8 describes the interface between deMon and Vu, a configurable visualization
tool. It also explains basic operations within Vu.
Chapter 9 describes the interface between deMon and MAG, a program for the calculation of NMR properties. It also describes the use of MAG.
Chapter 10 is a quick reference guide to all input keywords.
Chapter 6 is a more in-depth introduction to QM/MM features implemented in deMon.
Chapter 7 gives additional hints on running molecular dynamics simulations.
Chapter 4.6 gives all DFTB and CP-DFTB related commands.
Chapter 11 describes troubleshooting of common problems.
Appendix A describes the automatic generation of auxiliary function sets.
Appendix B describes the format of the platform configuration files, used by the deMon
build system.
Appendix C describes the structure test suite for deMon, and how to expand it.
1.4
How to Use deMon
In deMon linear combinations of atomic Gaussian–type orbitals (LCGTO) are used for
representing Kohn-Sham orbitals. In this ansatz the Kohn–Sham orbitals ψi (r) are given
by:
ψi (r) =
X
cµi µ(r)
(1.1)
µ
Here µ(r) represents an atomic orbital (build from contracted Gaussians) and cµi the
corresponding molecular orbital coefficient. With this expansion we find for the electronic
density:
ρ(r) =
X
µ,ν
Pµν µ(r) ν(r)
(1.2)
4
1 Getting Acquainted
Pµν represents an element of the (closed-shell) density matrix, defined as:
occ
X
Pµν = 2
cµi cνi
(1.3)
i
Using the LCGTO expansions for the Kohn–Sham orbitals (1.1) and the electronic density
(1.2), the Kohn–Sham self-consistent field (SCF) energy expression [9] can be calculated
as:
ESCF =
X
Pµν Hµν +
µ,ν
1 XX
Pµν Pστ h µν k στ i + Exc [ρ]
2 µ,ν σ,τ
(1.4)
The total energy is the sum of ESCF and the nuclear repulsion energy, which can be
calculated analytically. In (1.4), Hµν represents matrix elements of the core Hamiltonian.
They are built from the kinetic and nuclear attraction energy of the electrons and describe the movement of an electron in the nuclear framework. The second term in (1.4)
represents the Coulomb repulsion energy of the electrons. In contrast to Hartree–Fock
theory, the calculation of the Coulomb and exchange energies are separated in DFT. For
the calculation of the exchange–correlation energy Exc [ρ] a numerical integration has to
be performed. In deMon, the calculation of the N 4 scaling Coulomb repulsion energy
is avoided by introducing an auxiliary function expansion for the electron density. This
¯
approximated density ρ˜(r) is expanded in primitive Hermite Gaussians k(r)
which are
centered at the atoms:
ρ˜(r) =
X
¯
xk k(r)
(1.5)
k
With the LCGTO expansion for ρ(r) and ρ˜(r) we obtain the following approximate SCF
energy:
ESCF =
X
Pµν Hµν +
µ,ν
X
xk
X
µ,ν
k
Pµν h µν k k i −
1 X
xk xl h k k l i + Exc [ρ]
2 k,l
(1.6)
Therefore, only three–center electron repulsion integrals (ERIs) are necessary for the
SCF and energy calculation in deMon. This represents the most accurate energy model
available in deMon. It is activated by the keyword VXCTYPE BASIS (see Section 4.4.9
for more details about the VXCTYPE keyword). By default (VXCTYPE AUXIS), the
approximated energy is also used for the calculation of the exchange–correlation energy:
ESCF =
X
µ,ν
Pµν Hµν +
X
k
xk
X
µ,ν
Pµν h µν k k i −
1.4 How to Use deMon
5
1 X
xk xl h k k l i + Exc [˜
ρ]
2 k,l
(1.7)
This approximation has proven very accurate (∼ 1 kcal/mol) in combination with GENA* auxiliary functions sets (see Section 4.2.2 and Appendix A for the selection and automatic generation of GEN-A* auxiliary function sets). Because it represents a considerable
saving in computational time we suggest to use this approximation at the beginning of
each study. If high accuracy is requested the systems optimized with VXCTYPE AUXIS
(1.7) should be further optimized with VXCTYPE BASIS (1.6). Usually, convergence will
be achieved within a few cycles. If sensitive properties like frequencies, polarizabilities
etc. are intended to be calculated, the use of VXCTYPE BASIS is recommended. The
differences between VXCTYPE AUXIS and VXCTYPE BASIS for geometries and bond
energies are in the range of the accuracy of the methodology. An exception should be
made for very weak bonds, where the geometry may be significantly affected by the use
of fitted density. It should be noted that the default setting for the auxiliary functions is
A2, independent which energy expression is used (see Section 4.2.2). For all theoretical
models available in deMon VXCTYPE AUXIS results can be used as a restart guess
(GUESS RESTART; see Section 4.4.4) for VXCTYPE BASIS calculations.
The most often encountered problem in DFT calculations is the failure of the SCF convergence. This is usually due to the small gap between the highest occupied (HOMO) and
lowest unoccupied (LUMO) molecular orbital. In deMon the DIIS procedure (Section
4.4.7) is activated by default. For a small HOMO–LUMO gap DIIS may be counterproductive and should be switched off. Several alternative methods are available in deMon to enforce SCF convergence. Most important are modification in the choice of the
starting GUESS (Section 4.4.4) and the MIXING (Section 4.4.5) procedure of the old and
new densities and the enlarging of the HOMO–LUMO gap by the level SHIFT (Section
4.4.6) procedure. In all cases it is recommended to check the orbital energies and occupations using the PRINT keyword (Section 4.12.3). Other relevant keywords to alter or
enforce SCF convergence are MOEXCHANGE (Section 4.3.3), FIXMOS (Section 4.3.4)
and SMEAR (Section 4.12.3). For atomic calculation the CONFIGURATION keyword
(Section 4.3.6) should be used in order to ensure SCF convergence.
By default the electron repulsion integrals (ERIs) are calculated at the beginning of the
SCF and kept in main memory (RAM). For larger systems the ERIs may not fit into the
RAM. In this case deMon will automatically switch to a direct SCF procedure, which
may also be activated manually, by using option DIRECT of the ERIS keyword (see
6
1 Getting Acquainted
4.4.3). It is also possible to store ERIs on disk, by using ERIS CONVENTIONAL input
keyword. However, this results in an I/O bottleneck, which will slow down the calculation
on most computer architectures. For very large systems the option MULTIPOLE of the
ERIS keyword is recommended. It activates an asymptotic expansion of the ERIs, which
results in a considerable time saving. It should also be noted that for very large systems
the linear algebra steps in deMon may become a bottleneck. The keyword PRINT RAM
may be useful in resolving such issues. The amount of memory required in the matrix
diagonalization step (which dominates memory requirements for large systems) can be
adjusted with the keyword MATDIA (see 4.12.6).
If the calculation has to be restarted the new input file produced in each deMon run may
be useful. The input may be modified and extended. However, the molecular geometry
specified in this file should not be changed if the restart file will be used. However, the
new input file may change the ordering of the atoms in the molecule. As a result, it should
not be used for continuing molecular dynamics simulations.
2 Getting Started
2
7
Getting Started
2.1
Before you Begin
Before you begin the installation process, please make sure to consult operating system
specific instructions, which came with the distribution. They are found in the subdirectory
“doc/os” of the deMon distribution. The list of system-specific notes, available at the
time this manual was prepared, is given in table1.
Table 1: Platform-specific installation notes in doc/os.
README.forcheck
Forchek (http://www.forcheck.nl/)
README.ifc
Intel Fortran compiler on Linux/x86
README.pathscale
README.pgi
PathScale Fortran-90 compiler on Linux/x86-64 (AMD64)
Portland Group Fortran-90 compiler on Linux/x86 and
Linux/x86-64
README.SunOS
SUN OS and Solaris hosts
README.Tru64
OSF/1, Digital Unix, and Tru64 hosts
README.windows
MicroSoft Windows
These notes summarise many days of porting and debugging, and may save you countless
hours of frustration!
2.2
2.2.1
How to Install deMon
Contents of the Distribution Package
deMon distribution package contains the following:
basis/
-
deMon basis sets.
bin/
-
Executable directory
doc/
-
Documentation.
examples/
-
Some examples, and the test suite.
include/
-
Include files.
makefiles/
-
Make files (see section 2.2.4).
objects/
-
Scratch area for intermediate build objects.
8
2 Getting Started
source/
-
deMon source code.
timings/
-
Some representative timing data.
tools/
-
Additional deMon programs.
database/
-
Unsupported CREX configuration files
unsupported/ -
Various unsupported tools and scripts
Please keep in mind that many subdirectories contain specific “README.txt” files. You
may find those useful.
2.2.2
Installation Prereqisites
You will need a computer system capable of running deMon. At the absolute minimum,
you will need 1 Gbyte of free disk space to compile deMon, and run the installation
test set. To run the full test set, you will need at least 512 Mbytes of RAM installed in
your computer. Your computer must run a Unix-like operating system to install deMon
successfully. (We may be able to supply precompiled binaries for non-Unix platforms).
You will also need a working Fortran-90
1
compiler. Not all compilers claiming to be
Fortran-90 actually are, or are capable of producing correct machine code. Unless deMon
has already been tested with your compiler (see 2.1), you very likely will see compilation
problems, ranging from compiler crashes, to run-time aborts, to incorrect results. On
average, we discover 2-5 compiler bugs (and a few bugs in deMon itself) every time
deMon gets ported to a completely new Fortran compiler. You have been warned.
Additionally, for best performance, you will need optimized BLAS libraries. If your vendor
does not provide such, you will have to obtain and install one of the publicly available
BLAS libraries yourself. A good place to start is: http://www.netlib.org/atlas/. Having
an optimized LAPACK is also helpful, but less important.
If your BLAS and LAPACK are installed in non-standard directories, you can specify
their location by setting the LIBPATH environment variable. For example, if your BLAS
libraries are in the directory “/usr/local/lib64”, you should say:
export LIBPATH="-L /usr/local/lib64/" # for bash
setenv LIBPATH "-L /usr/local/lib64/" # for csh
1
Many people insist on spelling this language name as Fortran. In fact, the correct spelling is simply
“Fortran”. It is specified by an international standard (ISO/IEC 1539).
2.2 How to Install deMon
9
If you require an unusual command line option, or a linker parameter to make your BLAS
work, it can be included in LIBPATH as well. This command has to be executed each
time you build deMon - so you may find it useful to put it in your shell profile.
2.2.3
Express Installation
If deMon has been previously ported to the operating system and compiler you are
using, and all needed libraries are installed in standard locations, you can use the express
installation procedure, namely:
1. Unpack deMon distribution files in subdirectory ”deMon” of your home directory.
2. Set deMon environment variable to point to the installation directory, for example:
export deMon=$HOME/deMon # For bash
setenv deMon $HOME/deMon # For csh
3. Run ”install.sh” script:
cd $deMon
./install.sh
4. Follow instructions on your screen.
If no errors are reported, you have a working deMon installation. Otherwise, proceed to
the advanced installation instructions in section 2.2.4.
2.2.4
Advanced Installation
The procedure outlined in this section is functionally identical to the express installation
(2.2.3). However, you will have an opportunity to take immediate corrective action if any
problems arise.
1. Set deMon environment variable to point to the installation directory, for example:
export deMon=$HOME/deMon # for bash
setenv deMon $HOME/deMon # for csh
The rest of this section will assume, what deMon is installed in the directory $deMon.
10
2 Getting Started
2. If you intend to study very large systems, with more than 5000 atoms, or more than
10000 contracted basis functions, you will need to modify some of the settings in
file “$deMon/include/parameter.h”. See section 3.1 for the instructions.
3. Go to the $deMon/makefiles/ subdirectory, and issue the command:
make makesys
The installation script will attempt to find combinations of options, which are
compatible with your platform. Please sevect the most specific set of parameters,
which applies to your configuration. If the only platform choice you are given is
“Generic settings”, deMon does not have a pre-defined set of compilation options
for your system. In this case, consult section 2.3 below. Please set the environment
variable(s) suggested by the platform detection script. Usually, the only required
variable is “deMonPlatform”. To make these settings permanent, please add them
to your shell initialization file (usually .cshrc or .profile, found in your home
directory). The deMonPlatform setting will also affect the default choice of the
deMon executable, selected by the deMon driver script (see Section 2.6).
4. It is usually a good idea to review the default compilation settings, particularly the
library locations and optimization flags. The file sys-$deMonPlatform.mak, which
is located in the directory $deMon/makefiles/sys contains this information.
If you change any of the compilation options in the platform-specific
makefile (sys-$deMonPlatform.mak), please execute the command “make
clean” before rebuilding your program. Otherwise, your build may be
inconsistent.
5. Build main deMon binary, by executing the command:
make clean
make build
The first command (make clean) removes all old makefiles, objects, and binaries
for this platform. The second command (make build) builds the main deMon executable, which is placed in the directory “$deMon/bin/”.
The build script takes care to keep all platform-specific data in unique files. It is
therefore safe to perform several builds at the same time, as long as the deMonPlatform
settings of the builds are different.
2.3 Porting deMon to a new platform
11
6. Build addional tools (MASTER and MAG), by issuing the command:
make tools
(You can combine the last two steps by using the command “make clean build
tools”).
7. Go to the $deMon/examples/ directory, and run deMon test suite:
cd $deMon/examples
make clean
# remove old test results
make cheap
# execute minimal test set
make all
# execute the complete test set
(You can also say “make clean all”, which is equivalent).
The test suite (Appendix C) can take a while to run. If all goes right, you should see
the message ”deMon installation tests completed successfully” after each of the test
phases. Otherwise, do not use the deMon executable you have compiled.
Currently, the minimal test set requires about ten minutes on a mid-range PC
(2.4GHz P4). The complete test set requires about two CPU-days. If you have a
queuing system (such as PBS, SGE, or LSF) installed, you can run the test set in
parallel. To activate parallel execution, set the environment variable QSUB COMMAND
to “qsub” (or whatever the name of your job submission command it):
export QSUB_COMMAND=qsub # for bash
setenv QSUB_COMMAND qsub # for csh
before starting the test set. Please consult $deMon/examples/README.txt for more
information on running deMon test set with batch queuing systems.
If the test set completes successfully, you are now ready to use deMon.
2.3
Porting deMon to a new platform
Compiling deMon on a previously upsupported platform involves a bit more work.
1. Examine the existing platform definition files in the “$deMon/makefiles/sys/”
directory. Select a platform definition file, which is most similar to your system. If
none of the platforms appear similar, choose “sys-generic.mak”.
12
2 Getting Started
2. Create a copy of the prototype platform definition file, using a (short) descriptive
name, corresponding to your platform.
3. Edit the new platform definiton file, following the instruction within the file itself.
Also see the appendix B.
4. Set the environment variable deMonPlatform to reflect the name of your configuration file. For example, if your new platform configuration file is called “sys-babayaga.mak”,
the appropriate shell command is:
export deMonPlatform=babayaga # for bash
setenv deMonPlatform babayaga # for csh
5. Execute the commands:
cd $deMon/makefiles
make clean build tools
6. Run the test suite (see sections 2.2.4 and C).
7. If any of the tests fail, iterate to 3 above.
2.4
Tuning deMon for a specific host
You can almost always improve the preformance of deMon by using higher optimization
levels, and/or supplying vendor-optimized mathematical libraries. The tuning procedure
is very similar to the porting procedure outlined section 2.3 above. The only difference
is:
1. In the directory $deMon/makefiles/sys/, choose a platform definition file which
is already compatible with your system (If you do not remember which platforms are compatible, you can execute command make makesys in the directory
$deMon/makefiles).
2.5 What to do when things go wrong
Hint:
13
If you would like the platform file to apply to a single computer,
the appropriate test in the checkplatform: rule is:
[ "‘hostname‘" = "whatever_the_name_is" ]
(Don’t forget to start this line with a tab - or make will refuse
to accept your configuration file).
Hint:
You can choose the location to search for the optimized libraries
by setting the LIBPATH environment variable, either in your
shell initialization file, or in the platform configuration file itself.
2.5
What to do when things go wrong
First of all, Don’t Panic. Installation of any complex software package is rarely entirely
problem-free – and deMon consists of more than 1,200 source files and almost 200,000
lines of code.
There several ways of getting assistance. However, there are a few things you can do
yourself before asking for help. First of all, please make sure that you are using the most
up-to-date version of deMon – the problem you see may have already been resolved.
Please also check that your Fortran compiler and numerical libraries are up-to-date. If
this does not resolve your problem, here are a few more things you may want to try.
deMon build scripts rely on having a large subset of tools specified by the POSIX Unix
standard (IEEE Std 1003.1-2001). If you need to build deMon on a platform lacking full
POSIX support (such as Microsoft Windows or some exotic variants of Unix), you can
do the following:
1. Find a POSIX-compliant Unix system. Linux or SGI Irix will work very well. Most
recent Unix systems should also work.
2. Unpack deMon distribution, and follow advanced installation instructions in section
2.2.4. The platform you will need to build for is called “flatsource”. It cannot be
invoked through the express installation script. The build script will populate the
$deMon/flatsource directory with symbolic links to all relevant source code files.
3. In the directory $deMon/flatsource, execute shell script $deMon/unsupported/flat-makedepend
It will create a rudimentary Makefile for building deMon.
14
2 Getting Started
4. Transfer all files collected in the flatsource dirctory to your target system. You
may find this command:
tar chf ~/deMon-source.tar *
useful for following symbolic links.
5. Build deMon on your target system, using the minimum necessary force.
If you experience errors while running deMon shell scripts, check your operating system
documentantion for instructions on enabling POSIX conformance. Some hints for the
Tru64 and Solaris operating systems are found in $deMon/doc/os/.
If you experience mysterious crashes running compiled binary, try some of the following:
• Try increasing process limits, such as the stack size limit, data segment size limit,
or memory size limit. If you use Bourne shell or derivative (sh, ksh, bash, ...), the
appropriate command is ulimit. For C-shell or its ilk (csh, tcsh, ...), the right
command is limit. On some operating systems tuning kernel parameters may be
required to adjust the limits.
• Try building deMon without using any external libraries – the libraries may be incompatible with the compilation options in the deMon makefile, or may not support
functions needed by deMon properly. Debugging and generic versions typically do
not require external libraries.
• Try decreasing optimization level – see section 2.2.4 and appendix B for the instructions on modifying platform configuration files.
If your compiled binary runs, but fails some of the tests, you can try some of the following:
• Try decreasing optimization level – see section 2.2.4 and appendix B for the in-
structions on modifying platform configuration files. The following routines are
particularly numerically sensitive, and may be affected by aggessive compiler optimizations:
2.5 What to do when things go wrong
15
tensor.f
Calculation of the molecular inertia tensor
rs.f
EISPACK diagonalization routine
jacobi.f
Jacobi diagonalization routine
pythag.f
auxiliary routine for jacobi.f
dsyev.f
LAPACK diagonalization routine
dsyevd.f
dsyevr.f
LAPACK diagonalization routine (divide and conquer)
LAPACK
diagonalization
routine
(relatively
robust
representations)
dgelss.f
LAPACK least-squares fit routine
dpptrf.f
dpptri.f
LAPACK routine for Cholesky factorization
LAPACK routine for matrix inversion
using Cholesky
factorization
dtptri.f
LAPACK routine for triangular matrix inversion
blas1.f
BLAS Level 1 routines
blas2.f
BLAS Level 2 routines
dgemm.f
BLAS Level 3 DGEMM routine
dsyrk.f
BLAS Level 3 symmetric rank k update
• Try enforcing strict conformance to the IEEE-754 floating point standard. Your
compiler documentation should provide the instructions for enforcing the conformance.
• If you are building deMon with an external BLAS or LAPACK libraries, try building using Fortran routines supplied with deMon itself. Sometimes, vendor-supplied
libraries are trading off accuracy for extra speed.
• If you are building deMon using Fortran BLAS or LAPACK, included with the
deMon distribution, try building with vendor-supplied numerical libraries. Sometimes, vendor compilers have issues with generating numerically accurate code for
these routines, but hand-optimized numerical libraries work around these issues.
If none of these suggestions help, try searching the archives of the deMon users mailing list,
and deMon developers mailing list. Both archives are found at http://www.deMon-software.com/.
If you cannot find the solution to your problem in the archives, you can try sending your
questions to one of the mailing lists. The addresses of the lists can be found on the same
web site.
16
2 Getting Started
Finally, if everything else fails, you can ask deMon developers for assistance. Important:
Please do not bother telling to the gatekeeper, or to any of the developers: “Your stuff
does not work, and it is just a piece of junk”. This kind of approach will only pi** off
people who put a lot of effort into making sure that everything works as smoothly as it
possibly could.
Instead, please provide answers to these questions:
• Which version of deMon are you using? (Check $deMon/source/deMon.f).
• Which hardware platform are you using? The processor and system model are
important.
• Which operating system are you using? (Try uname -a shell command).
• Which compiler are you using? Please give both the name and the version.
• Which numerical libraries (if any) are you using? Both name and version are important.
• Which options for the build platforms are you offered by the ./install.sh script?
• Which build platform did do you choose?
• If your build fails while running ./install.sh, what are the contents of the the
build-*.log file it produced in the makefiles subdirectory?
• If you experience failures in the test set, which test cases fail?
• If you experience failures in a specific test case, please include both the *.out and
*.err files from the corresponding subdirectory of examples directory.
While doing all this, please keep in mind that people you are asking for assistance are
just as busy as you are. Please try to help them to resolve your problem.
2.6
How to Run deMon
The use of deMon requires the preparation of the input file (see Chapter 4), job execution
and interpretation of the output (see Chapter 5). To run the program, three files are
necessary:
2.6 How to Run deMon
17
deMon.inp is the input file.
AUXIS contains the Gaussian auxiliary function.
BASIS contains the Gaussian orbital basis functions.
AUXIS and BASIS are ASCII files which are provided with the program. It is possible to
run deMon executable binaries, found in $deMon/bin/*.x, directly. In this case the above
three files have to be present in the current directory. The job can be executed by the
command $deMon/bin/*.x & or, without hangups, by nohup $deMon/bin/*.x & in the
background. The current directory is used as working directory and all necessary scratch
files are created here. A second possibility, less cumbersome, is to use the standard deMon
driver script, provided in $deMon/bin/deMon. You may wish to add the $deMon/bin
directory to your command search path, to avoid repeatedly typing the complete path to
the script. The deMon driver script is invoked as follows:
deMon [option ...] job name
The driver script will attempt to determine the location of the deMon installation directory, and will supply reasonable defaults for most options. Additionally, the following
options can be specified:
-a
auxis
Specify alternate basis set file. The default fiting basis set file
is looked up in the current directory, then in deMon installation
-b
basis
directory.
Specify alternate basis set file. The default is looked up exactly
as for the auxis.
-d
-e
debugger
ecp
Specify an alternative debugger (the default is gdb)
Specify alternate ECP file. The default is looked up as described
for auxis.
-g
Start deMon interactively, in a debugger
-p
Print the name of deMon binary to be used, and exit
-h
-r
Print summary of the available options
Use “directory” to store restart files, and other large files, pro-
directory
duced by the simulation. By default, it is the current directory
-s
directory
Use “directory” for scratch files
-v
Be verbose
-q
Be quiet
18
2 Getting Started
-x program Use an alternative deMon binary
The deMon driver expects to find the job input (using the syntax, described in Section 3.3)
in the file specified on the command line, with the extension .inp attached. Depending
on the job type, additional data have to be provided in files with the same base name,
and extensions .cub, .lat, .rst, and .qmd.
The main deMon output is stored in a file with the extension .out. Additionally, some
system information is provided in a file with the same base name, and extension .err.
Depending on the nature of the job, additional output files, with extensions .asc, .bin,
.lat, .mkl, .mol, .new, .pie, .rst, .nmr11, .nmr70, .nmr71 and .nmr72 may be generated. The driver script takes care not to delete any previously existing output files. If
any of the output files with these extensions exist at the time execution begins, the driver
will create backup copies. These copies will have an additional extension .bak, appended
to their full name (if any of the files with these extensions exist, they will be removed).
Note that most of the deMon output files are originally created in the scratch area, and
are only copied to the job starting directory after the job completes. The only exception is
the system information file (with the .err extension). If you need to examine the output
of a running deMon job, you will find the exact location of the scratch directory in the
system information file.
3 Carrying On
3
3.1
19
Carrying On
Parameters
Table 2 summarizes the limits of the shipped deMon version for memory and disk space
requirements, number of atoms as well as basis and auxiliary functions. These limits are
specified in the section user defined parameters of the parameter.h file. (This file is
found in the directory $deMon/include.) If a calculation violates these limits the program stops and prints PARAMETER EXCEEDED statement(s) for the parameters that
have to be changed. In this case, please change only the first exceeded parameter (the following ones may be corrupted by array bound violations) to the suggested (USED) value
and recompile the program. Repeat this procedure until all PARAMETER EXCEEDED
statements have disappeared. It is recommended not to increase the MAXRAM value over
the physical memory (RAM) available, because otherwise it may result in large paging
overheads during program execution.
This configuration is optimised for DFTB and MM calculations. For large-scale DFT
calculations we encourage the interested user to obtain a recent version of deMon2K at
http://www.demon-software.com. The disk requirements are dominated by the size of
the electron repulsion integral scratch file ioeri.scr, in the case of a conventional SCF
calculation, and by the files ioscf.scr, iocdf.scr and iogrd.scr in the case of direct
SCF calculations (see 4.4.1).
3.2
Files
All files in deMon are opened with explicit filenames, rather than using operating systemspecific compiler defaults. Throughout this guide, the name used in the Fortran OPEN
statement is referred to as the logical filename. These logical filenames are read by the
program or created in the working directory.
The logical files are connected to Fortran I/0 units in the file fileio.h. Table 3 contains
the logical and physical filenames used in the program.
Besides these files, special plot files may be created for the visualization of molecular
fields, like the electronic density, electrostatic potential, etc. with VU (see Chapter 8).
These files are named according to the molecular field to be visualized. For example, a
file RHO.bin is created for the visualization of the electronic density. The deMon.pie file
is the VU control file and refers in this case to the RHO.bin file. If the job is executed
20
3
Carrying On
Table 2: Parameter settings in the shipped deMon version.
Parameter
MAXDISK
MAXRAM
Setting Description
8192 Maximum disk size for scratch files in Mbytes
32768 Maximum RAM size for program kernel in Mbytes
MAXATOM
MAXAUX
MAXAUXSET
MAXAUXSHL
MAXCON
MAXCUBE
MAXGTO
MAXECPGTO
MAXECPSHL
MAXLAUX
MAXLBAS
MAXLECP
MAXSHL
MAXSTO
MAXTBSTO
MAXADJ
20000
5000
2000
2000
21
100000
5000
2500
500
6
5
5
3000
3000
3000
25
Maximum
Maximum
Maximum
Maximum
Maximum
Maximum
Maximum
Maximum
Maximum
Maximum
Maximum
Maximum
Maximum
Maximum
Maximum
Maximum
number of atoms
number of auxiliary functions
number of auxiliary function sets
number of auxiliary function shells
degree of contraction
number of cube (embedding) points
number of primitive Gaussian functions
number of ECP Gaussian functions
number of ECP shells
L quantum number for auxiliary functions
L quantum number for basis functions
L quantum number for ECP functions
number of orbital basis shells
number of contracted (STO) orbitals
number of tight-binding (STO) orbitals
number of close, bond-line contacts per atom
by a script, care has to be taken that these files are copied from the working directory
to the output path and that the filenames are consistent with the ones in the deMon.pie
file. These entries can also be modified manually in order to use the same VU control file
for different molecular fields.
3.3
Input Syntax
The input of deMon is easy and mnemonic and at the same time offers high flexibility.
The input file contains keywords, options and keyword bodies. A keyword forms together
with its options and the keyword body a keyword block. The ordering of these blocks is
free. All keywords, with exception of GEOMETRY, have associated default values, which
are used, if the keyword is not explicitly specified in the input. Table 4 summarizes the
special symbols allowed in the job input file deMon.inp.
The input lines in the job input file deMon.inp are restricted to 160 characters. The input
3.3 Input Syntax
21
Table 3: Logical and physical filenames in deMon. The I/0 units are given in parentheses.
Logical
Filename
AUXIS
BASIS
ECPS
deMon.cub
deMon.inp
deMon.out
deMon.new
deMon.mola
deMon.mkla
deMon.pie
deMon.rst
deMon.lat
deMon.qmd
deMon.nmr11
deMon.nmr70
deMon.nmr71
deMon.nmr72
deMon.mag
a
Physical
Filename
AUX
(1)
BAS
(2)
ECP
(3)
CUB
(4)
INP
(5)
OUT
(6)
NEW
(7)
MOL
(8)
MOL
(8)
PIE
(9)
RST
(10)
LAT
(11)
MDRST
(12)
NMR11
(13)
NMR70
(14)
NMR71
(15)
NMR72
(16)
MAGTAPE (17)
Description
Auxiliary function file
Basis set file
ECP file
Embedding file
Input file
Output file
New input file for restart
MOLDEN input file
MOLEKEL input file
VU control file
RESTART file
Plot lattice file
MD restart file
NMR grid file
NMR dimension file
NMR pointer file
NMR orbital file
Alternative NMR file
Type
ASCII
ASCII
ASCII
ASCII
ASCII
ASCII
ASCII
ASCII
ASCII
ASCII
BINARY
ASCII / BINARY
ASCII
BINARY
BINARY
BINARY
BINARY
BINARY
The I/O unit 8 is connected either to the MOLDEN or MOLEKEL input file depending on
the VISUALIZATION (4.11.1) keyword. For the XYZ output the MOLDEN file (deMon.mol)
is used in any case.
Table 4: Special symbols of the deMon input.
Symbol
&
#
=
Space
,
END
Description
Continuation of a keyword line
First character of a comment line
Assignment of numerical values to options
Combination of keyword options
Separation of keyword and options
Separation of different options (optional)
End of a keyword data block
is not case sensitive. A keyword line can be continued by the & symbol at the very end of
the line. At most, five continuation lines are permitted. The TITLE line cannot be
22
3
Carrying On
continued and the length of the title string is restricted to 60 characters.
The symbol # indicates a comment line. Comment lines can only occur between complete
keyword data blocks including the keyword body, if existing. Comments inside a keyword
line or body result in unpredictable read operations. The end of a keyword block can be
indicated by an END statement or comment symbol # in a new input line or simply by a
new keyword line.
The keyword options have to be given directly after the keyword, separated by space(s).
If they are not present the default option setting for the keyword is used. Numerical
values are assigned by the = sign to keyword options (e.g. SCFTYPE Tol = <Real>, where
<Real> indicates a real number). For the combination of keyword options the - sign is
reserved (e.g. VXCTYPE B88-LYP). The keyword body starts in a new input line directly
after the keyword line(s), and must have the appropriate structure given in Chapter 4,
”Keywords”.
4 Keywords
4
23
Keywords
The description of the Keywords is grouped according to their functionality into geometry input, basis set input, electronic state control, self-consistent field (SCF) control,
optimization control, frequency control, property control, environmental control and interface control keywords. Keywords which don’t fit into these groups are listed at the
end of this chapter under miscellaneous keywords. Keyword options which exclude each
other are listed in one line separated by / signs. If these options are not specified in the
input the underlined one is used by default. If more than one of these options are given
the last one will override the previous ones. The boldface printed part (first five letters)
of the keywords and options are mandatory for the input. Therefore, OPTIMIZATION,
OPTIMIZE and OPTIM are all allowed input forms for the keyword OPTIMIZATION.
4.1
Geometry Input
4.1.1
Keyword GEOMETRY
This keyword is mandatory! It specifies the molecular geometry.
Options:
CARTESIAN / ZMATRIX / MIXED
CARTESIAN
The molecular structure is given in Cartesian coordinates.
ZMATRIX
The molecular structure is defined by a Z-Matrix.
MIXED
The first atoms of the molecular structure are defined by
Cartesian coordinates and the following ones are defined
by a Z-Matrix.
ANGSTROM / BOHR
ANGSTROM
Coordinates or bond distances are given in ˚
Angstr¨om.
BOHR
Coordinates or bond distances are given in atomic units.
Description:
The geometry is read in the keyword body of GEOMETRY in free format, one line for
each atom. In the case of a CARTESIAN input the atomic symbol (e.g. H), which may
carry an identification number (e.g. H1), and the x, y and z coordinates of each atom of
the system have to be specified. As an example the geometry of H2 O may be specified
as:
24
GEOMETRY
O 0.00
H 0.76
H -0.76
4 Keywords
CARTESIAN ANGSTROM
0.00 0.00
0.00 0.52
0.00 0.52
The Cartesian coordinates of an individual atom, defined by the atomic symbol (e.g.
H2), or an atom group, defined by the element symbol (e.g. H), can be frozen during the
geometry optimization. In order to keep all three degrees of freedom of an atom fixed the
corresponding atomic symbol or element symbol has to be specified with the string XYZ
in the keyword body of CONSTANTS. The string X will only freeze the x coordinate of
the atom, the string XY the x and y coordinates, and so on. In the following example the
oxygen atom is kept frozen and the movements of all hydrogen atoms are restricted to
the xz-plane (y coordinate frozen) during the optimization.
GEOMETRY CARTESIAN ANGSTROM
O 0.00 0.00 0.00
H 0.76 0.00 0.52
H -0.76 0.00 0.52
#
CONSTANTS
O XYZ
H Y
As usual, the comment line (#) is only given for clarity. The effective nuclear charge (in
atomic units), which is identical to the number of electrons counted for the atom, and the
nuclear mass (in atomic mass units; amu) can be specified by an integer and real number
after the coordinates, respectively. Therefore, the following input for water consists of a
frozen oxygen atom, a hydrogen atom, H1, fixed in the xz-plane with the effective nuclear
charge 0 (no electrons are counted for this hydrogen) and another hydrogen atom, H2,
with the effective nuclear charge 2 and the nuclear mass 2.014 (Deuterium).
GEOMETRY CARTESIAN ANGSTROM
O
0.00 0.00 0.00
H1 0.76 0.00 0.52 0
H2 -0.76 0.00 0.52 2 2.014
#
CONSTANTS
O XYZ
H1 Y
If quantities are not specified the default values are used (e.g. the O atom in the above
example has a nuclear charge from 8 and a mass of 15.999 amu).
If the geometry of the system is given by a Z-Matrix, using the option ZMATRIX, each
line in the keyword body of GEOMETRY describes the connectivity of the atom, which
4.1 Geometry Input
25
again is defined over its atomic symbol. The first atom (C) of the Z-Matrix defines the
origin of the input (see Figure 1). No connectivity information should be given for this
atom.
Figure 1: Orientation of the first three atoms defined in the Z-Matrix.
z
B
RAB
TABC
RBC
A
C
y
x
The second atom (B) lies on the z-axis of the input coordinate system and is found at
the distance RBC from atom C. The third atom (A) lies in the xz-plane, at a distance RAB
from atom B. The two atoms A and B form angle TABC with the first atom (C).
All the remaining atoms are defined by input line(s) in the free format, containing the
information:
LABEL
NB
RAB
NC
TABC
ND
PABCD
Here LABEL is the atomic label of atom A. NB, NC and ND are the Z-Matrix line
numbers or the atomic labels of the atoms B, C and D over which atom A is defined
by a length (RAB), angle (TABC) and dihedral angle (PABCD). These are the so-called
˚ngstr¨om or Bohr),
internal coordinates of the system. There values are given by RAB (in A
TABC and PABCD (in degrees), respectively. The dihedral angle is defined by the angle
between the planes spanned by the atoms A, B, C and B, C, D, respectively. The sign of
26
4 Keywords
the dihedral angle is defined according to the Newman projection shown in Figure 2. If
the projection angle is oriented clockwise, the dihedral angle is taken to be positive.
Figure 2: Definition of positive (top) and negative (bottom) dihedral angles in deMon.
Structure
ϕ
A
C
D
Newman Projection
A
D
B
D
C
B
D
ϕ
A
A
Instead of giving the values for RAB, TABC and PABCD directly in the Z-Matrix they
may also be represented by symbolic strings, up to eight characters long. These labels
have to be defined in the CONSTANT and VARIABLE input sections (see 4.1.2 and
4.1.3). As an example, ethene C2 H4 can be defined as:
GEOMETRY ZMATRIX
C1
C2 C1 rCC
H1 C1 rCH C2
H2 C1 rCH C2
H3 C2 rCH C1
H4 C2 rCH C1
#
VARIABLES
rCC
1.3390
ANGSTROM
aCCH
aCCH
aCCH
aCCH
H1
H2
H1
dHCCH
dHCCH
dHCCH
4.1 Geometry Input
27
rCH
1.0850
aCCH 121.09
#
CONSTANT
dHCCH 180.0
The internal coordinates listed under VARIABLES may change during geometry optimization, whereas the coordinates given under CONSTANTS are kept frozen.
If one uses internal coordinates, dummy atoms are often useful. The atomic symbol X
has been reserved for them. The following example shows the use of a dummy atom in
the definition of HCN.
GEOMETRY ZMATRIX ANGSTROM
H
C 1 R1
X 2 1.0 1 90.0
N 2 R2
3 A1
1 180.0
#
VARIABLES
R1 1.089
R3 1.166
A1 90.00
Here the dummy atom is used to avoid a 180◦ angle which may cause problems in a
geometry optimization. Example 5.1 shows the use of a dummy atom for the definition of
a ring system. In order to facilitate the input of the internal coordinates for some special
cases, alternative geometrical parameters can be given instead of angles and dihedral
angles in the Z-Matrix input. The substitutions are indicated by string codes, which may
appear after the connectivity information in each Z-Matrix line. The possible substitution
codes are listed in Table 5.
Table 5: Substitution codes for the internal coordinate substitution in the Z-Matrix.
Subst. Code
RAD
RRD
RAA
RRA
RRR
Internal Coordinates
RAB TABC PABCD
RAB RAC PABCD
RAB TABC TABD
RAB RAC
TABD
RAB RAC
RAD
Description
Default: distance-angle-dihedral
Angle ABC is substituted by length AC
Torsion ABCD is substituted by ABD angle
Combination of RRD and RAA
Atom specified by three distances
Similar to the CARTESIAN input, the atomic mass and effective nuclear charge of an
atom can be optionally specified at the end of each Z-Matrix line. The examples 5.2 and
28
4 Keywords
5.3 show some common applications of Z-Matrix inputs.
If the geometry of the system is given in the MIXED format, the first atoms are defined
using their Cartesian coordinates. At least the first three atoms (some which may be
dummy atoms), have to be defined in this way. The following atoms can then be specified
in the Z-Matrix format, as described above. The reference atoms can be defined using
either Cartesian or internal coordinates. The coordinate system is defined by the first
three atoms. The following example (see Figure 3), shows a MIXED input of a CO
molecule adsorbed on a MgO cluster:
GEOMETRY MIXED ANGSTROM
Mg1
0.000
0.000
0.000
O2
0.000
2.105
0.000
O3
2.105
0.000
0.000
O4
0.000 -2.105
0.000
O5
-2.105
0.000
0.000
Mg6
2.105
2.105
0.000
Mg7
2.105 -2.105
0.000
Mg8
-2.105 -2.105
0.000
Mg9
-2.105
2.105
0.000
O10
0.000
0.000 -2.105
Mg11
0.000
2.105 -2.105
Mg12
2.105
0.000 -2.105
Mg13
0.000 -2.105 -2.105
Mg14 -2.105
0.000 -2.105
O15
2.105
2.105 -2.105
O16
2.105 -2.105 -2.105
O17
-2.105 -2.105 -2.105
O18
-2.105
2.105 -2.105
C19
Mg1
RAD
O2
X20
C19
1.0
Mg1
O21
C19
RCO
X20
#
CONSTANTS
Mg
XYZ
O2
XYZ
O3
XYZ
O4
XYZ
O5
XYZ
O10 XYZ
O15 XYZ
O16 XYZ
O17 XYZ
O18 XYZ
#
VARIABLES
AAD
90.0
90.0
O3
O2
Mg1
DAD
180.0
180.0
4.1 Geometry Input
RAD
AAD
DAD
RCO
29
2.0
90.0
90.0
1.4
In the example above, geometry of the MgO cluster is fixed by freezing its Cartesian
coordinates (see 4.1.2). The carbon atom of the CO molecule is defined with respect to
the Mg1, O2 and O3 plane of the Cluster (see Figure 3) by the internal coordinates RAD,
AAD and DAD. These coordinates optimized because they are declared as variables (see
4.1.3). The same holds for the CO bond length that is defined by the internal coordinate
RCO. The dummy atom X20 has to be used in order to avoid an ill-defined dihedral angle
if the O21, C, Mg1 arrangement becomes linear. Therefore, the above example describes
an input for the free optimization of an CO molecule, including internal relaxation, on a
frozen MgO cluster surface.
Figure 3: Coordinate definition of the above MIXED input.
O21
C
X
RAD
Mg
DAD
O
AAD
Mg
O2
Mg1
O
Mg
O3
Mg
Mg
Mg
O
O
Mg
O
Mg
O
30
4 Keywords
For a hybrid QM/MM calculation, additional parameters can be added at the end of
each geometry specification line. These parameters are in the keyword form (parameter=value), and do not need to appear in any particular order. QM/MM keywords are
recognized, and behave identically in all three input formats. The QM/MM-related geometry keywords are:
Keyword
Value
Description
Specifies handling of this atom when QM/MM is activated:
QMMM=
MM
This atom is a part of the molecular-mechanical subsystem
QM
This atom is a part of the quantum-mechanical subsystem
LINK
This a bond-terminating atom, at the QM/MM boundary.
The default is ’MM’ if QM/MM is activated, and ’QM’ if it
is not.
Specifies choice of the capping potential for a link atom:
CAP=
MMTYPE=
ZLY1
“3-21G” pseudohalogene capping potential of Ref. [10]
ZLY2
“6-31G*” pseudohalogene capping potential of Ref. [10]
DHC1
“QCP1” pseudohydrogen capping potential of Ref. [11]
DHC2
“QCP2” pseudohydrogen capping potential of Ref. [11]
DHC3
“QCP3” pseudohydrogen capping potential of Ref. [11]
DHC4
“QCP4” pseudohydrogen capping potential of Ref. [11]
(ecp)
Use specified potential from the ECP library
type
Specifies forcefield atom type. Atom types are forcefieldspecific. The list of the known atom types for the built-in
UFF force field is given in section 6.
MMTYPE is relevant for all types of atoms, including QM
GUESS
and LINK atoms.
A special atom type GUESS will instruct deMon to try to
guess the appropriate force field type (see below). This is
the default.
4.1 Geometry Input
Q=
charge
31
This option can be used to specify partical charge on the
MM atom, in proton charge units. It is up to the forcefield
GUESS
to use, or not to use this charge.
A special charge GUESS (which is the defalt) will instruct
deMon to guess the appropriate value.
BOND=
list
Specifies the list of explicit bonds for this atom. Bond specification contains a list of colon-separated atom ranges, each
followed by optional slash sign and the bond order. Both
atom numbers (in the input order) and atom labels are recognized; atom labels must refer to a previously defined atom.
There are no restrictions on atom numbers.
A few examples:
3
Bond to atom 3; the bond order will be guessed
automatically.
C PHEN/1.5
Bond of the order 1.5 to a previously defined atom with the
label C PHEN
15-17
Bonds to atoms 15 through 17, with automatically defined
orders
1/2:2-3/1.0
Double bond to atom 1; single bonds to atoms 2 and 3
It is OK to specify a bond for only one of the connected
atoms. If FORCEFIELD BONDS=AUTO is in effect, explicit bonds
take precedence over the guessed bonds.
NOBOND=
list
Specifies that no bond should be formed to these atoms. The
syntax of the atom ranges is identical to the BOND= option.
NOBOND= overrides automatically guessed bonds.
32
4 Keywords
Keywords QMMM=, CAP=, MMTYPE=, and Q= can only appear once for each atom. Multiple
instances of the BOND= and NOBOND= keywords are allowed. Their effect is cumulative.
QM/MM keywords are recognized, but ignored by deMon in a QM-only calculation (QMMM
QM, see section 4.5.1).
An example of a QM/MM geometry input is:
GEOMETRY Z-MATRIX
C1
O1
1 RCO
H1
1 RH1
2 AH1
H2
1 RH2
2 AH2
O_W1 1 RCW1
2 ACW1
H3
1 RH3
2 AH3
H_W11 5 RH1W1 1 AH1W1
H_W12 5 RH2W1 7 AH2W1
O_W2 1 RCW2
2 ACW2
H_W21 9 RH1W2 1 AH1W2
3
3
3
2
1
3
2
DH2
DCW1
DH3
DH1W1
DH2W1
DCW2
DH1W2
H_W22 9 RH2W2 10 AH2W2 1 DH2W2
#
VARIABLES
RCO
1.37257
RH1
1.11599
...
AH2W2
104.0
DH2W2
0.0
QMMM=QM
QMMM=QM Q=-1.0 BOND=C1/1.0
QMMM=QM
QMMM=QM
Q=-0.834
QMMM=QM
Q=+0.417 BOND=O1/0.03:O_W1/0.90
Q=+0.417
QMMM=QM Q=-0.834
QMMM=QM Q=+0.417 &
BOND=O1/0.03:O_W2/0.90
QMMM=QM Q=+0.417
The full input for this caluclation is discussed in section 5.4.
4.1.2
Keyword CONSTANTS
This keyword assigns constant Cartesian and internal variables.
Options: None
Description:
In the body of CONSTANTS all atoms and internal coordinates which are kept frozen
during the geometry optimization are listed. The atomic or element symbol with the string
XYZ is used to freeze the atom or group. The string X will only freeze the x coordinate of
the atom, the string XY the x and y coordinates, and so on.
For internal coordinate inputs the symbolic coordinate string of the Z-Matrix and its
value (real number) is given in order to freeze this degree of freedom. In the case of a
4.1 Geometry Input
33
MIXED input this two syntax forms can be mixed (see Figure 3 and the corresponding
input example).
Constants
H
XYZ
A 120.0
Here the first line freezes all hydrogen atoms to their positions. In the second line the
internal coordinate A is set to 120◦ and kept frozen. Note that all parameters listed in
CONSTANTS section are treated as VARIABLEs in a molecular dynamics simulation.
4.1.3
Keyword VARIABLES
This keyword assigns variable internal coordinates.
Options: None
Description:
The body of VARIABLES collects all symbolic coordinate strings of the Z-Matrix with
their values (real number) which should be optimized (see 4.1.1 for examples).
4.1.4
Keyword SYMMETRY
This keyword controls the symmetry analysis of the molecular structure.
Options:
OFF / ON
OFF
No symmetry analysis is performed.
ON
A symmetry analysis is performed.
DETECT=ON
Symmetry detection is performed. This is the default
DETECT=OFF
No symmetry detection is performed.
Description:
If a symmetry analysis is performed the point group of the molecule is automatically
detected and the molecular INPUT ORIENTATION is transformed into the STANDARD
ORIENTATION defined by the center of mass and the main axes of the tensor of inertia.
All further calculations are done in this orientation. If no symmetry analysis is performed
the INPUT ORIENTATION is used for the SCF calculation. However, if a geometry
optimization or a frequency analysis is requested the STANDARD ORIENTATION is
34
4 Keywords
used in any case (see 5.1). For atoms, the use of the SYMMETRY keyword enforces
the zeroing of the off-diagonal elements of the Kohn-Sham matrix for different angular
momenta.
Currently, deMon symmetry analyser supports Cn , Cnv , Cnh , Dn , Dnd , and Dnh point
groups with n ≤ 6. The rotation-inversion groups Sn are supported for n ≤ 8. If a higher
group is present in a molecule, deMon will abort the calculation, even when SYMMETRY OFF
is specified. Symmetry analysis can be suppressed altogether by including DETECT=OFF
on the symmetry line (See example 5.5.
If you are uncertain about a symmetry assignment, made by deMon, or are wondering
whether your system is approximately symmetric, you may find an unsupported program
$deMon/unsupported/symmetry.c useful.
4.2 Basis Set Input
35
4.2
Basis Set Input
4.2.1
Keyword BASIS
This keyword specifies the basis set.
Options:
<Basis>
The basis set string <Basis> defines the global basis set.
If absent, the DZVP basis set is used by default.
Description:
In the BASIS keyword body basis sets can be assigned to individual atoms according to
the atomic symbol (e.g. H4) or to atom groups over based on the element symbol (e.g.
H). The global basis set is used for all atoms which are not defined by either of these
mechanisms. Specifying QM/MM capping potential will override basis set selection, made
with the BASIS keyword (see section 6). More specific definition of the basis set overrides
a less specific one, so that:
QM/MM CAP=
overrides:
<atomic symbol> overrides:
<element symbol> overrides:
<global basis set>
Thus any basis set definition for an atom can be overriden by the explicit assignment of
the basis set using the atomic symbol. E.g. for C2 H4 (see input in 4.1.1) the following
basis set definition:
Basis (DZVP)
C
(TZVP)
C1
(STO-3G)
assigns a STO-3G basis to atom C1, a TZVP to C2 and the DZVP basis to all other
atoms. Instead of using basis set strings, the basis set of an atom may also be specified
by the Huzinaga notation given in the BASIS file. Using this notation the above basis set
definition would read as:
Basis (DZVP)
C
(7111/411/1*)
C1
(33/3)
The basis set file BASIS contains the basis sets listed in Table 6. Other basis sets can be
obtained from the Extensible Computational Chemistry Environment Basis Set Database
[12] at http://www.emsl.pnl.gov/forms/basisform.html by choosing the deMon basis set
36
4 Keywords
format. Instead of reading the basis set from the BASIS file the user can also define it
directly in the input file following the format:
Basis
SYMBOL Read
N L K
EXPONENT
:
EXPONENT
COEFFICIENT
:
COEFFICIENT
where SYMBOL can be an element or atomic symbol, N and L are the main and angular
momentum quantum number of this shell and K is the contraction degree. Here, a shell
collects all contracted orbitals of the same angular momentum quantum number like
px , py , pz or dxx , dxy , dxz , dyy , dyz and dzz . The contracted orbitals, µ(r), are linear
combination of (atom centered, normalized) Gaussian type orbitals (LCGTO–Ansatz),
which are called the primitive orbitals, g(r):
µ(r) =
K
X
dµk gk (r)
(4.1)
k=1
2
gk (r) = (x − Ax )ax (y − Ay )ay (z − Az )az e−ζk (r−A)
(4.2)
The exponents ζk and contraction coefficients dµk are listed in free format under the shell
definition line, one line for each primitive orbital (EXPONENT and COEFFICIENT). A
user defined basis set input is shown in Example 5.6.
4.2.2
Keyword AUXIS
This keyword specifies the auxiliary function set.
Options:
<Auxis>
The auxiliary function set string <Auxis> defines the global
auxiliary function set. If absent, the A2 auxiliary function
set is used by default.
Description:
Similar to the basis sets the auxiliary function sets can also be assigned to individual
atoms according to the the atomic symbol or to atom groups based on the element
symbol in the keyword block of AUXIS. The syntax and hierarchy of these assignments
are similar to the basis set assignments (see 4.2.1). For C2 H4 (see input 4.1.1) the following
auxiliary function definition:
4.2 Basis Set Input
37
Table 6: Basis sets available in the deMon BASIS file.
Basis Set
Elements
Description
DZV
H, C
Double ζ basis set for saturation only
DZVP
H–Xe
LDA double ζ polarizationa basis set [13]
DZVP2
Be–F, Al–Ar, Sc–Zn
Modified DZVP basis set
TZVP
H, Li, C–F, Si–Cl
LDA optimized triple ζ polarization basis set
TZVP-FIP1
H, C–F, S, Cl
TZVP with field-induced polarization [15,16]
TZVP-FIP2
H, C–F, S, Cl
for α, β (FIP1) and γ (FIP2) calculations
SADLEJ-FIP H, C, N, O, F
EPR-III
H–F
EPR basis set [17]
IGLO-II
H–F, Si, Cr, Fe
NMR basis set [18]
IGLO-III
H–F, Si, Cr, Fe
NMR basis set [18]
STO-3G
H–I
STO–3G basis set for testing only [19]
SAD
H, C–F
Sadlej FIP basis set [20]
LIC
H–Ne
Lie-Clementi basis set [21]
WACHTERS
Sc–Cu
Wachters basis set without f functions [22]
DZ-ANO
Ahlrichs
H–Zn
H–Kr
Double ζ ANO basis set from Roos [23]
Ahlrichs basis sets: A-VDZ, A-PVDZ, A-VTZ, APVTZ
ECP-SD
EECP-SD
QECP-SD
a
Modified for Li and Na [14].
AUXIS (A2)
C
(GEN-A2*)
C1
(GEN-A2)
assigns a GEN-A2 auxiliary function set to atom C1, a GEN-A2* to atom C2 and the
A2 auxiliary function set to all other atoms. The AUXIS file of deMon contains only the
A2 auxiliary functions set. The other auxiliary function sets used above (GEN-A2 and
GEN-A2*) are automatically generated according to the procedure described in Appendix
A. Additionally, the auxiliary function sets can be directly specified in the input file using
38
4 Keywords
the format:
Auxis
SYMBOL Read
LMAX EXPONENT
:
:
LMAX EXPONENT
where LMAX denotes the maximum angular momentum quantum number of the auxiliary
function set and EXPONENT the exponent which is shared [24,25] by all functions in the
set. Thus an auxiliary function set with LMAX = 2 contains ten functions, namely one s,
three p and six (Cartesian) d functions. In deMon these functions are primitive Hermite
Gaussians of the form (without normalization):
a
¯(r) =
∂
∂Ax
!a¯x
∂
∂Ay
!a¯y
∂
∂Az
!a¯z
2
e−ζ(r−A)
(4.3)
The input of an user defined auxiliary function set is described in Example 5.6.
4.2.3
Keyword ECPS
This keyword specifies the effective core potentials (ECPS).
Options:
<ECP>
The ECP string <ECP> defines the global effective core
potentials. If absent, an all-electron calculation is assumed
by default.
The keyword ECPS is very similar to the keyword BASIS (Section 4.2.1). Different ECPs
can be assigned to individual atoms according to the atomic symbol (e.g. Au1) or to
atom groups based on the element symbol (e.g. Au). The global ECP is used for all
atoms which are not explicitly defined. Requesting a QM/MM capping potential (see
section 6 overrides ECP section in the ECPS keyword. If ECPS is not specified for an
atom but the substring ”ECP” is present in the basis set definition, then an ECP with
the same name will be automatically assigned to the atom. The assignment of the ECP
based on atomic symbols, element symbols and global ECPS definition possesses the
hierarchy:
QM/MM CAP=
<atomic symbol>
<element symbol>
<global ECPS>
<basis set name>
overrides:
overrides:
overrides:
overrides:
4.2 Basis Set Input
39
Thus any ECPS definition for an atom can be overwritten by the explicit assignment of
the ECP using the atomic symbol. E.g. for Au2 O the following ECPS definition:
ECPS (ECP|SD)
Au
(ECP19|SD)
Au1 (ECP1|SD)
assigns the one-electron ECP denoted by (ECP1|SD) to the atom Au1, the 19 electron
ECP denoted by (ECP19|SD) to the gold atom and the global defined (ECP|SD) ECP
to the oxygen atom. The file ECPS contains the ECPs from Stuttgart [26] in the deMon
format. Also the effective core potentials can be directly specified in the input file using
the format:
ECPS
SYMBOL Read ELECTRONS
N L K
EXPONENT COEFFICIENT
:
:
EXPONENT COEFFICIENT
where SYMBOL can be an element or atomic symbol, ELECTRONS is an integer number
specifying the number of valence electrons, N denotes the radial power of the operator,
L the angular momentum of the effective potential and K the contraction degree. The
exponents and contraction coefficients are listed in free format under the ECP block
definition line, one line for each Gaussian (EXPONENT and COEFFICIENT). A user
defined ECP input is shown in Example 5.7.
Effective core potentials in deMon are defined according to the functional form:
U
ECP
= UL +
L−1
X
l
X
l=0 m=−l
Ul (r) =
X
Cli r n e−ζli r
[Ul (r) − UL (r)] |l, m >< l, m|
2
i=1
The radial power n must satisfy n ≥ −2. Positive coefficients Cli correspond to a repulsive
contribution to the potential.
40
4.3
4.3.1
4 Keywords
Electronic State Control
Keyword MULTIPLICITY
This keyword specifies the multiplicity of the system.
Options:
<Integer>
Multiplicity of the system.
Description:
The default multiplicities are 1 for closed and 2 for open shell systems. The program will
check automatically if the defined multiplicity is allowed for a given molecular system
and its charge.
4.3.2
Keyword CHARGE
This keyword specifies the charge of the system.
Options:
<Integer>
Charge of the system.
Description:
The default charge is 0.
4.3.3
Keyword MOEXCHANGE
This keyword alters the molecular orbital ordering in the start or restart density.
Options:
<Integer1>
Number of spin-α molecular orbital exchanges.
<Integer2>
Number of spin-β molecular orbital exchanges.
Description:
The numbers of the molecular orbitals to be exchanged are read in pairs of integers in
the keyword body of MOEXCHANGE, one line for each pair. First all spin-α exchanges
are read, and then, in the case of an unrestricted calculation (see UKS in 4.4.1), spin-β
exchanges are read. The application of MOEXCHANGE is shown in Example 5.8.
4.3 Electronic State Control
4.3.4
41
Keyword FIXMOS
This keyword fixes a molecular orbital occupations during the SCF by projection.
Options:
FIXED / ITERATIVE
FIXED
The molecular orbital configuration at the beginning of the
projection is used as reference configuration.
ITERATIVE
The molecular orbital configuration of the previous SCF
cycle is used as reference configuration.
SCF=<Integer>
SCF cycle at which the projection procedure is turned on.
By default the projection starts with the first SCF cycle.
TOL=<Real>
SCF convergence tolerance after which the projection
procedure is turned on.
Description:
The projection of the molecular orbital configuration is based on the maximum overlap
criterion Ωaj in the form:
Ωaj =
X
old
cnew
µa Sµν cνj
(4.4)
µ,ν
old
where {cnew
µa } are the new molecular orbital coefficients and {cνj } are MO coefficients
of the reference configuration. Occupations of a new MOs a is choosen to match the
occupation of the most similar MO j of the reference cycle, which occurs when |Ωaj | is at
maximum. Note that FIXMOS and SMEAR keywords are mutually exclusive. An application
of FIXMOS is shown in 5.9.
4.3.5
Keyword SMEAR
This keyword allows fractional orbital occupations close to the Fermi level.
Options:
<Real>
Energy range ∆E [in a.u.] around the Fermi
level in which orbitals are fractionally occupied.
Description:
The SMEAR keyword effects only the molecular orbitals within the specified energy interval [EHOMO − ∆E/2 , EHOMO + ∆E/2]. Therefore, the ∆E value should be selected based
42
4 Keywords
on the orbital energy spectrum (see MOS option of PRINT in 4.12.3). The converged
fractional orbital occupation is used in any further step of the calculation (optimization,
frequencies, properties etc.). SMEAR will assign higher occupations to lower-energy orbitals, and will give identical occupations to degenerate MOs. An application of SMEAR
is shown in Example 5.10.
4.3.6
Keyword CONFIGURE
This keyword controls the atomic configuration and is therefore only applicable for atoms.
Options:
MAX=<Integer>
Maximum number of SCF cycles for which the configuration is
used. By default it is used for all SCF cycles.
OCCUPY
The (fractional) occupation of the specified atomic configuration
will be given explicitly.
Description:
The orbital configuration is defined in the first line of the keyword body of CONFIGURE.
In the case of OKS and ROKS calculations (see 4.4.1) two lines, the first for the α-orbital
configuration and the second for the β-orbital configuration, are requested. If spherical
orbitals (see 4.4.2) are used one integer number for s, three for p, five for d, etc. have to be
specified. Empty shells (see 4.2.1 for the definition of a shell) can be omitted. In deMon
spherical orbitals are defined over real spherical harmonic Gaussians [27] as (omitting
normalization):
2
φlm (r) = r l e−ζr Slm
(4.5)
Here Slm are real spherical harmonics which have been constructed from complex spherical
harmonics by [28]:
1 m
√
Yl + Yl−m
2
i m
Yl − Yl−m
= −√
2
Slm =
Sl−m
(4.6)
(4.7)
It should be noted that the Slm are not eigenfunction of the ˆlm operator and, therefore,
m is no longer a good quantum number for these orbitals.
The ordering of the integer numbers in the configuration line follows the l (shell) and m
index of the real spherical harmonics:
4.3 Electronic State Control
Shell
l
m
Orbital
s
0
0
s
-1
py
43
p
1
0
pz
1
px
-2
dxy
-1
dyz
d
2
0
dz 2
1
2
dxz
dx2 −y2
...
...
...
...
As an example the open-shell (OKS or ROKS) triplet ground state configuration of the
carbon atom, 1s2 2s2 2p2 , can be defined in the keyword block of CONFIGURE as:
2 1 1 0
2 0 0 0
The first line defines the α-orbital configuration with 2 electrons in the most stable sorbital (1s and 2s) and 1 electron each in the most stable py (m = −1) and pz (m = 0)
α-orbital. The second line assigns 2 electrons to the two most stable β s-orbitals. Not
specified shells (here d and higher) are not considered. If Cartesian orbitals are used only
the number of electrons in the s, p, d, etc. shells have to be specified. Thus, the above
configuration changes for Cartesian orbitals to:
2 2
2 0
Because of the p-orbital degeneracy the chosen Cartesian configuration has to be stabilized during the SCF procedure (see 4.3.4 or 4.4.6). The calculation of the triplet carbon
ground state with spherical and Cartesian orbitals using CONFIGURE is described in
the examples 5.11 and 5.12, respectively.
In order to access excited atomic states or to use fractional occupation numbers the
option OCCUPY has to be selected. In this case an explicit definition of the orbital
occupation is expected after the orbital configuration line(s). In the case of spherical
atomic orbitals an input line describing the occupation of each real spherical harmonic
orbital type is requested. The occupation numbers can be given as real or integer values.
To access the excited triplet state of carbon, 1s2 2s1 2p2 3s1 , the following keyword body
of CONFIGURATION, using the option OCCUPY, is needed:
3
2
1
1
1
1
1 1 0
0 0 0
0 1
1
Here, the 3 in the first configuration line indicates that the three lowest s type α-orbitals
have to be occupied according to the occupation scheme given in the third line ( 1 0 1 ).
44
4 Keywords
Based on this scheme the lowest α s-orbital is occupied with 1 electron, the next is empty
and the following one is again occupied with 1 electron. Therefore, a hole in the α s-orbital
occupation is produced. The next two occupation lines assign 1 electron to the py and
to the pz α-orbital. The two lowest β s-orbitals are occupied with 1 electron each by the
last occupation line. Please also note that zero (0) entries in the configuration line(s) do
not have corresponding occupation lines. Example 5.13 describes the calculation of the
excited 1s2 2s1 2p2 3s1 carbon triplet state.
As already mentioned the OCCUPY option may also be used to generate fractional
occupation, e.g. for the calculation of spherical atoms. In the case of the triplet carbon
ground state the following CONFIGURATION keyword body produces a spherical atom:
2 1 1 1
2 0 0 0
1 1
0.6666
0.6666
0.6666
1 1
The first configuration line describes the occupation of 2 α s-orbitals and of all three porbitals. According to the defined occupation (third line) the two s-orbitals are occupied
each by 1 electron. The three α p-orbitals are homogeneously occupied by 2/3 of an
electron (0.6666 in line 4, 5 and 6). Finally, the β s-orbital occupation is given by the
last line. (see Example 5.14).
Because the Kohn-Sham method is a one-determinant approach atomic states have to be
approximated by one configuration (see however [29] for a multi-determinant approach).
This can be done as in spatially unrestricted Hartree-Fock calculations [30]. For s and p
occupations all possible configurations yield the correct spatial symmetry. However, for
d occupation this is not the case and care has to be taken to the occupation scheme. The
following d configuration have to be used:
d1 (2 D) : (dz 2 )1
d2 (3 F ) : (dz 2 )1 (dx2 −y2 )1
d3 (4 F ) : (dxy )1 (dxz )1 (dyz )1
d4 (5 D) : (dx2 −y2 )1 (dxy )1 (dxz )1 (dyz )1
d5 (6 S) : (dz 2 )1 (dx2 −y2 )1 (dxy )1 (dxz )1 (dyz )1
d6 (5 D) : (dz 2 )2 (dx2 −y2 )1 (dxy )1 (dxz )1 (dyz )1
d7 (4 F ) : (dz 2 )2 (dx2 −y2 )2 (dxy )1 (dxz )1 (dyz )1
4.3 Electronic State Control
45
d8 (3 F ) : (dz 2 )1 (dx2 −y2 )1 (dxy )2 (dxz )2 (dyz )2
d9 (2 D) : (dz 2 )1 (dx2 −y2 )2 (dxy )2 (dxz )2 (dyz )2
For the d1 , d4 , d6 and d9 configurations the orbital occupancies are purely arbitrary, but for
the d2 , d3 , d7 and d8 configurations the orbital occupancies must be as given above to yield
the correct spatial symmetry. Note, however, that for truly multiconfigurational cases,
either the ∆SCF method, or time-dependent density functional theory are preferable,
and usually lead to more accurate results.
46
4.4
4 Keywords
SCF Control
4.4.1
Keyword SCFTYPE
This keyword selects SCF method and convergence thresholds.
Options:
RKS / UKS / ROKS
RKS
The restricted Kohn-Sham method will be used. This is the
default for closed-shell systems.
UKS
The unrestricted Kohn-Sham method will be used. This is the
default for open-shell systems.
ROKS
The spin-restricted open-shell Kohn-Sham method will be used.
MAX=<Integer>
Maximum number of SCF iterations. Default is 100.
TOL=<Real>
Minmax SCF convergence criterion. Default is 10−5.
CDF=<Real>
Fit convergence criterion. Default is max(10−4 , 102 ∗ TOL).
Description:
The Kohn-Sham SCF methods are in their implementation similar to the corresponding
Hartree-Fock methods [31–35]. Therefore, they have similar advantages and disadvantages. Importantly, in the case of UKS calculations spin-contamination may appear. In
deMon the approximated spin contamination [36,37] is calculated and printed before of
the converged UKS SCF energy. This value has proven to be a good guide to judge the
spin-contamination of UKS calculations. Please keep in mind, however, that this value is
calculated for the Kohn-Sham reference wavefunction, and not for the true wavefunction
of the system. The physical significance of this spin-contamination parameter is therefore
not entirely clear.
Due to the variational fitting of the Coulomb potential [38,39] a Minmax SCF procedure
is implemented in deMon [40]. In the variational Minmax procedure strict convergence
from above is not guaranteed. Therefore, it is possible to obtain energies below the converged energy during the SCF iterations in deMon. Fitting can also lead to small-scale
noise (on the order of the SCF convergence criterion) in the energy during the geometry
optimization.
The maximum number of SCF iterations is specified with the MAX option. With MAX=0
an ”energy only” calculation with the molecular orbital coefficients from the restart file is
4.4 SCF Control
47
performed. No SCF iteration is done! The MAX=0 option automatically triggers GUESS
RESTART (see 4.4.4) and, therefore, fails if no adequate restart file deMon.rst exists.
The ordering and, thus, the occupation of the molecular orbitals in the restart file can
be altered with the MOEXCHANGE keyword (see Section 4.3.3).
The SCF convergence criterion can be modified using the TOL option. During geometry
optimization, user-defined SCF convergence criterion is applied to the first single point
SCF calculation. In the course of optimization, the convergence criterion is automatically tightened according to the residual forces. However, if the user defined convergence
criterion is smaller than the actual requested value it is used instead. If self-consistent
perturbation calculation is performed the default SCF convergence criterion is increased
to 10−7 . It may be overwritten by a smaller value with the TOL option, if desired. Note
that tight SCF convergence criteria are largerly meaningless, unless tight numerical criteria a used for the the numerical integration of the exchange-correlation potential and
ECPs.
In addition to the standard, orbital-based convergence criterion, deMon also monitors
the convergence of the fit density. The default convergence criterion is adequate for an
overwhelming majority of applications, and should not be modified. In a few rare
cases (for example, when very accurate values are needed for the multipole moments of
the molecular charge distribution), the fit convergence threshold can be modified using
the CDF= option.
4.4.2
Keyword ORBITALS
This keyword controls the atomic orbital choice.
Options:
SPHERICAL / CARTESIAN
SPHERICAL
Spherical atomic orbitals (5d, 7f ) are used. This is the default.
CARTESIAN
Cartesian atomic orbitals (6d, 10f ) are used.
Description:
In deMon the spherical (see 4.3.6 for the definition of real spherical harmonic Gaussians)
and Cartesian atomic orbitals have the general form (omitting the normalization):
2
φlm (r) = r l e−ζr Slm
φijk (r) = xi y j z k e−ζr
(4.8)
2
(4.9)
48
4 Keywords
Because spherical orbitals do not include contaminants with lower angular momenta
(l′ = l − 2, l − 4, ...), they are the recommended choice for most applications.
4.4.3
Keyword ERIS
This keyword controls the calculation method for the three-center electron repulsion
integrals (ERIs).
Options:
CONVENTIONAL / DIRECT / MULTIPOLE / MEMORY
CONVENTIONAL
All ERIs are calculated at the beginning of the SCF procedure
and stored.
DIRECT
All ERIs are recalculated at each SCF iteration.
MEMORY
Chooses between CONVENTIONAL and DIRECT, based on the
available memory. This is the default.
MULTIPOLE
An asymptotic multipole expansion for long-range ERIs is
performed. The DIRECT option is activated, too.
TOL=<Real>
Threshold for ERI screening. The default is 10−14 for
CONVENTIONAL, and 10−8 otherwise.
Description:
Calculation of the three-center ERIs is one of the most time critical steps in a deMon
calculation. Therefore, the calculation method should be carefully selected. The CONVENTIONAL method calculates ERIs before of the SCF procedure and, if possible, stores
them in memory (RAM). This so-called in-core method is extremely fast as long as all
integrals fit into the RAM. The RAM size is controlled with the MAXMEM keyword (see
Section 4.12.1). If the RAM space is not sufficient deMon will write ERIs to the scratch
file ioeri.scr. Thus, the ERIs have to be read from the disk in each SCF step. For larger
systems this disk I/O becomes the bottleneck of the calculation.
The DIRECT method [41] avoids the disk I/O by recalculating the ERIs in each SCF
step. This is, therefore, the method of choice for larger systems where the ERIs do not
fit in the available RAM space.
The default MEMORY method will choose between a CONVENTIONAL and DIRECT
calculations, based on the amount of the available memory (see 4.12.1). This method will
typicall lead to the lowest wall-clock time for a calculations. On computer systems with
4.4 SCF Control
49
an unusually high I/O capacity, or when CPU time, rather than wall-clock time is of
interest, CONVENTIONAL approach may be preferable.
If the calculated system has a large extension (≥ 10 ˚
A) asymptotic expansions for longrange ERIs [42] can be used. This method is activated with the MULTIPOLE option.
Because asymptotic expansions are only implemented in the DIRECT method the MULTIPOLE option also triggers the DIRECT option. This is the method of choice for large
extended systems.
Table 7 shows timings for the ERI calculation of small alkenes with the three methods.
The DZVP basis and A2 auxiliary function set was used. The SCF convergence (10−5 )
was reached for all systems within 10 cycles. Thus, the ERIs hat to be calculated ten
times in the DIRECT and MULTIPOLE method. Table 7 shows the total time for these
two methods and the time for the (one) ERI calculation with the CONVENTIONAL
method. In brackets the difference between the real and CPU time is given, too. All
calculations were performed on a single SGI R14000 node with 2 Gbyte of RAM and a
Ultra SCSI disk.
Table 7: Timing [sec] for the ERI calculation of alkenes with the CONVENTIONAL, DIRECT and MULTIPOLE method. The number of basis and auxiliary functions
is given by NBasis and NAuxis . The difference of real and CPU time is given in
brackets.
Alkene
C24 H50
C36 H74
C48 H98
C60 H122
C72 H146
a
NBasis
610
910
1210
1510
1810
NAuxis
1016
1520
2024
2528
3032
CONVENTIONAL
132 ( 14)
306 ( 220)
572 (2374)a
914 (2516)a
1297 (2558)a
DIRECT
345 ( 5)
862 ( 434)
1522 ( 493)
2375 ( 465)
3428 ( 466)
MULTIPOLE
266 ( 5)
567 ( 300)
898 ( 264)
1294 ( 255)
1762 ( 260)
Disk I/O is performed.
As Table 7 shows the CONVENTIONAL method is most economic for the two smallest
alkenes. In these cases in-core calculations can be performed and the disk I/O is reduced
to a minimum. The ERI calculation with the DIRECT method takes almost three times
more. Remarkable is also the relative large overhead of the DIRECT method (434 sec)
for C36 H74 . The reason is the incremental build of the Kohn-Sham matrix which involves
I/O operations (reading and writing of the Kohn-Sham matrix) not present in the in-core
2
calculation. However, the Kohn-Sham matrix is only a NBasis
object and, therefore, the
50
4 Keywords
I/O overhead of the DIRECT method becomes less important for large systems.
In the case of the C48 H98 and larger alkenes the ERIs do not fit any longer in the RAM.
Now I/O ERI operations are necessary with the CONVENTIONAL method. This immediately increases the overhead of the calculations. Altogether, the real time for the
CONVENTIONAL calculation in C48 H98 (2946 sec) is larger than for the DIRECT calculation (2015 sec). A further reduction is obtained with the MULTIPOLE method (1162
sec). As can be seen from Table 7 the MULTIPOLE method becomes more favorable with
1.7
increasing size of the alkenes. This is due to the reduced scaling (NBasis
) of this method
even for such small systems as the alkenes in Table 7.
The TOL option controls screening of the ERIs. The screening threshold τ is calculated
as:
TOL
(4.10)
Number of Electrons
ERIs with an orbital overlap smaller than τ are not calculated (screened). The threshτ=
old τ also enters into the asymptotic expansion radii for the MULTIPOLE method.
The density screening threshold for the numerical integration of the exchange-correlation
potential is given by τ , too. The default settings for TOL are 10−14 and 10−8 for the
CONVENTIONAL and DIRECT/MULTIPOLE method, respectively. Please note that
the screening of the ERIs can deteriorate the SCF convergence, particularly for systems
with extensive electron delocalization.
4.4.4
Keyword GUESS
This keyword specifies the SCF start density.
Options:
TB / CORE / FERMI / RESTART
TB
A tight-binding SCF start density is calculated. This is the
default.
FTB
Use “Free” tight-binding guess. This is a synonym for TB.
HTB
Use “Harmonic” tight-binding guess.
CORE
The SCF start density is obtained from the diagonalization
of the core Hamiltonian.
FERMI
The start density is obtained by quenching a fractional
occupied SCF solution to integer occupation numbers.
4.4 SCF Control
51
RESTART
The start density is read from the restart file deMon.rst.
ONLY
The program stops after the generation of the start density.
HARM=
Radius of the harmonic restraining potential, in units of
covalent atomic radius. The default is 2.0
FREE=
Radius of the restraining potential for a “free” atom.
The default is 100.0
SCC
Perform self-consistent charge DFTB computation.
Description:
The choice of the start density can be crucial for the SCF convergence. In most cases the
tight-binding start density is the recommended choice. For metal clusters the core start
density may be sometimes advantageous. If the molecular orbitals of the start density (for
the printing of the MOs see Section 4.12.3) show only a very small HOMO-LUMO gap
the FERMI option may be used to obtain a better start density. It should be noted that
this option implies a SCF calculation and, therefore, is much more time consuming than
the other start density options. If a FERMI start density is used the OMA option of the
keyword MIXING (see 4.4.5) should not be applied! The start density can also be read
from the restart file deMon.rst of a previous run. With the keyword MOEXCHANGE
(see 4.3.3) the molecular orbital ordering of the start density can be altered. The option
ONLY stops the program after the start density is generated and written to the restart
file. This option is recommended if a FERMI start density was requested or the start
density should be altered after inspection (see example 5.14). Note that GUESS RESTART
is currently not supported for geometry optimizations. To print guess molecular orbitals,
use a combination of the keywords SCFTYP MAX=1 and PRINT MOS.
4.4.5
Keyword MIXING
This keyword controls the charge density mixing.
Options:
+<Real>
Fixed charge density mixing parameter between 0 and 1
-<Real>
Dynamic charge density mixing parameter. The default is -0.3.
OMA
The optimal mixing algorithm is activated.
Description:
52
4 Keywords
Hartree damping [44] of the charge density fitting coefficients is performed, using the
specified the charge density mixing parameter. Smaller mixing parameters correspond to
stronger damping. For some systems mixing parameter smaller than default (e.g. 0.1)
may be necessary to achieve convergence. If a dynamic mixing parameter is chosen, the
mixing ratio is reduced during the SCF iterations if the Minmax SCF error increases.
The OMA option activates the optimal mixing algorithm of Canc´es [43]. This option is
recommended if the dynamical mixing fails or is converging slowly. It should be noted
that OMA possesses a considerable overhead and exhibits bad convergence for small
HOMO-LUMO gaps.
4.4.6
Keyword SHIFT
This keyword activates the level-shift procedure.
Options:
+<Real>
Fixed level-shift value [in a.u.].
-<Real>
Dynamic level-shift value [in a.u.].
Description:
The level-shift procedure allows artificially expanding the HOMO-LUMO gap during
SCF iterations [45]. This stabilizes the initial SCF configuration. The shift value is subtracted from the orbital energies once SCF procedure has converged. Because of the small
HOMO-LUMO gap in DFT calculations this procedure is very valuable to improve SCF
convergence. However, with a fixed level-shift calculation may converge to an excited
state. Therefore, the converged orbital occupation should be examined using the MOS
option of the PRINT keyword (see 4.12.3). This problem can be reduced if the dynamical
level-shift procedure [14] is used. Here the shift value is adapted to the Minmax SCF error. While the error is large, usually at the beginning of the SCF procedure, the full shift
value is used. When the error decreases the shift value is decreased, too (see Example
5.15). Usually a small shift value remains at the end of the SCF procedure. The dynamical level-shift procedure has proven very valuable for the SCF convergence of transition
metal clusters and systems [46].
If the SHIFT keyword is used, the level-shift procedure is applied in all calculation steps
(optimization, frequencies, properties, etc.). If this is not desirable, a single point SCF
can to be performed with the SHIFT keyword. The converged density can then to be
4.4 SCF Control
53
used as the restart density (see 4.4.4) in the following calculations without the SHIFT
keyword (See example 5.16).
A cautionary note: Neither having the aufbau orbital occupations at the end of the
SCF, nor using dynamis level shifting procedure provide a guarantee that your calculation
converged to the ground electronic state. Conversely, in approximate density-functional
theory it is also possible (although extremely rate) to have a non-aufbau ground state.
Caveat user!
4.4.7
Keyword DIIS
This keyword activates the DIIS procedure.
Options:
ON / OFF
ON
The DIIS procedure is switched on. This is the default.
OFF
The DIIS procedure is switched off.
TOL=<Real>
The DIIS procedure is switched on after the SCF energy
error is smaller than <Real>.
CMAX=<Real>
Largest allowed expansion coefficient. Default is 100.0
DMIN=<Real>
Smallest allowed determinant. Default is 10−14
Description:
By default the DIIS (direct inversion in the iterative subspace) procedure [47] is switched
on if the energy error is less than 0.1 a.u. in a SCF step. For most systems this result in a
considerable speed-up of the SCF convergence. However, for some systems this threshold
is too large and a SCF convergence failure will occur. In these cases the start of the DIIS
procedure should be manipulated with the TOL option. If the HOMO-LUMO gap of a
system is very small DIIS can be counter productive for the SCF convergence. In these
cases the DIIS procedure can be switched off with the OFF option. It should be noted
that in deMon different DIIS algorithms are used for the AUXIS and BASIS option of
the VXCTYPE keyword (see 4.4.9). Both are based on the energy gradient with respect
to the charge density fitting coefficients. In the case of the AUXIS option only charge
density fitting coefficients are used in the DIIS step. Therefore, no extra I/O is necessary.
In the case of the BASIS option also density matrices are used in the DIIS step, leading
to an increase in the I/O overhead.
54
4 Keywords
DIIS procedure will be restarted if any of the extrapolation coefficients exceeds CMAX
in absolute magnitude, or if the determinant of the DIIS linear problem drops below
DMIN. The defaults for these values are adequate in an overwhelming majority of cases,
and should not be tinkered with.
4.4.8
Keyword BROYDEN
This keyword activates the BROYDEN convergence procedure.
Options:
TOL=<Real>
The BROYDEN procedure is switched on after the SCF energy
error is smaller than <Real>.
MAX=<Integer>
Number of iterations stored in BROYDEN procedure. The default is 10.
Description:
The BROYDEN procedure is a special form of mixing, which tries to minimize the error,
i.e. the difference between a guess vector and a returned vector of the SCF procedure
by inversion of an error matrix. The error matrix is defined in the standard way, as,
for example, in DIIS. The new guess depends on the former iterations as well as on the
MIXING coefficient. The method is applied only on charge density coefficients.
4.4.9
Keyword VXCTYPE
With this keyword the exchange-correlation potential is selected. It also controls the
density used for the exchange-correlation energy and potential calculation.
Options:
VWN / PZ81 / PW92 / PW86 / BLYP / PW91 / PW91SSF / PBE / PBESSF
/ OLYP / XALPHA / NONE
VWN
Dirac exchange with local VWN correlation.
PZ81
Dirac exchange with local PZ81 correlation.
PW92
Dirac exchange with local PW92 correlation.
PW86
PW86 GGA exchange with P86 GGA correlation.
BLYP
B88 GGA exchange with LYP GGA correlation.
OLYP
OPTX GGA exchange with LYP GGA correlation.
4.4 SCF Control
55
PW91
PW91 GGA exchange and correlation.
PW91SSF
PW91 with full spin scaling function.
PBE
PBE GGA exchange and correlation.
PBESSF
PBE with full spin scaling function.
XALPHA
Xα calculation. The default α value is 0.75. A user defined
α value can be selected with the X = <Real> option.
NONE
No exchange-correlation functional is used.
(ie Hartree calculation is done).
AUXIS / BASIS
AUXIS
The auxiliary function density is used for the calculation of the
exchange-correlation energy and potential. This is the default.
BASIS
The orbital density is used for the calculation of the
exchange-correlation energy and potential.
Description:
The above options represent the most common combinations of exchange and correlation
functionals. Besides these combinations the exchange and correlation functionals listed
in Table 8 can be combined by the user. The syntax for the user defined potentials is
<Exchange> - <Correlation> (e.g. B88 - P86). With NONE the exchange or correlation
part can be omitted. In contrast to many examples in the literature, the PW86 exchange
functional in deMon is implemented with a cutoff. The local contribution of the P86
correlation functional is calculated using the VWN functional. All other functionals are
implemented according to the cited references. Where possible, second derivatives with
respecy to density were eliminated by integration by parts [48].
The AUXIS and BASIS options specify the density that is used for the calculation of
the exchange-correlation energy and potential. By default the auxiliary function density
is used for the calculation of the exchange-correlation energy and potential. Using the
option BASIS results in a significant slow-down of the SCF calculation. However, for
sensitive property calculation the BASIS option is recommended. For more details and
recommendations see Section 1.4, ”How to Use deMon”.
For the PW91 and PBE functional two different spin scaling functions are implemented
in deMon. By default a numerical more stable cutoff function is used. The suffix “SSF”
(options PW91SSF and PBESSF) selects the original form of the spin scaling function.
56
4 Keywords
This may change the orbital energies considerably. For the total energies the effect is
usually negligible. For other functionals, the default choice of the spin scaling function
can be overriden by specifying SSF=FULL or SSF=G98 options of the VXTYPE keyword.
Table 8: Exchange and correlation functionals available in deMon. The minimal abbreviations are given in bold.
Exchange:
DIRAC
PW86
B88
OPTX
PW91
PBE96
PBE98
PBE99
Local Dirac exchange [49]
Perdew and Wang GGA exchange from 1986 [50]
Becke GGA exchange from 1988 [51]
Handy and Cohen GGA exchange from 2001 [52]
Perdew and Wang GGA exchange from 1991 [53]
Original Perdew, Burke and Ernzerhof GGA exchange from 1996 [54]
PBE GGA exchange modified by Y. Zhang and W. Yang [55]
PBE GGA exchange modified by B. Hammer et al. [56]
Correlation:
VWN
PZ81
PW92
PZ86
P86
LYP
PW91
PW91SSSF
PBE
PBESSF
Local Vosko, Wilk and Nusair correlation [57]
Local Perdew and Zunger correlation [58]
Local Perdew and Wang correlation from 1992 [59]
Perdew GGA correlation from 1986 [60]
Perdew GGA correlation from 1986 with VWN local correlation [61]
Lee, Yang and Parr GGA correlation from 1988 [62–64]
Perdew and Wang GGA correlation from 1991 [53]
PW91 with full spin scaling function
Perdew, Burke and Ernzerhof GGA correlation form 1996 [54]
PBE with full spin scaling function
Note: In the literature, combinations of the PBE correlation functional, and the PBE98 and
PBE99 exchange functionals are often refered to as “revPBE” and “RPBE” functionals,
respectively.
4.4.10
Keyword GRID
This keyword specifies the grid for the numerical integration of the exchange-correlation
energy and potential.
Options:
4.4 SCF Control
57
ADAPTIVE / FIXED
ADAPTIVE
An adaptive grid is used for the integration. This is the default.
FIXED
A fixed grid is used for the integration.
MEDIUM / COARSE / FINE
MEDIUM
A medium grid accuracy is requested. This is the default.
COARSE
A coarse grid accuracy is requested.
FINE
A fine grid accuracy is requested.
TOL=<Real>
Adaptive grid tolerance.
SCF / GUESS
SCF
The converged SCF density is used for the adaptive grid
generation. This is the default.
GUESS
The start density is used for the adaptive grid generation.
Description:
By default, deMon uses an adaptive grid [65] with a tolerance of 10−5 (MEDIUM) for the
numerical integration of the exchange-correlation energy and potential. This tolerance
corresponds to the accuracy of the numerical integration of the diagonal elements of the
exchange-correlation potential matrix. For the converged SCF energy an accuracy better
than 100 µHartree is usually obtained with this setting [66]. The COARSE and FINE
option for the adaptive grid refer to grid tolerances of 10−4 and 10−6 , respectively. Thus,
the stability of the numerical integration can be easily checked by choosing different grid
tolerances. The COARSE adaptive grid should not be used for final energy or property
calculations. User-defined grid tolerances can be specified using TOL option. This option
is only applicable for an adaptive grid. The same holds for the SCF and GUESS options,
which specify the trial density used for the grid generation. By default, the converged
SCF density is used. This implies a second SCF cycle for the initial energy calculation.
If the option GUESS is specified the adaptive grid is build using the start density for the
calculation of the grid generating exchange-correlation potential.
Molecular dynamics simulations, where a good guess density is tipically available on each
step, typically run faster with GUESS grids, due to smaller computational overhead due
to grid generation.
58
4 Keywords
Fixed grids available in deMon can be selected by the option FIXED in combination with
the options COARSE, MEDIUM and FINE, with MEDIUM being the default. Options
COARSE and MEDIUM select (50,194)p and (75,302)p pruned grids, respectively. In
this notation the first value refer to the radial shells. The second value is the size of the
largest Lebdev grid [68] on each of these shells. The p stands for pruned [67] indicating
that smaller Lebedev grids have been used for some radial shells. The FIXED FINE
option requests the unpruned (200,1202) reference grid. Due to its size its application is
impractical for all but smallest systems.
4.5 Molecular mechanics and QM/MM Control
4.5
4.5.1
59
Molecular mechanics and QM/MM Control
Keyword QMMM
This keyword activates molecular mechanics or hybrid QM/MM calculation.
Options:
QM / MM / QM/MM
QM
Treat the whole system using quantum mechanics; parse but ignore
all MM- and QM/MM-related input keywords. This is the default.
MM
Treat the whole system using molecular mechanics.
QM/MM
Partition system into QM and MM parts, as specified by the
QM/MM keywords in the geometry section. This keyword can also
be given as simply ON.
COUPLING=
Specifies coupling Hamiltonian, used to describe interactions between the QM and MM parts. The default is MECHANICAL.
RUNTYPE=
Specifies time synchronization between the partitions. The default
is SYNCHRONOUS.
QMPBC
Both, QM part and MM part are treated within periodic boundary
conditions. This option is only available for DFTB as QM part. As
default, the QM region is treated as a cluster (molecule) and PBC
are applied only to the embedding MM part.
Description:
With the default QMMM QM setting, all keywords and flags related to molecular mechanics
and hybrid calculations are recognized, but ignored in the input file. The whole system is
therefore treated using DFT Hamiltonian. The opposite setting, QMMM MM will use molecular mechanics forcefield (specified by the FORCEFIELD keyword, see 4.5.2) to treat the
entire system. If this option is specified, deMon will behaive as if QMMM=MM has been given
for each atom in the GEOMETRY section (see 4.1.1). This option is particularly useful for the
initial geometry optimization, or for the initial part of a molecular dynamics equilibration
run. Finally, the QMMM QM/MM will honour all QM/MM-related input options.
The two options supported for the QM/MM coupling Hamiltonian (COUPLING=) are:
60
4 Keywords
Use simple mechanical embedding. The total energy of the system
MECHANICAL
in this case is given by:
Etot = EM M (QM + MM) − EM M (QM) + EQM (QM).
Apart from bond termination, the molecular mechanics subsystem
ELECTROSTATIC
does not directly influence the electronic structure of the QM part.
Also include cross-system electrostatic polarization terms. This op-
tion is not yet implemented.
Regardless of the coupling scheme used, bond termination at the partition boundary is
handled using capping effective potential approach, using either pseudo-halogen[10] or
pseudo-hydrogen[11] pseudopotentials.
The dynamical relationship between the QM and MM partitions can be controlled with
the RUNTYPE= keyword. The recognized options are:
SYNCHRONOUS Time steps are identical across the partitions. This is the default.
RELAX MM
The MM partition is completely relaxed for every time step in the
FREE ENERGY
QM partition. This option is not yet implemented.
Time runs faster in the MM partition; the QM partition moves on
the free-energy surface of the MM partition. This option is not yet
implemented.
If your application requires a currently unimplemented coupling scheme or RUNTYPE,
you are encouraged to contact: [email protected]
Note: Currently, the RESTART option is not supported for calculations where a change
in the system Hamiltonian occurs. Molecular dynamics simulations can be restarted (see
4.8.1) from a simulation using different system partitioning.
4.5.2
Keyword FORCEFIELD
This keyword is used to specify molecular mechanics forcefield, and to set forcefieldrelated computational options. Options:
TYPE=
Chooses molecular mechanics forcefield. This can be either UFF
or EXTERNAL. Only UFF, which is the default, is currently
implemented.
BONDS=
Chooses the bond matrix definition. This can be either EXPLICIT
or AUTO. AUTO is the default.
PROG=
Name of an external molecular mechanics program. There is no
default.
Description:
4.5 Molecular mechanics and QM/MM Control
61
The default molecular mechanics forcefield, used in molecular mechanics and QM/MM
calculations is the Universal Force Field (UFF) by Rappe et al.[69–74] (see also section
6). While this force field is not particularly accurate for any specific system, it covers the
entire periodic table, and often produces reasonable structures. This forcefield implementation is integrated in deMon, and carries minimal overhead in applications.
The option BONDS= selects handling of the connection matrix, which is needed in most
MM force fields. The two options are:
EXPLICIT All bonds and bond orders must be specified explicitly, in the geometry input (see keywords BOND= and NOBOND= in section 4.1.1.
deMon will try to guess the connection matrix, suitable for the
AUTO
specified starting geometry. It is still possible to override the guess
(partially or completely) by supplying BOND= and NOBOND= flags in
the GEOMETRY section. This is the default.
Finally, PROG= chooses the name of the external executable (or a shell script), which will
be called to evaluate MM energies and forces, if an external force field is activated. This
must be a absolute path name - paths relative to the current directory will not work.
There is no default for this option.
4.5.3
Keyword MMOPTIONS
This keyword sets various technical parameters for the MM simulation part. Not all of
those parameters have a meaning for all run types.
Options:
TSTEP=
MM MD time step, in atomic time units.
STEPS=
Number of MM MD time steps, or maximum number of MM geometry optimization steps.
TOLE=
Energy convergence criterion for MM optimizations, in Hartree.
TOLG=
Gradient
convergence
criterion
for
MM
optimizations,
in
Hartree/Bohr.
TOLX=
Displacement convergence criterion for MM optimizations, in Bohr.
TOL=
Set TOLE, TOLG, and TOLX to values suitable for energy convergence specified here, in Hartree.
TAU=
Berendsen tau parameter, in atomic time units.
T=
MM simulation temperature, in Kelvin.
62
4.5.4
4 Keywords
Keyword VDWAALS
This keyword sets defaults for van-der-Waals non-bonded interaction summation cutoffs
(used in DFTB, MM, and QM/MM).
Options:
ECUT=
Energy cutoff in Hartree. The default is 10−6 .
SMOOTH
The van-der-Waals energy is shifted for long-range interactions by
a value ∆E, as specified by the X= keyword, and repulsive contributions are set to zero. This allows tighter cutoffs and hence faster
calculations, but shifts the potential energy upwards.
X=
Specifies the energy shift, which is X · dij , dij specifying the depth
of the van-der-Waals minimum of the interaction potential between
centres i and j. The default is X=0.01.
Description:
All van-der-Waals type non-bonded pair interactions smaller than ECUT will be neglected.
This parameter affects both periodic and non-periodic calculation. Although this cut-off
looks extremely tight, it is necessary to guarantee total energy conservation to within
0.1 kcal/mol in a typical periodic MD simulation.
4.5.5
Keyword MADELUNG
This keyword sets defaults for the summation of the Madelung contribution to the electrostatic potential (used in DFTB, MM and QM/MM). Options:
EWALD
The Ewald summation technique is used for the Madelung potential. This is the default.
SCREEN
The real space summation, described below, is applied.
KAPPA=
The κ parameter for the Ewald summation. If κ is negative, deMon
chooses it such that the real-space part can be calculated within
the minimum image convention. This is the default.
D=
d parameter of the screening function, in Bohr. The default is 6.0
R0=
r0 parameter of the screening function, in Bohr. The default is 70.0
ECUT=
Description:
Energy cutoff, in Hartee. The default is 10−6
In periodic MM, QM/MM, and DFTB calculation on systems with non- negligible atomic
charges, deMon must evaluated the electrostatic potential of the infinite periodic charge
4.5 Molecular mechanics and QM/MM Control
63
distribution (Madelung potential). This can be done using the Ewald technique, or alternatively employing real-space summation, with a screened interaction potential[75]. In
this technique, the usual Coulomb interaction potential is replaced with:
v scr (r) =
1
1
∗
r 1 + exp ((r − r0 )/d)
(4.11)
The screening parameters r0 and d must be selected such that:
• The screening function is not distinguishable from 1 within the unit cell surrounding
the origin.
• The fall-off region of the screening function is wider than the typical extent of
non-neutral charge distrubutions in the system.
The defaults are appropriate for simulation of liquids of small polar molecules. Simulations
of systems with long-range charge correlations may require large values of d and r0 .
4.5.6
Keyword PERIOIC
This keyword activates periodic boundary conditions in MM and QM/MM calculations,
and specifies the initial extent of the unit cell.
Options:
CUBIC / RECTANGULAR / GENERAL
CUBIC
Cubic simulation box is used.
RECTANGULAR
Rectangular simulation box is used.
GENERAL
General simulation box is used. This is the default.
FULL
Use true periodic boundary conditions. This is the default.
MINIMAL
Use minimal-image (cyclic cluster) periodic boundary conditions in the molecular mechanics part.
Description:
The PERIODIC keyword requires additional input, which must follow it in the input,
with no intervening comments or blank lines. For the CUBIC box, just one parameter is
required - the extent of the box edge, in the same units as in the coordinate section.
For the RECTANGULAR box, the extents along the X, Y , and Z spatial directions must
be specified. In either case, the lattice vectors are taken to be parallel to the coordinate
axes: a||X, b||Y , and c||Z. For a GENERAL unit cell, the Cartesian coordinates of the a, b,
and c lattice vectors must be specified, with one vector per line.
64
4 Keywords
If MINIMAL keyword is specified, for each pair of atoms inside the unit cell only interactions
with the closest image of the pair will be considered. This is equivalent to the cyclic cluster
periodic model. Calculations using minimal image convention can be considerably faster
than full periodic boundary conditions, at the expense of the more approximate treatment
of long-range interactions.
Note that in a molecular dynamics run, the unit cell parameters may be affected by the
pressure bath (see keyword MDPRESSURE in section 4.8.8).
4.6 DFTB Control
4.6
DFTB Control
4.6.1
Keyword DFTB
65
NOTE: The usage of DFTB requires a little understanding of the method. The predicitivity of this method is somewhat between MM and DFT, therefore you should always know
what you are doing. At the moment, both variants, non-self-consistent DFTB[118] (sometimes referred to as NSCC-DFTB) and self-consistent charge DFTB (SCC-DFTB)[119]
are used. For systems where London dispersion forces are relevant, a UFF-based[69] dispersion correction can be used.[120]. Recently, a Car-Parrinello version has been developed and can be used.[121] Note that this method is still in an early stage, and use it with
particular caution. An Open Access review of DFTB has been published recently.[122]
The default method for electron structure calculations is DFT. The DFTB keyword
activates the alternative the Density-Functional Tight binding (DFTB) method. This
method runs, similarly as molecular mechanics, principally independent from the SCF
core of the deMon code. In future, a hybrid run DFT + DFTB is planned, but not realised
yet.
In the present installation the DFTB keyword disables all DFT features and a DFTB
computation is performed. DFTB is running as single point, with optimisation and molecular dynamics. QM/MM is available with DFTB/UFF. You can run it for finite molecules
as well as for periodic boundary conditions. Various features of DFTB are still missing,
the most important list is
• Frequency analysis
• LSDA extension
• TD-DFTB
For the moment, the following features are working and tested:
• standard DFTB and self-consistent-charge (SCC)-DFTB in molecular and periodic
representation
• London dispersion correction for DFTB and SCC-DFTB
• geometry optimisation for molecules
• geometry optimisation for solids with fixed unit cells
66
4 Keywords
• molecular dynamics
DFTB uses parameters which are stored in Slater-Koster (SlaKo) integral tables. For
the moment, there are two independent sets of these tables given, and their location is
specified in $deMon/basis/SLAKO for standard DFTB and in $deMon/basis/SCC-SLAKO
for self-consistent charge DFTB. The parameter tables differ slightly from those which
are available at http://www.dftb.org, so before you use your own tables please contact
one of the deMon-Nano authors. Moreover, deMon-Nano comes with two sets of SCCDFTB parameters. The default parameters have been developed for materials and should
be cited as follows:
J. Frenzel, A. F. Oliveira N. Jardillier, T.Heine, and G. Seifert. Semi-relativistic, selfconsistent charge Slater-Koster tables for density-functional based tight-binding (DFTB)
for materials science simulations. TU-Dresden 2004-2009.
The content of the default SCC-DFTB parameter file $deMon/basis/SCC-SLAKO file is
identical to $deMon/basis/MAT-SCC-SLAKO. If you study organic or biological systems we
recommend to use a parameter set which has been optimised for these applications. An
excellent choice is the set of Marcus Elstner, to be cited as in ref. [119], which is available
in BIO-SCC-SLAKO. In order to activate this, you may want to copy BIO-SCC-SLAKO to
SCC-SLAKO.
Options:
DFTB / SCC / DISP / DIAG=method / / BS / SIMPLE / MIX=value /
MAX=value TOL=value / L-DEP / FERMI / FERMI= / ETOL / SPLA /
NOSPLA
SCC
The self-consistent charge variant of DFTB. The default is nonSCC.
DISP
London dispersion energy and gradients are included. By default
this option is switched off.
DIAG=
User-specified diagonalisation routine for the generalised Eigen
value systems. The method is specified after the ’=’. Legal options
are DSYGVD, DSYGV, DSYGVR, SOM. The default is DSYGVD.DSYGV,
DSYGVD and DSYGVR are standard LAPACK subroutines (for documentation see http://www.netlib.org.) SOM uses the built-in mechanism to orthogonalise the secular equations and makes use of the
global variable MATDIA.
4.6 DFTB Control
BS
67
All Eigen values and vectors are computed. This option is useful
when using the subspace diagonaliser, and per default it is switched
off. The option makes only sense for DSYGVR.
SIMPLE
Simple mixing instead of Broyden convergence acceleration is performed for SCC. This is per default switched off.
MIX=
Mixing value for simple and Broyden mixing. The default is 0.2.
This value is independent from the DFT mixing value.
MAX=
Maximum number of SCC cycles. The default is 100.
TOL=
The SCC convergence criterion in electron charges. The default is
1.0E-8.
L-DEP
Angular momentum resolved SCC. Here, the self-consistent charge
procedure is applied to Mulliken shell charges rather than to Mulliken atomic charges. Otherwise, the formalism remains as in standard SCC-DFTB. This option might be important when treating
transition metals where d and s,p shells have significantly different hardness parameters. By default, this option is deactivated.
FERMI
The electron occupation follows a Fermi distribution as described
in Ref. [123]. The Fermi temperature is 5 K, so in practise degenerate orbitals are occupied fractionally. Note that no spinpolarised calculation is carried out. The FERMI keyword overwrites
the MULTIPLICITY of the system.
FERMI=
As FERMI, but with used-specified Fermi temperature (in K) following the keyword.
ETOL
Requested energy tolerance for iterative solvers of the total energy
(so far only used for CPBO, see CARPAR section.
SPLA
Use Sparse Linear Algebra for overlap and Hamiltonian matrix as
well as for density and energy-weighted density matrices. This option is faster for large problems and requires only a small amount
of memory. By default, the program selects this mode if the overlap
matrix has less than 5 % occupation.
NOSPLA
Contrary to SPLA, this mode uses full matrices and allows the
usage of BLAS throughout the program. Requires significantly more
memory than SPLA. Might be a good option for computers with
very high BLAS efficiency.
68
4 Keywords
Some remarks on the diagonalisation:
The default diagonaliser is a self-written DSYGVR. This is LAPACK’s DSYGV routine, which
is solving the generalised Eigen value system of a symmetric, positively defined problem,
using Cholesky decomposition and a standard diagonalisation tool DSYEV. The computationally inefficient DSYEV routine is replaced by DSYEVR, which is LAPACK’s most performant diagonaliser, and which uses only a small work field. This allows Born-Oppenheimer
DFTB computations with only 2 matrices in memory plus a little work field which is Order(N).
DFTB parameter files:
The DFTB method requires interaction parameters. These are stored in so-called SlaterKoster tables. With this distribution, we provide three sets of tables which can be used
for different types of applications. If you have your own parameter files or if you obtained
them through the www.dftb.org website you need to make them compatible to deMon.
In the $deMon/basis directory are two files which the program uses to identify the
necessary parameter files. These files are called SLAKO and SCC-SLAKO. These files has the
following structure:
First, it gives the atomic numbers of the pair of interaction, so for exampe for H−H 1 1.
Then, it gives the full string to the location of the parameter file. At your installation,
it uses variables which are resolved in the deMon script. If you want to use your own
parameter tables you will have to adopt these files.
All parameter files are in the $deMon/basis/skf directory.
4.7 Optimization Control
4.7
4.7.1
69
Optimization Control
Keyword OPTIMIZATION
This keyword controls the geometry optimization. It differs considerably from the deMon2K version as it is optimised for large systems. deMon-nano only uses robust optimisers with low memory consumption and with moderate numerical effort.
Options:
MAX=
TOL=<Real>
Maximum number of optimisation steps. The default is 100.
Optimization convergence criterion for the energy. Default is 3 ×
10−4 Hartree.
GRADTOL=<Real>
Optimization convergence criterion for the gradients. Default is 3 ×
10−6 Hartree.
STEP=<Real>
Maximum step size in optimization. Default is 0.3 Bohr
CGRAD
Uses the conjugate gradient method. This is the default.
SDC
Uses the steepest descent method for geometry optimisation.
SDC=<Real>
Uses the steepest descent method for geometry optimisation with
specified multiplied for the gradient. The default is 0.1.
OUT=<Integer> Interval to output structure during optmimisation.
NUCFRIC
Perform geometry optimisation via MD with additional friction
term for degrees of freedom where velocity and gradient have opposite sign. The multiplier of the dissipative term is by default 0.6.
NUCFRIC=<Real>
Perform optimisation via MD with additional friction term with a
user-defined mulitplier to the dissipative term.
INTQMO=
Interval when MOs are quenched during simultaneous optmimisation procedure of MOs and coordinates. The default is 0 (all steps).
Description:
deMon-nano uses simple, robust optimisation techniques which are scalable to large systems. They include the steepest descent method, a conjugate gradient scheme, and molecular dynamics with an additional friction term.
70
4.8
4 Keywords
MD Control
4.8.1
Keyword MDYNAMICS
This keyword activates the Born-Oppenheimer molecular dynamic (BOMD) simulation.
Options:
ZERO / RANDOM / RESTART / READ / RESET
RESTART
Restart previous molecular dynamics simulation using information
of deMon.qmd file
ZERO
Start a trajectory with initial velocities of the nuclei set to zero.
This is the default.
RANDOM
Start a tajectory using random initial velocities, which have no
net momentum or angular momentum and give the requested
temperature.
READ
Start a new MD trajectory. Velocities of the nuclei are given
in Cartesian form in the input file, immediately following the
MDYNAMICS keyword.
RESET
This option resets all averages and the MD step of the trajectory.
It only makes only sense with the RESTART option, otherwise it
has no effect.
Description:
During an MD restart, the following processing occurs: Atom types and masses are taken
from the input file. All other information, including coordinates, velocities, last gradients,
unit cell parameters, and information for computing averages are taken from the MD
restart file, which has a .qmd extension. Running averages of the temperature, total and
potential energy, pressure, etc. are taken from the MD restart file as well. All other
computational parameters (SCF, MD, QM/MM, etc) are taken from the input file. If
the molecular dynamics time step or QM/MM partitioning scheme are different in the
restart file and the input file, all running averages will be reset, and simulation will start
at time t = 0. WARNING: You should always use the same atom ordering in the input
file used to produce the MD restart file, and in the file used to contue the simulation. In
particular, you should never use the .new file generated by deMon for performing
an MD restart – there is no guarantee that atom ordering is identical in the original input
file and the .new file.
When MDYNAMICS RANDOM is specified, the initial velocity components of an atom i are
4.8 MD Control
71
calculated as:
4
sin (2π(rand − 0.2))
mi
4
sin (2π(rand − 0.2))
vy (i) = k
mi
4
vz (i) = k
sin (2π(rand − 0.2))
mi
vx (i) = k
where rand is a random number, uniformly distributed on the interval [−1 : +1], and
mi is mass of atom i. Scaling coefficient k is calculated such that the initial kinetic
energy matches the desired temperature T . Important: this initial velocity distribution
is non-physical, and must be equilibrated before any production sampling runs.
MDYNAMICS READ expects velocity specification in the format:
I
VX
VY
VZ
where I is the sequential atom number (in the input order), and VX, VY, and VZ are initial
velocity components. Velocities must be specified in Bohr/au[time] or in ˚
A/femtosecond,
depending on the units used to specify the geometry. Atoms not mentioned in MDYNAMICS
READ input section are given zero initial velocities.
4.8.2
Keyword MDSTEPS
This keyword controls the steps of the molecular dynamic (MD) simulation.
Options:
MAX= / OUT= / SOUT= / MDRST= / TSIM=
MAX=<Integer>
Maximum number of MD steps. The default is 1.
TSIM=<Real>
Maximum MD run time in ps.
OUT=<Integer>
Step interval to update the deMon.mol file. The default is 10.
SOUT=<Integer>
Step interval to update the deMon.out file. The default is 1.
MDRST=<Integer> Step interval to update the deMon.qmd file. The default is 1.
Description:
MDSTEPS controls the total runtime of the MD simulation, either by the number of MD
steps or by specifying a requested runtime.
72
4 Keywords
In addition, parameters which control the I/O load of the calculation can be specified.
The output to deMon.mol and to deMon.qmd (trajectory and restart) might be timeconsuming, in particular when running in parallel. Therefore, it makes sense to select
larger intervals for saving restart information, and to write the trajectory only at intervals
where they are meaningful.
4.8.3
Keyword MDTEMP
This keyword specifies the desired temperature of the molecular dynamic (MD) simulation.
Options:
<Real>
Desired MD temperature in Kelvin. The default is 300.
Description:
This keyword may affect a molecular dynamics run in two ways. For MDYNAMICS RANDOM,
the desired temperature is used to rescale initial velocities. Otherwise, in a normal MD
run, MDTEMP has no effect unless one of the thermostat options is activated - see MDBATH
(section 4.8.5).
4.8.4
Keyword TIMESTEP
This keyword specifies the time step of the molecular dynamic (MD) simulation.
Options:
<Real>
MD time step in femtoseconds [fs]. The default is 1
Description:
The default MD time step is determined by the most light weighted atom and the simulation temperature and rounded down to 1/4 fs units. The determination is done using
the following formula: dt = 0.25 ∗
q
mmin /mH ∗
q
300K/MAX(T0 , 5K). This assumes
a time step of 0.25 fs for a system containing protons at 300K. In systems where hard
wall-like parts of the potential energy surfaces are probed (e.g. gases at high pressures),
time steps smaller than the default may be required to maintain total energy constant.
An excessively large time step will manifest itself in poor energy conservation during an
MD simulation.
4.8 MD Control
73
Negative time steps allow reverse-time dynamics. However, note that this option is extremely sensitive to numerical errors, which include time steps, SCF convergence and
alike.
If Car-Parrinello molecular dynamics is requested, the default time step is 0.05 fs.
4.8.5
Keyword MDBATH
This keyword specifies the temperature bath used in the molecular dynamic (MD) simulation.
Options:
NONE / SCALING / BERENDSEN / LOCAL
NONE
No temperature bath. This is the default.
SCALING
The velocities are scaled in order to control the temperature. This
option is only meaningful for very large systems of for the first
equilibration steps of a trajectory.
BERENDSEN
Berendsen thermostat is applied to the system as a whole.
LOCAL
Berendsen thermostat is applied individually to each atom.
VAL=<Integer>
SCALING: number of steps between velocity resets.
VAL=<Real>
BERENDSEN or LOCAL: Thermostat time constant τ in picoseconds
[ps]. Default is τ = 0.5 ps.
Description:
Free-running MD equations generate trajectories with constant energy, to within the
numerical error of the time evolution algorithm (velocity Verlet in deMon) and of the
energy gradients. In other words, molecular dynamics simulation is performed for the microcanonical (N,V,E) ensemble. However, it is often of interest to perform simulations at
constant temperature, which requires modifications to the standard equations of motion.
There are a number of different approaches for performing constant temperature MD.
In deMon, the temperature can be controlled by scaling the velocities, i.e. at each time
step the velocities are scaled according to v ′ = kv. The three temperature control methods, implemented in deMon, differ in how this constant k is calculated.
With the SCALING thermostat, the velocities of all atoms are scaled (by the same factor),
so that the instantaneous temperature Tkin matches desired temperature T0 set by the
MDTEMP keyword (see Section 4.8.3). In this case k is simply:
k=
s
T0
Tkin
(4.12)
74
4 Keywords
If the Berendsen thermostat [92] is used the scaling factor k is calculated using the Berendsen parameter τ and the time step, desired temperature and instantaneous temperature
from the dynamics.
Because rescaling introduces a large perturbation in system’s dynamics, MDBATH SCALING
should only be used for very rough, initial equilibration of the system.
With BERENDSEN thermostat [92], the scaling coefficient is calculated as:
k=
s
δt T0
−1
1+
τ Tkin
(4.13)
where δt is the MD time step. That is, the average temperature in this case will gradually
approach the desired simulation temperature, with a characteristic time constant τ .
Finally, MDBATH LOCAL invokes local version of the Berendsen thermostat. In this case,
scaling coefficient k is still given by eq. 13. However, it is now calculated per-atom,
from the individual atomic kinetic energies. As a result, local Berendsen thermostat will
approach equilibrium faster that the global version, at the cost of introducing additional
dynamical aftefacts in the system.
Typically, a molecular dynamics simulation would use SCALING thermostat for the initial
equilibration, then complete the equilibration with local Berendsen thermostat. The production runs, used to collect statistical information on the system then will be done using
either free-running dynamics, or global Berendsen thermostat with a large time constant
τ . In a production run, Berendsen τ must be large, compared to the characteristic time
of the process of interest.
For MD runs where the initial configuration is already well equilibrated (eg taken from
a previous MD run using an empirical forcefield), the equilibration should start with the
LOCAL thermostat, and omit SCALING altogether.
4.8.6
Keyword CONSERVE
The keywords requests imposition of constraints on some or all mechanical constants of
the overall motion of the system.
Options:
NONE / POSITION / MOMENTUM / ANGULAR / ALL
4.8 MD Control
NONE
75
Do not impose constraints. This is the default, except for MDYNAMICS
RANDOM.
POSITION
Reset position of the centre of mass of the system to (0,0,0) on each
MD step.
MOMENTUM
Reset overall momentum of motiom of the system to zero on each
MD step.
ANGULAR
Reset overall angular momentum of the system to zero on each MD
step. This constraint makes no sense, and is ignored, for periodic
calculations.
ALL
Equivalent to POSITION MOMENTUM ANGULAR. This is the default for
MDYNAMICS RANDOM
Description:
For infinitely accurate arithmetics, and an infinitely small time step, overall constants
of motion must be conserved automatically. Unfortunately, in the presence of numerical
noise and residual errors in the time evolution algorithms, mechanical constants of motion
are no longer conserved in practice. For example, it is easy to show that the residual error
of the velocity Verlet algorithm (used in deMon) with lead to an unbound amplification
of the angular momentum of a diatomic molecule.
Additionally, in many cases, we are interested in the time evolution of a system which is
at rest with respect to the laboratory frame. However, constructing an initial guess with
zero initial momenta may be cumbersome.
Therefore, the CONSERVE keyword provides a band-aid for both problems. deMon keeps
track of the total energy loss due to these constraints.
4.8.7
Keyword MDCONSTRAINTS
This keyword imposes atomic constraints during a molecular dynamics simulation.
Description:
Atomic positions or separate X, Y, Z Cartesian coordinates of atoms can be constrained,
using the following syntax:
76
or
4 Keywords
POSITION atom1
XYZ coordinates of atom1 are fixed
POSITION atom1, atom3
XYZ coordinates of atom1 and atom3 are
fixed
or
POSITION atom1-atom3
XYZ coordinates of atom1 through atom3
are fixed
or
Γ atom1, atom3
Γ, i.e. X or Y or Z coordinate of atom1 and
atom3 fixed
or
Γ atom1-atom3
Γ, i.e. X or Y or Z coordinate of atom1
through atom3 are fixed
The default is no constraints.
Other constraint types will be added in the future.
4.8.8
Keyword MDPRESSURE
Activates constant-pressure molecular dynamics. Options:
P=
External pressure in MPa. The default is 0.1
TAU=
Berendsen Coupling constant in ps. The default is 1.0.
COMP=
Compressibility of the system in 1/MPa. The default is set to be
reasonable for water.
Description:
This keyword requires periodic boundary conditions to be active (see section 4.11.6). Because periodic boundary conditions are currently implemented only for MM and QM/MM
runs (see section 4.5.1), constant-pressure dynamics necessarily implies either MM or
QM/MM.
With MDPRESSURE on, deMon will adjust the size of the periodic box to maintain the desired pressure. The periodic box parameters are controlled as described for the Berendsen
thermostat (see eq. 13 in section 4.8.5). The corresponding time constant is defined by the
TAU= parameter. Practical implementation of the Berendsen pressure bath also requires
an estmation of the compressibility of the system, which is specified with the COMP= keyword. This value need not to be exact, as it enters only the calculation of the coordinare
scaling factor for coordinates.
Compressibilities of some representative systems are:
4.8 MD Control
77
System
gases at 0.1 Mpa
water
solids
Compressibility, MPa−1
≈ 10
≈ 500 × 10−6
≈ 1-20 × 10−6
Warning: Constand-pressure molecular dynamics only makes sense if the system is sufficiently large (thousands of particles) and/or time scales are reasonably long (tens of
picoseconds).
4.8.9
Keyword HEATPIPE
Activates calculation of heat conductivity, using non-equilibrium heatpipe molecular dynamics. Options:
LENGTH=
MASKS=
If periodic boundary conditions are not used, sets the length of the
heat-transfer axis, in ˚
Angstrom. Not allowed in PBC simulations.
Number of temperature probes in the simulation. Must be at least
four (see below).
GUESS=
Expected order of magnitude of heat conductivity, in Watts/meterKelvin. The default (0.1 W/m-K) is appropriate for amorthous
solids and liquids.
FLUX=
Desired heat flux between the cold and the hot zones, in Hartree
per time step. It is usually safer to use GUESS= instead.
AXIS=
Orientation of the heat transfer axis. Only three choices are allowed for periodic calculations: [AB], [AC], and [BC], choosing a
direction normal to one of the sides of the periodic box. For nonperiodic calculations, Cartesian direction must be provided, in the
form [dx:dy:dz].
BASE=
Position of the heat transfer axis’ origin, which corresponds to the
centre of zone 1. Can be specified as an integer (giving the atom
located at the origin) or as Cartesian coordinates in the format
[x:y:z].
AREA=
If PBCs are not used, area of the section through which heat flows,
˚ngstrom squared. Not allowed in PBC simulations.
A
78
4 Keywords
n
1
2
3
n−2
n−1
n
1
....
Cold
Base
Hot
Length (periodic)
Length (non−periodic)
Figure 4: Set up of heatpipe zone masks
COLD=
Index of the cold zone of the heatpipe. Must be between 1 and
number of masks.
HOT=
Index of the hot zone of the heatpipe. Must be between 1 and
number of masks. Hot and cold zones must not overlap.
OFF
Description:
Turn the heatpipe off.
The structure of this keyword is rather complex, which matches the complex nature of the
numerical experiment. The essential parameter of the simulation is the direction of the
heat transfer, which is established by the AXIS= keyword. In a non-periodic simulation,
any direction of the heat transfer is allowed (although not every direction is sensible!).
If periodic boundary conditions are used, there exist only three heat transfer directions,
which do not break the periodicity of the system. Namely, directions which are perpendicular to one of the sides of the simulation box ([AB], [AC], or [BC]) are allowed. If you
are interested in heat conductivity along a different crystallographic direction, you will
have to choose an appropriate supercell.
In the periodic case, the unit cell and the direction of heat transfer also establish the
length of the heat transfer axis and the crossection for the heat flow. In the non-periodic
case, these parameters have to be specified using LENGTH= and AREA= keywords.
Once the direction of heat transfer is chosen, deMon will set up a set of overlapping
temperature probes (called zones). The zones are infinite slabs in space, oriented perpendicular to the heat transfer direction. The number of zones is specified using the MASKS=
4.8 MD Control
79
keyword. Each zone is associated with a masking function (Fig. 4). Masking functions
are unit at the zone centre, and decrease to zero at the zone edges. Two of the zones
also serve as heat source and sink for the heatpipe. These zones are specifies by HOT= and
COLD= keywords. In the absence of the source and sink specification, deMon will place the
cold zone in position 1. The hot zone will then be placed as far away as possible, taking
periodicity of the system into account.
On each time step, the heat pipe transfers a constant amount of kinetic energy from the
cold zone to the hot zone, by scaling velocities of particles:
v~′ i = (1 + αc fc (~ri ) + αh fh (~ri )) ~vi + (αc fc (~ri ) + αh fh (~ri )) ~v0
(4.14)
where fc (~r) and fh (~r) are masking functions of the hot and cold zones. Parameters αc ,
αh , and ~v0 are chosen such that the total energy and total momentum of the system are
conserved.
The amount of energy (in Hartrees) transfered per time step can be specified using the
FLUX= keyword. Usually, it is more convenient to let deMon to choose the flux which will
lead to the difference of ≈10K between the hot and the cold zones. In addition to the heat
transfer length and cross-section, estimation of the necessary heat flux requires a guess
of the heat conductivity of the system. It can be specified (in Watts per meter-Kelvin)
using GUESS= keyword. The default is 0.1 W/m-K.
Heat conductivities of some representative system are:
System
Argon gas, 300K, 0.1MPa
Argon liquid, 85K, 0.1MPa
Argon solid, 8K
NaCl, 373K
NaCl, 20K
Asbestos, 273K
Copper solid
Diamond, 300K
Graphite, 300K
Heat conductivity, W/m-K
≈ 0.02
≈ 0.13
≈6
≈5
≈ 300
≈ 0.09
≈ 400
≈ 900 − 2300
≈ 2000 (in-plane)
≈ 6 (across planes)
(Note that the heatpipe approach implemented in deMon calculates only the lattice
contribution to heat conductivity).
Once the the heatpipe simulation reached the steady state (which may require 104 –105
time steps), heat conductivity λ can be calculated by fitting the phenomenological heat
80
4 Keywords
transfer equation:
dE
S
= λ ∆T
(4.15)
dτ
l
to the observed temperature profile. The resulting heat conductivity may be influenced
by the errors in temperature measurement. These errors decrease with increasing number
of particles per slab and for longer simulation times. Larger values of heat flux will also
reduce the impact of the uncertainties in temperature measurements (but see below).
Another potential source of errors is the finite width of the zones, which introduces the
uncertaintly in the length of heat transport. This error decreases with increasing number
of zones. It is suggested that zones should contain at least 20-100 particles each, using at
least 6-8 zones.
Yet another potential source of errors is fractionation of the system and non-linear heat
conductivity effects, induced by excessive heat flux. To guard against this possibility,
one should perform several heat conductivity runs, with thermal fluxes varying over 1-2
orders of magnitude.
Meaningful calculation of heat conductivities requires that the overall position and orientation of the system remain constant over the simulation time. Overall momentum of
the system must be zero, to eliminate the mass flow contribution. These constraints can
be enforced by using the CONSERVE keyword (see section 4.8.6).
Heatpipe dynamics is not compatible with constant-pressure dynamics (MDPRESSURE section 4.8.8). The only allowed thermostat is global Berendsen (MDBATH - section 4.8.5).
If possible, heatpipe dynamics should be performed with the thermostat off (MDBATH
NONE).
4.8.10
Keyword CARPAR
Activates a Car-Parrinello Molecular Dynamics simulation. This option is only available
for DFTB. By default, this option is off.
Options:
FOM=
Ficticious orbital mass for Car-Parrinello DFTB (in atomic units).
The default is 25 ∆t2 (in atomic units). The higher the value, the
stronger are the deviations from the Born-Oppenheimer surface,
and also the electronic vibrations get larger periods (so large values
allow larger time steps).
4.8 MD Control
LGTOL=
81
The requested tolerance to ensure orthonormality of the orbitals
using a Lagrange multiplier. The default is 10−10 in atomic units.
LGPROP
Requests propagation of the Lagrange multiplier matrix. This computational step might be time demanding, so extrapolating the multiplier matrix might be of advantage. The performance of this option needs to be tested for each individual system. The default is
false.
BO
The Car-Parrinello approach is used to reach the BornOppenheimer surface at each point. For this purpose, the positions
of the atoms are frozen until the electronic system reaches a stationary point. This option might be useful for very large system, as
it avoids diagonalisation.
DELMO=
Multiplier for MO energy gradient for the MO optimisation using
the Car-Parrinello approach in conjunction with the steepest descent technique (requires CARPAR BO). The default is 0.1.
MOMD
Requests a molecular dynamics-like optimisation of the molecular
orbitals: A Car-Parrinello MD is carried out. Whenever gradient
and velocity of an orbital have opposite signs an additional friction
term is added to the equations of motion.
MOMD=
Same as the MOMD keyword, but specifying the friction parameter.
The default is 10−2 .
Description: deMon-DFTB is able to perform Car-Parrinello MD simulations. The CarParrinello technique (see ref. [121]) is interesting for systems with band (HOMO-LUMO)
gap, with protons, and, in particular, if SCC-DFTB is carried out. In this case, CPMD
may significantly improve the performance.
The Car-Parrinello approach does not require SCC cycles, and avoids the diagonalisation
of the secular equations. By these means, the scalability of the method is much superior compared to Born-Oppenheimer molecular dynamics. Note that this is still a young
functionality of deMon-nano.
82
4.9
4.9.1
4 Keywords
Property Control
Keyword POPULATION
This keyword requests population analysis. In the absence of the POPULATION keyword,
the default is to skip population analysis.
Options:
MULLIKEN / LOEWDIN / BADER
MULLIKEN
A Mulliken population analysis is performed. This is the default.
LOEWDIN
A L¨owdin population analysis is performed.
BADER
A Bader population analysis is performed (unimplemented).
Description:
In all three cases, deMon calculates atomic charges and, in the case of open-shell systems,
atomic spin charges. For the Mulliken [93] and L¨owdin [94] population analysis, the bond
order [95] or valence matrix [96] are calculated as well (closed-shell systems only). In the
case of the population analysis of Bader [97] critical points of the electronic density and
the molecular graph are calculated. These quantities can be visualized with Vu (see 8)
using the deMon.pie file.
4.9.2
Keyword DIPOLE
This keyword activates the calculation of molecular electrostatic moments.
Description:
With the DIPOLE keyword the calculation of the electrostatic dipole, quadrupole and
octupole moment is activated. If higher moments are required the parameter MAXMOM
in the parameter.h file has to be adjusted. The electrostatic moments are calculated from
the orbital density (see Section 1.4) independent from the chosen energy approximation.
If the electrostatic moments should be calculated in STANDARD ORIENTATION (see
Section 4.1.4 for the definition) a single point calculation with the symmetry keyword
activated (SYMMETRY ON) has to be performed.
4.9.3
Keyword MAGOUT
This keyword requests writing of the interface files for the NMR/EPR program MAG.
Options:
4.9 Property Control
83
MASTER / MAG
MASTER
Produce output for MASTER NMR programs (see section ??).
MAG
Description:
Produce output for MAG NMR program (see section 9).
If MAGOUT MASTER is given, four interface files, deMon.nmr11, deMon.nmr70, deMon.nmr71
and deMon.nmr72 are written. These are data files for the MASTER program (see Section ??. The file deMon.nmr11 contains the grid information (points and weights). The
other files, deMon.nmr70, deMon.nmr71 and deMon.nmr72 contain dimension, pointer and
molecular (coordinates, basis, etc.) informations, respectively. In the case of an EPR calculation the file deMon.nmr11 is not produced. It should be noted that for NMR and
EPR calculations the angular momentum for the auxiliary functions is limited to l = 2.
If MAGOUT MAG is given, deMon will create two interface files, deMon.nmr11 and deMon.nmr50,
in the format expected by the MAG program (see Section 9). Note that the version of
the MAG program, included in this distribution, lacks some of the key features due to licensing restrictions on it’s redistribution. An up-to-date version of the MAG code should
be requested directly from Vladimir Malkin.
The use of the MASTER and MAG programs is described in Sections ?? and 9.
4.9.4
Keyword POLARIZABILITY
This keyword activates calculation of the polarizabilities and hyperpolarizabilities.
Options:
ALPHA / BETA / GAMMA
ALPHA
The polarizability tensor α is calculated. This is the default.
BETA
The first hyperpolarizability β is calculated.
GAMMA
The second hyperpolarizability γ is calculated.
FFS=<Real>
Finite field strength used in the polarizability calculation.
EFISH
Activate the EFISH orientation.
Description:
The polarizabilities are calculated by the finite field method [98] using field strengths of
0.003 a.u. for α and β and 0.01 a.u. for γ. The field value can be modified by the option
FFS. The hyperpolarizabilities β and γ are always calculated in the so-called EFISH
(Electric Field Induced Second Harmonic generation) orientation. In this orientation the
84
4 Keywords
z axis of the molecule is oriented along the permanent dipole moment of the system. The
mean first and second hyperpolarizabilities are defined in EFISH orientation as:
3X
βiiz
(4.16)
β =
5 i
1X
γ =
γiijj
(4.17)
5 i,j
For the calculation of the polarizability tensor α no special orientation is used. With the
option EFISH the EFISH orientation can be enforced. The mean polarizability α and
polarizability anisotropy |∆α|2 are calculate as:
1
(αxx + αyy + αzz )
(4.18)
α =
3
3tr α2 − (tr α)2
|∆α|2 =
(4.19)
2
Requesting β or higher polarizabilities changes default for the grid generating function
to GUESS (See section 4.4.10).
4.9.5
Keyword FREQUENCY
This keyword activates frequency analysis.
Options:
RAMAN
The Raman intensities are calculated.
RESTART
The frequency analysis is restarted from the deMon.rst file.
VIB=<Real>
Scaling factor for the numerical step size.
Description:
By default only the infra-red adsorption intensities of the harmonic normal modes are
calculated, using the dipolar approximation. With the option RAMAN, the Raman intensities (in atomic units) and the depolarization ratio [99] are calculated, too. This is
considerably more time consuming conpared to the standard frequency analysis. The
RESTART option permits the restart of a frequency analysis. The Hessian matrix elements already calculated are read from the restart file deMon.rst and the analysis is
continued. The VIB option allows adjustments to the numerical step size, used for differentiation of the ananytical gradients. By default, step size of 0.001 Bohr is used for
VXCTYP BASIS (see section 4.4.9). If VXCTYP AUXIS is used, the default displacement is
changed to 0.025 Bohr. For examples, with VIB=2 (and VXCTYP BASIS) the step size
becomes 0.01 Bohr.
4.9 Property Control
4.9.6
85
Keyword THERMO
This keyword activates the calculation of thermodynamic functions. The THERMO keyword can only be used in combination with the FREQUENCY keyword (see Section
4.9.5).
Options:
MAX=<Real>
Maximum temperature. The default is 2000 K.
MIN=<Real>
Minimum temperature. The default is 100 K.
INT=<Real>
Temperature interval. The default is 100K.
Description:
The thermodynamic functions are calculated in the approximation of the ideal polyatomic
gas [100] and harmonic vibrational approximation. For diatomic molecules a ro-vibronic
correction term is included (see code). The electronic contributions are neglected, apart
from the trivial contribution due to the multiplicity of the ground electronic state. The
options MAX and MIN allow specification of the temperature range. The step size of the
temperature interval is defined by the option INT. The output contains the heat capacities
cp , the entropy S, the enthalpy H and the inner energy U for each temperature. Also
the natural logarithm of the partition function z = ztrans zrot zvib zel is given. Imaginary
normal modes are not included in the summation.
4.9.7
Keyword FNMC
Requests finite nuclear mass correction. Options:
OFF / ON
Description:
With this keyword, the finite nuclear mass correction to the self-atomic elements of the
kinetic energy matrix are switched on. Before you use this you should read the literature,
e.g. ref. [101].
4.9.8
Keyword HARDNESS
Compute orbital-dependent reactivity indices hardness, softness and Fukui function.
Options:
86
4 Keywords
DELN=
Energy perturbation (in electrons). The defalt is 0.01
FPMO=
First perturbed molecular orbital. The default is 1
TOL=
Description:
Tolerance for degeneracy detection (in Hartree). The default is 10−5
Compute orbital-dependent reactivity indices hardness, softness and Fukui function, as
given in a supplied preprint by Mineva and Heine.
4.10 Solvent Effects
4.10
87
Solvent Effects
WARNING: THE IMPLEMENTATION OF SOLVENT EFFECTS IS NOT
FINAL, AND MAY BE SIGNIFICANTLY MODIFIED OR EVEN
REMOVED IN A FUTURE RELEASE
4.10.1
Keyword SOLVENT
This keyword activates calculation of the solvent effects.
Options:
ONSAGER / IT-PCM / CLS-PCM / BEM-PCM
ONSAGER
The calculations are performed using the Onsager
Model.
IT-PCM
The calculations are performed using the Polarizable
Continuum Model (PCM) (Iterative Procedure).
CLS-PCM
The calculations are performed using the PCM (Partial
Closure Approximation).
BEM-PCM
The calculations are performed using the PCM (Boundary Element Method).
BONDI
Bondi’s atomic van der Waals radii.
MERZ
Merz-Kollman’s atomic van der Waals radii.
˚ngstr¨oms (only
Value of the spherical cavity radius in A
RAD=
for the Onsager Model).
SCALR=
The radius of each sphere is determined by multiplying
the van der Waals radius by a scaling factor.
DIPOLE
Dipolar perturbative reaction field.
QUADRUPOLE
Dipolar and Quadrupolar perturbative reaction field.
OCTUPOLE
Dipolar, Quadrupolar and Octupolar perturbative reaction field.
CLS=
ITER
1
First level of truncation in CLSPCM procedure.
2
Second level of truncation in CLSPCM procedure.
Self-Consistent
procedure.
Reaction
Field
(SCRF)
iterative
88
4 Keywords
TOL=
Tolerance for SCRF iterative procedure (only for Onsager Model).
CAVITATION
The cavitation energy contribution is calculated.
DISPERSION
The dispersion and repulsion energy contributions are
calculated.
COMP=
0
No charge compensation.
1
The difference between calculated and theoretical polarized total charge is distributed on each tessera proportionally to its area.
2
The calculated charge on each tessera is scaled by a
costant factor.
SOLV=
Solvent, one of: WATER, H2O, DMSO, DIMETHYLSULFOXIDE, NITROMETHANE, METHANOL, CH3OH,
ETHANOL, CH3CH2OH, ACETONE, CH3COCH3,
DICHLOROETHANE,
CH2CH2CL2,
DICHLOROMETHANE, CH2CL2, THF, TETRAHYDROFURAN, ANILINE, CHLOROBENZENE, CHLOROFORM,
CHCL3,
DIETHYLETHER, ETHER, TOLUENE, BENZENE,
C6H6, CARBONTETRACHLORIDE, CCL4, CYCLOHEXANE, HEPTANE, ACETONITRILE
4.10 Solvent Effects
89
EPS=
Value of dielectric constant.
DIRECT
The PCM integral blocks are calculated at each SCF
cycle.
Sets the minimum radius in ˚
Angstr¨oms of the added
RMIN=
spheres to build the cavity.
TSNUM=
Description:
Number of tesserae on each sphere.
For the generation of the cavity the PCM and the Onsager Model require the definition of
the atomic radii. For the PCM the van der Waals radii can be used to build up the cavity
putting a sphere around each atom. The cavity is constructed by overlapping spheres with
radii equal to the van der Waals radii. For the Onsager Model the solute is enclosed in
a spherical cavity whose radius, by default, is calculated as maximum distance between
atoms of the solute plus the respective BONDI van der Waals radii. The value of the
spherical cavity radius can be modified by the option RAD=value.
The options DIPOLE, QUADRUPOLE, and OCTUPOLE are valid only for Onsager
Model and define the level of the expansion of the potential. Basically, only the core
Hamiltonian is modified as:
0
Hµν = Hµν
+
3
P
k
fkl hφµ xk iφν +
1
2
3
P
kn
l
fkn
hφµ xk xn iφν +
1
2
3
P
knp
l
fknp
hφµ xk xn xp iφν
l
l
Where Hµν is the core Hamiltonian matrix, fkl , fkn
and fknp
are the dipolar, quadrupolar
and octupolar reaction field factors, and hφµ xk iφν , hφµ xk xn iφν and hφµ xk xn xp iφν are the
dipolar, quadrupolar and octupolar integral blocks, respectively.
The Partial Closure procedure (CLS keyword) permits to avoid the iterative self-polarization
internal cycles to calculate the virtual charge over the surface of each tessera. One of the
two levels of truncation can be selected by the option CLS=1 or CLS=2. By default the
first level of truncation approach is switched on.
An iterative SCRF procedure can be performed, by specifying the IT keyword. For the
Polarizable Continuum Model the electronic component of the electric field and the apparent surface charge :
el(I)
Ek
=
P
µν
h
1
(I)
iφν
Pµν
hφµ |r−(rk1+δk )| iφν − hφµ |r−r
k|
h
el(I)
Ek
σ (I,0) = − ε−1
4πε
+ EkN
i
i
1
δk
(I)
are calculated at each (I) Self-Consistent Field (SCF) cycle by the Pµν
density matrix. If
90
4 Keywords
this option is not active, the electronic component of the electric field and the apparent
(0)
surface charge are calculated once by the gas phase density matrix Pµν
.
For the Onsager Model more than one SCF convergence step is performed. Each step
yields a new set of reaction field factors to be used for the next SCF convergence step. This
iterative procedure is performed until energy convergence between two steps is achieved.
The tolerance value can be given by input. By default TOL=10*SCFTOL.
By default, CAVITATION and DISPERSION options are switched off, and only electrostatic energy contributions are calculated.
Formaly the bf COMP option is used to specify the charge compensation mode. By option
COMP=1 the virtual charge over each tessera is normalized as:
Ask
qkcomp = qk + (Qttot − Qtot ) As
tot
while, by the option COMP=2 as:
Qt
qkcomp = qk Qtot
tot
where qk is the virtual charge over the surface of each tessera before of the compensation,
Qttot is the theoretical virtual charge, Qtot is the total virtual charge over the cavity before
of the compensation, Ask is the area of each tessera and Astot is the area of the cavity
surface. The default option is COMP=1. For uncharged systems Qttot = 0. In this case, if
the option COMP=2 is selected for the Iterative and the Partial Closure procedures, it
will be automatically set equal to 1.
The DIRECT option can be activated with the keywords ITPCM and CLSPCM. The
DIRECT procedure avoids the disk I/O by calculating at each SCF cycle the PCM
integral blocks. This procedure is recommended for large molecular systems and should
be carefully selected.
The solvent dielectric constant value can be given by input with the EPS keyword. For
PCM calculations the program will automatically check if the corresponding solvent is
available in the database.
If the RMIN value is increased, the number of the added spheres will be decreased.
RMIN=100 inhibits the creation of the added spheres which is recommended for large
and complicated molecular systems. The default value is RMIN=0.2
The increase of the number of tesserae for each sphere is not recommended for large molecular system except if DIRECT procedure is selected. The default value is TSNUM=60.
4.10 Solvent Effects
91
Figure 5: Representation of the benzene cavity surface.
As an example, in Figure 1 the cavity surface of the benzene using 1000 tesserae for each
sphere is shown.
4.10.2
Keyword PCM-CARD
In the keyword body PCM-CARD the radius of each initial sphere (RAD), which initially
is centered at the corresponding atom, a scaling factor (SCALRAD) and an order number
(NORD) can be read. Each column of values is identified by the corresponding string. So
that, under RAD, for example, the values of the radii must be tabulated. As an example,
the PCM-CARD of H2 O could be:
PCM-CARD
RAD
1.52
1.10
1.10
SCALRAD
1.00
1.00
1.20
NORD
1
2
3
The numbers under NORD coincide with the position number of the atoms in the Z-
92
4 Keywords
MATRIX or CARTESIAN input. If one of the columns is suppressed in the PCM-CARD,
the corresponding values are set to be equal to the default ones:
[1)] If the column ORD is suppressed the order numbers are determined
by the Z-MATRIX or CARTESIAN input coordinates.
If the RAD column is suppressed the radii of the spheres are determined
by the option BONDI or MERZKOLLMAN.
3) If the SCALRAD column is suppressed the scaling factors are set equal
2)
to 1.0.
In this way, only the parameters of the desired spheres can be modified:
NORD
2
RAD
1.10
SCALRAD
1.20
4.11 Visualization and Topology
4.11
93
Visualization and Topology
The keywords PLOT, ISOSURFACE and GEOSURFACE of this section are exclusive
and cannot be combined. The same holds for the keywords BOX and POINTS.
4.11.1
Keyword VISUALIZATION
This keyword activates the interface to the visualization programs MOLDEN [102] and
MOLEKEL [103]. Options:
MOLDEN / MOLEKEL
MOLDEN
The MOLDEN interface is activated. The MOLDEN input is
written to the file deMon.mol. This is the default.
MOLEKEL
The MOLEKEL interface is activated. The MOLEKEL input
is written to the file deMon.mkl.
XYZ / OPT / FULL
XYZ
Only the molecular structures and energies of each optimization
step are written to the deMon.mol file. This is the default if the
option MOLDEN or MOLEKEL is missing.
OPT
The molecular structures as well as the forces and step sizes of
each optimization step are written to the deMon.mol file. This
option triggers the MOLDEN interface.
FULL
A full MOLDEN or MOLEKEL input is written. This is the
default if the option MOLDEN or MOLEKEL is set.
Description:
For large systems the FULL option of the VISUALIZATION keyword produces very
large MOLDEN and MOLEKEL input files. Therefore, it is recommended to use the VU
interface for the plotting of molecular fields of large systems (see the keyword PLOT and
ISOSURFACE below). In the case of a frequency analysis the FULL option is activated
by default in order to animate the vibrations.
For molecular dynamics runs, only VISUALIZATION MOLDEN XYZ is supported. This option
will be used, regardless of the parameters supplied here.
94
4.11.2
4 Keywords
Keyword PLOT
This keyword controls the calculation and plotting of molecular fields. See the keywords
BOX (4.11.6) and POINTS (4.11.7) for the definition of the plot support.
Options:
hFIELDi
Molecular field specification. The available field acronyms are
given in Table 9. This option is mandatory!
BASIS / AUXIS
BASIS
The orbital density is used for the construction of the plot
function. This is the default.
AUXIS
The auxiliary function density is used for the construction of
the plot function. This option is incompatible with the
READ option.
BINARY / ASCII / TABLE
BINARY
A binary file FIELD.bin is written containing the coordinates
and plot function values. The VU file format is used. The VU
control file deMon.pie is written, too. This is the default.
ASCII
An ascii file FIELD.asc is written containing the coordinates
and plot function values.
TABLE
A function table is written in the output file deMon.out.
READ
Specifies that an orbital list is read in the keyword body of
PLOT. This keyword is incompatible with the EFIDS field.
Description:
The definition of the molecular field is mandatory for the PLOT keyword. The available
molecular fields are listed in Table 9. If the BOX and POINTS keywords are missing
the default BOX setting is used for the plot. Only one molecular field at a time can
be calculated. If several fields have to be calculated restart calculations with SCFTYPE
MAX=0 and GUESS RESTART (see 4.4.1 and 4.4.4) may be performed. In any case,
no more than 20 molecular field entries at a time can be calculated. A field with its first
and second derivatives counts for 10 entries (1 + 3 + 6).
4.11 Visualization and Topology
95
The AUXIS and BASIS option specify the density (auxiliary function or orbital density,
respectively) used for the calculation of the molecular field. The AUXIS option is incompatible with the READ option. If the auxiliary function density is used for the calculation
of the molecular field a considerable speed up will be achieved. However, the field values
may show deviations with up to 10% compared to the orbital density. Nevertheless, the
field topology is in most cases qualitatively correct.
By default, the PLOT keyword produces two additional output files, the binary file
FIELD.bin and the VU control file deMon.pie. With these files a VU session can be
started (see Section 8). With the ASCII option only the file FIELD.asc is created. The
coordinates and molecular field values are listed in this file. With the TABLE option a
function table is created in the output file.
With the READ option a set of molecular orbitals for the calculation of the molecular
field can be selected. This option cannot be used for the EFIDS field. In the case of the
PSI and D1PSI field the READ option specifies directly the calculated molecular orbitals.
Again, not more than 20 orbitals at a time can be calculated (in the case of D1PSI not
more than 5). If for the other fields orbitals are selected than the field is constructed from
the sum of the selected orbitals. E.g. the following input specifies the calculation of the
density using the molecular orbitals 1, 2 and 4:
PLOT RHO READ
1 2 4
Virtual orbitals can be included. They will be occupied by one electron. In the case of an
unrestricted calculations (UKS option of keyword SCFTYPE; see 4.4.1) α and β-orbitals
can be selected, each set in a separate input line. In the following example the α-orbitals
1, 2, 3, 4, 5, 6, 7 and the β-orbitals 1, 2, 3, 4, 5, 6 are selected for the calculation of the
molecular electrostatic potential:
PLOT ESP READ
1 2 3 4 5 6 7
1 2 3 4 5 6
If only the β-orbitals should be selected the first line in the above example must have a
0 entry:
PLOT ESP READ
0
1 2 3 4 5 6
96
4 Keywords
Table 9: Molecular fields available in deMon
Acronym
PSI
Description
Molecular orbitals. By default the five highest occupied and lowest
unoccupied orbitals are calculated.
D1PSI
Molecular orbitals and first derivatives. By default the HOMO
and LUMO are calculated
RHO
Electron density. By default the density of the occupied orbitals
is calculated.
D1RHO Electron density and first electronic derivatives. By default the density
and its derivatives of the occupied molecular orbitals are calculated.
D2RHO Electron density, first and second electronic derivatives. By default the
density and its derivatives of the occupied orbitals are calculated.
SPIN
Spin density. By default the spin density of the occupied molecular
orbitals is calculated.
D1SPIN Spin density and first electronic derivatives. By default the spin density
and its derivatives of the occupied molecular orbitals are calculated.
LAP
Electron density Laplacian. By default the Laplacian of the occupied
molecular orbitals is calculated.
ESP
Molecular electrostatic potential. By default the MEP of the occupied
molecular orbitals is calculated. A minus sign in front indicates that
only the electronic part is calculated.
D1ESP Molecular electrostatic potential and its first derivatives. By default
the MEP and its derivatives of the occupied orbitals are calculated.
D2ESP Molecular electrostatic potential and its first and second derivatives.
By default the MEP and its derivatives of the occupied orbitals
are calculated.
EFI
Molecular electric field. By default the electric field of the occupied
molecular orbitals is calculated.
D1EFI
Molecular electric field and its first derivatives. By default the electric
field and its derivatives of the occupied orbitals are calculated
EFG
Molecular electric field gradients. Identical to D1EFI.
EFIDS
Anion surface generation [104].
ELF
Electron localization function [105,106].
4.11.3
Keyword CPSEARCH
This keyword activates the critical point search of scalar molecular fields. For the definition of the search area see the keywords BOX (4.11.6) and POINTS (4.11.7).
Options:
4.11 Visualization and Topology
hFIELDi
97
Scalar molecular field specification. The available field acronyms
are RHO and ESP. See Table 9 for the acronym meanings. This
option is mandatory!
BASIS / AUXIS
BASIS
The orbital density is used for the calculation of the scalar
molecular field. This is the default.
AUXIS
The auxiliary function density is used for the calculation of
the scalar molecular field.
Description:
The critical points of the density (RHO) and molecular electrostatic potential (ESP)
can be searched with this option. By default, the critical point search is performed in
a marching cube like style [107], embedding the molecule in a rectangular box (BOX
LARGE; see Section 4.11.6), which is in turn divided into subboxes. Each subbox is then
searched for a critical point [108]. The box size and shape can be manipulated by the BOX
keyword. As an alternative, start points for the critical point search can be supplied by the
POINTS keyword (see Section 4.11.7). With the POINTS option POLYGON start points
between atom pairs and triples are automatically generated. This option is recommended
for the critical point search of the density. In addition to the critical point search the
molecular connectivity over (3,-1) critical points is generated [97].
EFFECT OF AUXIS ON THE CPS -> SIGFRIDO
4.11.4
Keyword ISOSURFACE
This keyword controls the calculation and plotting of molecular field isosurfaces. Only
one isosurface at a time can be generated. For the generation of multiple isosurfaces the
restart description of the keyword PLOT (Section 4.11.2) can be applied. See the keyword
BOX (4.11.6) for the definition of the isosurface boundary.
Options:
hFIELDi
Molecular field specification. The available field acronyms are
PSI, RHO, SPIN, LAP, ESP, EFIDS and ELF. See Table 9
for the acronym meanings. This option is mandatory!
BASIS / AUXIS
98
BASIS
4 Keywords
The orbital density is used for the construction of the isosurface.
This is the default.
AUXIS
The auxiliary function density is used for the construction of the
isosurface. This option is incompatible with the READ option.
LINEAR / BILINEAR / LOGARITHMIC
LINEAR
Linear interpolation scheme for the isosurface construction.
This is the default.
LOGARITHMIC
Logarithmic interpolation scheme for the isosurface construction
BINARY / ASCII / TABLE
BINARY
A binary output of the isosurface is written in the file LAT.bin
using the VU file format. The VU control file deMon.pie is
written, too. This is the default.
ASCII
An ascii output of the isosurface is written in the file deMon.lat.
TABLE
A function table of isosurface coordinates is written in the output
file deMon.out.
READ
Specifies that an orbital list is read in the body of ISOSURFACE.
ISO=<Real>
Isosurface value. This value is mandatory!
TOL=<Real>
Tolerance for data reduction.
Description:
The options AUXIS, BASIS and READ of the ISOSURFACE keyword are identical to
the corresponding options of the PLOT keyword (4.11.2). The BINARY option specifies
that the isosurface coordinates and connectivities are written into the binary file LAT.bin.
With the VU control file deMon.pie, that is also generated, the isosurface grid can be
visualized in VU (see 8). It is important to note that the LAT.bin file can be used as
an input file for the plotting of a molecular field on the isosurface! For this purpose this
file has to be in the same directory as the deMon.inp file and the POINTS keyword (see
Section 4.11.7) has to be used to define the isosurface in LAT.bin as plot support. With
the ASCII option the ascii file deMon.lat is generated. This file contains the following
data:
RHO ( .100000E+00 A.U. ) ISOSURFACE COORDINATES IN ANGSTROM
4.11 Visualization and Topology
99
NUMBER OF VERTICES: 1814
NUMBER OF FACETS: 3624
VOLUME: .163076E+04
AREA: .176664E+03
X
.312196E+00
.243158E+00
.243158E+00
...
CONNECTIVITY
1
4
6
...
Y
-.729474E+00
-.765796E+00
-.729474E+00
...
2
5
7
...
Z
.105260E+01
.105260E+01
.917119E+00
...
3
6
8
...
The first line is the file header including the field information (RHO), the isovalue (0.1
a.u.) and the units (ANGSTROM) used for the coordinates, area and volume data. The
NUMBER OF VERTICES correspond to the number of interpolation points for the isosurface. The NUMBER OF FACETS is the number of surface elements of the isosurface.
The volume (here in ˚
A3 ) and the surface area (here in ˚
A2 ) are then given. It should be
noted that the volume includes all volume elements with field values under the given
threshold. Therefore, the volume of an 0.1 a.u. isosurface of a field with positive and
negative values includes all volume elements with positive field values less than 0.1 a.u.
and all volume elements with negative field values. The next block represents the coordi˚). The last block, entitled CONNECTIVITY,
nates of the interpolation points (here in A
describes the triangulation of the interpolation points. With the option TABLE the same
data as with the ASCII option are written into the file deMon.out.
The options LINEAR and LOGARITHMIC specify the interpolation scheme for the
construction of the isosurface using the marching tetrahedron algorithm [107]. The value
of the isosurface is specified with the ISO option. A value in vertical delimiters (e.g. |0.1|)
indicates an absolute isosurface value. It can be used to generated simultaneously two
isosurfaces with the same positive and negative threshold (e.g. for molecular orbitals). If
the threshold of the two isosurfaces should be different (e.g. for molecular electrostatic
potentials) the two values can be specified as:
ISO = 0.1/-0.05
The ISO option has no default setting and, therefore, is mandatory! With the TOL
100
4 Keywords
option the amount of data reduction is specified. Allowed values range from 10−3 to 0.5,
specifying the severity of the data reduction. A value of 10−3 indicates no data reduction
at all. This is the default. With increasing TOL values the data reduction increases, thus
decreasing the memory request and at the same time the resolution.
4.11.5
Keyword GEOSURFACE
This keyword controls the calculation and plotting of geometrical surfaces like spheres or
ellipsoids. See the keyword BOX (4.11.6) for the definition of the geosurface boundary.
Options:
SPHERE / ELLIPSOID
SPHERE
A sphere with the radius a around the origin ~ro is constructed.
ELLIPSOID
An ellipsoid with the semi-axes a, b and c around the origin ~ro
is constructed.
BINARY / ASCII / TABLE
BINARY
A binary output of the geometrical surface is written in the file
LAT.bin using the VU file format. The VU control file deMon.pie
is written, too. This is the default.
ASCII
An ascii output of the geometrical surface is written in the file
deMon.lat.
TABLE
A function table of the geometrical surface coordinates is written
in the output file deMon.out.
TOL=<Real>
Tolerance for data reduction.
Description:
The options BINARY, ASCII, TABLE and TOL are identical to the corresponding options of the ISOSURFACE keyword. The option SPHERE or ELLIPSOID is mandatory
for the GEOSURFACE keyword. The origin and radius or semi-axes of the sphere or
ellipsoid have to be defined in the keyword body of GEOSURFACE. For a sphere with
the origin at ~ro = (1, 3, 4) and a radius of 7 the input has the form:
GEOSURFACE SPHERE
4.11 Visualization and Topology
101
1.0 3.0 4.0
7.0
The units are defined by the unit definition in the GEOMETRY keyword (see Section
4.1.1). For an ellipsoid the following input form is used:
GEOSURFACE ELLIPSOID
1.0 3.0 4.0
7.0 1.0 0.5
The definition of all three semi-axes is mandatory.
4.11.6
Keyword BOX
With this keyword the box around a molecule for plotting (4.11.2), critical point search
(4.11.3) and iso- and geosurface generation (4.11.4 and 4.11.5) is defined.
Options:
STANDARD / SMALL / LARGE
STANDARD
A box enclosing all defined atoms with an extension of two times
there covalent radii. This is the default.
SMALL
A box enclosing all defined atoms with an extension of 1.2 times
there covalent radii.
LARGE
A box enclosing all defined atoms with an extension of four times
there covalent radii. This is the default for the critical point search.
MEDIUM / COARSE / FINE
MEDIUM
The mesh spacing of the box is 0.3 Bohr. In the case of the
critical point search it is 0.5 Bohr. This is the default.
COARSE
The mesh spacing of the box is 0.9 Bohr. In the case of the
critical point search it is 1.0 Bohr.
FINE
The mesh spacing of the box is 0.1 Bohr.
READ
The box definition is read from the keyword body.
Description:
With the options STANDARD, SMALL and LARGE the size of the box is specified. The
box will be every time created in STANDARD ORIENTATION (see 4.1.4) and, therefore,
is symmetry adapted to the system. The options MEDIUM, COARSE and FINE define
102
4 Keywords
the mesh spacing within the box. The FINE mesh spacing will generate huge data files
and, thus, is only applicable for small systems. With the READ option the box can be
explicitly defined in the keyword body of BOX. If only a part of the molecule should be
considered the box can be defined simply by the atoms that should be included. In this
case the STANDARD, SMALL and LARGE, as well as the MEDIUM, COARSE and
FINE options remain active. For a SMALL box with FINE resolution containing the H1
and O atom of the water molecule the input has the form:
BOX SMALL FINE READ
O H1
Please note that the atoms are all defined in one input line. Instead of the explicit
definition of each atom, elements can be used for the box definition, too. Therefore a
SMALL box with MEDIUM resolution containing only the hydrogen atoms of water is
defined as:
BOX SMALL MEDIUM READ
H
Alternatively, the plot box can be defined by vectors in the keyword body of BOX. In this
case the options STANDARD, SMALL and LARGE, as well as MEDIUM, COARSE and
FINE loose their meaning. The following format is used for the box definition with vectors:
BOX
NI
X0
XI
XJ
READ
NJ NK
Y0 Z0
YI ZI
YJ ZJ
XK YK ZK
~
rk
~
rj
~
ri
~
r0
O
where NI, NJ and NK specify the number of points in the corresponding ~ı, ~ and ~k directions. The ~r0 = (x0 , y0, z0 ) position vector defines the origin of the box. The other three
orthogonal vectors ~ri = (xi , yi , zi ), ~rj = (xj , yj , zj ) and ~rk = (xk , yk , zk ) span the box as
sketched above. In the case of a plane (two-dimensional box) the definition of NK, XK,
YK and ZK are obsolete. For a line (one-dimensional box) only NI, ~r0 and ~ri have to be
defined. The explicit definition of a plot box is described in Example 5.17.
4.11 Visualization and Topology
4.11.7
103
Keyword POINTS
With this keyword points for the plotting of molecular fields (4.11.2) and start points for
the critical point search (4.11.3) can be specified.
Options:
BINARY / ASCII / POLYGON / READ
BINARY
The plot point coordinates and connectivities are read from the
external binary file LAT.bin. This is the default.
ASCII
The plot point coordinates and connectivities are read from the
external ascii file LAT.asc.
POLYGON
Starting points for the critical point search are generated by
the polygon algorithm.
READ
The plot point coordinates are read from the input file
deMon.inp.
Description:
With the option BINARY coordinates and connectivities are read from the file LAT.bin.
Usually this file has been produced by a previous run calculating isosurfaces or geometrical
surfaces (see Sections 4.11.4 and 4.11.5). In this way molecular fields can plotted on
isosurfaces. Example 5.17 describes the plotting of the electrostatic potential on the
isodensity surface of benzene. The LAT.bin file and the corresponding plot output can be
visualized with VU (see 8). With the ASCII option plot point coordinates, and optional
connectivities, can be read from the external ascii file LAT.asc. The file format is:
1814
.312196E+00
.243158E+00
.243158E+00
...
1
4
6
...
-.729474E+00
-.765796E+00
-.729474E+00
...
2
3
5
6
7
8
...
...
.105260E+01
.105260E+01
.917119E+00
...
The first integer number (here 1814) denotes the number of points. For each point one
line with its coordinates (x, y and z) is then given. The integer triples after the coordinate
specification are the connectivities. This part of the input is optional. The LAT.asc file
can be easily generated from the deMon.lat file (see 4.11.4 and can directly be used for
the visualization with VU.
104
4 Keywords
With the POLYGON option starting points for the critical point search are automatically generated. In the case of the electronic density this algorithm usually generates
a sufficient set of starting points to find all critical points. This is not the case for
the critical points of other molecular fields. The READ option can be used to read
point coordinates from the input file deMon.inp. The coordinates are given in free format
in the keyword body of POINTS, one input line for each point.
4.12 Miscellaneous Keywords
4.12
Miscellaneous Keywords
4.12.1
Keyword MAXMEM
105
This keyword specifies the amount of memory available to the job.
Options:
<integer>
Amount of memory available. The default is 256 Mbytes.
KBYTE
Units: Kilobytes.
MBYTE
Units: Megabytes. This is the default.
GBYTE
Units: Gigabytes.
Description:
deMon can use the MAXMEM value for choosing the appropriate handling of ERIS MEMORY
(see 4.4.3) and the value of MATDIA THRESHOLD= (see 4.12.6).
The amount of memory specified in MAXMEM cannot exceed the MAXRAM compilation-time
parameter.
4.12.2
Keyword TITLE
This keyword specifies the job title.
Options: None
The job title is limited to 60 characters and the title line cannot be continued over more
than one line.
4.12.3
Keyword PRINT
This keyword controls optional printing.
Options:
106
4 Keywords
ALWAYS
Print requested output for all calculation phases.
ATOMMAP
Print atom ordering maps.
AUXIS
Print auxiliary function table.
BASIS
Print molecular orbitals in long format.
CGTO
Print GTO contraction table.
COORD
Print primitive internal coordinates.
DE2
Print Hessian matrix.
DEBUG
Generate debug output.
ECP
Print ECP table.
EMBED
Print embedding charges.
ERIS
Print three-center electron repulsion integrals.
FLUX
Prints the program flux (for debugging purpose).
G
Print auxiliary function Coulomb matrix.
GRID
Generate full grid output.
GTO
Print primitive GTO table.
KS
Print Kohn-Sham and core Hamiltonian matrix.
MAX=<Integer> Maximum number of SCF cycles with print output.
MD
Print detailed MD output.
MDEXTRA
Print insanely detailed MD output .
MM
Print additional MM output.
MOE
Print molecular orbital energies and occupations.
MOS
Print molecular orbital energies, occupations and coefficients.
OPT
Generate full optimization output.
ORTHO
Print orthogonalization matrix S −1/2 .
P
Print density matrix.
POPAN
Generate full population analysis output.
PBC
Print periodic bondary conditions information.
QMMM
Print additional QM/MM information.
RAM
Print RAM allocation table.
S
Print overlap matrix.
SYMMETRY
Generate full symmetry output.
T
Print kinetic energy matrix.
TB
Print tight-binding matrices.
XCE
Generate full exchange-correlation energy output.
XCV
Generate full exchange-correlation potential output.
4.12 Miscellaneous Keywords
107
Description:
With the MAX option the number of SCF cycles with active print output can be specified.
All matrices that are changing during the SCF (e.g. P, KS, MOS, etc.) can be printed
for each cycle. The following example activates the printing of the density matrix in the
first 10 SCF cycles:
PRINT MAX=10 P
Printing of molecular orbitals can be limited with the MOS option. The input MOS =
<First> - <Last> specifies that only the MOs in the range from <First> to <Last>
are printed. Here <First> and <Last> refer to the (integer) numbers of the molecular
orbitals. With the MOE option the printing of the molecular orbital coefficients can be
suppressed. In order to print the orbital energies and occupation numbers of the orbitals
10 to 25 the following input can be used:
PRINT MOS=10-25 MOE
Without MOE the MO coefficients would be printed, too.
With the DEBUG option a complete printing of all matrices is activated! Of course such
a output is huge and should only be used for debugging.
The CAVITY option produces a deMon.cav output that can be used as MOLDEN input
to visualize the cavity surface (Figure 1). For large systems the FULL option produces a
very large deMon.pcm output.
4.12.4
Keyword EMBED
This keyword specifies the embedding into point charges.
Options:
FILE
FILE / READ
The embedding point coordinates and charges are read from the
deMon.cub file. This is the default.
READ
The embedding point coordinates and charges are read from the
input file.
Description:
The coordinates and charges are given in free format as real numbers either in the embedding file deMon.cub (see Table 3 or in the keyword body of EMBED. The Cartesian
coordinates of the embedding points refer to the input orientation of the molecule as
defined in the input file. If a Z-Matrix input is used the embedding point coordinates are
defined with respect to the Z-Matrix input orientation (see Figure 1). The point charge
108
4 Keywords
coordinate unit (˚
Angstr¨om or atomic units) correspond to the unit system selected for
the input coordinates (see keyword GEOMETRY in 4.1.1). Optional, an element symbol
or a radius (real number) may be specified after the point charge coordinates and the
charge values. This information will be used for the bond drawing in a Vu output of
the embedding region (see Section 8). The following input specifies an embedding in a
(distorted) octahedron of positive point charges:
EMBED READ
5.0 0.0 0.0
-5.0 0.0 0.0
0.0 5.0 0.0
0.0 -5.0 0.0
0.0 0.0 2.0
0.0 0.0 -2.0
1.0
1.0
1.0
1.0
1.0
1.0
H
H
H
H
1.5
1.5
The first line is only necessary if the embedding points are defined within the input file.
In this example, a radius of the hydrogen atom is assigned to the first four embedding
points. For the two others the radius is explicitly defined to 1.5 ˚
Angstr¨om or atomic units
depending on the unit system specification in the GEOMETRY keyword (see Section
4.1.1).
4.12.5
Keyword CHOLESKY
With this keyword a Cholesky decomposition for the inversion of the auxiliary Coulomb
matrix can be selected.
Options:
OFF / ON
OFF
Singular value decomposition is used. This is the default.
ON
Cholesky decomposition is used.
SVDTOL=
Tolerance for singular value decomposition. Default is 10−5 .
SVDTO2=
Description:
The Cholesky decomposition of the auxiliary function Coulomb matrix is considerably
faster than the singular value decomposition (SVD). However, it may introduce numerical
instabilities for large auxiliary functions sets. This may result in convergence failures of
the SCF procedure. The Cholesky and singular value decomposition results are identical
if the SVD nullspace is empty. By default, the SVD procedure will eliminate eigenvalues
of the auxiliary Coulomb matrix smaller than 10−5 . This cut-off is appropriate in the vast
majority of cases. However, if necessary, it can be modified with the SVDTOL= parameter.
4.12 Miscellaneous Keywords
109
Specifying SVDTO2= will activate gradual quenching of the eigenvalues (see svdmat.f).
4.12.6
Keyword MATDIA
This keyword chooses matrix diagonalization technique.
Options:
RS
EISPACK Household
DSYEV
LAPACK block Hou
D&C
This is another nam
DSYEVD
LAPACK divide and
RS / DSYEV / D&C / DSYEVD / DSYEVR / JACOBI DSYEVR
LAPACK Relatively
JACOBI
Jacobi diagonalizatio
SMALL=
Diagonalizer to use f
LARGE=
Diagonalizer to use f
THRESHOLD= Maximum side dime
Description:
The fastest diagonalizer available in deMon in the divide and conquer diagonalization
from LAPACK [109]. However, it might fail in special situations. Fail safe diagonalizers
in deMon are based on the Householder algorithm. For large systems, the DSYEV diagonalizer is considerably faster than the RS diagonalizer. However, it is also more memory
demanding. Therefore, it should only be used if enough RAM space is available. The
JACOBI diagonalizer is much slower but produces very pure eigenvectors. It is used by
default for atomic calculations.
Normally, deMon will make the optimal diagonalizer choice based on the amount of
memory available (see MAXMEM, section 4.12.1).
4.12.7
Keyword WEIGHTING
This keyword specifies the weight function used for the grid partitioning.
Options:
SCREENED / BECKE
SCREENED
A screened weight function is used. This is the default.
BECKE
The original Becke weight function is used.
Description:
The BECKE weight function [110] introduces a cubic scaling in the grid generation step
[111]. By introducing a new piece wise defined weight function [66] near linear scaling
for the grid generation can be obtained. This weight function is specified by the option
110
4 Keywords
SCREENED.
4.12.8
Keyword QUADRATURE
Numerical quadrature scheme can be modified using this keyword.
Options:
GAUSS
GAUSS / EULER
A Gauss-Chebyshev radial quadrature is selected. This is the
default.
EULER
A Euler-MacLaurin radial quadrature is selected.
RANDOM
Random rotation of the Lebedev grids is activated.
REFERENCE
Description:
A reference grid is generated.
TO BE WRITTEN
4.12.9
Keyword ECPINTEGRATION
This keyword chooses numerical integration scheme, used in calculation of the effective
core potentials.
Options:
COARSE / MEDIUM / FINE
COARSE
Use 67 radial shells, (50,194)p fixed grid. Nominal accuracy 10−6
MEDIUM
Use 99 radial shells, (75,302)p fixed grid. Nominal accuracy 10−8
FINE
Description:
Use 131 radial shells, (200,1202) fixed grid. Nominal accuracy 10−10
TO BE WRITTEN
4.12.10
Keyword GENAO
Generate atomic orbitals. This is an experimental keyword. Do not use.
Options:
CUTOFF=
Cutoff energy in Hartree. The default is 0.0.
AO=
indices of atoms, in a following list of integers, whose basis will be
transformed to AO’s.
NOT=
indices of atoms, in a following list of integers, whose basis be excluded from the AO transformation.
Description:
4.12 Miscellaneous Keywords
111
Default: GENAO is switched off. If GENAO is switched on, the default is that all atoms
are transformed. CUTOFF = 0.0 by default.
If GENAO is specified, the GTO basis will be transformed to an atomic basis. For the
forthcoming SCF computation, not all AO’s need to be considered and hence the matrix
size may be reduced. The number of AO’s which will be considered in the SCF depends
on the cutoff energy: Orbitals with higher atomic energy will be left out. If the cutoff
energy is higher than the highest atomic energy this is a unitary transformation and gives
exactly the same result as the original GTO computation.
112
5
5
Examples
THIS SECTION NEEDS TO BE WRITTEN. VOLUNTEERS?
5.1
Example ps18
5.2
Examples amk5, amk6, amk7, and amk8
5.3
Examples amk29, amk30, and amk31
5.4
Example ps86
5.5
Example th6
5.6
Example amk17
5.7
Example ps84
5.8
Example amk19
5.9
Example amk20
5.10
Example ps15
5.11
Example ps6
5.12
Example amk24
5.13
Example amk26
5.14
Example amk33
5.15
Example amk35
5.16
Example ps88
5.17
Example ps90
Examples
6 Using QM/MM in deMon
6
Using QM/MM in deMon
TO BE WRITTEN
113
114
7
7 Using MD in deMon
Using MD in deMon
TO BE WRITTEN
8 Vu
8
115
Vu
This chapter contains an introduction to the basic operating instructions of the Vu graphical interface which is part of the visualization program Vu. It constitutes a subset of all
the available options in the graphical interface of Vu. This chapter shows only the most
important features required to start using Vu. Prior to its use, it is highly recommended to
read Chapter 4.11 of this guide. For a complete description of the structure and operation
of Vu, please consult:
• The User Manual of the Configurable Scientific Visualization Program Vu. Which
contains full description of architecture, capabilities and limitations of Vu.
• Description of the file format. Describes the structure of .pie files which are written
by deMon and read by the Vu graphical interface.
• Frequently Asked Questions on the Utilization of Vu. Contains answers to many of
the doubts involved with the use of Vu.
• http://www3.sympatico.ca/chantal.pic/vu/eng/index.html. This is a temporarily address.
8.1
What is Vu?
Vu is a configurable visualization software tool for the display and analysis of numerical
solutions. It is used for exploring and interpreting results from simulation programs or
experimental measurements. Based on a dictionary and a grammar common to researchers
of various domains, it finds applications in fluid mechanics, civil engineering, applied
mathematics, manufacturing, injection molding, combustion, structures, computational
chemistry, etc. Vu is available on computers of various sizes, from laptops using Linux to
virtual reality immersion environments (CAVE, ImmersaDesk, Reality Center, etc.), on
all Unix platforms.
Vu uses three ingredients to construct an image: a support, an entity and a mode. The
support is the place where the image resides: a plane, sphere, cylinder, geometric, etc.
The entity is what it is represented graphically: a set of vectors like velocity, a scalar
field like electron density, etc. There are eight types of entities: mesh, graph, iso, vector,
tensor among others.
The mode is the way the user wants to see the image: static, dynamic, injection or
animation.
116
8.2
8 Vu
Vu and deMon
As a part of deMon, there are a set of routines that build the required files for visualization
with Vu. This allows the users to display many scalar fields, critical points, etc. and
select the way of visualize it. Before using Vu, the user has to run deMon with the
proper visualization keywords in order to generate the two required files for the graphical
interface:
deMon.pie: The control file for Vu. It contains all the definitions for construct the images.
FIELD.bin: This file is referred by the deMon.pie file. Contains the values of the scalar
field selected for plotting.
Most supporting definitions for chemistry used by deMon reside in the deMonMacros.vu
file which is included automatically at run time. So, deMonMacros.vu file should reside
in the location where Vu stands.
8.3
Running Vu
To start Vu first export the display to your current window, then, on the command line
type:
Vu
and press Enter key. The program starts without any file and then you can get your own
deMon.pie file from the menu or type:
Vu deMon.pie
The program starts with the structure associated to the .pie file that you typed. There
are other ways to read files. See Vu reference manual.
When Vu starts, three windows appear on the screen:
1. Menu Window. Used to build images.
2. Visualization Window (Empty at startup or displaying the molecular structure if a
.pie file was selected) Used to visualize and modify images.
3. Message Window. Used by Vu to show information extracted of the files and error
messages.
The user can adjust the size and position of these windows with the mouse.
Menu window. This window is composed of:
1. a menu bar (File, Zones, Info, Options, Stereo, Help),
2. a list of images proposed by Vu,
8.3 Running Vu
117
3. the lists of Supports, Entities and Windows (components of the image)
See figure ??.
Through the File option of the menu bar the user can open a .pie file among others. Or
can import files in other formats (Gaussian, PDB, etc) or save profiles, data and images.
The Images part of this Menu window lets the user select the image to be displayed in the
visualization window. By default the Geometry image is active at startup. To select other
image just click with the mouse on the box at the left of the image desired. An unlimited
number of images can be displayed simultaneously on the Visualization window.
Each time the user selects an image on the Image section, the corresponding Support,
Entity and Window associated with is shaded in the bottom section of this window.
By clicking with the mouse these shaded Supports, Entities and Windows, the user open
new windows where their attributes can be changed to modify the way the image is
displayed. See chapter 8 of the Vu Reference Manual.
118
9
9 MAG
MAG
The deMon program prepares the data needed to run the MASTER program [115,116] for
NMR (shielding tensor) and EPR (Fermi term and anisotropic hyperfine tensor) properties. The deMon program provides the ground-state electronic structure of a system in
the absence of an external magnetic field and then writes out information (atomic coordinates, basis set, grid, MO coefficients, orbital energies) to different files. The MASTER
program, in a second step, calculates the magnetic properties. The NMR calculation is
valid for closed shell and EPR for open shell systems. Recall that if a ROKS calculation
has been performed, there will be no spin polarization possible and the Fermi term will
then be zero, except if the unpaired electron is in an s-type orbital. The interface routine
prepares 4 output files (deMon.nmr11, deMon.nmr70, deMon.nmr71, deMon.nmr72) for
NMR and 3 output files (deMon.nmr70, deMon.nmr71, deMon.nmr72) for EPR.
The MASTER program has been successfully used for various applications: NMR screening constants for all magnetic nuclei for closed shell systems and hyperfine structure calculations for open-shells. The IGLO bases are recommended to obtain good NMR results
[117]. The README files available in the MASTER program provide useful information
to the users.
NMR calculations are only possible within DFT. It might be possible to recompile the
code with (much) larger values for specifying the maximum number of basis functions
and auxiliary basis functions and depending variables. For this purpose, inspect the file
parameter.h in the $deMon/include folder.
10 Quick Keyword Reference
10
Quick Keyword Reference
TO BE WRITTEN. VOLUNTEERS?
119
120
11
11 Troubleshooting
Troubleshooting
NEED TO BE WRITTEN. VOLUNTEERS?
A Automatic Generation of Auxiliary Functions
A
121
Automatic Generation of Auxiliary Functions
With the auxiliary function definition GEN-An and GEN-An*, n = 1, 2 and 3, automatically generated auxiliary function sets are selected in deMon. The GEN-An sets consist of
s, p and d Hermite Gaussian functions. The GEN-An* sets possess also f and g Hermite
Gaussians. Because the auxiliary functions are used to fit the electronic density they are
grouped in s, spd and spdfg sets. The exponents are shared within each of these sets [24,
25]. Therefore, the auxiliary function notation (3,2,2) describes 3 s sets with together 3
functions, 2 spd sets with together 20 functions and 2 spdfg sets with together 42 functions (see also 4.2.2). The range of exponents of all auxiliary functions is determined by
the smallest, ζmin , and largest, ζmax , primitive Gaussian exponent of the chosen basis set.
Therefore, the GEN-An and GEN-An* automatically generated auxiliary function sets
are different for different basis sets. The number of exponents N (auxiliary function sets)
is given by:
N = Int
ln( ζmax/ζmin )
ln(6 − n)
(A.1)
Here n is 2, 3 or 4 according to the chosen GEN-An or GEN-An* set. The exponents are
generated (almost; as it is explained below) even tempered and splitted into s, spd and,
if a GEN-An* set is requested, spdfg sets, The tightest (largest) exponents are assigned
to the s sets, followed by the spd and, if exist, spdfg sets. The basic exponent from which
the generation starts is defined as:
ζo = 2 ζmin (6 − n)(N −1)
(A.2)
From this exponent the two tightest s set exponents, ζ1 and ζ2 are generated according
to the formulas:
ζ1
ζ2
n
= 1+
12 − 2n
ζo
=
6−n
ζo
(A.3)
(A.4)
The other s set exponents are generated according to the even tempered progression:
ζi+1 =
ζi
6−n
(A.5)
The ζo exponent of the following spd sets is also generated according to the progression
(A.5). Based on this ζo exponent the exponents of the first two spd sets are calculated
122
A Automatic Generation of Auxiliary Functions
with the formulas (A.3) and (A.4). The following spd set exponents are then calculated
again according to the even tempered progression (A.5). In the same way the spdfg set
exponents are calculated. In the case of 3d elements an extra diffuse s auxiliary function
set is added.
B Format of a platform description file
B
123
Format of a platform description file
B.1
Philosophy of deMon configuration files
On most of the supported platforms, deMon comes with three sets of compilation flags
(these are designated by stars in the table above). These sets are: dbg, std, and opt. The
idea behind these options is as follows:
The opt (for optimized) version is intended to give maximum performance. To achieve
this, this set of compilation options uses aggressive optimization flags. Compiling an
opt version may also require vendor-optimized BLAS and LAPACK libraries (such as
ESSL or SCS), which are not supplied with deMon. Often, such vendor-supplied libraries
increase computational efficiency by sacrificing numerical accuracy. Because The deMon
Developers don’t have any control over the trade-off, made by the library’s vendor, the opt
version may either fail altogether, or produce wrong answers. Additionally, the location
of the vendor-optimized libraries on different computers may vary. For this reason, you
may have to adjust the opt makefile to reflect these location.
Although we make an effort to test the opt settings, this configuration is very sensitive to
minor changes in compiler optimizations and vendor’s libraries. If the opt version fails,
please try repeating the same calculation with the std version, before submitting a bug
report.
The std (for standard) version does not use any external numerical libraries, and employs
conservative optimization setting. These settings are expected to result in good, but
not the best possible, performance. Usually, compiling the std version does not require
anything beyond the deMon distribution and Fortran compiler, to get it running.
Finally, the dbg (for debug) version allows meaningful symbolic debugging of the executable. In order to achieve high performance, the compiler will rearrange the execution
order of the Fortran statements in the opt and std versions. The dbg version requests the
compiler to generate machine code, which is as close as possible to the original Fortran
source.
B.2
Structure of a platform configuration files
Platform configuration files in deMon are processes by the standard Unix make utility.
To be accepted by make, these files have to follow a few simple syntactic rules:
124
B Format of a platform description file
• Lines beginning with the hash mark (“#”) are treated as comments.
• A backslash character (“\”) in the last column of a line indicates that a line is to
be continued on the following line.
• An alphanumeric sequence, followed by the equals sign (“=”) and a value indicates
a variable assignment.
• As alphanumeric sequence, followed by a colon (“:”) introduces a rule. A rule may be
followed by zero or more actions. Each action must consist of a tabulation character,
followed by a shell command.
Variables and rules, which deMon build scripts expect to find in a platform configuration
files, are summarized in Table 10.
Table 10: Structure of a platform configuration file.
SHELL
Name of a POSIX-compliant version of Bourne shell
F90
Command used to invoke Fortran-90 compiler
FTN FIXED
Command line option, which selects the fixed (Fortran-77 style)
source form
FTN FREE
Command line option, which selects the free source form
F90COMMON
Common compilation flags, which will be applied to all Fortran
source files. In order to compile the deMon source code correctly,
you should instruct the compiler to treat REAL variables and constants, as well as all untyped floating-point literals, in at least
64-bit precision. On many Unix systems, the -r8 compilation
option will do this. You should also instruct the compiler to look
in the deMon include directory. On many systems, -I$(INCL) is
the right option. Please do NOT add optimization options
to this line
B.2 Structure of a platform configuration files
F90OPT
125
“Normal” optimization flags, which should apply to most source
files. On many UNIX systems, the -O compilation option will
correspond to safe optimization setting (please consult your compiler’s documentation)
F90OPTLOW
“Low” optimization settings, which may be required to compile
a few of the source code files, which are not treated correctly at
the default level. You’ll be able to specify the affected files later,
in FILES LOW
F90OPTHIGH
“Extra-high” optimization settings, not suitable for most files,
but required for a few of them. You’ll be able to specify the
affected files later, in FILES HIGH
LIBPATH
Use this entry to specify non-standard locations of the object
libraries, required to link the code. On most UNIX systems, the
correct syntax is -Lpath. Several locations may be specified, if
necessary
LINKLIB
Specify any additional libraries here (usually BLAS). If you include an optimized BLAS or LAPACK libraries, please make sure
to exclude generic versions of these routines, supplied with deMon
in FILES EXCLUDE
FILES EXCLUDE
Source code files (including the .f extension, but without the
path), mentioned in this list, will not be compiled. Instead,
you are supposed to provide optimized versions of these routines (see LINKLIB above). If you supply an optimized BLAS
library (such as Atlas, or a vendor-provided library), please list
the following source files in FILES EXCLUDE: blas1.f blas2.f
dgemm.f dsyrk.f lsame.f xerbla.f. If you link with an optimized LAPACK library (such as MKL, complib.sgimath, or
LAPACK with Atlas, also list these files: dpptrf.f dpptri.f
dtptri.f dsyev.f dsyevd.f dsyevr.f dgelss.f. Finally, if
you link with a vendor-optimized EISPACK library, also exclude:
pythag.f rs.f. All these files are located in subdirectories of
$deMon/source/math
126
FILES PLATFORM
B Format of a platform description file
This entry gives the list of platform-specific source code files.
These source code files implement functionality, which is impossible, or difficult, to provide in a platform-independent manner. The corresponding source code files are found in the
$deMon/source/platform directory, and its subdirectories. You
need to specify the implementation for the functions DEFLUSH and
DETIME (see below). The complete list of the platform-dependent
implementations included with deMon is given in Table 11
FILES LOW
The list of files (with the .f or .f90 extension, but without
paths), which require low optimization settings (see F90OPTLOW
above)
FILES HIGH
The list of files, which require extra-high optimization settings
(see also F90OPTHIGH above
FILES SPECIAL
The list of files, which require special treatment such as using
a different compiler, or special preprocessing). Any files, mentioned on this line, will be excluded from the standard compilation process. Instead, you should provide complete compilation
instructions, using Makefile rules.
ARCMD
The command, required to add (or replace) an object file to (in)
an archive library. The command, will be invoked as: “$(ARCMD)
library name object file name”. On most UNIX systems, you
can leave the default (ar r) unchanged.
RANLIBCMD
The command, required to make an archive library usable by the
system linker. This command will be invoked as: “$(RANLIBCMD)
library name” Usually, this is not required, and can be set to
RANLIBCMD=true. However, on some UNIX systems, this should
be set to RANLIBCMD=ranlib
B.2 Structure of a platform configuration files
127
deMon object files, which must be included while build-
TOOLOBJ
ing auxiliary tools. The usual set is: “$(OBJ)/deflush.o
$(OBJ)/detime.o”. However, if you intend to build the MAG
tool, you may also need to include some of the files from the
generic BLAS library.
checkplatform:
This make target should succeed, if this platform definition file
is appropriate for the current system. The actual command line
used by the driver script is: “make -s -f platform makefile
checkplatform”. This target will typically check the operating
system name, CPU model, compiler name and version. Please
keep in mind that all platform-specific makefiles get invoked in
this way by make makesys, on every platform (this is how make
makesys determines which makefiles are appropriate)
describeplatform:
This target should produce a one-line description of the platformspecific makefile, on standard output. You should always put
something here
commentplatform:
This target should produce any platform-specific installation instructions, or warnings, on standard output. If there are no specific instructions, leave this empty.
For a successful compilation, you need to provide platform-specific implementations of
the following two functions:
DETIME
This REAL function must return CPU time in seconds, used by the program
since it started.
DEFLUSH
This subroutine, taking one integer argument, the Fortran LUN number, must ensure that all outstanding write operations on that LUN have
completed.
deMon comes with several pre-defined implementations for these routines, which are listed
in Table 11.
128
B Format of a platform description file
Table 11: Platform specific functions in the shipped deMon version.
Function Path
DETIME
platform/Generic/detime-f95.F
Description
Portable
Fortran-95
DETIME
CPU TIME() intrinsic
Use system library subroutine ETIME
platform/Linux/detime.F
routine,
using
(Linux, HP-UX, OSF1, SunOS, IRIX, and
DETIME
DETIME
platform/AIX/detime.F
many others)
Use system library function MCLOCK()
platform/Generic/detime-f90.F
(IBM AIX and many others)
Portable Fortran-90 routine, using SYSTEM CLOCK() intrinsic. This routine returns real time, rather than the CPU time,
DEFLUSH
DEFLUSH
DEFLUSH
platform/Generic/qflush-POSIX.F
and should only be used as the last resort.
Use POSIX function PXFFFLUSH (IEEE
platform/Linux/deflush.F
1003.9-1992, ISO/IEC 9945-1:1990)
Use one-argument FLUSH library routine
platform/IRIX/deflush.F
(Linux, HP-UX, OSF1, SunOS, many others)
Use two-argument FLUSH library routine
(IRIX)
DEFLUSH
platform/AIX/deflush.F
Use FLUSH subroutine (IBM AIX)
C Working with the deMon Test Suite
C
Working with the deMon Test Suite
129
130
D
D Global Counters, Limits and Pointers
Global Counters, Limits and Pointers
The following tables list the most important global counter, limit and pointer variables
in deMon.
Table 12: Counters.
Counters
NATOM
NAUX
NAUXSET
NAUXSHL
NCN
NELEC
NGTO
NSEC
NSES
NSHL
NSIG
NSN
NSTO
Description
Number of atoms
Number of auxiliary functions
Number of auxiliary function sets
Number of auxiliary function shells
Degree of C axis of highest degree
Number of electrons
Number of primitive Gaussian functions
Number of n-fold C axes
Number of n-fold S axes
Number of shells
Number reflection planes
Degree of S axis of highest degree
Number of orbitals
→
→
→
→
→
→
→
→
→
→
→
→
→
IATOM, JATOM, ...
IAUX, JAUX, ...
IAUXSET, JAUXSET, ...
IAUXSHL, JAUXSHL. ...
ICN, JCN, ...
IELEC, JELEC, ...
IGTO, JGTO, ...
ISEC, JSEC, ...
ISES, JSES, ...
ISHL, JSHL, ...
ISIG, JSIG, ...
ISN, JSN, ...
ISTO, JSTO, ...
D Global Counters, Limits and Pointers
131
Table 13: Limits.
Limit
LL(IATOM)
UL(IATOM)
LLAUX(IAUXSHL)
ULAUX(IAUXSHL)
LLAUXSET(IATOM)
ULAUXSET(IATOM)
LLAUXSHL(IAUXSET)
ULAUXSHL(IAUXSET)
LLGP(IATOM)
ULGP(IATOM)
LLGTO(ISHL)
ULGTO(ISHL)
LLSHL(IATOM)
ULSHL(IATOM)
LLSTO(ISHL)
ULSTO(ISHL)
Description
Number of the first orbital (STO) of atom IATOM
Number of the last orbital (STO) of atom IATOM
Number of the first auxiliary function of auxiliary
function shell IAUXSHL
Number of the last auxiliary function of auxiliary
function shell IAUXSHL
Number of the first auxiliary function set of atom IATOM
Number of the last auxiliary function set of atom IATOM
Number of the first auxiliary function shell of
auxiliary function set IAUXSET
Number of the last auxiliary function shell of
auxiliary function set IAUXSET
Number of the first grid point of atom IATOM
Number of the last grid point of atom IATOM
Number of the first primitive Gaussian function of shell
ISHL
Number of the last primitive Gaussian function of shell
ISHL
Number of the first shell of atom IATOM
Number of the last shell of atom IATOM
Number of the first orbital of shell ISHL
Number of the last orbital of shell ISHL
132
D Global Counters, Limits and Pointers
Table 14: Pointers.
Pointer
Description
AUXPTR(IAUX,I)
Auxiliary function pointer
I = 1 : Origin of auxiliary function IAUX (number of atom)
I = 2 : Auxiliary function shell
I = 3 : AX of auxiliary function IAUX
I = 4 : AY of auxiliary function IAUX
I = 5 : AZ of auxiliary function IAUX
AUXSETPTR(IAUXSET,I) Auxiliary function set pointer
I = 1 : Origin of auxiliary function set IAUXSET (number of atom)
I = 2 : Maximum angular quantum number L in the set IAUXSET
AUXSHLPTR(IAUXSHL,I) Auxiliary function shell pointer
I = 1 : Origin of auxiliary function shell IAUXSHL (number of atom)
I = 2 : Auxiliary function set of auxiliary function shell IAUXSHL
I = 3 : Angular quantum number L of auxiliary function shell IAUXSHL
SHLPTR(ISHL,I)
Shell pointer
I = 1 : Origin of shell ISHL (number of atom)
I = 2 : Main quantum number N of shell ISHL
I = 3 : Angular quantum number L of shell ISHL
I = 4 : Contraction of shell ISHL
STOPTR(ISTO,I)
Orbital pointer (STO pointer)
I = 1 : Origin of orbital (number of atom)
I = 2 : Shell of orbital ISTO
I = 3 : AX of orbital ISTO
I = 4 : AY of orbital ISTO
I = 5 : AZ of orbital ISTO
References
133
References
[1] P. Hohenberg, W. Kohn, Phys. Rev. 136, B864 (1964).
[2] W. Kohn, L.J. Sham, Phys. Rev. 140, A1133 (1965).
[3] R.M. Dreizler, E.K.U. Gross, Density Functional Theory (Springer-Verlag, Berlin,
1990).
[4] A. St-Amant, Ph.D. Thesis, Universit´e de Montr´eal, 1992.
[5] M.E. Casida, C. Daul, A. Goursot, A.M. K¨oster, L.G.M. Pettersson, E. Proynov,
A. St-Amant, D.R. Salahub, H. Duarte, N. Godbout, J. Guan, C. Jamorski, M.
Leboeuf, V. Malkin, O. Malkina, F. Sim, A. Vela, deMon-KS Version 3.4, deMon
Software, Montr´eal, 1996.
[6] M.E. Casida, C. Daul, A. Goursot, A.M. K¨oster, L.G.M. Pettersson, E. Proynov,
A. St-Amant, D.R. Salahub, H. Duarte, N. Godbout, J. Guan, K. Hermann, C.
Jamorski, M. Leboeuf, V. Malkin, O. Malkina, M. Nyberg, L. Pedocchi, F. Sim, L.
Triguero, A. Vela, deMon Software, 2001.
[7] A.M. K¨oster, M. Krack, M. Leboeuf, B. Zimmermann, AllChem, Universit¨at Hannover, 1998.
[8] A.M. K¨oster, R. Flores-Moreno, G. Geudtner, A. Goursot, 16292 T. Heine, J.U.
Reveles, A. Vela, D.R. Salahub, deMon, NRC, Canada, 2003.
[9] R.G. Parr, W. Yang, Density–Functional Theory of Atoms and Molecules (Oxford
University Press, New York, 1989).
[10] Y. Zhang, T.-S. Lee, W. Yang, J. Chem. Phys. 110, 46 (1999).
[11] G.A. DiLabio, M.M. Hurley, P.A. Christiansen, J. Chem. Phys. 116, 9578 (2002).
[12] Extensible Computational Chemistry Environment Basis Set Database, Version
4/05/02, Molecular Science Computing Facility, Environmental and Molecular Sciences Laboratory, Pacific Northwest Laboratory, P.O. Box 999, Richland, Washington 99352, USA.
[13] N. Godbout, D.R. Salahub, J. Andzelm, E. Wimmer, Can. J. Phys. 70, 560, (1992).
134
References
[14] A.M. K¨oster, Habilitation Thesis, Universit¨at Hannover, 1998.
[15] J. Guan, P. Duffy, J.T. Carter, D.P. Chong, K.C. Casida, M.E. Casida, M. Wrinn,
J. Chem. Phys. 98, 4753 (1993).
[16] D.P. Chong (private communication).
[17] N. Rega, M. Cossi, V. Barone, J. Chem. Phys. 105, 11060 (1996).
[18] S. Huzinaga, J. Chem. Phys. 42, 1293 (1965).
[19] W.J. Hehre, R.F. Stewart, J.A. Pople, J. Chem. Phys. 51, 2657 (1969).
W.J. Hehre, R. Ditchfield, R.F. Stewart, J.A. Pople, J. Chem. Phys. 52, 2769
(1970).
[20] A.J. Sadlej, Collection Czech. Chem. Commun. 53, 1995 (1988).
[21] G.C. Lie, E. Clementi, J. Chem. Phys. 60, 1275 (1974).
[22] A.J.H. Wachters, J. Chem. Phys. 52, 1033 (1970).
[23] P.-O. Widmark, B.J. Persson, B.O. Roos, Theor. Chim. Acta 79, 419 (1991).
R. Pou-Am´erigo, M. Merch´an, I. Nebot-Gil, P.-O. Widmark, B.O. Roos, Theor.
Chim. Acta 92, 149 (1995).
[24] J. Andzelm, E. Radzio, D.R. Salahub, J. Comput. Chem. 6, 520 (1985).
[25] J. Andzelm, N. Russo, D.R. Salahub, J. Chem. Phys. 87, 6562 (1987).
[26] http:://www.theochem.uni-stuttgart.de
[27] H.B. Schlegel, M.J. Frisch, Int. J. Quantum Chem. 54, 83 (1995).
[28] H.-J. Glaeske, J. Reinhold, P. Volkmer, Quantenchemie, Band 5, Eds. W. Haberditzl, M. Scholz, L. Z¨
ulicke (Dr. Alfred H¨
uhtig Verlag, Heidelberg, 1987).
[29] C. Daul, Int. J. Quantum Chem. 52, 867 (1994).
[30] P.J. Hay, J. Chem. Phys. 66, 4377 (1977).
[31] C.C.J. Roothaan, Rev. Mod. Phys. 23, 69 (1951).
References
135
[32] G.G. Hall, Proc. Roy. Soc. Ser. A 205, 541 (1951).
[33] J.A. Pople, R.K. Nesbet, J. Chem. Phys. 22, 571 (1954).
[34] C.C.J. Roothaan, Rev. Mod. Phys. 32, 179 (1960).
[35] J.S. Binkley, J.A. Pople, P.A. Dobosh, Mol. Phys. 28, 1423 (1974).
[36] T. Amos, L.C. Snyder, J. Chem. Phys. 41, 1773 (1964).
[37] A.D. Bacon, M.C. Zerner, Theor. Chim. Acta 53, 21 (1979).
[38] B.I. Dunlap, J.W.D. Connolly, J.R. Sabin, J. Chem. Phys. 71, 4993 (1979).
[39] J.W. Mintmire, B.I. Dunlap, Phys. Rev. A 25, 88 (1982).
[40] A.M. K¨oster, in preparation.
[41] J. Alml¨of, K. Faegri, Jr., K. Korsell, J. Comput. Chem. 3, 385 (1982).
[42] A.M. K¨oster, J. Chem. Phys., 118, 9943 (2003).
[43] E. Canc´es, J. Chem. Phys. 114, 10616 (2001).
[44] D.R. Hartree, The Calculation of Atomic Structures (Wiley, New York, 1957).
[45] V.R. Saunders, I.H. Hillier, Int. J. Quantum Chem. 8, 699 (1973).
[46] A.M. K¨oster, P. Calaminici, Z. G´omez, U. Reveles in Reviews of Modern Quantum
Chemistry, A Celebration of the Contribution of Robert G. Parr, Ed. K. Sen (World
Scientific Publishing Co., Singapore, 2002).
[47] P. Pulay, J. Comput. Chem. 3, 556 (1982).
[48] B. Miehlich, A. Savin, H. Stoll, H. Preuss, Chem. Phys. Lett. 157, 200 (1989).
[49] P.A.M. Dirac, Proc. Cambr. Phil. Soc. 26, 376 (1930).
[50] J.P. Perdew, Y. Wang, Phys. Rev. B 33, 8800 (1986); 34, 7406(E) (1986).
[51] A.D. Becke, Phys. Rev. A 38, 3098 (1988).
[52] N.C. Handy, A.J. Cohen, Mol. Phys. 99, 403 (2001).
136
References
[53] J.P. Perdew, J.A. Chevary, S.H. Vosko, K.A. Jackson, M.R. Pederson, D.J. Singh,
C. Fiolhais, Phys. Rev. B 46, 6671 (1992).
[54] J.P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 77, 3865 (1996).
[55] Y. Zhang, W. Yang, Phys. Rev. Lett. 80, 890 (1999).
[56] B. Hammer, L.B. Hansen, J.K. Norskov, Phys. Rev B 59, 7413 (1999).
[57] S.H. Vosko, L. Wilk, M. Nusair, Can. J. Phys. 58, 1200 (1980).
[58] J.P. Perdew, A. Zunger, Phys. Rev. B 23, 5048 (1981).
[59] J.P. Perdew, Y. Wang, Phys. Rev. B 45, 13244 (1992).
[60] J.P. Perdew, Phys. Rev. B 33, 8822 (1986).
[61] B. Zimmermann, Ph.D. Thesis, Universit¨at Hannover (1999).
[62] C. Lee, W. Yang, R.G. Parr, Phys. Rev. B 37, 785 (1988).
[63] R. Colle, D. Salvetti, Theor. Chim. Acta 37, 329 (1975).
[64] R. Colle, D. Salvetti, J. Chem. Phys. 79, 1404 (1983).
[65] M. Krack, A.M. K¨oster, J. Chem. Phys. 108, 3226 (1998).
[66] A.M. K¨oster, R. Flores-Moreno, J.U. Reveles, in preparation.
[67] P.M.W. Gill, B.G. Johnson, J.A. Pople, Chem. Phys. Lett. 209, 506 (1993).
[68] V.I. Lebedev, Russian Acad. Sci. Dokl. Math. 50, 283 (1995).
[69] A.K. Rappe, C.J. Casewit, K.S. Colwell, W.A. Goddard III, W. M. Skiff, J. Am.
Chem. Soc. 114, 10024 (1992).
[70] A.K. Rappe, W.A. Goddard III, J. Phys. Chem. 95, 3358 (1991).
[71] M. O’Keeffe and N.E. Brese, J. Am. Chem. Soc. 113, 3226 (1991).
[72] S.L. Mayo, B.D. Olafson, W.A. Goddard III, J. Phys. Chem. 94, 8897 (1990).
[73] C.J. Casewit, K.S. Colwell, A.K. Rappe, J. Am. Chem. Soc. 114, 10046 (1992).
References
137
[74] C.J. Casewit, K.S. Colwell, A.K. Rappe, J. Am. Chem. Soc. 114, 10035 (1992).
[75] G. te Velde, E.J. Baerends, Phys. Rev. B 44, 7888 (1991).
[76] J. Baker, A. Kessi, B. Delley, J. Chem. Phys. 105, 192 (1996).
[77] A. Banerjee, N. Adams, J. Simons, R. Shepard, J. Phys. Chem. 89, 52 (1985).
[78] J. Baker, J. Comput. Chem. 7, 385 (1986).
[79] M.J.D. Powell, Math. Prog. 1, 26 (1971).
[80] T.H. Fisher, J. Alml¨of, J. Phys. Chem. 96, 9768 (1992).
[81] R. Lindh, A. Bernhardsson, G. Karlstr¨om, P. Malmqvist, Chem. Phys. Lett. 241,
423 (1995).
[82] J. Nichols, H. Taylor, P. Schmidt. J. Simons, J. Chem. Phys. 92, 340 (1990).
[83] R. Fletcher, Practical Methods of Optimization, Second Edition (Wiley, New York,
1987).
[84] C.G. Broyden, J. Inst. Maths. Applns. 6, 76 (1970); ibid 6, 222 (1970).
[85] R. Fletcher, Computer J. 13, 317 (1970).
[86] D. Goldfarb, Maths. Comp. 24, 23 (1970).
[87] D.F. Shanno, Maths. Comp. 24, 647 (1970).
[88] H.B. Schlegel, J. Comput. Chem. 3, 214 (1982).
[89] W.C. Davidon, AEC Res. & Dev. Report ANL-5990 (1959).
[90] R. Fletcher, M.J.D. Powell, Computer J. 6, 163 (1963).
[91] J.M. Bofill, J. Comp. Chem. 15, 1 (1994).
[92] H.J.C. Berendsen, J.P.M. Postma, W.F. van Gunsteren, D. DiNola, J.R. Haak, J.
Chem. Phys. 81, 3684 (1984).
[93] R.S. Mulliken, J. Chem. Phys. 23, 1833 (1955).
138
References
[94] P.O. L¨owdin, J. Chem. Phys. 18, 365 (1950).
[95] I. Mayer, Chem. Phys. Lett. 97, 270 (1983). Gonvalves, Mohallem, Chem. Phys.
Lett. 280, 378 (2003).
[96] M.S. Gopinathan, K. Jug, Theor. Chim. Acta 63, 497 (1983).
[97] R.F.W. Bader, Atoms in Molecules, A Quantum Theory (Clarendon, Oxford, 1990).
[98] P. Calaminici, K. Jug, A.M. K¨oster, J. Chem. Phys. 109, 7756 (1998).
[99] W.F. Murphy, W. Holzer, H.J. Bernstein, Appl. Spectrosc. 23, 211 (1969).
[100] K. Jug, F. Janetzko, A.M. K¨oster, J. Chem. Phys. 114, 5472 (2001)
[101] J. R. Mohallem, T. de O. Coura, L. G. Diniz, G. de Castro, D. Assafrao, T. Heine
J. Phys. Chem. A 112 (2008) 8896-8901.
[102] G. Schaftenaar, Molden3.2, CAOS/CAMM Center, University of Nijmegen (1997).
[103] S. Portmann, P.F. Fluekiger, CSCS/ETHZ and CSCS/UNI Geneva (2002).
[104] R.K. Pathak, S.R. Gadre, J. Chem. Phys. 93, 1770 (1990).
[105] A.D. Becke, K. Edgecomb, J. Chem. Phys. 92, 5397 (1990).
[106] B. Silvi, A. Savin, Nature 371, 683 (1994).
[107] W.E. Lorensen, H.E. Cline, Comp. Graph. 21, 163 (1987).
[108] M. Leboeuf, A.M. K¨oster, K. Jug, D.R. Salahub, J. Chem. Phys. 111, 4893 (1999)
[109] LAPACK driver routine (version 3.0) –
Univ. of Tennessee, Univ. of California Berkeley, NAG Ltd.,
Courant Institute, Argonne National Lab, and Rice University, 1999
[110] A.D. Becke, J. Chem. Phys. 88, 2547 (1987).
[111] R.E. Stratmann, G.E. Scuseria, M.J. Frisch, Chem. Phys. Lett. 257, 213 (1996).
[112] H.B. Jansen, P. Ross, Chem. Phys. Lett. 3, 140 (1969).
[113] P. Fuentealba, Y. Sim´on-Manson, Chem. Phys. Lett. 314, 108 (1999).
References
139
[114] K. Jug, B. Zimmermann, P. Calaminici, A.M. K¨oster, J. Chem. Phys. 116, 4497
(2002).
[115] V.G. Malkin, O.L. Malkina, M.E. Casida, D.R. Salahub, J. Am. Chem. Soc. 116,
5898 (1994).
[116] V.G. Malkin, O.L. Malkina, L.A. Eriksson, D.R. Salahub in Theoretical and Computational Chemistry, Eds. P. Politzer and J.M. Seminario (Elsevier, Amsterdam,
1995)
[117] W. Kutzelnigg, U. Fleischer, M. Schindler in NMR Basic Principles and Progress,
Vol. 23 (Springer-Verlag, Heidelberg, 1990).
[118] G. Seifert, D. Porezag, T. Frauenheim, Int. J. Quantum Chem. 58, 185 (1996).
[119] M. Elstner, D. Porezag, G. Jungnickel, J. Elsner, M. Haugk, T. Frauenheim, S.
Suhai, G. Seifert, Phys. Rev. B 58, 7260 (1998).
[120] L. Zhechkov, T. Heine, S. Patchkovskii, G. Seifert, H. A. Duarte, J. Chem. Theor.
Comput. 1, 841 (2005)
[121] M. Rapacioli, R. Barthel, T. Heine, G. Seifert, J. Chem. Phys. 126 124103 (2007).
[122] O. F. Oliveira, G. Seifert, T. Heine, H. A. Duarte, J. Braz. Chem. Soc. 20 1193
(2009).
[123] M. Springborg, R. C. Albers, K. Schmidt, Phys. Rev. B 57, 1427 (1998).
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement