DALTON2015 – lsDalton Program Manual

V. Bakken, R. Bast, P. Baudin, S. Coriani, R. Di Remigio,

P. Ettenhuber, J. J. Eriksen, T. Helgaker, S. Høst, I.-M. Høyvik,
R. Izsák, B. Jansı́k, P. Jørgensen, J. Kauczor, T. Kjærgaard,
A. Krapp, K. Kristensen, P. Merlot, C. Nygaard,
J. Olsen, S. Reine, V. Rybkin, P. Salek,
A. Teale, E. Tellgren, A. Thorvaldsen,
L. Thøgersen, M. Watson, and M. Ziolkowski
Contents

Preface iv

I lsDalton Installation Guide 1

1 Installation of lsDalton 2
1.1 Installation instructions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Hardware/software supported . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Source files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.4 Installing the program using the Makefile . . . . . . . . . . . . . . . . . . . 3
1.5 The Makefile.config . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.5.1 Intel compiler . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.5.2 Gfortran/gcc compiler . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.5.3 Using 64-bit integers . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.5.4 Using Compressed-Sparse Row matrices . . . . . . . . . . . . . . . . 13

II lsDalton User’s Guide 14

2 Getting started with lsDalton 15

2.1 The Molecule input file . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.1.1 BASIS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.1.2 ATOMBASIS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.1.3 User defined basis sets . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.2 The LSDALTON.INP file . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

3 Interfacing to lsDalton 24
3.1 The Grand Canonical Basis . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.2 Basisset and Ordering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.3 Normalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

i
CONTENTS ii

3.4 Matrix File Format . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

3.5 The lsDalton library bundle . . . . . . . . . . . . . . . . . . . . . . . . . . 27

III lsDalton Reference Manual 29

4 List of lsDalton keywords 30

4.1 **GENERAL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.1.1 *TENSOR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.2 **INTEGRAL basic keywords . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.3 **INTEGRAL advanced keywords . . . . . . . . . . . . . . . . . . . . . . . 35
4.3.1 *FMM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.4 **WAVE FUNCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.4.1 *DFT INPUT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.4.2 *DENSOPT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.4.3 $INFO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.5 **OPTIMIZE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.6 **DYNAMI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.7 **LOCALIZE ORBITALS . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.8 **RESPONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.8.1 *SOLVER . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.8.2 *DIPOLE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.8.3 *DIPOLEMOMENTMATRIX . . . . . . . . . . . . . . . . . . . . . 60
4.8.4 *ALPHA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.8.5 *BETA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.8.6 *DAMPED TPA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.8.7 *QUASIMCD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.8.8 *SHIELD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.8.9 *INASHIELD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.8.10 *ESGRAD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.8.11 *GAMMA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.8.12 *MOLGRA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.8.13 *TPA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.9 **DEC and **CC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.10 **PLT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.11 **PLTGRID . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.12 **PCM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
CONTENTS iii

IV Index 84
Preface

This is the manual for the lsDalton quantum chemistry program — Release DALTON2015
— for computing Hartree-Fock and DFT wave functions, energies, and molecular properties.
For correlated models, MP2 geometry optimizations and CCSD energies (not linear-scaling)
are available. Most parts of lsDalton employ linear scaling and massively parallel imple-
mentations, which makes it suitable for calculations on large molecular systems, in particular
when the calculations are carried out on large super computer architectures.
If results obtained with this code are published the following paper should be cited:

K. Aidas, C. Angeli, K. L. Bak, V. Bakken, R. Bast, L. Boman, O. Christiansen, R. Cimi-

raglia, S. Coriani, P. Dahle, E. K. Dalskov, U. Ekström, T. Enevoldsen, J. J. Eriksen, P. Et-
tenhuber, B. Fernández, L. Ferrighi, H. Fliegl, L. Frediani, K. Hald, A. Halkier, C. Hättig,
H. Heiberg, T. Helgaker, A. C. Hennum, H. Hettema, E. Hjertenæs, S. Høst, I.-M. Høyvik,
M. F. Iozzi, B. Jansik, H. J. Aa. Jensen, D. Jonsson, P. Jørgensen, J. Kauczor, S. Kirpekar,
T. Kjærgaard, W. Klopper, S. Knecht, R. Kobayashi, H. Koch, J. Kongsted, A. Krapp,
K. Kristensen, A. Ligabue, O. B. Lutnæs, J. I. Melo, K. V. Mikkelsen, R. H. Myhre, C. Neiss,
C. B. Nielsen, P. Norman, J. Olsen, J. M. H. Olsen, A. Osted, M. J. Packer, F. Pawlowski,
T. B. Pedersen, P. F. Provasi, S. Reine, Z. Rinkevicius, T. A. Ruden, K. Ruud, V. Rybkin,
P. Salek, C. C. M. Samson, A. Sánchez de Merás, T. Saue, S. P. A. Sauer, B. Schim-
melpfennig, K. Sneskov, A. H. Steindal, K. O. Sylvester-Hvid, P. R. Taylor, A. M. Teale,
E. I. Tellgren, D. P. Tew, A. J. Thorvaldsen, L. Thøgersen, O. Vahtras, M. A. Watson,
D. J. D. Wilson, M. Ziolkowski, and H. Ågren, ”The Dalton quantum chemistry program
system”, WIREs Comput. Mol. Sci. (doi: 10.1002/wcms.1172)

We emphasize the conditions under which the program is distributed. It is furnished

for your own use, and you may not redistribute it further, neither in whole nor in part. Even
though lsDalton is completely separate from the original Dalton program, the program
packages are distributed together in the DALTON2015 suite. Thus, anyone interested in ob-
taining lsDalton should check out the Dalton homepage at http://daltonprogram.org.
Visit http://dalton-installation.readthedocs.org for comprehensive installation in-

iv
CONTENTS v

structions, and visit the Forum http://daltonprogram.org/forum for tutorials on how to

run lsDalton and other useful information. The Forum is also open for general discus-
sions, and is the place to look/ask for assistance if you do not find the answers you are
looking for in this manual.
Finally note that the program represents experimental code that is under constant
development. No guarantees of any kind are provided, and the authors accept no responsi-
bility for the performance of the code or for the correctness of the results.
 Part I

lsDalton Installation Guide

1
Chapter 1

Installation of lsDalton

1.1 Installation instructions

lsDalton is configured using CMake, typically via the setup script, and subsequently
compiled using make or gmake.
Please consult http://dalton-installation.readthedocs.org for details. Help
and support from the Dalton community is available at http://daltonprogram.org/forum.

1.2 Hardware/software supported

lsDalton can be run on a variety of systems running the UNIX operating system. The
current release of the program supports Linux, Cray, SGI, and MacOS using GNU, Intel,
PGI and XL compilers.
The program is written in FORTRAN 90 and C, with machine dependencies isolated
using C preprocessor directives. All floating-point computations are performed in 64-bit
precision.
The program should be portable to other UNIX platforms. Users who port the codes
to other platforms are encouraged to communicate any required changes in the original
source with the appropriate C preprocessor directives to the authors.

1.3 Source files

The DALTON2015 program suite is distributed as a tar file obtainable from the Dalton
homepage at http://www.daltonprogram.org. After you have downloaded the file extract
it with:

tar xvzf DALTON-2015.*-Source.tar.gz

2
CHAPTER 1. INSTALLATION OF LSDALTON 3

Most of the extracted subdirectories under LSDALTON/* contain source code for the different
sections constituting the lsDalton program. Furthermore, there is a directory containing
various public domain routines (pdpack), a set of test jobs including reference output files
(test), a directory containing some useful pre- and post-processing programs supplied to
us from various users (tools), and finally this documentation (doc_LSDALTON).
In addition to the above directories, the main dalton directory contain several files
that support the CMake build mechanism (CMakeLists.txt and cmake/*) as well as a
directory containing all the basis sets supplied with this distribution (basis).

1.4 Installing the program using the Makefile

The lsDalton program can, in addition to the standard standard CMake method (see
section 1.1), be built using a distributed Makefile and configure script.
The main .../LSDALTON directory contains a shell script (configure). Based
on the automatic detection of the architecture, the script will try to build a suitable
Makefile.config on the basis of what kind of mathematical libraries are found, and user
input. Thus, to execute the script, type
> ./configure
The script will now try to detect the architecture, which usually works fine for most
common platforms. However, if for some reason this is not possible, you may choose one
of the supported architectures from a list (currently comprising only three architectures:
aix,linux, and darwin):
Although this script is in most cases capable of making a correct Makefile.config,
we always recommend users to check the created Makefile.config against local system
set-up. In particular, check that Makefile.config contains the path(s) to the proper math-
ematical libraries. The program runs many times slower if not linked to any mathematical
libraries (i.e. using lsDalton’s own “hand coded” matrix multiplications etc.)
During the execution of the configure script, you will be asked to decide which
compiler you would like to use.
The configure script searches for compilers in a ordered fashion, and in each instance
you can choose to accept the compiler. When a suitable compiler has been found, you will
be asked a few questions:

1. Do you want to use debug compiler options ? [y/N]

Specifies whether or not you would like to use low optimization level and debug
options, useful for developments. The y and N in the square braket indicate the two
options. y=Yes, N=No. The Capital letter is the default. In this case the default
(chosen by pressing enter) is No.
CHAPTER 1. INSTALLATION OF LSDALTON 4

2. Do you want to install the program with OpenMP parallelization? [Y/n]

Specifies whether or not you would like to compile the code using openMP (utilizing
several cores on a single node.
Note that the configure script do not support MPI parallelization and the resulting
Makefile.config have to be manually changed to use MPI.
Default is Y=Yes.

3. Does the Fortran compiler support the Fortran 2003 features like proce-
dure pointers and iso c bindings?

4. Use current directory as installation directory for binaries and scripts?

Denotes the directory where the executable and the run script will be moved to.

5. Use basis as default basis set directory?

This defines the directory where the program will look for the basis sets supplied with
the distribution, and this need to be changed according to the local directory structure.
We recommend that the basis sets in this directory are not changed. Changes to the
basis set should be done in a separate directory, and you may then supply this basis
set directory to the program at execution time using the command.

> export BASDIR=path/to/your/basissetdir

In order to ensure that your basis set is read correctly we recommend to use the

**INTEGRALS
.BASPRINT
6

Which will print the exponents and contraction coefficients read from file as well as
the normalized basis used in lsDalton.

6. Default scratch space

Determines the default head scratch directory where temporary files will be placed.
This value will be put in the lsdalton run script. However, note that jobs will be
run in a subdirectory of this head scratch-directory, according to the name of the job
files. If /work or /scratch is defined in the local directory structure, the script will
normally suggest /work/$USER or /scratch/$USER as default head scratch space.
CHAPTER 1. INSTALLATION OF LSDALTON 5

Compiler options will be supplied in Makefile.config. When Makefile.config

has been properly created and checked to agree with local system set-up, all that is needed
in order to to build an executable version of the code is to type (in the same directory as
the Makefile.config file):

> make

1.5 The Makefile.config

In this section we give a brief introduction to the Makefile.config, which contains compiler
instructions for the compilation of the program. The introduction is directed to new users
with a limited understanding of Makefiles and we only discuss subjects that may be relevant.
Note that the configure script will in most cases provided a Makefile.config which works and
no modifications are required from the user. Depending on the version of math libraries,
the configure script may not be able to find these and it may be required to manually add
paths to mathematical libraries.

1.5.1 Intel compiler

The first lines of a Makefile.config using the intel ifort/icc compilers may look like this

1 ARCH = linux
2 FMMDIR = mm
3 #
4 #
5 CPPFLAGS = -DSYS_LINUX -D_FILE_OFFSET_BITS=64
-D’INSTALL_BASDIR="/home/user/lsdalton/basis/"’ -DVAR_LINSCA
-DVAR_OMP -DCOMPILER_UNDERSTANDS_FORTRAN_2003 -DVAR_IFORT
6 F77 = ifort
7 F90 = ifort
8 FLNK = ifort
9 CC = icc
10 CXX = icpc
11 RM = rm -f
12 FFLAGS = -O3 -ip -w
13 F90OPTFLAGS = -O3 -ip -w -fpp1 -openmp
14 SAFEFFLAGS = -O2 -w
15 CFLAGS = -O3 -ip -restrict -DRESTRICT=restrict -openmp -DUSE_UNDERSCORES
16 CXXFLAGS = -Wall -g -wd981 -wd279 -wd383 -vec-report0 -wd1572 -wd177 -fno-rtti
-fno-exceptions -O3 -openmp -DUSE_UNDERSCORES
CHAPTER 1. INSTALLATION OF LSDALTON 6

17 INCLUDES =
18 LIBS = -lstdc++ -openmp -liomp5 -lpthread
19 INSTALLDIR = /home/user/lsdalton
20 PDPACK_EXTRAS = linpack.o eispack.o cholesky.o invroutines.o
gp_dlapack.o gp_zlapack.o gp_dblas3.o gp_dblas2.o gp_dblas1.o gp_zblas.o

We now describe this Makefile.config file line by line, before we discuss a number of issues.

• Line 1: ARCH is set to the architecture. (for an aix architecture ARCH=rs6000, for
a Mac computer ARCH=darwin, these will be discussed shortly)

• Line 2: The directory of the Fast multipole moment code

• Line 5: CPPFLAGS is a list of precompiler flags which is required for the program to
compile correctly

-DSYS LINUX tells the lsDalton program that this is a linux architecture
-DINSTALL BASDIR The basisset installation directory
-DVAR LINSCA Is a mandatory keyword!
-DVAR OMP tells the lsDalton program that OpenMP is used
-DCOMPILER UNDERSTANDS FORTRAN 2003 tells the lsDalton program that
the Fortran compiler support the Fortran 2003 features like procedure pointers
and iso c bindings
-DVAR IFORT tells the lsDalton program that the Fortran compiler is an Intel
ifort compiler

Usually this line works out of the box

• Line 6: specifies the compiler used to compile fortran 77 files (obsolete)

• Line 7: specifies the compiler used to compile fortran 90 files

• Line 8: specifies the compiler used for linking

• Line 9: specifies the compiler used to compile c files

• Line 10: specifies the compiler used to compile c files

• Line 11: specifies the command to be used when cleaning

• Line 12-16: specifies compiler flags that should be used to compile fortran files. Nat-
urally these flags depend on the compiler. Here we use a intel fortran compiler (ifort)
CHAPTER 1. INSTALLATION OF LSDALTON 7

-O3 Specifies the code optimization for applications. Possible options are O0, O1,
O2, or O3, with O2 as default
-ip Additional interprocedural optimizations for single-file compilation are enabled.
-w Disables all warning messages (same as -nowarn)
-fpp1 Runs the Fortran preprocessor on source files before compilation. Syntax is
fpp[n], where n can be 0, 1, 2, or 3, where 1, 2, or 3 tells the compiler to run the
preprocessor
-openmp Enables the parallelizer to generate multi-threaded code based on the
OpenMP directives. This option sets option automatic, which causes all lo-
cal, non-saved variables to be allocated to the run-time stack, this may result in
a lack of stack-memory as discussed below.

• Line 15: specifies compiler flags that should be used to compile C files. Naturally
these flags depend on the compiler. Here we use a intel fortran compiler (icc). Some
of the compiler options have already been explained under the ifort options and they
will not be explained again.

• Line 16: specifies compiler flags that should be used to compile C++ files.

-restrict Pointer disambiguation is enabled with the restrict qualifier.

-DRESTRICT=restrict Precompiler flag used in DFT code
-DUSE UNDERSCORES required to define the communication between fortran and
C files.

• Line 17: Empty by default. Used if for some reason, any additional directories should
be included.

• Line 18: Libraries you wish to include. -openmp (used to be -lguide) and -lpthread are
required for openmp parallelization. See Sections 1.5.1.2 and 1.5.2.1 for more details
on how to link to mkl and other libraries. If you use the old version of ifort you need
to replace -openmp with -lguide.

• Line 19: the install directory

• Line 20: lsDalton provides its own blas and lapack libraries, located in the pdpack
directory. These should only be used if no mathematical libraries are available, since
they slow down the code significantly.
In case you link to an external library, like the intel mkl library, you only need to
include ”linpack.o eispack.o cholesky.o invroutines.o”.
CHAPTER 1. INSTALLATION OF LSDALTON 8

In case you do not link to an external library, you need to include ”linpack.o eispack.o
cholesky.o invroutines.o gp dlapack.o gp zlapack.o gp dblas3.o gp dblas2.o gp dblas1.o
gp zblas.o.”

1.5.1.1 OpenMP

Note that when compiling using the -openmp option it may be required to increase the
stack memory. On a linux machine the size of the stack-memory can be viewed using the
command

> ulimit -a

and changed by

> ulimit -s newstacksize

The newstacksize should be a large number or unlimited.

1.5.1.2 Math Library: MKL

Linking to a math library is crucial for optimal performance. The linking depends on the
library available. We encourage to first compile with the default Makefile.config composed
by the configure script. This ensures that your chosen compiler works with the lsDalton
source code, and then you can try to link to a math library. There is a number of different
ways that can be used to link lsDalton with the mkl library. An example of a brute force
way is the following:

16 LIBS =
/opt/intel/Compiler/11.1/056/mkl/lib/em64t/libmkl_solver_lp64_sequential.a
-Wl,--start-group /opt/intel/Compiler/11.1/056/mkl/lib/em64t/libmkl_intel_lp64.a
/opt/intel/Compiler/11.1/056/mkl/lib/em64t/libmkl_sequential.a
/opt/intel/Compiler/11.1/056/mkl/lib/em64t/libmkl_core.a -Wl,--end-group
-lguide -lpthread
17 INSTALLDIR = /home/user/lsdalton
18 PDPACK_EXTRAS = linpack.o eispack.o

for the sequential linking (no OpenMP), assuming that the intel compiler is located at
/opt/intel/Compiler/11.1/056 and that the mkl library is located at /opt/intel/Compil-
er/11.1/056.
This link line have been obtained from the ”Intel Math Kernel Library Link Line
Advisor” at http://software.intel.com/en-us/articles/intel-mkl-link-line-advisor.
The OpenMP version of the library can be used in the following way.
CHAPTER 1. INSTALLATION OF LSDALTON 9

16 LIBS =
/opt/intel/Compiler/11.1/056/mkl/lib/em64t/libmkl_solver_lp64_sequential.a
-Wl,--start-group /opt/intel/Compiler/11.1/056/mkl/lib/em64t/libmkl_intel_lp64.a
/opt/intel/Compiler/11.1/056/mkl/lib/em64t/libmkl_intel_thread.a
/opt/intel/Compiler/11.1/056/mkl/lib/em64t/libmkl_core.a -Wl,--end-group
-lguide -lpthread
17 INSTALLDIR = /home/user/lsdalton
18 PDPACK_EXTRAS = linpack.o eispack.o

A somewhat more elegant way is to add the MKL directory to the INCLUDE line

15 INCLUDES = -I/home/user/lsdalton/include
16 LIBS = -I/opt/intel/Compiler/11.1/056/mkl/include -lmkl_intel_lp64
-lmkl_sequential -lmkl_core
17 INSTALLDIR = /home/user/lsdalton
18 PDPACK_EXTRAS = linpack.o eispack.o

or a threaded version

15 INCLUDES = -I/home/user/lsdalton/include
16 LIBS = -I/opt/intel/Compiler/11.1/056/mkl/include -lmkl_intel_lp64
-lmkl_intel_thread -lmkl_core -lguide -lpthread
17 INSTALLDIR = /home/user/lsdalton
18 PDPACK_EXTRAS = linpack.o eispack.o

1.5.2 Gfortran/gcc compiler

The first lines of a Makefile.config using the gfortran/gcc (version 4.5.1) on a linux archi-
tecture may look like this

1 ARCH = linux
2 FMMDIR = mm
3 #
4 #
5 CPPFLAGS = -DSYS_LINUX -D_FILE_OFFSET_BITS=64
-D’INSTALL_BASDIR="/home/user/lsdalton/basis/"’ -DVAR_LINSCA
-DVAR_OMP -DCOMPILER_UNDERSTANDS_FORTRAN_2003 -DGFORTRAN=472
6 F77 = gfortran
7 F90 = gfortran
8 FLNK = gfortran
9 CC = gcc
CHAPTER 1. INSTALLATION OF LSDALTON 10

10 CXX = g++
11 RM = rm -f
12 FFLAGS = -march=x86-64 -O3 -ffast-math -funroll-loops
-ftree-vectorize -ffloat-store
13 F90OPTFLAGS = -march=x86-64 -O3 -ffast-math -funroll-loops
-ftree-vectorize -ffloat-store -I. -x f95-cpp-input -ffloat-store -fopenmp
14 SAFEFFLAGS = -march=x86-64 -O3 -ffast-math -funroll-loops
-ftree-vectorize -ffloat-store
15 CFLAGS = -march=x86-64 -O3 -ffast-math -funroll-loops
-ftree-vectorize -std=c99 -DRESTRICT=restrict -DFUNDERSCORE=1 -ffloat-store
-DUSE_UNDERSCORES
16 CXXFLAGS = -march=x86-64 -g -O3 -Wall -fno-rtti -fno-exceptions -fopenmp -DUSE_UNDERS
17 INCLUDES = -I/home/user/lsdalton/include
18 LIBS = -lstdc++ -lgomp
19 INSTALLDIR = /home/user/lsdalton
20 PDPACK_EXTRAS = linpack.o eispack.o cholesky.o invroutines.o gp_dlapack.o
gp_zlapack.o gp_dblas3.o gp_dblas2.o gp_dblas1.o gp_zblas.o
On a MAC computer a Makefile.config using gfortran/gcc could look like this
1 ARCH = darwin
2 FMMDIR = mm
3 #
4 #
5 CPPFLAGS = -DSYS_LINUX -D_FILE_OFFSET_BITS=64
-DVAR_SPLITFILES -DVAR_G77 -DGFORTRAN=472 -DVAR_LINSCA
-D’INSTALL_BASDIR="/home/user/lsdalton/basis/"’
-DVAR_OMP -DCOMPILER_UNDERSTANDS_FORTRAN_2003
6 F77 = gfortran
7 F90 = gfortran
8 FLNK = gfortran
9 CC = gcc
10 CXX = g++
11 RM = rm -f
12 FFLAGS = -O3 -ffast-math -fexpensive-optimizations -funroll-loops
-ftree-vectorize
13 F90OPTFLAGS = -O3 -ffast-math -fexpensive-optimizations -funroll-loops
-ftree-vectorize -I. -x f95-cpp-input
14 SAFEFFLAGS = -O2 -ffast-math -fexpensive-optimizations -funroll-loops
-ftree-vectorize
CHAPTER 1. INSTALLATION OF LSDALTON 11

15 CFLAGS = -O3 -ffast-math -fexpensive-optimizations -funroll-loops

-ftree-vectorize -std=c99 -DRESTRICT=restrict -DFUNDERSCORE=1 -DUSE_UNDERSCORES
16 CXXFLAGS = -O3 -Wall -fno-rtti -fno-exceptions -fopenmp -DUSE_UNDERSCORES
17 INCLUDES = -I/home/user/lsdalton/include
18 LIBS = -lstdc++ -lpthread -lm -lgfortranbegin -lgfortran
19 INSTALLDIR = /home/user/lsdalton
20 PDPACK_EXTRAS = linpack.o eispack.o cholesky.o invroutines.o gp_dlapack.o
gp_zlapack.o gp_dblas3.o gp_dblas2.o gp_dblas1.o gp_zblas.o

We now describe these Makefile.config files in comparison to section 1.5.1.

Line 1 to 5 is unchanged for the linux architecture, but on the MAC machine -
DGFORTRAN must be included in the CPPFLAGS.
Line 6-9 specifies the use of the gfortran/gcc compilers.
Line 11-13 specify compiler flags and are naturally different for different compilers

-march=x86-64 This specifies the name of the target architecture. GCC uses this name
to determine what kind of instructions it can emit when generating assembly code.
In this case it is a 64 bit machine. The use of this compiler flag is not required, but
recommended.

-O3 Specifies the code optimization for applications. Possible options are O0, O1, O2, or
O3, with O2 as default

-ffast-math A Compiler option to optimize floating-point arithmetic

-funroll-loops Unroll loops whose number of iterations can be determined at compile time
or upon entry to the loop

-ftree-vectorize Perform loop vectorization on trees. This flag is enabled by default at -O3

fexpensive-optimizations Perform a number of minor optimizations that are relatively ex-

pensive.

-ffloat-store This may be required depending on the gfortran version, but usually not. it
enforces IEEE compliance

-fopenmp Enables the parallelizer to generate multi-threaded code based on the OpenMP
directives.

-std=c99 Determines the language standard, in this case C 99 standard.

Line 16 are libraries you wish to include. -lgomp are required for openmp parallelization.
CHAPTER 1. INSTALLATION OF LSDALTON 12

1.5.2.1 Math Library: Linux MKL

Linking to a math library is crucial for optimal performance. The linking depends on the
library available. We encourage to first compile with the default Makefile.config composed
by the configure script. This ensures that your chosen compiler works with the lsDalton
source code, and then you can try to link to a math library. lsDalton provides its own blas
and lapack libraries, which is placed in the pdpack directory. However it is encouraged to
link to an external library, like the intel mkl library, then you only need to include linpack.o
eispack.o. In the case that you do not link to an external library, you need to include
all. linpack.o eispack.o. gp dlapack.o gp zlapack.o gp dblas3.o gp dblas2.o gp dblas1.o
gp zblas.o.
Note that linking to mkl library using OpenMP requires to use -liomp5 in stead of
-lgomp

17 INCLUDES = -I/home/user/lsdalton/include
-I/opt/intel/Compiler/11.1/056/mkl/include
18 LIBS = -lmkl_intel_lp64 -lmkl_intel_thread -lmkl_core
-lguide -lpthread -liomp5
19 INSTALLDIR = /home/user/lsdalton
20 PDPACK_EXTRAS = linpack.o eispack.o

or

17 INCLUDES =
18 LIBS =
/opt/intel/Compiler/11.1/056/mkl/lib/em64t/libmkl_solver_lp64_sequential.a
-Wl,--start-group /opt/intel/Compiler/11.1/056/mkl/lib/em64t/libmkl_intel_lp64.a
/opt/intel/Compiler/11.1/056/mkl/lib/em64t/libmkl_intel_thread.a
/opt/intel/Compiler/11.1/056/mkl/lib/em64t/libmkl_core.a -Wl,--end-group
-lguide -lpthread -liomp5
19 INSTALLDIR = /home/user/lsdalton
20 PDPACK_EXTRAS = linpack.o eispack.o

1.5.3 Using 64-bit integers

For large scale calculations, it may be necessary to compile the program using 64-bit inte-
gers instead of the default 32-bit integers. To do this, you need to include ”-DVAR INT64”
in the CPPFLAGS (in addition to what is already there). Furthermore, the FFLAGS,
F90OPTFLAGS, SAFEFFLAGS, and CFLAGS must be changed depending on you com-
piler. For ifort/icc, you need to add ”-i8”, and for XLF/XLC, you need to add ”-qintsize=8”.
The use of 64-bit integers has not been tested for gfortran. Note that you also have to change
CHAPTER 1. INSTALLATION OF LSDALTON 13

the mathematical libraries in LIBS to their 64-bit counterparts. This is not always a simple
task, and you may need to ask your system administrator.

1.5.4 Using Compressed-Sparse Row matrices

When using Compressed-Sparse Row (CSR) matrices, only non-zero elements in matrices
are stored. When matrices are sparse enough, linear scaling in both memory and cpu time
is obtained. However, for small or non-sparse systems, there is a considerable overhead in
cpu time compared to using standard dense matrices, so don’t use it unless you need it.
You enable CSR by putting ”-DVAR MKL” and ”-DVAR CSR” in your CPPFLAGS (in
addition to what is already there. You also need to link to MKL’s CSR library (what we
have in lsDalton is an interface to MKL CSR). Furthermore, to run a calculation using
CSR matrices, you always need .CSR under *DENSOPT in the input file LSDALTON.INP
(described in detail in chapters 2 and 4).
 Part II

lsDalton User’s Guide

14
Chapter 2

Getting started with lsDalton

In this chapter we give an introduction to the two input files needed for doing a calculation
with the lsDalton program (MOLECULE.INP and LSDALTON.INP), as well as the shell
script file that is supplied with the program for moving relevant files to the scratch-directory
and back to the home directory after a calculation has completed. A couple of examples of
input files are also provided. Finally, the different output files generated by the program
are discussed.
Note! The Dalton and lsDalton programs use different input keywords!
Some MOLECULE.INP files used for lsDalton can also be used for Dalton –
but NOT vice versa! The DALTON.INP input file of Dalton is different from
the LSDALTON.INP input file of lsDalton. Thus, Dalton keywords CANNOT
be assumed to work in lsDalton or vice versa.
Before going into detail with the MOLECULE.INP and LSDALTON.INP files in
chapters 2.1 and 2.2, we provide a simple input example for the impatient reader. The
example below will yield the HF energy for water and the two lowest lying excitation
energies and their associated transition strengths:

MOLECULE.INP:

BASIS
cc-pVDZ Aux=Ahlrichs-Coulomb-Fit
water R(OH) = 0.95Aa , <HOH = 109 deg.
Distance in Aangstroms
Atomtypes=2 Angstrom
Charge=8.0 Atoms=1
O 0.00000 0.00000 0.00000
Charge=1.0 Atoms=2

15
CHAPTER 2. GETTING STARTED WITH LSDALTON 16

H 0.55168 0.77340 0.00000

H 0.55168 -0.77340 0.00000

LSDALTON.INP:

**WAVE FUNCTIONS
.HF
**RESPONS
.NEXCIT
2
*END OF INPUT

To run this simple example, you may (i) create a new directory, (ii) copy the two
files above into this directory, and (iii) run the calculation from your newly created direc-
tory using the lsdalton.x executable located in your build directory (as mentioned above,
installation details are given at http://dalton-installation.readthedocs.org).
The converged HF energy can be found in the output file LSDALTON.OUT by
grepping for ”Final HF energy”:

Final HF energy: -76.026476337669

while the two excitation energies and associated transition strengths are given at the end
of LSDALTON.OUT:

***********************************************************************************
* ONE-PHOTON ABSORPTION RESULTS (in a.u.) *
***********************************************************************************
Excitation Transition Dipole Moments Oscillator
Energies x y z Strengths
====================================================================================
0.33897330 -0.94893963E-15 -0.30390933E-15 0.36349099 0.29858056E-01
0.42634048 0.57831041 -0.15070410E-14 -0.14539976E-14 0.95057706E-01

Note that if you run this test your results may vary slightly from the above due to numerical
inaccuracies.

2.1 The Molecule input file

Basically, the Molecule file contains the following information:
CHAPTER 2. GETTING STARTED WITH LSDALTON 17

• The Cartesian coordinates and nuclear charges of all atoms

• The basis set(s) to be used

Please note the following:

• Supplying the molecular geometry in the form of a Z-matrix is not supported by

lsDalton since this format is ill suited for large scale calculations.

• Molecular symmetry is not exploited by lsDalton (since the target systems of the
program are large biomolecules, which hardly ever contains any symmetry).

• The lsDalton program does not support the use of effective core potentials.

There are two different types of molecule files, using either BASIS or ATOMBASIS.
ATOMBASIS is relevant only if different basis sets on individual atoms are required. In the
first two subsections, we describe these two types of Molecule files using simple examples.
The basis can be selected from the provided basis-set library or can be specified by the user
(as described under section 2.1.3).

2.1.1 BASIS
Here we describe our example water molecule input file line by line using the BASIS
Molecule format:

1 BASIS
2 cc-pVDZ Aux=Ahlrichs-Coulomb-Fit
3 water R(OH) = 0.95Aa , <HOH = 109 deg.
4 Distance in Aangstroms
5 Atomtypes=2 Angstrom
6 Charge=8.0 Atoms=1
7 O 0.00000 0.00000 0.00000
8 Charge=1.0 Atoms=2
9 H 0.55168 0.77340 0.00000
10 H 0.55168 -0.77340 0.00000

• Line 1: The word ”BASIS”, indicating that the same basis set should be used for all
atoms.

• Line 2: The name of the basis set, in this case cc-pVDZ. Optional: The name of the
auxiliary basis set, in this case Ahlrichs-Coulomb-Fit (only referenced if density-fitting
is requested in LSDALTON.INP)
CHAPTER 2. GETTING STARTED WITH LSDALTON 18

• Lines 3-4: Comments (may be left blank)

• Line 5: Atomtypes= Number of different atoms - in the case of H2 O, there are

two different atom types, H and O. Angstrom If this keyword is omitted, the pro-
gram will assume that coordinates are given in atomic units (1bohr=0.5291772083Å).
Other options in this line are: Charge= Molecular charge - assumed to be zero if
omitted. Nosymmetry This have no effect as no pointgroup symmetry have been
implemented. SubSystems Use Subsystem labels for doing Counter Poise Corrected
interaction energies

• Lines 6 and 8: A line of this type has to come before the Cartesian coordinates of
each atom type, indicating a) Charge= the nuclear charge of the atom type, and b)
Atoms= the number of atoms of this type. Other options in this line are: SubSys-
tem=A Use Subsystem label A for these atoms phantom Place Phantom atom (add
Basis functions but no actual atom) pointcharge place pointcharge without Basis
functions

• Lines 7, 9-10: Cartesian coordinates of the atoms (in Angstrom, since this was specified
in line 5). An atom name (1-4 characters long) must be placed at the beginning of
these lines.

2.1.2 ATOMBASIS
Our example water molecule may also be specified as an ATOMBASIS Molecule file :
1 ATOMBASIS
2 water R(OH) = 0.95Aa , <HOH = 109 deg.
3 Distance in Aangstroms
4 Atomtypes=2 Angstrom
5 Charge=8.0 Atoms=1 Basis=cc-pVDZ Aux=Ahlrichs-Coulomb-Fit
6 O 0.00000 0.00000 0.00000
7 Charge=1.0 Atoms=2 Basis=cc-pVDZ Aux=Ahlrichs-Coulomb-Fit
8 H 0.55168 0.77340 0.00000
9 H 0.55168 -0.77340 0.00000
As seen from the example, there is little difference between the two formats. Line 2 in the
BASIS file has been removed, and instead the basis sets (and optionally, the auxiliary basis
sets) are now given for each atom. In this example they are the same, as it would probably
not make much sense to use different basis sets for the atoms in water. This can make
sense, though, for larger molecules. An example could be a biomocule containing an active
site with one or more transition metal atoms. In may be useful to describe such an active
site with a more sophisticated basis set than the surroundings.
CHAPTER 2. GETTING STARTED WITH LSDALTON 19

2.1.3 User defined basis sets

The user may supply his or her own basis set in two ways. Both ways require that the
MOLECULE.INP file is argumented with a basis as in the following two examples

1 BASIS
2 USERDEFINED
3 water R(OH) = 0.95Aa , <HOH = 109 deg.
4 Distance in Aangstroms
5 Atomtypes=2 Angstrom
6 Charge=8.0 Atoms=1
7 O 0.00000 0.00000 0.00000
8 Charge=1.0 Atoms=2
9 H 0.55168 0.77340 0.00000
10 H 0.55168 -0.77340 0.00000
11
12 USERDEFINED BASIS
13 a 1
14 $ HYDROGEN (4s,1p) -> [2s,1p]
15 $ HYDROGEN segmented contracted
16 $ S-TYPE FUNCTIONS
17 4 2 0
18 18.7311370 0.03349460 0.00000000
19 2.8253937 0.23472695 0.00000000
20 0.6401217 0.81375733 0.00000000
21 0.1612778 0.00000000 1.00000000
22 $ P-TYPE FUNCTIONS
23 1 1 0
24 1.1000000 1.00000000
25 a 8
26 $ OXYGEN (10s,4p) -> [3s,2p]
27 $ S-TYPE FUNCTIONS
28 10 3 0
29 5484.6717000 0.00183110 0.00000000 0.00000000
30 825.2349500 0.01395010 0.00000000 0.00000000
31 188.0469600 0.06844510 0.00000000 0.00000000
32 52.9645000 0.23271430 0.00000000 0.00000000
33 16.8975700 0.47019300 0.00000000 0.00000000
34 5.7996353 0.35852090 0.00000000 0.00000000
CHAPTER 2. GETTING STARTED WITH LSDALTON 20

35 15.5396160 0.00000000 -0.11077750 0.00000000

36 3.5999336 0.00000000 -0.14802630 0.00000000
37 1.0137618 0.00000000 1.13076700 0.00000000
38 0.2700058 0.00000000 0.00000000 1.00000000
39 $ P-TYPE FUNCTIONS
40 4 2 0
41 15.5396160 0.07087430 0.00000000
42 3.5999336 0.33975280 0.00000000
43 1.0137618 0.72715860 0.00000000
44 0.2700058 0.00000000 1.00000000
45 $ D-TYPE FUNCTIONS
46 1 1 0
47 0.8000000 1.00000000

Line 13 to 24 contain the Basis set information for the Hydrogen atom. Line 13 denote the
nuclear charge. Line 14 through 16 are comments. Line 17 states that there are 4 primitive
functions and 2 contracted functions. Lines 18 to 21 contain information about the primitive
functions. The first number is the exponent and the next 2 numbers are the coefficients
for the 2 contracted functions. Line 22-24 contain info about p orbitals on hydrogen, while
Line 25-47 contain similar information about oxygen.
Alternative the userdefined basis may be used in connection with the ATOMBASIS

ATOMBASIS
water R(OH) = 0.95Aa , <HOH = 109 deg.
Distance in Aangstroms
Atomtypes=2 Angstrom
Charge=8.0 Atoms=1 Basis=cc-pVDZ
O 0.00000 0.00000 0.00000
Charge=1.0 Atoms=2 Basis=USERDEFINED
H 0.55168 0.77340 0.00000
H 0.55168 -0.77340 0.00000

USERDEFINED BASIS
a 1
$ HYDROGEN (4s) -> [2s]
$ HYDROGEN (1p)
$ S-TYPE FUNCTIONS
4 2 0
18.7311370 0.03349460 0.00000000
CHAPTER 2. GETTING STARTED WITH LSDALTON 21

2.8253937 0.23472695 0.00000000

0.6401217 0.81375733 0.00000000
0.1612778 0.00000000 1.00000000
$ P-TYPE FUNCTIONS
1 1 0
1.1000000 1.00000000

Finally the basis set files from the Basis Set exchange (EMSL) https://bse.pnl.gov/bse/portal
may be downloaded and put into the basis set directory (the directory called ”basis” lo-
cated at the root of the Dalton library) and used directly. The name specified by the
input should match the name of the file in the basis set directory (see testcase LSDAL-
TON/test/linsca/linsca emsl for an example).

2.2 The LSDALTON.INP file

In this section, we show two typical examples of a LSDALTON.INP file, one for small
molecules and one for large molecules. Many other examples of input files for requesting
e.g. specific molecular properties may be found at http://daltonprogram.org, and a
complete reference manual of all keywords is found in Chapter 4.
In general, the LSDALTON.INP is divided in different sections under the headlines
(see Section 4 for details):

• **GENERAL contains general settings (optional)

• **INTEGRAL contains settings for the calculation of integrals (optional)

• **WAVE FUNCTION contains information about the wave function (e.g. HF/DFT)
and settings for the optimization of the wave function (mandatory)

• **OPTIMIZE contains settings for geometry optimization (optional)

• **DYNAMI contains settings for Born-Oppenheimer molecular dynamics (optional)

• **LOCALIZE ORBITALS contains settings for orbital localization procedure (op-

tional)

• **RESPONS contains information about requested molecular properties for HF and

DFT (optional)

• **DEC or **CC contains info about MP2 and coupled-cluster calculations (optional)

• **PLT contains information about construction of *.plt files which may be used to
visualize densities and orbitals using e.g. the Chimera program [1] (optional).
CHAPTER 2. GETTING STARTED WITH LSDALTON 22

• **PLTGRID contains information about grid used for construction of *.plt files (op-
tional).

• **PCM contains information about inclusion of a continuum description of the solvent

via the polarizable continuum model [2, 3].

Each of these sections may contain subsections, indicated by a single asterisk. The LSDAL-
TON.INP file should always end with *END OF INPUT.
A typical example of LSDALTON.INP for a small molecule could look like:

**WAVE FUNCTIONS
.HF
*DENSOPT
.RH
.DIIS
.CONVTHR
1.0D-6
**RESPONS
.NEXCIT
5
*END OF INPUT

Here, we have requested a Hartree-Fock optimization under **WAVE FUNCTION. The

subsection *DENSOPT contains setting for how the density should be optimized. The .RH
(.i.e. Roothaan-Hall) and .DIIS keywords request a standard diagonalization combined
with the DIIS scheme for convergence acceleration. With the keyword .CONVTHR (i.e.
convergence threshold) it is requested that iterations are terminated when the Frobenius
norm of the SCF gradient, divided by the square root of the number of occupied orbitals,
is smaller than 10−6 . Finally, we have under **RESPONS requested the calculation of the
five lowest excitation energies with .NEXCIT (i.e. number of excitation energies).
A typical example of LSDALTON.INP for a large molecule could look like:

**INTEGRAL
.DENSFIT
**WAVE FUNCTIONS
.DFT
BP86
*DENSOPT
.START
TRILEVEL
.ARH
CHAPTER 2. GETTING STARTED WITH LSDALTON 23

.CONVDYN
STANDARD
*END OF INPUT

Under **INTEGRAL, we have requested the use of density-fitting to speed up the calcula-
tion (the use of J-engine is default). Under **WAVE FUNCTIONS, we have requested a
DFT calculation using the BP86 exchange-correlation functional.
Since this is a calculation on a large molecule, we request the trilevel starting guess
(atomic, valence, and then full optimization) and the Augmented Rothaan-Hall (ARH)
method for density optimization. Note that ARH is default, so this keyword is actually
unnecessary. ARH is more robust than the standard diagonalization/DIIS scheme. Fur-
thermore, the ARH scaling with system size is asymptotically linear provided sparse-matrix
algebra is employed. You may use sparse-matrix algebra by putting .CSR (i.e. Compressed-
Sparse Row) under *DENSOPT (NB: requires linking to the MKL library). We have used
a dynamic SCF convergence threshold (.COVNDYN, see 4.4.2). This is suitable for calcu-
lations on large molecules, since the standard SCF convergence threshold is based on the
Frobenius norm of the SCF gradient, which is not size-extensive. The use of the STAN-
DARD dynamic threshold corresponds to 10−5 times the square root of the number of
electrons in the system.
Chapter 3

Interfacing to lsDalton

In this chapter we give an introduction to the normalization, basis set and ordering of
atomic orbitals used in the lsDalton program.
This is provided in case you want to interface to the lsDalton program or want to
read one of the files written by the lsDalton program, for instance the local orbitals. Cer-
tain functionality can also be incorporated into other programs by linking to the lsDalton-
library bundle, as outlined in section 3.5.

3.1 The Grand Canonical Basis

Unlike all other Quantum Chemistry Programs the lsDalton program, uses the so called
grand canonical basis [4, 5] as the internal default basis. The results therefore cannot
directly be compared to other programs unless the

.NOGCBASIS

keyword is specified. This keyword deactivates the use of the grand canonical basis and
uses the input basis instead. An exception is Dunning’s correlation consistent basis sets
(cc-pVDZ, cc-pVTZ, ..., aug-cc-pVDZ, etc.) for which the .NOGCBASIS keyword is used
by default.

3.2 Basisset and Ordering

In lsDalton the default basis set are real-valued spherical harmonic Gaussian type orbitals
(GTOs)
2
Gilm (r; ai ; A) = Slm (xA ; yA ; zA ) exp(−ai rA ) (3.1)
where Slm (xA ; yA ; zA ) is a real solid harmonic introduced in section 6.4.2 of Ref. [6] (see
Table 6.3). Concerning the ordering of basis functions, it is one atomtype after the other

24
CHAPTER 3. INTERFACING TO LSDALTON 25

as given in the input, and for each atomtype, it is one atom after another as given in input.
For each atom increasing with angular momentum l, within one angular momentum l, the
ordering is first the components (e.g. px , py , pz ), and then the contracted functions.
So for an atom consisting of 3 s functions 2 p functions and 1 d function, the ordering
is {1s; 2s; 3s; 1px ; 1py ; 1pz ; 2px ; 2py ; 2pz ; 1dxy ; 1dyz ; 1dz2−r2 ; 1dxz ; 1dx2−y2 } In lsDalton the
ordering of the basis functions are m = {−l, · · · , 0, · · · , l} for a given angular momentum
l. Although for computational efficiency the p orbitals are treated as a special case with
p = {px , py , pz } which corresponds to m = {1, −1, 0}.

3.3 Normalization
For an unnormalized primitive spherical harmonic GTO, the overlap is given by

π 3/2 1
 

GTO GTO  p 2
χilm |χilm = (Ni ) (3.2)
2ai (4ai )l
(where the last term corresponds to 1/(2p)l according to for example Eq. 9.3.8 of Ref. [6].
This gives the primitive normalization factor N p
(4ai )l/2+3/4
Nip = (3.3)
π 3/4
Naturally the default in lsDalton is not primitive GTOs but contracted GTOs written as
linear combinations of primitive spherical-harmonic GTOs of different exponents
X
Gµlm (r; a; A) = Gilm (r; ai ; A)dnorm
iµ (3.4)
i

in lsDalton we determine dnorm

iµ as

dnorm
iµ = Nip Niµ
c basis
diµ (3.5)

As explained, Nip is the normalization of the primitive GTOs, dbasis iµ are the unmodified
c
contraction coefficients read from the basis set file. Ni is the normalization of the contracted
functions and is determined using the overlap between two normalized primitive GTOs of
the same angular momentum, but different exponents
D E  p4ai aj  32 +l
GTO GTO
χai lm |χaj lm = (3.6)
ai + aj
so that the contraction coefficients for the contracted function Nic are found by first finding
the overlap
3
4ai aj 2 +l
X p
basis basis

GTO GTO  X basis basis
Sµ = diµ djµ χilm |χjlm = diµ djµ (3.7)
ai + aj
ij ij
 CHAPTER 3. INTERFACING TO LSDALTON 26

where dbasis
iµ are the unmodified coefficients and then construct the coefficients

basis   l +3  3
p diµ dbasis
iµ 2 4 1 4
dnorm
iµ = Nip Niµ
c basis
diµ = Ni p = p 4ai (3.8)
Sµ Sµ π

3.4 Matrix File Format

During an lsDalton calculation a number of files may be generated. This list of files
include

dens.restart - The density matrix

fock.restart - The fock matrix
overlapmatrix - The overlap matrix
cmo_orbitals.u - The canonical Molecular orbitals
lcm_orbitals.u - The local Molecular orbitals

the files are all written using the Fortran 90 code

1 OPEN(UNIT=IUNIT , FILE= ’ dens . r e s t a r t ’ ,STATUS= ’UNKNOWN’ ,FORM= ’UNFORMATTED’ ,IOSTAT
=IOS )
2 WRITE( i u n i t ) A%Nrow , A%Ncol
3 WRITE( i u n i t ) (A%elms ( I ) , I =1 ,A%nrow∗A%n c o l )
4 CLOSE(UNIT=IUNIT , ,STATUS= ’KEEP ’ )

Where A is a derived type containing the number of rows (A%nrow), the number of
columns (A%ncol), and the elements stored in a vector array (A%elms).
The dens.restart is different as it contains an additional logical
1 OPEN(UNIT=IUNIT , FILE= ’ dens . r e s t a r t ’ ,STATUS= ’UNKNOWN’ ,FORM= ’UNFORMATTED’ ,IOSTAT
=IOS )
2 WRITE( i u n i t ) A%Nrow , A%Ncol
3 WRITE( i u n i t ) (A%elms ( I ) , I =1 ,A%nrow∗A%n c o l )
4 WRITE( i u n i t ) GCBASIS
5 CLOSE(UNIT=IUNIT , ,STATUS= ’KEEP ’ )

important: When interfacing to other programs, the logical GCBASIS should always be
false, i.e. by specifying the keyword

.NOGCBASIS

in the LSDALTON.INP. If the logical GCBASIS is true the dens.restart is given in terms
of the grand canonical basis rather than the input basis.
CHAPTER 3. INTERFACING TO LSDALTON 27

3.5 The lsDalton library bundle

The lsDalton library bundle liblsdalton.a allows for porting lsDalton code function-
ality to other programs. Currently this functionalty is limited to the evaluation of certain
standard AO integrals and integral components: the overlap matrix, the one-electron in-
tegral matrix (combining kinetic energy and nuclear-electron attraction contributions), the
Coulomb, exchange and exchange-correlation matrices and the four-center two-electron re-
pulsion integrals. Also, the first derivatives of these integral components with respect to the
nuclear displacments are available, see the file LSDALTON/lsdaltonsrc/LSlib_tester.F90
for details.
After building lsDalton under a build directory (consult the installation guide
http://dalton-installation.readthedocs.org for assistance), the lsDalton library
liblsdalton.a can be found under build\lib. You can now link you program to the
lsDalton library. The use is perhaps best illustrated by a simple example. The file
lslib_test.F90

PROGRAM testlslib
implicit none
integer :: nbast,natoms,nelectrons,lupri,luerr
integer :: i,j
double precision, allocatable :: Smat(:,:)

call lsinit_all()

lupri = 6 !logical unit number of standard output

luerr = 0 !logical unit number of standard error
CALL LSlib_get_dimensions(nbast,natoms,nelectrons,lupri,luerr)
allocate(Smat(nbast,nbast))
CALL LSlib_get_overlap(Smat,nbast,lupri,luerr)

deallocate(Smat)
call lsfree_all()

END PROGRAM testlslib

calculates the AO overlap matrix (Smat). To compile lslib_test.F90 you have to link to
the external libraries required by lsDalton, typically lapack and blas. For an example
intel/mkl build

./setup --fc=ifort --cc=icc --cxx=icpc --mkl=parallel --omp

CHAPTER 3. INTERFACING TO LSDALTON 28

you can build the excecutable lslib_test.x from lslib_test.F90 according to

ifort -mkl=parallel -openmp -parallel -O3 lslib_test.F90 -o lslib_test.x

/dalton-path/build/lib/liblsdalton.a

Exaclty how to compile and link your code to libsldalton.a for other types of build varies,
and it can then be instrutive to see how lsDalton is buildt. To see how lsdalton.x is
compiled and linked, you can type

make VERBOSE=1 lsdalton.x

in the build directory. The actual build of lsdalton.x can be found towards the end of
the standard output.
The execution of lslib_test.x requires the precense of valid Molecule and ls-
Dalton input files in the directory where you execute your program, for example the
MOLECULE.INP

BASIS
STO-3G

Atomtypes=2
Charge=8.0 Atoms=1
O 0.00000 0.00000 0.00000
Charge=1.0 Atoms=2
H 0.00000 0.50000 0.00000
H 0.00000 0.00000 0.50000

and the LSDALTON.INP

**GENERAL
.NOGCBASIS
**WAVE FUNCTION
.DFT
BLYP
*END OF INPUT
 Part III

lsDalton Reference Manual

29
Chapter 4

List of lsDalton keywords

In this chapter, we describe all keywords for the different sections of the LSDALTON.INP
file. In general, LSDALTON.INP is divided into the following sections under the headlines:
• **GENERAL contains general settings (optional)

• **INTEGRAL contains settings for the calculation of integrals (optional)

• **WAVE FUNCTION contains information about the wave function (e.g. HF/DFT)
and settings for the optimization of the wave function (mandatory)

• **OPTIMIZE contains settings for geometry optimization (optional)

• **DYNAMI contains settings for Born-Oppenheimer molecular dynamics (optional)

• **LOCALIZE ORBITALS contains settings for orbital localization procedure (op-

tional)

• **RESPONS contains information about requested molecular properties for HF and

DFT (optional)

• **DEC or **CC contains info about MP2 and coupled-cluster calculations (optional)

• **PLT contains information about construction of *.plt files which may be used to
visualize densities and orbitals using e.g. the Chimera program [1] (optional).

• **PLTGRID contains information about grid used for construction of *.plt files (op-
tional).

• **PCM contains information needed to set up a polarizable continuum model calcu-

lation using the PCMSolver external module [7] (optional).
Each of these sections may contain subsections, indicated by a single asterisk. LSDAL-
TON.INP should always end with *END OF INPUT.

30
CHAPTER 4. LIST OF LSDALTON KEYWORDS 31

4.1 **GENERAL
This input module contains general settings.

.INTERACTIONENERGY Calculation of the Counter Poise Corrected Interaction energy (Hartree-

Fock, DFT, DEC-MP2 and CCSD).
NB: Requires the use of SubSystem labels in the MOLECULE.INP file

.SAMESUBSYSTEMS Assumes that the two subsystems of the Counter Poise Corrected Inter-
action energy calculation is identical.

.SUBSYSTEMDENSITY Constructs a density matrix for the subsystems of the Counter Poise
Corrected Interaction energy calculation using the full system density matrix.

.SCALAPACK Use SCALAPACK matrices.

NB: Requires linking to the scalapack lib (provided by MKL) and the use of the –
scalapack keyword during installation. This keyword will distribute the memory of
matrices among the MPI processes. This keyword will also use the SCALAPACK
library to parallize the matrix operations like matrix multiplication.

.SCALAPACKBLOCKSIZE
<ScalapackBlockSize>
The blocksize used in SCALAPACK matrices (not required)

.SCALAPACKNODES
<ScalapackNodes>
The number of MPI processes you wish to use for the SCALAPACK format. This
must be a subset of the total number of MPI processes used. It may be benficial
to chose a subset of the total MPI processes because the matrix operations scale
differently with MPI nodes than the Integral evaluation.

.CSR Use Compressed-Sparse Row matrices. NB: Requires linking to MKL library and the
use of the –csr keyword during installation!

.TIME Activate the printout of timings.

.NOGCBASIS Deactivate the use of the Grand Canonical basis [4, 5]

.FORCEGCBASIS The Grand Canonical basis [4, 5] is deactivated when the program detects
the use of a Dunning basis set like (cc-pVXZ) but the calculation will force the use of
the Grand Canonical basis using this keyword.

.PSFUN Use the older DFT functional library of Salek, instead of the XCFun library. See
the documentation of .DFT under **WAVE FUNCTION for more information.
CHAPTER 4. LIST OF LSDALTON KEYWORDS 32

4.1.1 *TENSOR
This input submodule of **GENERAL contains settings specific for setting the behaviour
of the underlying tensor library ScaTeLib.

.DEBUG Set debugging information and safe execution variables.

.SEGMENT LENGTH
<SegmentLength>
Set the max segment length for tiled distributed tensors in each of the tensor dimen-
sions.

4.2 **INTEGRAL basic keywords

This input module defines the way the integrals should be calculated.

.ADMM Use the Auxiliary Density Matrix Method (ADMM) [8, 9] to approximate the exact-
exchange contribution; by projecting the density to a smaller basis for calculation of
the exact exchange, and by including a GGA exchange correction between the regular
and smaller auxiliary basis. When using this keyword, an auxiliary basis set has to
be specified in the Molecule file such as:

1 BASIS
2 6-31+G* Aux=df-def2 ADMM=3-21G
...

where the ”ADMM” basis is used as an auxiliary basis set for ADMM (and where
”Aux” specify the auxiliary basis for density fitting (optional)). The .ADMM keyword
defaults to ADMM2 with Becke type exchange for the GGA correction term (identical
to using .ADMM2 together with .ADMM-FUNC specified to B88X). Note that although
calculations requesting ADMM and reponse properties are allowed, the response part
of the code does not actually apply the ADMM approximation. ADMM energy and
gradients are implemented.

.ADMM1 Use the ADMM1 approximation of Ref. [8]. Note: This approximation does not
work for gradients!

.ADMM2 Use the ADMM2 approximation of Ref. [8]. Default.

.ADMMS Use the ADMMS approximation of Ref. [9]. Reccomended.

CHAPTER 4. LIST OF LSDALTON KEYWORDS 33

.ADMMP Use the ADMMP approximation of Ref. [9].

.ADMMQ Use the ADMMQ approximation of Ref. [9]. Not reccomended.

.ADMM-FUNC Reads a GGA exchange functional from the next input line. Specifies the GGA
exchange functional to use for the correction term. List of possible functionals: B88X,
PBEX, RPBEX, REVPBEX, MPBEX, PW91X, KT1X, KT2X, KT3X, G96X, LG93X
and OPTX. KT3X together with ADMMS works well for basis-set combinations 6-
31G**/3-21G and cc-pVTZ/3-21G [9].
LDAX is also implemented, but is currently not working as expected. Its use is
therefore discouraged.

.AOPRINT
<Print level>
Print the Atomic Orbital information. The higher the print level, the more information
will be printed. Default value is 0.

.BASPRINT
<Print level>
Print the basis set information. The higher the print level, the more information will
be printed. Using a print level of 6 will print the input exponents and contraction
coefficients read from file as well as the normalized basis used in lsDalton.

.DENSFIT Use density-fitting for the Coulomb contribution (no exact exchange density-
fitting method have been introduced). When using this keyword, an auxiliary basis
set has to be specified in the Molecule file. For the Poples basis sets we recommend
df-def2 and for Dunnings basis sets use the corresponding cc-pVXZdenfit basis sets,
with X =T,Q, 5.

.LINSCAPRINT
<Print level>
Print level in the integral code. The higher the print level, the more information will
be printed. This keyword corresponds to

.AOPRINT
<Print level>
.BASPRINT
<Print level>
.MOLPRINT
<Print level>
CHAPTER 4. LIST OF LSDALTON KEYWORDS 34

See the individual keywords for more details. Default value is 0.

.MOLPRINT
<Print level>
Print information about the molecule. The higher the print level, the more information
will be printed. Default value is 0.

.NOJENGINE Turn off J-engine algorithm (which is default) for the calculation of the Coulomb
matrix (see Refs. [10] and [11]). Compute instead Coulomb-like contributions from
the explicitly calculated integrals.

.NOLINK Turn off the LinK algorithm (which is default) for the calculation of the exchange
matrix (see Ref. [12]).

.NO SCREEN Deactivate the use of all screening methods.

.NOFAMILY Deactivate the exploitation of family type basis set, basis set sharing exponents
for different angular momentums.

.PARI Use the Pair-Atomic Resolution-of-the-Identity approximation [13] for the Coulomb
and exact Exchange contribution. When using this keyword, an auxiliary basis set
has to be specified in the Molecule file.
We do not recommend to use this method for any production runs, and have included it
in case other method developers would like to explore it (see the reference for details).

.RUNMM Use the (Fast) Multipole Method.

.THRESH
<Threshold>
An overall screening threshold for integral evaluation. The default value is 10−8
The various integral-evaluation thresholds (below) are set according to this Threshold.

• The Cauchy-Schwarz screening threshold is set equal to Threshold (can be set

separately by .THR CS)
– The Screening threshold used for Coulomb is Threshold ·10−2 (Default:
10−10 )
– The Screening threshold used for Exchange is Threshold ·10−0 (Default:
10−8 )
– The Screening threshold used for One-electron operators is Threshold ·10−7
(Default: 10−15 )
• The Primitve Cauchy-Schwarz screening threshold is set equal to Threshold·10−1
(can be set separately by .THR PS)
CHAPTER 4. LIST OF LSDALTON KEYWORDS 35

• The Overlap-distribution distance-screening threshold is set equal to Threshold

·10−1 . When calculating overlap integrals, this threshold is used for setting up
AO extents. Overlap distributions (ODs) for which the distance between the two
AOs are larger than the sum of the extents are screened away.
• The Overlap-distribution extent-screening threshold is set equal to Threshold
·10−1 . When calculating overlap integrals, this threshold is used for setting up
OD extents. Overlap integrals between two ODs separated by more than the
sum of the extents are screened away.
• The FMM threshold is used to distinguish the Coulomb repulsion into classical
and non-classical interactions, based on the non-classical extent of the continuous
charge distribution (orbitals or orbital products). By default this threshold is
set equal to Threshold ·10−1 (can be set separately by .SCREEN under *FMM)

4.3 **INTEGRAL advanced keywords

This input module defines the way the integrals should be calculated. This section contains
some advanced keywords which is mostly useful for debugging or comparisons with other
codes.

.NO CS Deactivate the use of Cauchy-Schwarz screening. Note that this does not deactivate
the Primitive Cauchy-Schwarz screening.

.NO PS Deactivate the use of primitive Cauchy-Schwarz screening.

.THR CS
<CS_Threshold>
Cauchy-Schwarz screening threshold. The default value is 10−8 . Note that

• The Screening threshold used for Coulomb is CS Threshold ·10−2 (Default: 10−10 )
• The Screening threshold used for Exchange is CS Threshold ·10−0 (Default:
10−8 )
• The Screening threshold used for One-electron operators is CS Threshold ·10−7
(Default: 10−15 )

.THR PS
<PS_Threshold>
Primitive Cauchy-Schwarz screening threshold. The default value is 10−9
CHAPTER 4. LIST OF LSDALTON KEYWORDS 36

.CART-E Use cartesian E-coefficients instead of hermite E-coefficients for the McMurchie-
Davidson integral-evaluation scheme [14] used in lsDalton. The use of Cartesian E-
coefficients is default in most integral programs. We however use hermite E-coefficients
according to ref. [15].

.FTUVMAXPRIM
<Maxprim>
Sets the maximum number of primitives used in the FTUV batches in the J-engine
algorithm.

.LOW RJ000 ACCURACY Use a decreased accuracy for the calculation of the Boys function.

.MAXPASSES
<maxpass>
Change the maximum number of collected overlap distributions (default is 40).

.NO PASS Deactivate the use of Passes, the collection of overlap distributions which is used
as default in order to increase efficiency.

.NOSEGMENT Deactivate the use of segments. The use of segments is used to reduced the
number of primitive functions in the calculation of integrals. The basis is considered
to be a general contracted basis where all primitives contribute to all contracted
functions.

.NSETUV Use non-spherical E-coefficients.

.OVERLAP-DF-J Use the Overlap density fitting algorithm for the calculation of the Coulomb
matrix.

.UNCONT Treat the basis functions (given by input) as fully uncontracted basis functions
(i.e. ignore any contraction coefficients given in the basis and treat instead all the
primitives as separate basis functions).

4.3.1 *FMM
This input block is used to specify settings for the (continuous) fast multipole moment
(FMM) treatment of classical Coulomb interactions.
For gradients OpenMP parallelization is turned off. The reason for this is that a continuous
counter used to identify moments for orbital products has to be the same in both moments
and derivative moments. For the same reason screening [only during writing (see printmm
routines)!!] is turned off both for the serial and parallel cases. As a consequence of the
screening been turned off the moments have to be recalculated for the gradient and can not
CHAPTER 4. LIST OF LSDALTON KEYWORDS 37

be reused from the preceding energy evaluation. The gradient implementation has been
tested both for regular and density fitting Coulomb interactions. It works for the combined
N-e + e-e + N-N interactions (default), and for the e-e interaction only (invoked by .NOONE).
It has however not been tested for speed and efficiency.

.LMAX
<Lmax>
The maximum multipole-moment expansion order of a given orbital product. Default
Value is 8.

.TLMAX
<TLmax>
The maximum order used for translated multipole-moments. Default Value is 20.

.SCREEN
<Threshold>
The FMM threshold is used to distinguish the Coulomb repulsion into classical and
non-classical interactions, based on the non-classical extent of the contineous charge
distribution (orbitals or orbital products). By default this threshold is set equal to
·10−9 Defines the threshold used to determine if a contribution can be calculated
classically or non-classically.

.NOONE Do not use FMM for the nuclear-electron attraction.

.NOMMBU Do not use io-buffers for interfacing multipole-moments to the FMM driver.

4.4 **WAVE FUNCTION

.HF Hartree-Fock calculation.

.DFT
<FUNCTIONAL>
DFT calculation.
The exchange—correlation functionals in lsDalton can be divided into two groups:
generic and combined functionals. A number of generic functionals are included within
lsDalton. The combined functionals are linear combinations of generic ones.
It should be noted that the input is not case sensitive, although the notation employed
in this manual makes use of case to emphasise exchange or correlation functional
properties and to reflect the original literature sources.
CHAPTER 4. LIST OF LSDALTON KEYWORDS 38

4.4.0.1 Exchange Functionals

Slater Dirac-Slater exchange functional [16, 17, 18].

Becke 1988 Becke exchange GGA correction [19]. Note that the full Becke88 exchange
functional is given as Slater + Becke.
OPTX Handy’s 2001 exchange functional correction [20]. The full OPTX exchange
functional is given by 1.05151*Slater - 1.43169*OPTX.
PBEx Perdew, Burke and Ernzerhof 1996 exchange functional [21].
PW86x Perdew and Wang 1986 exchange functional (the PWGGA-I functional)[22].

4.4.0.2 Correlation Functionals

VWN3 correlation functional of Vosko, Wilk and Nusair, 1980 (equation III) [23]. This
is the form used in the Gaussian program.
VWN5 correlation functional of Vosko, Wilk and Nusair, 1980 (equation V – the rec-
ommended one). The VWN keyword is a synonym for VWN5 [23].
LYP correlation functional by Lee, Yang and Parr, 1988 [24, 25].
P86c non-local part of the correlation functional of the Perdew 1986 correlation func-
tional [26]. PZ81 (1981 Perdew local) is usually used for the local part of the
functional, with a total corelation functional of P86c + PZ81.
PBEc Perdew, Burke and Ernzerhof 1996 correlation functional, defined as PW91c
local and PBEc non-local correlation [21].
PW91c 1991 correlation functional of Perdew and Wang (the pwGGA-II functional)
[27]. This functional includes both the PW91c non-local and PW91c local (ie
PW92c) contributions. The non-local PW91c contribution may be determined
as PW91c - PW92c.
PZ81 local correlation functional of Perdew and Zunger, 1981 [28].

4.4.0.3 Standalone Functionals

LB94 asymptotically correct functional of Leeuwen and Baerends 1994 [29]. This
functional improves description of the asymptotic density on the expense of core
and inner valence.

4.4.0.4 Combined functionals

GGAKey is a universal keyword allowing users to manually construct arbitrary lin-

ear combinations of exchange and correlation functionals from the list above.
CHAPTER 4. LIST OF LSDALTON KEYWORDS 39

Even fractional Hartree–Fock exchange can be specified. This keyword is to

be followed by a string of functionals with associated weights. The syntax is
NAME=WEIGHT .... As an example, B3LYP may be constructed as:
.DFT
GGAKey HF=0.2 Slater=0.8 Becke=0.72 LYP=0.81 VWN=0.19
The following GGA and hybrid functional aliases are defined within lsDalton
and provide further examples of the GGAKey keyword.
SVWN5 is a sum of Slater functional and VWN (or VWN5) correlation functional. It
is equivalent to
GGAKey Slater=1 VWN5=1
SVWN3 is a sum of the Slater exchange functional and VWN3 correlation functional.
It is equivalent to the Gaussian program LSDA functional and can alternatively
be selected by following set of keywords
GGAKey Slater=1 VWN3=1
LDA A synonym for SVWN5.
BLYP is a sum of Slater functional, Becke88 correction and LYP correlation functional.
It is equivalent to
GGAKey Slater=1 Becke=1 LYP=1
B3LYP 3-parameter hybrid functional [30] equivalent to:
GGAKey HF=0.2 Slater=0.8 Becke=0.72 LYP=0.81 VWN=0.19
B3LYP-G hybrid functional with VWN3 form used for correlation—this is the form
used by the Gaussian quantum chemistry program. Keyword B3LYPGauss is a
synonym for B3LYPg. This functional can be explicitly set up by
GGAKey HF=0.2 Slater=0.8 Becke=0.72 LYP=0.81 VWN3=0.19
BP86 Becke88 exchange functional and Perdew86 correlation functional (with Perdew81
local correlation). The explicit form is:
GGAKey Slater=1 Becke=1 PZ81=1 P86c=1
B3P86 variant of B3LYP with VWN used for local correlation and P86 for the nonlocal
part.
GGAKey HF=0.2 Slater=0.8 Becke=0.72 P86c=0.81 VWN=1
B3P86-G variant of B3LYP with VWN3 used for local correlation and P86 for the non-
local part. This is the form used by the Gaussian quantum chemistry program.
GGAKey HF=0.2 Slater=0.8 Becke=0.72 P86c=0.81 VWN3=1
BPW91 Becke88 exchange functional and PW91 correlation functional. The explicit
form is:
GGAKey Slater=1 Becke=1 PW91c=1
CHAPTER 4. LIST OF LSDALTON KEYWORDS 40

CAMB3LYP Coulomb Attenuated Method Functional of Yanai, Tew and Handy [31].
This functional accepts additional arguments alpha, beta and mu to modify the
fraction of HF exchange for short-range interactions, additional fraction of HF
exchange for long-range interaction and the interaction switching factor µ. This
input can be specified as follows:
.DFT
CAMB3LYP alpha=0.190 beta=0.460 mu=0.330

KT1 Slater-VWN5 functional with the KT GGA exchange correction [32, 33].
GGAKey Slater=1 VWN=1 KT=-0.006
KT2 differs from KT1 only in that the weights of the Slater and VWN5 functionals
are from an empirical fit (not equal to 1.0) [32, 33].
GGAKey Slater=1.07173 VWN=0.576727 KT=-0.006
KT3 a hybrid functional of Slater, OPTX and KT exchange with the LYP correlation
functional [34]. The explicit form is
GGAKey Slater=1.092 KT=-0.004 LYP=0.864409 OPTX=-0.925452
OLYP is the sum of the OPTX exchange functional with the LYP correlation functional
[20, 24, 25].
GGAKey Slater=1.05151 OPTX=-1.43169 LYP=1
PBE0 a hybrid functional of Perdew, Burke and Ernzerhof with 0.25 weight of exact
exchange, 0.75 of PBEx exchange functional and the PBEc correlation functional
[35]. Alternative aliases are PBE1PBE or PBE0PBE.
GGAKey HF=0.25 PBEx=0.75 PBEc=1
PBE same as above but with exchange estimated exclusively by PBEx functional [21].
Alias of PBEPBE. This is the form used by CADPAC and NWChem quantum
chemistry programs.
GGAKey PBEx=1 PBEc=1
Note that the Molpro quantum chemistry program uses the PW91c non-local
correlation functional instead of PBEc, which is equivalent to the following:
GGAKey PBEx=1 PW91c=1 .

Note that combinations of local and non-local correlation functionals can also be gener-
ated with the GGAKey keyword. For example, GGAKey P86c=1 PZ81=1 combines the
PZ81 local and P86c non-local correlation functional, whereas GGAKey VWN=1 P86c=1
combines the VWN local and P86 non-local correlation functionals.
Linear combinations of all exchange and correlation functionals listed above are pos-
sible with the GGAKey keyword.
CHAPTER 4. LIST OF LSDALTON KEYWORDS 41

4.4.0.5 Functional implementation details

Starting with lsDalton1.0 the functionals are evaluated using the XCFun library[36]
by default, unless this is disabled at build time. The functional library of Pawel
Salek et al [37] is still available and can be activated using the .PSFUN keyword under
**GENERAL. For open-shell calculations the use of XCFun is strongly encouraged,
but for closed shell calculations the .PSFUN option may result in faster calculations.
Because of the small differences between the two different implementations it may
not be possible to obtain exactly the same results using XCFun or .PSFUN, but the
difference should be very small (i.e. energy differences of about 10−8 ).

4.4.1 *DFT INPUT

Controls the XC-integration grid. The grid is generated as follows: For each atom a radial
grid is created (different radial grids available, see below), which is multiplied with an
angular grid (Lebedev-grids). The angular grids are by default pruned for radial points
close to the nucleus, following Murray, Handy and Laming. [38] Grid weights for the grid
points are calculated following different schemes (see below). Grid points with weights
below 10−20 are disregarded. The default settings corresponds to .GRID4, but with the
BLOCKSSF rather than the BLOCK partitioning scheme (see details below). Please note
that the grid construction routines in the lsDalton program changed compared to the
Dalton program which may result in small deviations in the number of grid points etc.

.ANGINT
<ANGINT>
Determines the quality of the angular Lebedev grid – the angular integration of spher-
ical harmonics will be exact up to the specified order. Default value is 31. Maximum
value is 64. Note that the value of ANGINT is changed by the following keywords:
COARSE, NORMAL, FINE, ULTRAFINE. The GRIDX keywords also imply specific ANGINT
values.

.COARSE Shortcut keyword for radial integration accuracy 10−11 and angular expansion
order equal to 35.

.DFTELS
<DFTELS>
safety threshold – stop if the charge integration error becomes larger then this thresh-
old.

.DFTTHR
<DFTHR0, DFTHRL, DFTHRI, RHOTHR>
CHAPTER 4. LIST OF LSDALTON KEYWORDS 42

DFTHR0 is not used

DFTHRL is not used
DFTHRI threshold for screening product of gaussian atomic orbitals
RHOTHR threshold for screening of the electron density

.DISPER defaults to .DFT-D3BJ (see below)

.DFT-D2 switches on Grimmes DFT-D2 empirical dispersion correction [39]. The code will
attempt to assign the correct functional dependent parameters based on the chosen
DFT functional. Analytic gradient contributions are available.

.D2PAR READ (LUINP,*) D2 s6 inp, D2 alp inp, D2 rs6 inp

using this keyword user input values of the s6 , α and sr,6 DFT-D2 parameters may
be specified. If supplied these values override any values defined within the code.

.DFT-D3 switches on Grimmes DFT-D3 empirical dispersion correction [40]. The code will
attempt to assign the correct functional dependent parameters based on the chosen
DFT functional. Analytic gradient contributions are available.

.DFT-D3BJ switches on Grimmes DFT-D3 empirical dispersion correction with Becke-Johnson

damping [41]. The code will attempt to assign the correct functional dependent pa-
rameters based on the chosen DFT functional. Analytic gradient contributions are
available. This is the presently recommended version.

.3BODY keyword for adding 3-body terms to the DFT-D3 dispersion energy. Note that
gradients are not implemented for these corrections

.D3PAR READ (LUINP,*) D3 s6 inp, D3 alp inp, D3 rs6 inp, D3 rs18 inp, D3 s18 inp
keyword for specifying the s6 , α, sr,6 , sr,8 and s8 parameters of the DFT-D3 methods.
Note, take care to match the parameter values to the correct version of the DFT-D3
correction.

.FINE Shortcut keyword for radial integration accuracy 10−13 and angular expansion order
equal to 42.

.GRID TYPE
<GC2 LMG TURBO BECKE BECKEORIG SSF BLOCK BLOCKSSF>
GC2, LMG, TURBO define radial quadrature schemes, only one can be set. BECKE,
BECKEORIG, SSF, BLOCK, BLOCKSSF define the grid partitioning schemes, only one can
be set.

GC2 Gauss-Chebyshev quadrature of second kind.

CHAPTER 4. LIST OF LSDALTON KEYWORDS 43

LMG As proposed by Lindh, Malmqvist and Gagliardi (default). [42]

TURBO Treutler-Ahlrichs M4-T2 scheme. [43] Implies also that the angular integration
quality becomes Z-dependant (see also ANGINT) and that pruning is used (see
also NOPRUN).
BECKE Becke partitioning scheme with atomic size correction. [44]
BECKEORIG Becke partitioning scheme without atomic size correction (default). [44]
SSF Stratmann-Scuseria-Frisch partitioning scheme. [45]
BLOCK Becke partitioning scheme with atomic size correction [44] combined with a
blockwise handling of grid points [46]. Useful for large molecules.
BLOCKSSF Stratmann-Scuseria-Frisch partitioning scheme [45] combined with a block-
wise handling of grid points [46]. Useful for large molecules.

.GRID1 Shortcut for radial integration accuracy 1.0 ∗ 10−5 , angular expansion order equal
to 17, radial quadrature TURBO, grid partitioning scheme BLOCK.

.GRID2 Shortcut for radial integration accuracy 2.15447 ∗ 10−7 , angular expansion order
equal to 23, radial quadrature TURBO, grid partitioning scheme BLOCK.

.GRID3 Shortcut for radial integration accuracy 4.64159 ∗ 10−9 , angular expansion order
equal to 29, radial quadrature TURBO, grid partitioning scheme BLOCK.

.GRID4 Shortcut for radial integration accuracy 5.01187 ∗ 10−14 , angular expansion order
equal to 35, radial quadrature TURBO, grid partitioning scheme BLOCK.

.GRID5 Shortcut for radial integration accuracy 2.15443 ∗ 10−17 , angular expansion order
equal to 47, radial quadrature TURBO, grid partitioning scheme BLOCK.

.HARDNESS
<HARDNESS>
sets the hardness of the partitioning function in the Becke weighting scheme. Only
positive integer values allowed. The higher the hardness the more stepfunction like
the partitioning functions. Default value is 3 as proposed by Becke. [44]

.NOPRUN Supresses the default pruning of the atomic angular integration grids for radial
points close to the nucleus.

.NORMAL Shortcut keyword for radial integration accuracy 10−13 and angular expansion
order equal to 35.

.RADINT
<RADINT>
CHAPTER 4. LIST OF LSDALTON KEYWORDS 44

Determines the quality of the radial integration grid. Default radial integration accu-
racy is 10−11 . Note that the value of RADINT is changed by the following keywords:
NORMAL, FINE, ULTRAFINE. The GRIDX keywords also imply specific RADINT values.

.ULTRAF Shortcut keyword for radial integration accuracy 10−15 and angular expansion
order equal to 64.

.LB94 Uses the van Leeuwen-Baerends asymptotic correction. This correction can be com-
bined with all GGA type functionals

.CS00 Using the Casida-Salahub asymptotic correction[47, 48]

0
VXC (r) = max (VXC + ∆, VXC + (1 − HFexchange) · VLB94,correction ) , (4.1)

with a Zhan-Nichols-Dixon shift [47, 49] (HFexchange = 0.2 for B3LYP)

∆ = ZN D1 · HOM O − ZN D2 (4.2)

Where ZN D1 = 0.2332 and ZN D2 = 0.0116. Note that ZN D2 = 0.0116au. =

0.315eV . The parameters have been optimized with respect to a B3LYP functional
so these parameters can be changed using the following keywords. Note that this
keyword only works with GGA type functionals. The CS00 keyword only work with
a method which calculates the HOMO energy.

.CS00 ZND1
<ZND1 in au.>
Changes the parameter ZN D1 (default ZN D1 = 0.2332) see .CS00

.CS00 ZND1
<ZND2 in au.>
Changes the parameter ZN D2 (default ZN D1 = 0.0116) see .CS00

.CS00 SHIFT
<Delta in au.>
Using the Casida-Salahub correction with a shift of Delta see .CS00

4.4.2 *DENSOPT
This section contains the different options for obtaining the SCF wave function and energy.
The default is augmented Roothaan–Hall density optimization in combination with a con-
jugate residual optimal trialvectors (CROP) scheme for solving the (level-shifted) Newton
equations ((see ref. [50])) and using an atom-in-a-molecule starting guess.
CHAPTER 4. LIST OF LSDALTON KEYWORDS 45

.2ND ALL Use second order optimization in all SCF iterations.

.2ND LOC Use second order optimization in local SCF iterations.

.ARH Use Augmented Roothaan-Hall scheme for density optimization (default) [51, 52]
where the CROP solver ([50]) is used for solving the (level-shifted) Newton equation.

.ARH FULL Use Augmented Roothaan-Hall for density optimization - no truncation of the
reduced space, keep all micro vectors for the CROP algorithm.

.ARH DAVID Use Augmented Roothaan-Hall for density optimization where the (level-shifted)
Newton equations are solved using a reduced space solver based Davidson’s algorithm
[53]. This setting may be used if default settings converges to a saddle point.

.ARH(LS) DAVID The same procedure as described for .ARH DAVID augmented with a line
search to accelerate convergence for very large molecular systems, but is less efficient
than .ARH DAVID for smaller molecular systems.

.STABILITY Check stability of the optimized wave function by calculation of the lowest
Hessian eigenvalue.

.STAB MAXIT
<Max. number of stability iterations>
Max. number of iterations in calculation of lowest Hessian eigenvalue (default is 40).

.CHOLESKY Do Cholesky decomposition of overlap matrix (for density optimization in or-

thogonal AO basis) - default is Löwdin decomposition by diagonalization.

.CONTFAC
<Contraction factor>
For update of trust-radius (used in ARH and second order optimization). If trust-
radius should be contracted, contract it by this factor (default is 0.7).

.CONTRAC
<Contraction criterion>
For update of trust-radius (used in ARH and second order optimization). Contract
trust-radius if trust-radius ratio is smaller than this criterion (default is 0.25).

.CONVDYN
<Option>
Dynamic SCF convergence threshold. This is suitable for calculations on large molecules,
since the standard SCF convergence threshold is based on the Frobenius norm of
the SCF gradient, which is not size-extensive. Options are SLOPPY, STANDARD,
CHAPTER 4. LIST OF LSDALTON KEYWORDS 46

TIGHT, VTIGHT corresponding to 10−4 , 10−5 , 10−6 , 10−7 times the square root of
the number of electrons in the system.

.CONVTHR
<Threshold>
SCF convergence threshold - Frobenius norm of SCF gradient (default is 1.0d-4).
Note that this convergence criterion is not size-extensive. For large molecules, it is
recommended to use .CONVDYN instead.

.DIIS Pulay’s DIIS scheme is used to speed up SCF convergence [54, 55].

.DISK Store densities and Fock/KS matrices from previous iterations on disk instead of in
core when constructing the new Fock/KS matrix (memory-saving option).

.DISKSOLVER Store trial and sigma vectors on disk instead of in core when solving lin-
ear equations (memory-saving option). Only referenced using ARH , Direct Density
optimization or second order optimization.

.EXPAND
<Expansion criterion>
For update of trust-radius (used in ARH and second order optimization). Expand
trust-radius if trust-radius ratio is larger than this criterion (default is 0.75).

.EXPFAC
<Expansion factor>
For update of trust-radius (used in ARH and second order optimization). If trust-
radius should be expanded, expand it by this factor (default is 1.2).

.FIXSHIFT
<Shift>
Use a fixed level shift in all SCF iterations.

.HESVEC
<Number of Hessian eigenvalues>
Number of lowest Hessian eigenvalues to be calculated with keyword .STABILITY
(default is 1).

.LCV Compute the Least-Change Valence orbitals after valence density optimization step.
All subsequent calculations are computed using Least-Change Valence orbitals aug-
mented with atomic virtual orbitals of the Atomic density as basis set. Keyword only
effective with TRILEVEL starting guess.
CHAPTER 4. LIST OF LSDALTON KEYWORDS 47

.LCM Compute the Least-Change Molecular orbitals after the Full molecular density cal-
culation. All subsequent calculations are computed using these orbitals as basis set.
.LCV is automatically included. Keyword only meaningful with TRILEVEL starting
guess.

.LEVELSH
<N> <shift1> <shift2> ...<shiftN>
Use custom level shifts in the first N SCF iterations.

.LOW ACCURACY START Use low accuracy settings in the first couple of iterations until the
appropriate gradient threshold have been reached.

.MAXELM
<Trust radius (max. element)>
Absolute max. element of step allowed (default is 0.35). Used in ARH and second
order optimization.

.MAXIT
<Max. number of iterations>
Max. number of SCF iterations (default is 100).

.MAXSTEP
<Trust radius (Frobenius norm)>
Absolute max. Frobenius norm of step allowed (default is 0.6). Used in ARH and
second order optimization.

.MICTHRS
<Threshold for micro iterations>
The micro iterations of the CROP algorithm are converged to gradient norm times
this factor (default is 1.0d-2). Only referenced if ARH, Direct Density optimization
or second order optimization.

.MICROVECS
<Max. number of microvectors>
Max. number of microvectors to be kept in the CROP scheme (default is 2). Only
referenced if ARH, Direct Density optimization or second order optimization.

.MINDAMP
<Minimum damping>
Never allow level shift to be smaller than this.

.NOAV Turn off Fock matrix averaging in SCF iterations (this is default since the default
optimization scheme is ARH, which contains implicit averaging).
CHAPTER 4. LIST OF LSDALTON KEYWORDS 48

.NVEC
<Max. no of vectors for averaging>
Maximum number of previous Fock and density matrices to be stored and used for
ARH and DIIS (default is 10 for HF and 7 for DFT).

.NOPREC Turn off preconditioning of linear equations (ARH, TrFD, and second order opti-
mization).

.NOSHIFT Do no level shifting.

.RESTART The code automatically saves the density matrix in each SCF iteration in a file
called dens.restart. This option restart the calculation from this density matrix. If
the program cannot find the dens.restart file the keyword have no effect.

.SKIPSCFLOOP The code automatically saves the density matrix and the Fock/Kohn-Sham
matrix at the end of the SCF iterations in two files called dens.restart and fock.restart.
This option reads the files and skips the SCF iterations.

.RH Use standard Roothaan-Hall scheme (diagonalization) for density optimization.

.SOEO Use second order ensemble optimization. Remember to specify active space and
starting occupations. Use only in combination with .NOGCBASIS. Requires .HF or
.DFT as wave function. After an ensemble optimization, the orbitals are written to
file cmo.out and the AO density matrix is written to dao.out, which may be used to
plot molecular orbitals and charge density plots.

.SOEOSPACE
<core> <act>
Specification of the orbitals in the active space. <core> being the number of core
orbitals (occupations not optimized), and <act> the number of active orbitals, where
the orbital occupations are optimized.

.SOEOOCC
<fully occupied>
<fractionally occupied>
<n1> <n2> · · ·
Specification of the distribution of electron pairs in the starting orbitals in a sec-
ond order ensemble optimization. <fully occupied> is the number of fully occupied
orbitals and <fractionally occupied> is the number of fractionally occupied or-
bitals. <n1>, <n2> etc. are the fractional occupations of the fractionally occupied
spin-orbitals in the starting guess. The given occupations must lie between zero and
one. If they are either zero or one, this orbital occupation is not optimized. Equal
CHAPTER 4. LIST OF LSDALTON KEYWORDS 49

distribution may result in wrong results due to symmetry. The sum of the fully occu-
pied orbitals and the given orbital occupations <n1>, <n2> etc., must equal the total
number of electron pairs in the system. (NOTE that it is possible to specify an active
space of zero orbitals, but still include fractional occupations. The optimization will
thus optimize the shapes of the orbitals, without optimizing the orbital occupations.)

.SOEOSAVE Saves the AO Fock matrix (and the AO density matrix) to the file soeosave.out.
Used for restart.

.SOEORST Restart second order ensemble optimization from orbitals stored in soeosave.out
(so remember to copy this file to the working directory). The molecular geometry need
not be the same as in the calculation where the stored orbitals were created. The ac-
tive space and starting occupations still need to be specified using .SOEOSPACE and
.SOEOOCC.

.SOEOGC Grand-canonical ensemble optimization. Removes the restriction on the total

number of electrons from the second order ensemble optimization. The sum of the
starting occupations still must be correct.

.SOEOMATHR
<macro threshold>
Specifies the convergence threshold for the macro iterations in second order ensemble
optimization.

.SOEOMITHR
<micro threshold>
Specifies the convergence threshold for the micro iterations in second order ensemble
optimization.

.SOEOTRUST
<trust-radius>
Specifies the starting trust-radius in the ensemble optimization. Trust-radius will still
be updated throughout the optimization.

.SOEODIPOLE Use the ensemble density matrix to determine electric dipole moment and
static electric polarizability.

.START
<Option>
Starting guess for SCF optimization. Options are:

• H1DIAG Obtain initial guess by diagonalizing one-electron Hamiltonian

CHAPTER 4. LIST OF LSDALTON KEYWORDS 50

• ATOMS Obtain initial guess by atoms-in-molecules approach (default)

• TRILEVEL Optimization in three steps. 1. Atomic densities, 2. Valence
molecular density, 3. Full molecular density [4, 5].

Note that Huckel guess is not supported by lsDalton.

.VanLenthe Use Van Lenthe’s scheme for level shifting and averaging [56].

4.4.3 $INFO

For the different parts of the wave function optimization, it is possible to get very
detailed information. This section describes these keywords. They must be put under
*DENSOPT in a section beginning with $INFO and ending with $END INFO. Note
that these keywords do NOT start with a dot!

INFO CROP Detailed info from Conjugated Residual OPtimal vectors scheme (Direct
Density Optimization, ARH and second order optimization).
INFO DIIS Detailed info from Direct Inversion in the Iterative Subspace algorithm.
INFO LEVELSHIFT Detailed info from determination of dynamic level shift (Direct
Density Optimization, ARH and second order optimization).
INFO LINEQ Detailed info from solution of linear equations and trust-radius update
(Direct Density Optimization, ARH and second order optimization).
INFO RH Detailed info from Roothaan-Hall algorithm (diagonalization).
INFO RH DETAIL Even more detailed info from Roothaan-Hall algorithm (diagonal-
ization).
INFO STABILITY Detailed info from calculation of lowest Hessian eigenvalue (stability
analysis) and HOMO-LUMO gap.
INFO STABILITY REDSPACE Reduced space info from calculation of lowest Hessian
eigenvalue (stability analysis) and HOMO-LUMO gap.

4.5 **OPTIMIZE
This input module describes the keywords needed to optimize a molecular geometry. In
the current release no second-order algorithms are available since the molecular Hessian
is not implemented, but a number of quasi-Newton trust-radius level-shifted techniques
for finding equilibrium geometries are available (.BFGS, .PSB, .DFP). The methods differ
in the way approximate Hessians are built and updated: .INIMOD, .INIRED and .INITEV
CHAPTER 4. LIST OF LSDALTON KEYWORDS 51

provide different approximations for the initial Hessian, while .MODHES recomputes the
approximated Hessian at every geometry.
The geometry optimization generates a sequence of geometry steps. The coordinates
of each accepted step are stored in the files MOLECULE.XXX, where XXX is the geometry
step number, and the final converged geometry in MOLECULE.OUT.
Geometric structure can be optimized in redundant internal (.REDINT) or Cartesian
(.CARTES) coordinates with two sets of convergence criteria (two for the gradient and two
for the step). Convergence is obtained when the following four quantities are respectively
smaller than the four convergence criteria , 5, 3 and 15 [57]:
• the root-mean-square of the gradient,

• the maximum absolute value of the gradient,

• the root-mean square of the step vector,

• and the maximum absolute element (in internal or Cartesian coordinates) of the step
vector.
The default setting specifies the minimization in internal coordinates [58], with a
model Hessian [59] updated with BFGS formula. The atoms-in-molecule approach is used
to set up the initial density-matrix at each new geometry step. However, it is possible to
take as starting guess McWeeny-purified converged density-matrix at the previous geometry.
This is achieved by setting the .RESTART option under the *DENSOPT subsection. The
default value of  is 10−5 (Other predefined values for the convergence criteria can be set
using the keywords .VLOOSE, .LOOSE, .TIGHT and .VTIGHT). Alternatively the convergence
criteria of Baker [60] can be set using .BAKER.

Example of test cases for geometry optimization calculations

The test cases are located in LSDALTON/test/geomopt. These test cases illustrate the
main features of the geometry optimization:
• geomopt/cartes bfgs min: performs a DFT/LDA geometry optimization in Cartesian
coordinates, with BFGS update of an initial Hessian equal to a unit matrix and Baker
convergence criteria.

• geomopt/redint bfgs: performs a DFT/BLYP geometry optimization in redundant

internal coordinates with a model Hessian updated with BFGS formula and Baker
convergence criteria.

• geomopt/Excited state opt: performs a DFT/B3LYP geometry optimization of an ex-

cited state in redundant internal coordinates and Baker convergence criteria.
CHAPTER 4. LIST OF LSDALTON KEYWORDS 52

• geomopt/cartes psb min: performs a DFT/BLYP geometry optimization in Cartesian

coordinates with a diagonal initial Hessian updated by PSB formula, McWeenie’s
purification used for constructing the initial guess and Baker convergence criteria.

• geomopt/geoopt constrain: shows a constrained geometry optimization of ethane,

keeping the C-C bond distance fixed (Internal coordinate #1 as determined from
the use of .FINDRE).

• geomopt/cartes trilevel newconv : performs a geometry optimization in Cartesian co-

ordinates with .TRILEVEL starting guess and a default convergence criteria, a diagonal
initial Hessian updated by a loose DFP formula.

.BAKER Activates the convergence criteria of Baker [60]. The minimum is then said to be
found when the largest element of the gradient vector (in Cartesian or redundant
internal coordinates) falls below 3 · 10−4 and either the energy change from the last
iteration is less than 10−6 or the largest element of the predicted step vector is less
than 3 · 10−4 .

.BFGS Specifies the use of a first-order method with the Broyden-Fletcher-Goldfarb-Shanno

(BFGS) update formula for optimization. This is the preferred first-order method for
minimizations, as this update is able to maintain a positive definite Hessian.

.CARTES Specifies that Cartesian coordinates should be used in the optimization.

.CONSTR
<Number of contrained redundant internal coordinates>
<Indices of each contrained redundant internal coordinate, one per line>
Specifies a contrained geometry optimization. You will need to give the total number
of redundant internal coordinates to constrain and their indices as defined using the
.FINDRE keyword. For example fixing redundant internal coordinates 4 and 9 would
read like:

...
**OPTIMIZE
.CONSTR
2
4
9
...
**END OF INPUT
CHAPTER 4. LIST OF LSDALTON KEYWORDS 53

.DFP Specifies that a first-order method with the Davidon-Fletcher-Powell (DFP) update
formula should be used for optimization.

.FINDRE Using this keyword no geometry optimization is carried out, but only the redun-
dant internals coordinates are found.

.INIMOD Use a simple model Hessian [59] diagonal in redundant internal coordinates as the
initial Hessian. All diagonal elements are determined based on an extremely simplified
molecular mechanics model, yet this model provides Hessians that are good starting
points for most systems, thus avoiding any calculation of the exact Hessian. This is
the default for optimizations in redundant internal coordinates. Currently this is not
an option for optimization in Cartesian coordinates.

.INIRED Specifies that the initial Hessian should be diagonal in redundant internal coordi-
nates. The different diagonal elements are set equal to 0.5 for bonds, 0.2 for angles and
0.1 for dihedral angles, unless .INITEV has been specified. If the optimization is run
in Cartesian coordinates, the diagonal internal Hessian is transformed to Cartesians.

.INITEV
<Read EVLINI, the eigenvalues>
The default initial Hessian for first-order minimizations is the identity matrix when
Cartesian coordinates are used, and a diagonal matrix when redundant internal co-
ordinates are used. If .INITEV is used, all the diagonal elements (and therefore the
eigenvalues) are set equal to the value EVLINI.

.LOOSE Default convergence criterion  set to 1.0 · 10−4 .

.MAX IT Maximum number of iterations in geometry optimization scheme (default set to

100).

.MODHES Determine a new model Hessian (see .INIMOD) at every geometry without doing
any updating. The model is thus used in much the same manner as an exact Hessian,
though it is obviously only a relatively crude approximation to the analytical Hessian.

.PRINT
<Print level>
Print level for the output. The higher the print level, the more information is printed
to the output file.

.REDINT Specifies that redundant internal coordinates should be used in the optimization.
This is the default.
CHAPTER 4. LIST OF LSDALTON KEYWORDS 54

.PSB Specifies that a first-order method with the Powell-Symmetric-Broyden (PSB) update
formula should be used for optimization.

.TIGHT Default convergence criterion  set to 1.0 · 10−6 .

.VLOOSE Default convergence criterion  set to 1.0 · 10−3 .

.VTIGHT Default convergence criterion  set to 1.0 · 10−7 .

4.6 **DYNAMI
This input module is devoted to molecular dynamics simulations. lsDalton provides some
tools for performing direct Born-Oppenheimer molecular dynamics, i.e. trajectories are
integrated with forces calculated on the fly. Trajectory integration is available for SCF-type
wavefunctions (with dispersion correction for DFT) and for DEC wavefunctions. Properties
could be calculated along the trajectories and should be specified in the **RESPONSE
section. lsDalton simulates microcanonical(NVE) ensembles. Currently no sampling is
availabe and the starting coordinates and velocities must be provided as input. Trajectories
are integrated with gradient-based velocity-Verlet method [61].
lsDalton provides several options for the choice of starting guess for SCF-type
electronic theory methods. By default, the guess is generated anew each time step according
to the corresponding input. Alternatively, one can restart from the density optimized
at the previous time-step. This is achieved by setting the .RESTART option under the
*DENSOPT subsection. This reduces the number of SCF iterations per time-step and
accelerates the integration. Two more efficient algorithms for starting guess propagation
are implemented in lsDalton Fock matrix dynamics [62] and time-reversible propagator
[63].
Example test cases for **DYNAMI calculations:

These test cases illustrate the main features of direct dynamics. The test cases are located
in LSDALTON/test/ddynam:

• ddynam/H2O fmd : free water molecule with Fock matrix dynamics;

• ddynam/HCN ddyn: HCN dissociation with restart density;

• ddynam/HF vib: a vibrating HF molecule with time-reversible propagator.

.PRINT
<Print level>
CHAPTER 4. LIST OF LSDALTON KEYWORDS 55

Print level for the output. The higher the print level, the more information is printed
to the output file.

.TIMESTEP
Defines integration time step in femtoseconds which should be given in the next line
(real). The default i 0.5 fs.

.MAX ITER
Defines maximum number of time steps. Must be given in the next line (integer).
The default is 100.

.MASSWE
States that the integration is carried out in mass-weighted coordinates. The default
is FALSE.

.MWVEL
States that the input velocities are mass-weighted. The default is FALSE.

.VELOCI
Begins the input of the velocities. In the next line the number of atoms should be
given (integer), then the velocities grouped atom-wise: three(x,y,z) in a line (real).
The default is for all velocities to be zero.

.FOCKMD
Fock matrix dynamics. Below number of points for extrapolation N and polynomial
order M must be specified. Note that M > N (both integer). Works only with
.RESTART option under the *DENSOPT subsection

.TIMREV
Time-reversible density propagation. Below the number of points for extrapolation N
(integer) must be specified. The only thoroughly tested option is N = 2.

4.7 **LOCALIZE ORBITALS

This section describes keywords needed to perform localization of Hartree–Fock orbitals.
The optimization of the chosen localization function will be performed using a trust-region
algorithm [64], and can be performed for the occupied space and/or the virtual orbital
space. For the occupied space the core and valence spaces will be localized separately.
After running a Hartree-Fock calculation followed by an orbital localization two
files with molecular orbital coefficients will be present; cmo orbitals.u and lcm orbitals.u.
cmo orbitals.u contains coefficients for the canonical molecular orbitals (CMO) and lcm orbitals.u
CHAPTER 4. LIST OF LSDALTON KEYWORDS 56

contains coefficients for the localized molecular orbitals (LMO). The coefficient matrix row
index refers to the atomic orbital basis index, while the column index refers to the molecular
orbital basis. The format of the files are as described in the Matrix File Format part of
Section 3.
For the most efficient orbital localization for large molecular systems [65], the .ARH,
.START/TRILEVEL and .LCM options are recommended for use as described in Section
4.4.2, and .PSM with powers 2 2 are recommended for the **LOCALIZE ORBITALS sec-
tion.

.PSM
<m m>
Use powers m of the second moment (PSM) localization function [66]. A power 0
specifies that no localization should be carried out (e.g., 0 1 to only localize the
virtual space). m=1 is equivalent with the Boys localization function. For improved
locality m≥2 is recommended.

.PFM
<m m>
Use powers m of the fourth moment (PFM) localization function [67]. A power 0
specifies that no localization should be carried out (e.g., 0 1 to only localize the virtual
space). Using power 1 yields orbitals with good locality, but a power 2 is recommended
if locality of the orbitals is essential. Powers higher than 3 are not recommended.
The PFM localization function should be chosen if basis sets augmented with diffuse
functions are used.

.PipekMezey
Use Pipek-Mezey localization function using an implementation based on Löwdin
population analysis rather than the traditional Mulliken population analysis (see Ref.
[68]). Both occupied and virtual orbitals will be localized. Please note, the Pipek-
Mezey localization function suffers from system and basis set dependencies. For sys-
tems with complicated bonding structures or when using basis sets augmented with
diffuse functions the PSM or PFM options should be used if good locality is essential.

.OccPipekMezey
Pipek-Mezey localization of only the occupied space.

.VirtPipekMezey
Pipek-Mezey localization of only the virtual space.

.PipekM(MULL)
Use Pipek-Mezey localization function using the Mulliken population analysis. A
CHAPTER 4. LIST OF LSDALTON KEYWORDS 57

power 0 specifies that no localization should be carried out. Otherwise the same
considerations as for the .PipekMezey option.

.OccPipekM(MULL)
Pipek-Mezey localization (using Mulliken population analysis) of occcupied space.

.VirtPipekM(MULL)
Pipek-Mezey localization (using Mulliken population analysis) of virtual space.

.No Level2 Localization

Should only be used in combination with the .START/TRILEVEL and .LCM options
for the *DENSOPT section. The keyword is used to skip orbital localization after the
second level in the three level procedure. Not recommended for general use, but can
be used to save computational time if only the virtual space is to be localized.

.Only Loc
Provided the file with the molecular orbital coefficients (e.g., cmo orbitals.u or lcm orbitals.u)
using this keyword the Hartree-Fock calculation will be skipped, and the orbitals will
be read in and localized. Important: rename or copy the orbital coefficient file to
orbitals in.u, only this file will be read. After the localization procedure the molecu-
lar orbital coefficients corresponding to the localized basis is written to orbitals out.u.
Thus, for this keyword to be used, orbitals in.u must exist and a localization function
must be chosen.

.Orbital Locality
This keyword will make sure second and fourth moment orbital spreads are printed
for all orbitals, not just the least local ones.

.MACRO IT
¡option¿ Set maximum number of macro iterations for orbital localization procedure.
Default is 200.

.LOOSE MICRO THRESH

Looser threshold for solving level-shifted Newton equations in davidson solver in or-
bital localization procedure. Should only be used if a calculation fails and the error
message tells you to use this keyword.

.Orbital Plot
<orbital specification>
Generates .plt file for either the least local or the most local orbitals. To generate
.plt files for the least local orbitals use orbital specification LEASTL, whereas to
generate .plt files for the most local orbitals use specification MOSTL. If both least
CHAPTER 4. LIST OF LSDALTON KEYWORDS 58

local and most local are wanted, make two .Orbital Plot sections with each of
the specifications. The .plt files may be plotted using for example the UCSF Chimera
program package [1]. Note: for plotting other orbitals, use the **PLT option described
in section 4.10. Also note that the grid dimensions may be modified as described in
Section 4.11.

4.8 **RESPONS
This section describes the keywords needed for obtaining molecular properties for HF and
DFT based on the atomic orbital based response formulation [69]. The response section
begins with **RESPONS. General response information related to more than one property
(e.g. how many excitation energies) is put directly under **RESPONS. After this, each
individual response property (e.g. the polarizability tensor) is labeled by an asterisk, and
the specific (optional) information related to that property (e.g. which optical frequencies)
follows below. All input values are given in atomic units. The general structure is thus:

**RESPONS

.General response information

*Label for property 1

.Specific (optional) information for property 1

*Label for property 2

.Specific (optional) information for property 2

Each of the properties below have a test case. The test cases may be found in the LS-
DALTON/test/LSresponse directory. To run a test case, go to the test/ directory and
type:

./TEST <test case>

The test case files contains simple examples of input files.

Below, the list of general response input keywords follows.

.NEXCIT
<Number of excitation energies>
Determines the number of excitation energies to be calculated. If this keyword is set,
the excitation energies and the corresponding one-photon dipole absorption strengths
for the lowest NEXCIT excited states (i.e. excited states 1 to NEXCIT) are always
calculated. Important: If properties involving excited states are to be requested
CHAPTER 4. LIST OF LSDALTON KEYWORDS 59

(e.g. two-photon absorption or the excited state gradient), this keyword must be
listed before any of these properties.
Test case: LSresponse/LSresponse HF opa

4.8.1 *SOLVER
General information related to the way response equations are solved to obtain a given
property. By default, standard and eigenvalue response equations are solved using the
algorithm with paired trial vectors [70], whereas damped (complex) response equations are
solved using the algorithm with symmetrized trial vectors [71].
This choice can be modified using the .SYM SOLVER and .PAIR SOLVER keywords, see below.

.SYM SOLVER
The solver with symmetrized trial vectors is used for solving standard response equa-
tions.
Test case: LSresponse/LSresponse HF alpha astv

.PAIR SOLVER
The solver with paired trial vectors is used for solving damped (complex) response
equations.
Test case: LSresponse/LSresponse DFT alpha aptv

.CONVTHR
<Convergence threshold>
Convergence threshold to which response equations are solved. Default value is set to
10−4 .

.MAXIT
<Maximum number of iterations>
Maximum number of iterations in the iterative procedure. Default value is set to 100.

.MAXRED
<Maximum size of the reduced space>
Maximum size of the reduced space. Default value set equal to 200.

.AOPREC
AO preconditioning is used (MO preconditioning by default). AO preconditioning
may only be used within the algorithm with paired trial vectors.

.NOPREC
No preconditioning is used.
CHAPTER 4. LIST OF LSDALTON KEYWORDS 60

.CONVDYN
<Option>
Dynamic convergence threshold suitable for large calculations. Options are TIGHT,
STANDARD, and SLOPPY.

.RESTEXC
<Number Of excitation Vectors on file>
This is a restart option, that allows the user to restart the calculation.
Assume for instance that you have already done one calculation where you requested 6
excitation energies but realize that you need 10. You can provide the file rsp eigenvecs
and use
.RESTEXC
6
option to restart the calculation from the previous. This option can also be used to
bypass the excitation vector step if the previous calculation failed for some property
evaluation which needed the excitation vectors.

.DTHR
<Threshold>
Threshold for when excited states are to be considered degenerate

Now follows a list of input keywords for the specific properties available in lsDalton.

4.8.2 *DIPOLE
Calculation of the permanent dipole moment

4.8.3 *DIPOLEMOMENTMATRIX
Calculation of the full dipole moment matrix:

 
µ00 µ01 µ02 ...
 µ10 µ11 µ12 ... 
µ=
 µ20 µ21 µ22
 (4.3)
... 
... ... ... ...

• µ00 denote the permanent dipole moment for the ground state

• µ0i denote the transition dipole moment between the ground state and the i’th excited
state
CHAPTER 4. LIST OF LSDALTON KEYWORDS 61

• µii denote the Permanent dipole moment for the i’th excited states

• µij denote the transition dipole moment between the i’th and j’th excited state

Requires specification of .NEXCIT!

4.8.4 *ALPHA
Calculation of the (possibly complex) electric dipole polarizability tensor αAB and its
isotropic average. The polarizability equals minus the linear response function hhµA ; µB iiωB ,
where ωB is in general allowed to be complex – that is, ωB = ωB R +iω I . Unspecified frequen-
B
cies are set to zero by default, i.e. if no frequencies are specified, the static polarizability is
calculated.
Test case: LSresponse/LSresponse HF alpha

.BFREQ
<Number of real frequencies for B operator>
<Freq1, freq2, ... , freqN>
R in the corresponding linear response function. The first line after
Real frequencies ωB
.BFREQ contains the number of frequencies and the second line contains all frequency
values (on one single line).

.IMBFREQ
<Number of imaginary frequencies for B operator>
<Freq1, freq2, ... , freqN>
I in the corresponding linear response function. The imagi-
Imaginary frequencies ωB
nary part of the complex polarizability describes a broadened one-photon absorption
spectrum. If .IMBFREQ is not specified, only the real polarizability is calculated.
Input as for .BFREQ.

For example, the following input is used to calculate complex polarizability tensors for the
frequencies ωB = 0.1 + 0.05i and ωB = 0.2 (in a.u.):

**RESPONS

*ALPHA

.BFREQ

0.1 0.2
CHAPTER 4. LIST OF LSDALTON KEYWORDS 62

.IMBFREQ

0.05 0.0

4.8.5 *BETA
Calculation of the (possibly complex) first electric dipole hyperpolarizability tensor βABC
and the corresponding averages βk and β⊥ . The first hyperpolarizability tensor equals minus
the quadratic response function hhµA ; µB , µC iiωB ,ωC , where the frequencies are in general
allowed to be complex – i.e. ωB = ωB R + iω I and ω = ω R + iω I . The input is analogous to
B C C C
the *ALPHA input. Unspecified frequencies are set to zero by default, i.e. if no frequencies
are specified, the static first hyperpolarizability is calculated.
Test case: LSresponse/LSresponse HF beta
.BFREQ <Number of real frequencies for B operator><Freq1, freq2, ... , freqN>Real
R in the corresponding quadratic response function. Input as for .BFREQ
frequencies ωB
under *ALPHA.

.IMBFREQ
<Number of imaginary frequencies for B operator>
<Freq1, freq2, ... , freqN>
I in the corresponding quadratic response function. Input as
Imaginary frequencies ωB
for .IMBFREQ under *ALPHA.

.CFREQ
<Number of real frequencies for C operator>
<Freq1, freq2, ... , freqN>
R in the corresponding quadratic response function. Input as for
Real frequencies ωC
.BFREQ under *ALPHA.

.IMCFREQ
<Number of imaginary frequencies for C operator>
<Freq1, freq2, ... , freqN>
I in the corresponding quadratic response function. Input as
Imaginary frequencies ωC
for .IMBFREQ under *ALPHA.

4.8.6 *DAMPED TPA

Damped two-photon absorption [72] where both individual photons have the same energy
(= half the excitation energy).
Test case: LSresponse/LSresponse HF dtpa
CHAPTER 4. LIST OF LSDALTON KEYWORDS 63

.OPFREQ
<Number of one-photon frequencies>
<Freq1, freq2, ... , freqN>
The frequencies in the input are one-photon frequencies (=half the excitation energy
in case of resonance).

.GAMMA
<Damping parameter>
The damping parameter γ is used in all response equations, i.e. iγ is effectively added
to all frequencies occurring in the response equations. Default value: 0.005 a.u.

4.8.7 *QUASIMCD
Calculation of the A and B terms of magnetic circular dichroism (MCD), based on the
method of Ref. [73], but reformulated using the formalism of Ref. [69]. The calculation of
Damped MCD spectra according to Ref. [74] is also possible with this keyword.
Note that one or more of the 3 keywords { .DAMPEDXCOOR , .MCDEXCIT,
.DAMPEDRANGE } must be specified
If .MCDEXCIT is specified, the default is to proceed in the following way:
1. Determine the excited states requested under .MCDEXCIT

2. Determine B terms for all allowed transitions. Both London atomic orbitals LAO (also
called Gauge including atomic orbitals GIAO) and non-LAOs are used to determine
the B terms. LAOs are superior to standards AOs, but the difference can give an
indication of how close the result is to the basis set limit.

3. Simulate the MCD spectrum based on the calculated B terms (sticks) by associating
a Lorentzian lineshape function (with lineshape parameter of 0.005 a.u.) to each B
term (See Ref. [73]). The simulated spectrum is written to a file MCDspectraAU.dat
(in atomic units). The first number is the frequency, the second the LAO ellipticity
and the third is the non-LAO ellipticity. A corresponding file is generated called
MCDspectra.dat using standard units (cm−1 for frequencies and molar ellipticity for
the ellipticity). The individual B term contribution to the MCD spectra is written
to the files BtermAU.dat and Bterm.dat. Per default 5000 points are used for the
2µ
simulated spectrum. The standard units for the B term are given as Dcm−1
B
, where D is
the unit Debye and µB is the Bohr magneton. The conversion factor between atomic
units and the standard units is 5.887641̇0−5 . The standard unit for the A term is
given as D2 µB , with a conversion factor of 12.920947806163007.

4. Perform damped calculations based on the standard MCD calculation. A number of

points around each individual MCD peak (default value: 10) are determined for which
CHAPTER 4. LIST OF LSDALTON KEYWORDS 64

to calculate the damped spectra. The result is printed to the files dampedMCDspec-
traRAU.dat (atomic units) and dampedMCDspectraR.dat.
Test case: LSresponse/LSresponse mcd calc, LSresponse/LSresponse mcd
LSresponse/LSresponse mcd2, LSresponse/LSresponse mcd3
Note that only the LSresponse/LSresponse mcd calc test case is a good example for a
general MCD calculation.

.MCDEXCIT
<Number of excited states to consider>
The B terms will be calculated for all states, unless they are zero due to selection
rules (dipole forbidden)

.MCDEXSTATES
<Number of specific excited states to consider in MCD calc>
<Specific excited states to consider>
This keyword only calculates the B and possible A terms for the selected excited
states. The first line after .MCDEXSTATES contains the number of specific excited
states for which the MCD is to be calculated, and the second line specifies such states
(identified by their index in the .MCDNEXCIT list). Important: The position of
the highest lying excited state specified by .MCDEXSTATES must be smaller than
or equal to the total number of excited states specified by .MCDNEXCIT!

.DEGENERATE
This keyword must be set in input if the molecule has a degenerate state so that A
terms are possible, otherwise all states are assumed non-degenerate and no A terms
will be calculated. When this keyword is set, the A terms are calculated and a deriva-
tive Lorentzian lineshape function is used to simulate the individual A term contri-
butions to the MCD spectrum (See Ref. [73]). The individual A terms contributions
to the MCD spectrum are written to the files AtermAU.dat and Aterm.dat

.DAMPEDXCOOR
<Number of frequencies>
<Freq1, freq2, ... , freqN>
The frequencies for which to calculate the damped MCD spectra (in atomic units).

.DAMPEDRANGE
<Freq1>
<FreqN>
<Number of frequencies in range between Freq1 and Freq2>
The frequencies for which to calculate the damped MCD spectra (in atomic units),
here specified as an interval.
CHAPTER 4. LIST OF LSDALTON KEYWORDS 65

.NO LONDON
Deactivates the use of LAOs (i.e., only standard, non-LAO, basis is used)

.NO NONLONDON
Deactivates the use of non-LAOs (i.e., only LAO are used)

.NO SIMULATE
Deactivates the simulation of the MCD spectrum from the individual MCD terms

.NO ATERM
Deactivates the calculation of A terms

.NO BTERM
Deactivates the calculation of B terms

.NO DAMPED
Deactivates the execution of damped MCD spectra

.GAUSSIAN
Uses Gaussian lineshape functions (with a lineshape parameter of 0.0070851079363103793)
instead of the default Lorentzian lineshape functions.

.LINESHAPEPARAM
<lineshape parameter>
Changes the lineshape parameter (default value is 0.005 a.u. for Lorentzian, and
0.0070851079363103793 a.u. for Gaussian) used to simulate the spectrum

.NSTEPS
<number of frequencies used for the simulated MCD spectra>
Changes the number of frequencies used for the simulation of the MCD spectra from
the individual A and B terms’ sticks (default 5000).

.NVECFORPEAK
<number of frequencies used around each MCD peak>
Changes the number of frequencies used around each MCD peak (default 10) in the
damped response calculation.

Note that the keyword .DTHR under *SOLVER (see section 4.8.1 ) can be used to define
the threshold below which an excited state is considered degenerate

4.8.8 *SHIELD
Nuclear magnetic shielding tensors and chemical shift can be calculated with this keyword
Test case: LSresponse/LSresponse NMR SHIELD
CHAPTER 4. LIST OF LSDALTON KEYWORDS 66

4.8.9 *INASHIELD
Individual atom nuclear magnetic shielding tensors and chemical shifts. NB: Requires the
use of SubSystem labels in the MOLECULE.INP file see testcase Test case: LSrespon-
se/LSresponse NMR IANS SHIELD NoOpenRSP

ESDIPOLE
Permanent dipole moment for excited states. Always requires specification of .NEX-
CIT!
If .EXSTATES is not specified, the excited state dipole moments for all excited states from
1 to NEXCIT are calculated.
Test case: LSresponse/LSresponse HF esd

.EXSTATES
<Number of specific excited states to consider>
<Specific excited states to consider>
Only calculate excited state gradient for selected excited states. The first line after
.EXSTATES contains the number of excited states for which the excited state dipole
moment is to be calculated, and the second line specifies such states (identified by
their index in the .NEXCIT list). Important: The position of the highest lying
excited state specified by .EXSTATES must be smaller than or equal to the total
number of excited states specified by .NEXCIT!

4.8.10 *ESGRAD
Calculation of the molecular gradient for excited states. Always requires specification
of .NEXCIT! If .EXSTATES is not specified, excited state gradients for all excited states
from 1 to NEXCIT are calculated.
Test case: LSresponse/LSresponse HF esg

.EXSTATES
<Number of specific excited states to consider>
<Specific excited states to consider>
Input is identical to .EXSTATES under *ESDIPOLE.

For example, the following input is used to calculate excited state gradients for excited
states number 3 and 6:

**RESPONS

.NEXCIT
CHAPTER 4. LIST OF LSDALTON KEYWORDS 67

*ESGRAD

.EXSTATES

3 6

4.8.11 *GAMMA
Calculation of the (possibly complex) second electric-dipole hyperpolarizability tensor γABCD
and the corresponding averages γk and γ⊥ . The second hyperpolarizability tensor equals
minus the cubic response function hhµA ; µB , µC , µD iiωB ,ωC ,ωD , where the frequencies are in
general allowed to be complex – i.e. ωB = ωB R + iω I , ω = ω R + iω I , and ω = ω R + iω I .
B C C C D D D
The input is analogous to those for *ALPHA and *BETA. Unspecified frequencies are set
to zero by default, i.e. if no frequencies are specified the static second hyperpolarizability
is calculated.
Test case: LSresponse/LSresponse HF gamma

.BFREQ
<Number of real frequencies for B operator>
<Freq1, freq2, ... , freqN>
R in the corresponding cubic response function.
Real frequencies ωB Input as for
.BFREQ under *ALPHA.

.IMBFREQ
<Number of imaginary frequencies for B operator>
<Freq1, freq2, ... , freqN>
I in the corresponding cubic response function. Input as for
Imaginary frequencies ωB
.BFREQ under *ALPHA.

.CFREQ
<Number of real frequencies for C operator>
<Freq1, freq2, ... , freqN>
R in the corresponding cubic response function.
Real frequencies ωC Input as for
.BFREQ under *ALPHA.

.IMCFREQ
<Number of imaginary frequencies for C operator>
<Freq1, freq2, ... , freqN>
CHAPTER 4. LIST OF LSDALTON KEYWORDS 68

I in the corresponding cubic response function. Input as for

Imaginary frequencies ωC
.IMBFREQ under *ALPHA.

.DFREQ
<Number of real frequencies for D operator>
<Freq1, freq2, ... , freqN>
R in the corresponding cubic response function.
Real frequencies ωD Input as for
.BFREQ under *ALPHA.

.IMDFREQ
<Number of imaginary frequencies for D operator>
<Freq1, freq2, ... , freqN>
I in the corresponding cubic response function. Input as for
Imaginary frequencies ωD
.IMBFREQ under *ALPHA.

4.8.12 *MOLGRA
Single point calculation of the molecular gradient.
Test case: LSresponse/LSresponse HF molgra

4.8.13 *TPA
Two-photon absorption where both individual photons have the same energy (= half the
excitation energy). Always requires specification of .NEXCIT! If .EXSTATES is not
specified, two-photon absorption for all excited states from 1 to NEXCIT are calculated.
Test case: LSresponse/LSresponse HF tpa

.EXSTATES
<Number of specific excited states to consider>
<Specific excited states to consider>
Input is identical to .EXSTATES under *ESDIPOLE.

4.9 **DEC and **CC

Correlated coupled-cluster (CC) calculations using the Divide-Expand-Consolidate (DEC)
scheme [75, 76, 77, 78, 79, 80, 81], where local orbitals are used to carry out the correlated
calculation on a large molecular system in terms of many small local calculations. For-
mally, the computational time for the DEC scheme scales linearly with system size. DEC
calculations are invoked using the **DEC keyword.
Furthermore, conventional coupled-cluster energies for the RI-MP2, MP2, CC2,
CCSD, and CCSD(T) models are available using the **CC keyword. The computational
CHAPTER 4. LIST OF LSDALTON KEYWORDS 69

time for these implementations scale with the system size to the fifth, fifth, fifth, sixth, and
seventh power for RI-MP2, MP2, CC2, CCSD, and CCSD(T), respectively.
The **DEC and **CC implementations uses minimal I/O and employ massive par-
allelism and will therefore only be effective if many computing nodes are available.
Special note: It is mandatory that the memory available for the calculation is
specified in the input, see .MEMORY keyword below.
The **DEC and **CC sections use local orbitals by default, although the orbitals
are converted to the (pseudo-)canonical basis when the MP2 and CC equations are solved
(as well as for CCSD(T)). It is encouraged to use the keyword .CANONICAL when using
**CC.
Below are keywords for the **DEC section and **CC section and relevant test cases
are given.

**DEC input keywords :

.MP2
DEC-MP2 model requested

.RIMP2
DEC-RI-MP2 model requested

.CC2
DEC-CC2 model requested

.CCSD
DEC-CCSD model requested

.CCSD(T)
DEC-CCSD(T) model requested

.FOT
<Threshold>
Fragment optimization threshold (FOT), it defines the precision of a DEC calculation
compared to a conventional calculation [75, 76, 77, 78, 79, 80]. Default value is
10−4 .a.u. The error in the total correlation energy is propotional to the FOT and the
size of the system.

.DECPRINT
<Print level>
Level of information printed to LSDALTON.OUT. The input is a integer and the
default value is 0.
CHAPTER 4. LIST OF LSDALTON KEYWORDS 70

.MEMORY
<Memory in gigabytes>
It is mandatory to specify the memory (in gigabytes) available for the calculation using
this keyword! For an MPI run this is the memory available for each MPI process. If
this keyword is not set, the program will stop with a quit statement informing about
an error in the LSDALTON.INP file. (This keyword must also be used in **CC
section).

.NOTABSORBH
By default orbitals originally assigned to hydrogen atoms are reassigned to the nearest
heavy atom. This reassigning can be turned off by invoking this keyword.

.DENSITY
Calculate DEC-MP2 density matrix [79] (stored in MP2.dens). The MP2 electric
dipole moment is also calculated if this keyword is invoked.

.GRADIENT
Calculate MP2 molecular gradient [79] at input geometry (for geometry optimization,
see test case examples below). The MP2 density and electric dipole moment is also
calculated if this keyword is invoked.

.KAPPATHR
<THR>
¯ multiplier equations solved when DEC-MP2 molecular gradient
Threshold for kappa
is calculation. Default value is 10−4 .

.MPIGROUPSIZE
<Size>
Size of local MPI groups (integer) in massively parallel DEC scheme [80]. The default
settings should suffice for most purposes, and use of the keyword is only recommended
for advanced users.

Example test cases for **DEC calculations:

These test cases illustrate the main features of the DEC scheme. The test cases are located
in the test/dectests directory.

• decmp2 energy: Calculate DEC-MP2 energy.

• decmp2 density: Calculate DEC-MP2 density matrix and electric dipole moment.
CHAPTER 4. LIST OF LSDALTON KEYWORDS 71

• decmp2 gradient: Calculate DEC-MP2 molecular gradient.

• decmp2 geoopt: Carry out DEC-MP2 geometry optimization. Note that the geom-
etry optimization is invoked by the **OPTIMIZE section, while the **DEC section
specifies that the geometry optimization should use DEC-MP2 energies and molecu-
lar gradients. The DEC-MP2 geometry optimization will automatically estimate the
intrinsic energy error of the DEC calculation. If this error is larger than the difference
between two geometry, the FOT level will be increased to get a more accurate energy
and gradient in the next iteration.

**CC input keywords:

.MP2
Use MP2 model. This gives the canonical MP2 energy and is not optimized for parallel
performance. It is only intended to be used for benchmarking DEC-MP2 calculations
on relatively small molecular systems.

.RIMP2
Use RI-MP2 model. This gives the canonical RI-MP2 energy and is optimized for
parallel performance. It is encourage to use with the .CANONICAL keyword

.CC2
Use CC2 model which is not an optimized code, but rather uses the CCSD optimized
code with some contributions removed. See CCSD documentation for details.

.CCSD
Use CCSD model. The code employed is a massively parallel version and flexible with
respect to memory. Due to the use of one-sided communication in these models it is
recommended to run CCSD with an asynchronous progress engine switched on (how
to do that can be found in the documentation of the respective MPI library you are
using). When running a CCSD calculation make sure that at least one quantity of
size V 2 O2 (O = number of occupied orbitals, V = number of virtual orbitals) fits into
memory in double precision floating point numbers (with some additional space of
course). Using this keyword automatically triggers checkpointing after each iteration.
The calculation can be restarted with the usual ..RESTART keyword.

.CCSD(T)
Use the CCSD(T) model. This gives the canonical CCSD(T) energy and is optimized
for parallel performance. The .CANONICAL keyword must be used for CCSD(T).
CHAPTER 4. LIST OF LSDALTON KEYWORDS 72

.MEMORY
<Memory in gigabytes>
It is mandatory to specify the memory (in gigabytes) available for the calculation using
this keyword! For an MPI run this is the memory available for each MPI process. If
this keyword is not set, the program will stop with a quit statement informing about
an error in the LSDALTON.INP file. (This keyword must also be used in **DEC
section).

.SNOOP
Calculate intermolecular interaction energies using the Same Number Of Optimized
Parameters (SNOOP) scheme [82]. This can be done for the RIMP2, MP2, CCSD,
and CCSD(T) models. It is important to specify the subsystems in the molecule input
file. Example inputs are given in the ”snoop” test cases in the test/dectests directory.

.SNOOPTHR
The threshold for the Hartree–Fock optimization of SNOOP monomers. Default value
is very tight (10−7 ), and it is recommended to keep it tight due to a simplistic imple-
mentation of SCF loop.

.SNOOPMAXITER
Maximum number of iterations in SNOOP monomer Hartree–Fock calculations.

.SNOOPMAXDIIS
Maximum number of DIIS vectors to store in SNOOP HF calculations (RH/DIIS
scheme). If the SNOOP monomer Hartree–Fock calculations do not converge, it might
be advantageous to increase this value (default value is 5).

.SNOOPRESTART
Restart SNOOP calculation. It requires SNOOP restart files (SNOOP*****.restart).

Example test cases for **CC calculations:

The test cases are located in LSDALTON/test/dectests.

• fullmp2 energy: Calculate conventional canonical MP2 energy.

• fullccsd high: Calculate conventional canonical CCSD energy.

Common **DEC and **CC input keywords:

CHAPTER 4. LIST OF LSDALTON KEYWORDS 73

.HFRESTART
Start DEC/CC calculation from HF restart files dens.restart, fock.restart, lcm orbitals.u,
and overlapmatrix

.RESTART
Restart unfinished DEC/CC calculation. Requires that HF files dens.restart, fock.restart,
lcm orbitals.u, and overlapmatrix are present in the folder where the calculation is
restarted. Additionally, if .info files are present it means that some of the fragment
calculations have been carried out and the .RESTART keyword will read information
about the finished fragment calculations from file and calculate the missing frag-
ments. Specifically, if the files atomicfragments.info and atomicfragmentsdone.info
are present, it means that some (or all) of the atomic fragment calculations are fin-
ished. If, in addition, the file fragenergies.info (energy calculation) or mp2grad.info
(MP2 density or gradient calculation) is also present, it means that some (or all) of
the pair fragments are also done.
Note 1: This keyword should NOT be combined with .RESTART in the
*DENSOPT section.
Note 2: This keyword cannot be used for a DEC-MP2 geometry opti-
mization. The only way to ”restart” a DEC-MP2 geometry optimiza-
tion is use the current geometry in a new DEC-MP2 geometry opti-
mization – i.e., copy the most recently generated MOLECULE.xxx file
to MOLECULE.INP and run the calculation again using the same LS-
DALTON.INP.

.FROZENCORE
Use the frozen core approximation.

.CANONICAL
Use canonical orbitals. This is recommended in combination with **CC while it is
only recommended for advanced users for testing purposes in connection with the
**DEC keyword. In general, the DEC scheme is only meaningful for local orbitals.

.CCMAXITER
Maximum number of iterations in CC solver.

.CCTHR
Convergence threshold for residual norm in CC solver.

.SUBSIZE
Number of previous vectors to store in CROP scheme [50] for CC solver.
CHAPTER 4. LIST OF LSDALTON KEYWORDS 74

.CCSDNOSAFE
Prevent saving of CCSD amplitudes as checkpoint files.

.INTEGRALTHRESHOLD
Integral screening threshold in the correlation treatment.

4.10 **PLT
Generation of .plt files which may be used to visualize densities, orbitals, and electrostatic
potentials by calculating these at specific points in space. The .plt files can be visualized
e.g. using the Chimera program [1].
Important note: This section assumes that an LSDALTON calculation has already
been carried out to generate files containing density matrix elements or molecular orbital
coefficients. The .plt files can then be generated by running a new calculation where the
LSDALTON.INP file has been modified as exemplified below.
The density matrix from HF and DFT calculation are saved in a file dens.restart,
while the MP2 density is saved in the file MP2.dens (if requested, see **DEC section).
The canonical MO coefficients are saved in the file cmo orbitals.u, while localized molecular
orbitals are saved in a file lcm orbitals.u (if requested). The definition of the grid points is
described in the **PLTGRID section.
Note that it is also possible to generate .plt files for orbitals on the run as described
in Section 4.7.

.INPUT
<Name of input file>
The name of the input file containing density matrix elements or orbital coefficients.
Possible input files include dens.restart, MP2.dens, cmo orbitals.u, and lcm orbitals.u.

.OUTPUT
<Name of output file>
The name of the output plt-file where the calculated values at each grid point are
saved. Examples are given below.

.DENS
Construct .plt file for electron density at grid points. For example to calculate elec-
tron density from the file dens.restart and save information in file dens.plt, insert the
following section in the LSDALTON.INP file used to generate the dens.restart file in
the first place:
**PLT
.INPUT
CHAPTER 4. LIST OF LSDALTON KEYWORDS 75

dens.restart
.OUTPUT
dens.plt
.DENS

.EP
Construct .plt file for electrostatic potential grid points based on input density matrix.
For example to calculate electrostatic potential from the density in the file dens.restart
and save information in file ep.plt, insert the following section in the LSDALTON.INP
file used to generate the dens.restart file in the first place:
**PLT
.INPUT
dens.restart
.OUTPUT
ep.plt
.EP

.ORB
<Orbital index>
Construct .plt file for specific orbital at grid points. For example to calculate values
of orbital number 8 at grid points from the file lcm orbitals.u and save information in
file orb8.plt, insert the following section in the LSDALTON.INP file used to generate
the lcm orbitals.u file in the first place:
**PLT
.INPUT
lcm orbitals.u
.OUTPUT
orb8.plt
.ORB
8

.CHARGEDIST
<Orbital index 1 Orbital index 2>
Construct .plt file for charge distribution of two orbitals at grid points. For example
to calculate charge distribution between orbitals 5 and 8 at grid points from the file
lcm orbitals.u and save information in file cd 5 8.plt, insert the following section in
the LSDALTON.INP file used to generate the lcm orbitals.u file in the first place::
**PLT
CHAPTER 4. LIST OF LSDALTON KEYWORDS 76

.INPUT
lcm orbitals.u
.OUTPUT
cd 5 8.plt
.CHARGEDIST
58

4.11 **PLTGRID
Definition of 3-dimensinal grid used for generation of .plt files in **PLT section. The
default grid choice should suffice for most visualization purposes. This section is only for
the advanced user who wishes to modify the default choice of grid. This also modifies the
grid used when .plt files for local orbitals are generated on the run, see Section 4.7.
The grid box is defined in the following manner:

• The first point in the grid box is (X1,Y1,Z1).

• The remaining grid points are then defined by going out in the X, Y, and Z directions
with step sizes deltax, deltay, and deltaz, until there are nX, nY, and nZ points in the
X,Y, and Z directions (giving a total number of gridpoints nX × nY × nZ).

There are two options to define the grid box: (i) manual gridbox where X1, Y1, Z1,
deltax, deltay, deltaz, nX, nY, nZ are defined explicitly in the input file; (ii) molecule-specific
gridbox where all atoms are contained within the grid box AND there is an additional buffer
zone around the outermost atoms in the X, Y, and Z directions. Option (ii) thus requires
that the deltax, deltay, deltaz, and buffer values are defined by the input, which implicitly
defines X1, Y1, Z1, nX, nY, and nZ based on the molecular structure.
The default gridbox uses option (ii) with deltax=deltay=deltaz=0.3 a.u. and buffer=6.0
a.u. User-defined grid box based on options (i) and (ii) use the following inputs:

.MANUAL
<X1 Y1 Z1>
<deltax deltay deltaz>
<nX nY nZ>
Manual definition of gridbox as described above. All input values are in a.u.

.MOLECULE
<buffer>
<deltax deltay deltaz>
Molecule-specific definition of gridbox as described above. All input values are in a.u.
CHAPTER 4. LIST OF LSDALTON KEYWORDS 77

4.12 **PCM
Keywords specified in this section set up the inclusion of a continuum description of the
solvent using the polarizable continuum model for solvation [2].

.SEPARATE
Triggers whether the nuclear and electronic electrostatic potential and apparent sur-
face charge should be treated separately. This is a debug option.
Bibliography

[1] E. F. Pettersen, T. D. Goddard, C. C. Huang, G. S. Couch, D.M. Greenblatt, E. C.

Meng, and T. E. Ferrin. J. Comp. Chem., 25:1605, 2004.

[2] Jacopo Tomasi, Benedetta Mennucci, and Roberto Cammi. Quantum Mechanical
Continuum Solvation Models. Chem. Rev., 105(8):2999–3094, 2005.

[3] Monica Bugeanu, Roberto Di Remigio, Krzysztof Mozgawa, Simen Sommerfelt Reine,
Helmut Harbrecht, and Luca Frediani. Wavelet Formulation of the Polarizable Con-
tinuum Model. II. Use of Piecewise Bilinear Boundary Elements. Phys. Chem. Chem.
Phys., 2015.

[4] B. Jansı́k, S. Høst, M. P. Johansson, J. Olsen, and P. Jørgensen. Robust and Reli-
able Multilevel Minimization of the Kohn—Sham Energy. J. Chem. Theory Comput.,
5:1027, 2009.

[5] B. Jansı́k, S. Høst, M. P. Johansson, J. Olsen, and P. Jørgensen. A stepwise atomic,

valence–molecular, and full—molecular optimisation of the Hartree—Fock/Kohn—
Sham energy. Phys. Chem. Chem. Phys., 11:5805, 2009.

[6] T. Helgaker, P. Jørgensen, and J. Olsen. Molecular Electronic Structure Theory, First
Edition. Wiley, 2000.

[7] PCMSolver, an Application Programming Interface for the Polarizable Continuum

Model electrostatic problem, written by R. Di Remigio, L. Frediani and K. Mozgawa,
(see http://pcmsolver.readthedocs.org).

[8] Manuel Guidon, Jürg Hutter, and Joost VandeVondele. Auxiliary density matrix meth-
ods for Hartree-Fock exchange calculations. Journal of Chemical Theory and Compu-
tation, 6(8):2348–2364, August 2010.

[9] Patrick Merlot, Róbert Izs’ak, Alex Borgoo, Thomas Kjærgaar, Trygve Helgaker, and
Simen Reine. Charge-constrained auxiliary-density-matrix methods for the hartree-
fock exchange contribution. Journal of Chemical Physics, 141:094104, August 2014.

78
BIBLIOGRAPHY 79

[10] Yihan Shao and Martin Head-Gordon. An improved j matrix engine for density func-
tional theory calculations. Chem. Phys. Lett., 323(5-6):425, 2000.

[11] Yihan Shao, Christopher A. White, and Martin Head-Gordon. Efficient evalua-
tion of the coulomb force in density-functional theory calculations. J. Chem. Phys.,
114(15):6572, 2001.

[12] Christian Ochsenfeld, Christopher A. White, and Martin Head-Gordon. Linear and
sublinear scaling formation of hartree–fock-type exchange matrices. J. Chem. Phys.,
109(5):1663, 1998.

[13] Patrick Merlot, Thomas Kjaergaard, Trygve Helgaker, Roland Lindh, Francesco
Aquilante, Simen Reine, and Thomas Bondo Pedersen. Attractive electron-electron
interactions within robust local fitting approximations. J. Comp. Chem., pages 1–11,
April 2013.

[14] L E McMurchie and E R Davidson. One- and two-electron integrals over cartesian
gaussian functions. J. Comp. Phys., 26:218–231, 1978.

[15] Simen Reine, Erik Tellgren, and Trygve Helgaker. A unified scheme for the calculation
of differentiated and undifferentiated molecular integrals over solid-harmonic gaussians.
Phys. Chem. Chem. Phys., 9:4771–4779, 2007.

[16] P. Hohenberg and W. Kohn. Phys. Rev., 136:B864, 1964.

[17] W. Kohn and L. J. Sham. Phys. Rev., 140:A1133, 1965.

[18] J. C. Slater. Quantum Theory of Molecular and Solids, volume 4, chapter The Self-
Consistend Field for Molecular and Solids. McGraw-Hill, New York, 1974.

[19] A. D. Becke. Phys. Rev. A, 38:3098, 1988.

[20] N. C. Handy and A. J Cohen. Mol. Phys., 99:403, 2001.

[21] J. P. Perdew, K. Burke, and M. Ernzerhof. Phys. Rev. Lett., 77(3865), 1996.

[22] J. P. Perdew and Y. Wang. Phys. Rev. B, 33:8800, 1986.

[23] S. H. Vosko, L. Wilk, and M. Nusair. Can. J. Phys, 58:1200, 1980.

[24] C. Lee, W. Yang, and R. G. Parr. Phys. Rev. B, 57:785, 1988.

[25] B. Miehlich, A. Savin, H. Stoll, and H. Preuss. Chem. Phys. Lett., 157:200, 1989.

[26] J. P. Perdew. Phys. Rev. B, 33:8822, 1986.

BIBLIOGRAPHY 80

[27] J. P. Perdew, J. A. Chevary, S. H. Vosko, K. A. Jackson, M. R. Pederson, and C. Fi-

olhais. Phys. Rev. B, 46:6671, 1992.

[28] J. P. Perdew and A. Zunger. Phys. Rev. B, 23:5048, 1981.

[29] R. van Leeuwen and E. J. Baerends. Exchange-correlation potential with correct

asymptotic behavior. Phys. Rev. A, 1994.

[30] A. D. Becke. J. Chem. Phys., 98:5648, 1993.

[31] T. Yanai, D. P. Tew, and N. C. Handy. A new hybrid exchange-correlation functional

using the Coulomb-attenuating method (cam-b3lyp). Chem. Phys. Lett., 393:51, 2004.

[32] T. W. Keal and D. J. Tozer. The exchange-correlation potential in kohn-sham nuclear

magnetic resonance shielding calculations. J. Chem. Phys., 119:3015, 2003.

[33] T. W. Keal D. J. Tozer and T. Helgaker. Giao shielding constants and indirect spin-
spin coupling constants: performance of density functional methods. Chem. Phys.
Lett., 391:374, 2004.

[34] T. W. Keal and D. J. Tozer. A semi-empirical generalised gradient approximation

exchange-correlation functional. J. Chem. Phys., 121:5654, 2004.

[35] C. Adamo and V. Barone. J. Chem. Phys., 110:6158, 1999.

[36] Ulf Ekström, Lucas Visscher, Radovan Bast, and Andreas J Thorvaldsen. Arbitrary-
Order Density Functional Response Theory. J. Chem. Theory Comput., 6:1971–1980,
2010.

[37] P. Salek and A. Hesselmann. A self-contained and portable density functional theory
library for use in Ab Initio quantum chemistry programs. J. Comput. Chem, 28:2569,
2007.

[38] N. C. Handy C. W. Murray and G. J. Laming. Mol. Phys., 78:997, 1993.

[39] S. Grimme. J. Comp. Chem., 27:1787, 2006.

[40] S. Grimme, J. Antony, S. Ehrlich, and H. Krieg. J. Chem. Phys., 132:154104, 2010.

[41] S. Grimme, S. Ehrlich, and L. Goerigk. J. Comp. Chem., 32:1456, 2011.

[42] P.-A. Malmqvist R. Lindh and L. Gagliardi. Molecular integrals by numerical quadra-
ture. i. radial integration. Theor. Chem. Acc., 106:178–187, 2001.

[43] O. Treutler and R. Ahlrichs. J. Chem. Phys., 102:346, 1995.

BIBLIOGRAPHY 81

[44] A. D. Becke. J. Chem. Phys., 88:2547, 1988.

[45] G. Scuseria R. E. Stratmann and M. J. Frisch. Chem. Phys. Lett., 257:213, 1996.

[46] P. Macak M. A. Watson, P. Salek and T. Helgaker. J. Chem. Phys., 121:2915, 2004.

[47] S. Hirata, C.G. Zhan, E. Apr, T.L. Windus, and D.A.Dixon. A new, self-contained
asymptotic correction scheme to exchange-correlation potentials for time-dependent
density functional theory. J. Phys. Chem. A, 107:10154, 2003.

[48] M.E. Casida and D.R. Salahub. Asymptotic correction approach to improving ap-
proximate exchange–correlation potentials: Time-dependent density-functional theory
calculations of molecular excitation spectra. J. Chem. Phys., 113:8918, 2000.

[49] C.G.Zhan, J.A. Nichols, and D.A. Dixon. Ionization potential, electron affinity, elec-
tronegativity, hardness, and electron excitation energy: molecular properties from den-
sity functional theory orbital energies. J. Phys. Chem. A, 107:4184, 2003.

[50] M. Ziólkowski, V. Weijo, Poul Jørgensen, and J. Olsen. An efficient algorithm for
solving nonlinear equations with a minimal number of trial vectors: Applications to
atomic-orbital based coupled-cluster theory. J. Chem. Phys., 128:204105, 2008.

[51] S. Høst, B. Jansı́k, J. Olsen, P. Jørgensen, S. Reine, and T. Helgaker. A ground-state-

directed optimization scheme for the Kohn—Sham energy. Phys. Chem. Chem. Phys.,
10:5344, 2008.

[52] S. Høst, J. Olsen, B. Jansı́k, L. Thøgersen, Poul Jørgensen, and T. Helgaker. The
augmented Roothaan—Hall method for optimizing Hartree—Fock and Kohn—Sham
density matrices. J. Chem. Phys., 129:124106, 2008.

[53] Ernest R. Davidson. J. Comput. Phys., 17:87, 1975.

[54] P. Pulay. Chem. Phys. Lett., 73:393, 1980.

[55] P. Pulay. J. Comp. Chem., 3:556, 1982.

[56] J. H. Van Lenthe, R. Zwaans, H. J. J. Van Dam, and M. F. Guest. Starting SCF
calculations by superposition of atomic densities. J. Comp. Chem., 27:926, 2006.

[57] S. Reine, A. Krapp, M. F. Iozzi, V. Bakken, T. Helgaker, F. Pawlowski, and P. Salek.

An efficient density-functional-theory force evaluation for large molecular systems. J.
Chem. Phys., 133:044102, 2010.

[58] V. Bakken and T. Helgaker. The efficient optimization of molecular geometries using
redundant internal coordinates. J. Chem. Phys., 117:9160, 2002.
BIBLIOGRAPHY 82

[59] R. Lindh, A. Bernhardsson, G. Karlström, and P.-Å. Malmqvist. On the use of a hessian
model function in molecular geometry optimizations. Chem. Phys. Lett., 241:423, 1995.

[60] J. Baker. Techniques for geometry optimization: A comparison of cartesian and natural
internal coordinates. J. Comp. Chem., 14:1085, 1993.

[61] L. Verlet. Computer ”experiments” on classical fluids. i. thermodynamics properties

of lennard-jones molecules. Phys. Rev., 159:98, 1967.

[62] P. Pulay and G. Fogarasi. Fock matrix dynamics. Chem. Phys. Lett., 386:272, 2004.

[63] A. M. N. Niklasson, C. J. Tymczak, and M. Challacombe. Time-reversible born-

oppenheimer molecular dynamics. Phys. Rev. Lett., 97:123001, 2006.

[64] I.-M. Høyvik, B. Jansı́k, and P.Jørgensen. Trust region minimization of orbital local-
ization functions. J. Chem. Theory Comput., 8:3137, 2012.

[65] I.-M. Høyvik, B. Jansı́k, K. Kristensen, and P.Jørgensen. Local hartree-fock orbitals
using a three-level optimization strategy for the energy. J. Comp. Chem., 34:1311,
2013.

[66] B. Jansı́k, S. Høst, K.Kristensen, and P.Jørgensen. Local orbitals by minimizing powers
of the orbital variance. J. Chem. Phys., 134:194104, 2011.

[67] I.-M. Høyvik, B. Jansı́k, and P.Jørgensen. Orbital localization using fourth central
moment minimization. J. Chem. Phys., 137:224114, 2012.

[68] I.-M. Høyvik, B. Jansı́k, and P.Jørgensen. Pipek-mezey localization of occupied and
virtual orbitals. J. Comp. Chem., page (DOI: 10.1002/jcc.23281), 2013.

[69] A. J. Thorvaldsen, K. Ruud, K. Kristensen, P. Jørgensen, and S. Coriani. A density

matrix-based quasienergy formulation of the kohn–sham density functional response
theory using perturbation- and time-dependent basis sets. J. Chem. Phys., 129:214108,
2008.

[70] S.Coriani, S. Høst, B. Jansı́k, L. Thøgersen, J. Olsen, P. Jørgensen, S. Reine,

F. Pawlowski, T. Helgaker, and P. Salek. Linear-scaling implementation of molecu-
lar response theory in self-consistent field electronic-structure theory. J. Chem. Phys.,
126:154108, 2007.

[71] J. Kauczor, P. Jørgensen, and P. Norman. On the efficiency of algorithms for solving
hartreefock and kohnsham response equations. J. Chem. Theory Comput., 7:1610,
2011.
BIBLIOGRAPHY 83

[72] Kasper Kristensen, Joanna Kauczor, Andreas J. Thorvaldsen, Poul Jørgensen, Thomas
Kjærgaard, and Antonio Rizzo. J. Chem. Phys., 134:214104, 2011.

[73] T. Kjærgaard, P. Jørgensen, P. Salek A. J. Thorvaldsen, and S. Coriani. A gauge-

origin independent formulation and implementation of magneto-optical activity within
atomic-orbital-density based hartree-fock and kohn-sham response theories. J. Chem.
Theory Comput., 5:1997, 2009.

[74] T. Kjærgaard, K. Kristensen, J. Kauczor, P. Jørgensen, S. Coriani, and A. J. Thorvald-

sen. Comparison of standard and damped response formulations of magnetic circular
dichroism. J. Chem. Phys., 135:024112, 2011.

[75] Marcin Ziólkowski, Branislav Jansı́k, Thomas Kjærgaard, and Poul Jørgensen. Linear
scaling coupled cluster method with correlation energy based error control. J. Chem.
Phys., 133:014107, 2010.

[76] Kasper Kristensen, Marcin Ziólkowski, Branislav Jansı́k, Thomas Kjærgaard, and Poul
Jørgensen. J. Chem. Theory Comput., 7:1677, 2011.

[77] Ida-Marie Høyvik, Kasper Kristensen, Branislav Jansı́k, and Poul Jørgensen. J. Chem.
Phys., 136:014105, 2012.

[78] Kasper Kristensen, Ida-Marie Høyvik, Branislav Jansı́k, Poul Jørgensen, Thomas Kjær-
gaard, Simen Reine, and Jacek Jakowski. Mp2 energy and density for large molecular
systems with internal error control using the divide-expand-consolidate scheme. Phys.
Chem. Chem. Phys., 14:15706, 2012.

[79] Kasper Kristensen, Poul Jørgensen, Branislav Jansı́k, Thomas Kjærgaard, and Simen
Reine. Molecular gradient for second-order møller-plesset perturbation theory using
the divide-expand-consolidate (dec) scheme. J. Chem. Phys., 137:114102, 2012.

[80] Kasper Kristensen, Thomas Kjærgaard, Ida-Marie Høyvik, Patrick Ettenhuber, Poul
Jørgensen, Branislav Jansik, Simen Reine, and Jacek Jakowski. The divide-expand-
consolidate mp2 scheme goes massively parallel. Mol. Phys., 111:1196, 2013.

[81] Janus J. Eriksen, Pablo Baudin, Patrick Ettenhuber, Kasper Kristensen, Thomas Kjær-
gaard, and Poul Jørgensen. Linear-Scaling Coupled Cluster with Perturbative Triple
Excitations: The Divide-Expand-Consolidate CCSD(T) Model. J. Chem. Theory Com-
put., 11:2984, 2015.

[82] Kasper Kristensen, Patrick Ettenhuber, Janus Juul Eriksen, Frank Jensen, and Poul
Jørgensen. The same number of optimized parameters scheme for determining inter-
molecular interaction energies. J. Chem. Phys., 142(11):114116, 2015.
Part IV

Index

84