CRYSTAL09

2.0.1
User’s Manual
February 6, 2013

R. Dovesi1 , V.R. Saunders1 , C. Roetti1 , R. Orlando1,3 ,

C. M. Zicovich-Wilson 1,4 , F. Pascale5 , B. Civalleri1 , K. Doll6 ,
N.M. Harrison2,7 , I.J. Bush8 , Ph. D’Arco9 , M. Llunell10

1
Theoretical Chemistry Group - University of Turin
Dipartimento di Chimica IFM
Via Giuria 5 - I 10125 Torino - Italy
2
Computational Science & Engineering Department - CCLRC Daresbury
Daresbury, Warrington, Cheshire, UK WA4 4AD
3
Università del Piemonte Orientale
Department of Science and Advanced Technologies
Via Bellini 25/6 - I 15100 Alessandria - Italy
4
Departamento de Fı́sica, Universidad Autónoma del Estado de Morelos,
Av. Universidad 1001, Col. Chamilpa, 62210 Cuernavaca (Morelos) Mexico
5
LCM3B - UMR 7036-CNRS
Université Henri Poincaré
Nancy BP 239, F54506 Vandœuvre-les-Nancy, Cedex, France
6
Max-Planck-Institut für Festkörperforschung
Heisenbergstrasse 1
D-70569 Stuttgart, Germany
7
Department of Chemistry, Imperial College
South Kensington Campus
London, U.K.
8
The Numerical Algorithms Group (NAG)
Wilkinson House - Jordan Hill Road
Oxford OX2 8DR - U.K.
9
Laboratoire de Pétrologie, Modélisation des Matériaux et Processus
Université Pierre et Marie Curie,
4 Place Jussieu, 75232 Paris CEDEX 05, France
10
Departament de Quı́mica Fı́sica, Universitat de Barcelona
Diagonal 647, Barcelona, Spain

1
2
Contents

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Typographical Conventions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Getting Started - Installation and testing . . . . . . . . . . . . . . . . . . . . . 12

1 Wave function calculation:

Basic input route 13
1.1 Geometry and symmetry information . . . . . . . . . . . . . . . . . . . . . . . . 13
Geometry input for crystalline compounds . . . . . . . . . . . . . . . . . . . . . 14
Geometry input for molecules, polymers and slabs . . . . . . . . . . . . . . . . 14
Geometry input for polymers with roto translational symmetry . . . . . . . . . 15
Geometry input from external geometry editor . . . . . . . . . . . . . . . . . . 15
Comments on geometry input . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.2 Basis set . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.3 Computational parameters, hamiltonian,
SCF control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

2 Wave function calculation - Advanced input route 26

2.1 Geometry editing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
equation of state . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.2 Basis set input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
Effective core pseudo-potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
Pseudopotential libraries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
2.3 Computational parameters, hamiltonian,
SCF control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
DFT Hamiltonian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

3 Geometry optimization 105

Searching a transition state . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123

4 Vibration calculations at Γ point 127

Harmonic frequency calculation at Γ . . . . . . . . . . . . . . . . . . . . . . . . 127
IR intensities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
Scanning of geometry along selected normal modes . . . . . . . . . . . . . . . . 136
IR reflectance spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
Anharmonic calculation for X-H stretching . . . . . . . . . . . . . . . . . . . . 142
Harmonic calculation of phonon dispersion . . . . . . . . . . . . . . . . . . . . . 144

5 Coupled Perturbed HF/KS calculation 145

6 Mapping of CRYSTAL calculations to model Hamiltonians 147

7 Calculation of elastic constants 149

3
8 Properties 152
8.1 Preliminary calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
8.2 Properties keywords . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
8.3 Spontaneous polarization and piezoelectricity . . . . . . . . . . . . . . . . . . . 189

9 Input examples 194

9.1 Standard geometry input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
CRYSTAL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
SLAB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
POLYMER . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
MOLECULE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
9.2 Basis set input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
ECP - Valence only basis set input . . . . . . . . . . . . . . . . . . . . . . . . . 202
9.3 SCF options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
9.4 Geometry optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206

10 Basis set 215

10.1 Molecular BSs performance in periodic systems . . . . . . . . . . . . . . . . . . 215
10.2 Core functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
10.3 Valence functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
Molecular crystals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
Covalent crystals. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
Ionic crystals. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
From covalent to ionics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
Metals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
10.4 Hints on crystalline basis set optimization . . . . . . . . . . . . . . . . . . . . . 218
10.5 Check on basis-set quasi-linear-dependence . . . . . . . . . . . . . . . . . . . . 219

11 Theoretical framework 221

11.1 Basic equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
11.2 Remarks on the evaluation of the integrals . . . . . . . . . . . . . . . . . . . . . 222
11.3 Treatment of the Coulomb series . . . . . . . . . . . . . . . . . . . . . . . . . . 223
11.4 The exchange series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
11.5 Bipolar expansion approximation of Coulomb and exchange integrals . . . . . . 225
11.6 Exploitation of symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
Symmetry-adapted Crystalline Orbitals . . . . . . . . . . . . . . . . . . . . . . 226
11.7 Reciprocal space integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
11.8 Electron momentum density and related quantities . . . . . . . . . . . . . . . . 227
11.9 Elastic Moduli of Periodic Systems . . . . . . . . . . . . . . . . . . . . . . . . . 229
Examples of  matrices for cubic systems . . . . . . . . . . . . . . . . . . . . . . 230
Bulk modulus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
11.10Spontaneous polarization through the Berry phase approach . . . . . . . . . . . 233
Spontaneous polarization through the localized crystalline orbitals approach . . 233
11.11Piezoelectricity through the Berry phase approach . . . . . . . . . . . . . . . . 234
Piezoelectricity through the localized crystalline orbitals approach . . . . . . . 234

A Symmetry groups 236

A.1 Labels and symbols of the space groups . . . . . . . . . . . . . . . . . . . . . . 236
A.2 Labels of the layer groups (slabs) . . . . . . . . . . . . . . . . . . . . . . . . . . 239
A.3 Labels of the rod groups (polymers) . . . . . . . . . . . . . . . . . . . . . . . . 240
A.4 Labels of the point groups (molecules) . . . . . . . . . . . . . . . . . . . . . . . 243
A.5 From conventional to primitive cells: transforming matrices . . . . . . . . . . . 244

B Summary of input keywords 245

C Printing options 254

D External format 258

4
E Normalization coefficients 269

F CRYSTAL09 versus CRYSTAL06 278

G Acronyms 281

Bibliography 282

Subject index 292

5
Introduction
The CRYSTAL package performs ab initio calculations of the ground state energy, energy
gradient, electronic wave function and properties of periodic systems. Hartree-Fock or Kohn-
Sham Hamiltonians (that adopt an Exchange-Correlation potential following the postulates of
Density-Functional theory) can be used. Systems periodic in 0 (molecules, 0D), 1 (polymers,
1D), 2 (slabs, 2D), and 3 dimensions (crystals, 3D) are treated on an equal footing. In each
case the fundamental approximation made is the expansion of the single particle wave functions
(’Crystalline Orbital’, CO) as a linear combination of Bloch functions (BF) defined in terms
of local functions (hereafter indicated as ’Atomic Orbitals’, AOs). See Chapter 11.
The local functions are, in turn, linear combinations of Gaussian type functions (GTF) whose
exponents and coefficients are defined by input (section 1.2). Functions of symmetry s, p, d
and f can be used (see page 21). Also available are sp shells (s and p shells, sharing the same
set of exponents). The use of sp shells can give rise to considerable savings in CPU time.
The program can automatically handle space symmetry: 230 space groups, 80 layer groups, 99
rod groups, 45 point groups are available (Appendix A). In the case of polymers it can treat
helical structures (translation followed by a rotation around the periodic axis).
Point symmetries compatible with translation symmetry are provided for molecules.
Input tools allow the generation of slabs (2D system) or clusters (0D system) from a 3D crys-
talline structure, the elastic distortion of the lattice, the creation of a super-cell with a defect
and a large variety of structure editing. See Section 2.1
A special input option allows generation of 1D structures (nanotubes) from 2D ones.
Previous releases of the software in 1988 (CRYSTAL88, [39]), 1992 (CRYSTAL92, [42]), 1996
(CRYSTAL95, [43]), 1998 (CRYSTAL98, [112]), 2003 (CRYSTAL03, [113]) and 2006 (CRYS-
TAL06) [44] have been used in a wide variety of research with notable applications in studies
of stability of minerals, oxide surface chemistry, and defects in ionic materials. See “Applica-
tions” in

http://www.crystal.unito.it

The CRYSTAL package has been developed over a number of years. For basic theory and
algorithms see “Theory” in:

http://www.crystal.unito.it/theory.html

The required citation for this work is:

R. Dovesi, R. Orlando, B. Civalleri, C. Roetti, V.R. Saunders, C.M. Zicovich-Wilson
CRYSTAL: a computational tool for the ab initio study of the electronic properties of crystals
Z. Kristallogr.220, 571–573 (2005)

R. Dovesi, V.R. Saunders, C. Roetti, R. Orlando, C. M. Zicovich-Wilson, F. Pascale, B. Cival-

leri, K. Doll, N.M. Harrison, I.J. Bush, Ph. D’Arco, M. Llunell
CRYSTAL09 User’s Manual, University of Torino, Torino, 2009
CRYSTAL09 output will display the references relevant to the property computed, when cita-
tion is required.
Updated information on the CRYSTAL code as well as tutorials to learn basic and advanced
CRYSTAL usage are in:
http://www.crystal.unito.it/news.html

6
CRYSTAL09 Program Features
New features with respect to CRYSTAL06 are in italics.

Hamiltonian
• Hartree-Fock Theory

– Restricted
– Unrestricted

• Density Functional Theory

– Local functionals [L] and gradient-corrected functionals [G]

∗ Exchange functionals
· Slater (LDA) [L]
· von Barth-Hedin (VBH) [L]
· Becke ’88 (BECKE) [G]
· Perdew-Wang ’91 (PWGGA) [G]
· Perdew-Burke-Ernzerhof (PBE) [G]
· Revised PBE functional for solids (PBEsol) [G]
· Second-order expansion GGA for solids (SOGGA) [G]
· Wu-Cohen ’06 (WCGGA) [G]
∗ Correlation functionals
· VWN (#5 parameterization) (VWN) [L]
· Perdew-Wang ’91 (PWLSD) [L]
· Perdew-Zunger ’81 (PZ) [L]
· von Barth-Hedin (VBH) [L]
· Lee-Yang-Parr (LYP) [G]
· Perdew ’86 (P86) [G]
· Perdew-Wang ’91 (PWGGA) [G]
· Perdew-Burke-Ernzerhof (PBE) [G]
· Revised PBE functional for solids (PBEsol) [G]
· Wilson-Levy ’90 (WL) [G]
– Hybrid HF-DFT functionals
∗ B3PW, B3LYP (using the VWN5 functional), PBE0
∗ User-defined hybrid functionals
– Numerical-grid based numerical quadrature scheme
– London-type empirical correction for dispersion interactions (Grimme scheme)

Energy derivatives
• Analytical first derivatives with respect to the nuclear coordinates and cell
parameters

– Hartree-Fock and Density Functional methods

– All-electron and Effective Core Potentials

7
Type of calculation
• Single-point energy calculation

• Automated geometry optimization

– Uses a quasi-Newton algorithm

– Optimizes in symmetry-adapted cartesian coordinates
– Optimizes in redundant coordinates
– Full geometry optimization (cell parameters and atom coordinates)
– Freezes atoms during optimization
– Constant volume or pressure constrained geometry optimization (3D only¡/i¿)
– Transition state search

• Harmonic vibrational frequencies

– Harmonic vibrational frequencies at Gamma point¡/li¿

– Phonon dispersion using a direct approach (efficient supercell scheme)
– IR intensities through either localized Wannier functions or Berry phase
– Calculation of the reflectance spectrum
– Exploration of the energy and geometry along selected normal modes

• Anharmonic frequencies for X-H bonds

• Automated calculation of the elastic tensor of crystalline systems (3D only)

• Automated E vs V calculation for equation of state (3D only)

• Automatic treatment of solid solutions

Basis set
• Gaussian type functions basis sets

– s, p, d, and f GTFs
– Standard Pople Basis Sets
∗ STO-nG n=2-6 (H-Xe), 3-21G (H-Xe), 6-21G (H-Ar)
∗ polarization and diffuse function extensions
– User-specified basis sets supported

• Pseudopotential Basis Sets

– Available sets are:

∗ Hay-Wadt large core
∗ Hay-Wadt small core
– User-defined pseudopotential basis sets supported

8
Periodic systems
• Periodicity

– Consistent treatment of all periodic systems

– 3D - Crystalline solids (230 space groups)
– 2D - Films and surfaces (80 layer groups)
– 1D - Polymers
∗ space group derived symmetry (75 rod groups)
∗ helical symmetry (up to order 48)
– 0D - Molecules (32 point groups)

• Automated geometry editing

– 3D to 2D - slab parallel to a selected crystalline face (hkl)

– 3D to 0D - cluster from a perfect crystal (H saturated)
– 3D to 0D - extraction of molecules from a molecular crystal
– 3D to n3D - supercell creation
– 2D to 1D - building nanotubes from a single-layer slab model
– Several geometry manipulations (reduction of symmetry; insertion,
displacement, substitution, deletion of atoms)

Wave function analysis and properties

• Band structure

• Density of states

– Band projected DOSS

– AO projected DOSS

• All Electron Charge Density - Spin Density

– Density maps
– Mulliken population analysis
– Density analytical derivatives

• Atomic multipoles

• Electric field

• Electric field gradient

• Structure factors

• Electron Momentum Density and Compton profiles (Enhanced version)

• Electrostatic potential and its derivatives

– Quantum and classical electrostatic potential and its derivatives

– Electrostatic potential maps

• Fermi contact

• Localized Wannier Functions (Boys method)

9
 • Dielectric properties

– Spontaneous polarization
∗ Berry Phase
∗ Localized Wannier Functions
– Dielectric constant
∗ New Coupled Perturbed HF(KS) scheme
∗ Finite-field approximation

Software performance
• Memory management: dynamic allocation

• Full parallelization of the code

– parallel SCF and gradients for both HF and DFT methods

– Replicated data version (MPI)
– Massive parallel version (MPI) (distributed memory)

10
Conventions
In the description of the input data which follows, the following notation is adopted:

- • new record

- ∗ free format record

- An alphanumeric datum (first n characters meaningful)

- atom label sequence number of a given atom in the primitive cell, as

printed in the output file after reading of the geometry input

- symmops symmetry operators

- ,[] default values.

- italic optional input

- optional input records follow II

- additional input records follow II

Arrays are read in with a simplified implied DO loop instruction of Fortran 77:
(dslist, i=m1,m2)
where: dslist is an input list; i is the name of an integer variable, whose value ranges from m1
to m2.
Example (page 32): LB(L),L=1,NL
NL integer data are read in and stored in the first NL position of the array LB.

All the keywords are entered with an A format (case insensitive); the keywords must not end
with blanks.

conventional atomic number (usually called NAT) is used to associate a given basis set
with an atom. The real atomic number is the remainder of the division NAT/100. See page
20. The same conventional atomic number must be given in geometry input and in basis set
input.

11
Acknowledgments
Embodied in the present code are elements of programs distributed by other groups.
In particular: the atomic SCF package of Roos et al. [6], the GAUSS70 gaussian integral
package and STO-nG basis set due to Hehre et al. [70], the code of Burzlaff and Hountas for
space group analysis [21] and Saunders’ ATMOL gaussian integral package [84].
We take this opportunity to thank these authors. Our modifications of their programs have
sometimes been considerable. Responsibility for any erroneous use of these programs therefore
remains with the present authors.

We are in debt with Cesare Pisani, who first conceived the CRYSTAL project in 1976, for
his constant support of and interest in the development of the new version of the CRYSTAL
program.

It is our pleasure to thank Piero Ugliengo for continuous help, useful suggestions, rigorous
testing.

Michele Catti significantly contributed to the implementation of the geometry optimizer with
discussion, suggestions, contributions to the coding.

We thank Giuseppe Mallia for useful contribution to test parallel execution and to develop
automathic testing procedures.

We kindly acknowledge Jorge Garza-Olguin for his invaluable help in testing and documenting
the compilation of parallel executables from object files.

Contribution to validate the new features by applying them to research problem is recognized
to all reseachers working in the Theoretical Chemistry Group from 2006 to 2009: Marta Corno,
Raffaella Demichelis, Alessandro Erba, Matteo Ferrabone, Anna Ferrari, Migen Halo, Valentina
Lacivita, Lorenzo Maschio, Alessio Meyer, Alexander Terentyev, Javier F. Torres, Loredana
Valenzano.
Specific contribution to coding is indicated in the banner of the new options.

Financial support for this research has been provided by the Italian MURST (Ministero della
Università e della Ricerca Scientifica e Tecnologica), and the United Kingdom CCLRC (Council
for the Central Laboratories of the Research Council).

Getting Started
Instructions to download, install, and run the code are available in the web site:
http://www.crystal.unito.it → documentation

Program errors
A very large number of tests have been performed by researchers of a few laboratories, that
had access to a test copy of CRYSTAL09. We tried to check as many options as possible, but
not all the possible combinations of options have been checked. We have no doubts that errors
remain.
The authors would greatly appreciate comments, suggestions and criticisms by the users of
CRYSTAL; in case of errors the user is kindly requested to contact the authors, sending a
copy of both input and output by E-mail to the Torino group ([email protected]).

12
Chapter 1

Wave function calculation:

Basic input route

1.1 Geometry and symmetry information

The first record of the geometry definition must contain one of the keywords:

CRYSTAL 3D system page 14

SLAB 2D system page 14
POLYMER 1D system page 14
HELIX 1D system - roto translational symmetry page 15
MOLECULE 0D system page 14
EXTERNAL geometry from external file page 15
DLVINPUT geometry from DLV [119] Graphical User Interface. page 15

Four input schemes are used.

The first is for crystalline systems (3D), and is specified by the keyword CRYSTAL.
The second is for slabs (2D), polymers (1D) and molecules (0D) as specified by the keywords
SLAB, POLYMER or MOLECULE respectively.
The third scheme (keyword HELIX) defines a 1D system with roto-translational symmetry
(helix).
In the fourth scheme, with keyword EXTERNAL (page 15) or DLVINPUT, the unit cell,
atomic positions and symmetry operators may be provided directly from an external file (see
Appendix D, page 264). Such an input file can be prepared by the keyword EXTPRT (crystal
input block 1, page 41; properties).
Sample input decks for a number of structures are provided in section 9.1, page 194.

13
Geometry input for crystalline compounds. Keyword: CRYSTAL

rec variable value meaning

• ∗ IFLAG convention for space group identification (Appendix A.1, page 236):
0 space group sequential number(1-230)
1 Hermann-Mauguin alphanumeric code
IFHR type of cell: for rhombohedral groups, subset of trigonal ones, only
(meaningless for non-rhombohedral crystals):
0 hexagonal cell. Lattice parameters a,c
1 rhombohedral cell. Lattice parameters a, α
IFSO setting for the origin of the crystal reference frame:
0 origin derived from the symbol of the space group: where there
are two settings, the second setting of the International Tables is
chosen.
1 standard shift of the origin: when two settings are allowed, the first
setting is chosen
>1 non-standard shift of the origin given as input (see test22)
• ∗ space group identification code (following IFLAG value):
IGR space group sequence number (IFLAG=0)
or
A AGR space group alphanumeric symbol (IFLAG=1)
if IFSO > 1 insert II
• ∗ IX,IY,IZ non-standard shift of the origin coordinates (x,y,z) in fractions of
the crystallographic cell lattice vectors times 24 (to obtain integer
values).

• ∗ a,[b],[c], minimal set of crystallographic cell parameters:

[α],[β] translation vector[s] length [Ångstrom],
[γ] crystallographic angle[s] (degrees)
• ∗ NATR number of atoms in the asymmetric unit.
insert NATR records II
• ∗ NAT “conventional” atomic number. The conventional atomic number,
NAT, is used to associate a given basis set with an atom. The real
atomic number is the remainder of the division NAT100
X,Y,Z atom coordinates in fractional units of crystallographic lattice vec-
tors
optional keywords terminated by END/ENDGEOM or STOP II

Geometry input for molecules, polymers and slabs. Keywords:

SLAB, POLYMER, MOLECULE

When the geometrical structure of 2D, 1D and 0D systems has to be defined, attention should
be paid in the input of the atom coordinates, that are expressed in different units, fractional
(direction with translational symmetry) or Ångstrom (non periodic direction).

translational unit of measure of coordinates

symmetry X Y Z
3D fraction fraction fraction
2D fraction fraction Ångstrom
1D fraction Ångstrom Ångstrom
0D Ångstrom Ångstrom Ångstrom

14
 rec variable meaning

• ∗ IGR point, rod or layer group of the system:

0D - molecules (Appendix A.4, page 243)
1D - polymers (Appendix A.3, page 240)
2D - slabs (Appendix A.2, page 239)
if polymer or slab, insert II
• ∗ a,[b], minimal set of lattice vector(s)- length in Ångstrom
(b for rectangular lattices only)
[γ] AB
d angle (degrees) - triclinic lattices only

• ∗ NATR number of non-equivalent atoms in the asymmetric unit

insert NATR records II
• ∗ NAT conventional atomic number 3
X,Y,Z atoms coordinates. Unit of measure:
0D - molecules: x,y,z in Ångstrom
1D - polymers : y,z in Ångstrom, x in fractional units of crystallographic
cell translation vector
2D - slabs : z in Ångstrom, x, y in fractional units of crystallographic cell
translation vectors
optional keywords terminated by END or STOP II

Geometry input for polymers with roto translational symmetry.

Keyword: HELIX

rec variable meaning

• ∗ N1 order of rototranslational axis
∗ N2 to define the rototranslational vector
• ∗ a0 lattice parameter of 1D cell - length in Ångstrom
• ∗ NATR number of non-equivalent atoms in the asymmetric unit
insert NATR records II
• ∗ NAT conventional atomic number 3
X,Y,Z atoms coordinates. Unit of measure:
1D - polymers : y,z in Ångstrom, x in fractional units of crystallographic
cell translation vector
optional keywords terminated by END or STOP II

A helix structure is generated: each atom of the irreducible part is rotated by an angle β =
n · 360/N1 degrees and translated by a vector ~t = n · a0 N 2
N 1 with n = 1, ....(N 1 − 1).

As an example let’s consider the α-helix conformer of polyglycine whose structure is sketched
in Figure 1.1.

The helix structure is characterized by seven glycine residues per cell. The order of the roto-
translational axis is therefore seven, N 1 = 7. In order to establish the value of N 2, look for
instance at the atom labeled 7 in the Figure. The top view of the helix shows that upon rota-
tion by β = 360/7 degrees, atom 7 moves to atom 4; the side view clarifies that this movement
implies a translational vector ~t = a0 74 : therefore N 2 = 4.

Geometry input from external geometry editor. Keywords:

EXTERNAL, DLVINPUT
The fourth input scheme works for molecules, polymers, slabs and crystals. The complete
geometry input data are read from file fort.34. The unit cell, atomic positions and symme-
try operators are provided directly according to the format described in Appendix D, page

15
 Figure 1.1: Side view (left) and top view (right) of an α-helix conformer of polyglycine

264. Coordinates in Ångstrom. Such an input file is written when OPTGEOM route for
geometry optimization is chosen, and can be prepared by the keyword EXTPRT (program
crystal, input block 1, page 41; program properties), or by the the visualization software
DLV (http://www.cse.scitech.ac.uk/cmg/DLV/).
The geometry defined by EXTERNAL can be modified by inserting any geometry editing
keyword (page 26) into the input stream after EXTERNAL.

Comments on geometry input

1. All coordinates in Ångstrom. In geometry editing, after the basic geometry definition, the
unit of measure of coordinates may be modified by entering the keywords FRACTION
(page 44) or BOHR (page 34).

2. The geometry of a system is defined by the crystal structure ([55], Chapter 1 of ref. [100]).
Reference is made to the International Tables for Crystallography [64] for all definitions.
The crystal structure is determined by the space group, by the shape and size of the unit
cell and by the relative positions of the atoms in the asymmetric unit.

3. The lattice parameters represent the length of the edges of the cell (a,b,c) and the angles
between the edges (α = bcc; β = acc; γ = acb). They determine the cell volume and
shape.

4. Minimal set of lattice parameters to be defined in input:

cubic a
hexagonal a,c
trigonal hexagonal cell a,c
rhombohedral cell a, α
tetragonal a,c
orthorhombic a,b,c
monoclinic a,b,c, β (b unique)
a,b,c, γ (c unique)
a,b,c, α (a unique - non standard)
triclinic a,b,c, α, β, γ

5. The asymmetric unit is the largest subset of atoms contained in the unit-cell, where
no atom pair related by a symmetry operator can be found. Usually several equivalent
subsets of this kind may be chosen so that the asymmetric unit needs not be unique.
The asymmetric unit of a space group is a part of space from which, by application of
all symmetry operations of the space group, the whole of space is filled exactly.

16
 6. The crystallographic, or conventional cell, is used as the standard option in input. It
may be non-primitive, which means it may not coincide with the cell of minimum volume
(primitive cell), which contains just one lattice point. The matrices which transform the
conventional (as given in input) to the primitive cell (used by CRYSTAL) are given in
Appendix A.5, page 244, and are taken from Table 5.1 of the International Tables for
Crystallography [64].
Examples. A cell belonging to the face-centred cubic Bravais lattice has a volume four
times larger than that of the corresponding primitive cell, and contains four lattice points
(see page 57, keyword SUPERCEL). A unit cell belonging to the hexagonal Bravais
lattice has a volume three times larger than that of the rhombohedral primitive cell (R
Bravais lattice), and contains three lattice points.
7. The use of the International Tables to identify the symmetry groups requires some prac-
tice. The examples given below may serve as a guide. The printout of geometry informa-
tion (equivalent atoms, fractional and Cartesian atomic coordinates etc.) allows a check
on the correctness of the group selected. To obtain a complete neighborhood analysis
for all the non-equivalent atoms, a complete input deck must be read in (blocks 1-3),
and the keyword TESTPDIM inserted in block 3, to stop execution after the symmetry
analysis.
8. Different settings of the origin may correspond to a different number of symmetry oper-
ators with translational components.
Example: bulk silicon - Space group 227 - 1 irreducible atom per cell.

setting of the origin Si coordinates symmops with multiplicity

translational component
2nd (default) 1/8 1/8 1/8 36 2
1st 0. 0. 0. 24 2

NB With different settings, the same position can have different multiplicity. For instance,
in space group 227 (diamond, silicon) the position (0., 0., 0.) has multiplicity 2 in 1st
setting, and multiplicity 4 in 2nd setting.
Second setting is the default choice in CRYSTAL.
The choice is important when generating a supercell, as the first step is the removal of the
symmops with translational component. The keyword ORIGIN (input block 1, page
52) translates the origin in order to minimize the number of symmops with translational
component.
9. When coordinates are obtained from experimental data or from geometry optimization
with semi-classical methods, atoms in special positions, or related by symmetry are not
correctly identified, as the number of significative digits is lower that the one used by
the program crystal to recognize equivalence or special positions. In that case the
coordinates must be edited by hand (see FAQ at www.crystal.unito.it).
10. The symbol of the space group for crystals (IFLAG=1) is given precisely as it appears
in the International Tables, with the first letter in column one and a blank separating
operators referring to different symmetry directions. The symbols to be used for the
groups 221-230 correspond to the convention adopted in editions of the International
Tables prior to 1983: the 3 axis is used instead of 3. See Appendix A.1, page 236.
Examples:

Group number input symbol

137 (tetragonal) P 42/N M C
10 (monoclinic) P 1 2/M 1 (unique axis b, standard setting)
P 1 1 2/M (unique axis c)
P 2/M 1 1 (unique axis a)
25 (orthorhombic) P M M 2 (standard setting)

17
 P 2 M M
P M 2 M

11. In the monoclinic and orthorhombic cases, if the group is identified by its number (3-74),
the conventional setting for the unique axis is adopted. The explicit symbol must be
used in order to define an alternative setting.

12. For the centred lattices (F, I, C, A, B and R) the input cell parameters refer to the
centred conventional cell; the fractional coordinates of the input list of atoms are in a
vector basis relative to the centred conventional cell.

13. Rhombohedral space groups are a subset of trigonal ones. The Hermann-Mauguin symbol
must begin by ’R’. For instance, space groups 156-159 are trigonal, but not rhombohedral
(their Hermann-Mauguin symbols begin by ”P”). Rhombohedral space groups (146-148-
155-160-161-166-167) may have an hexagonal cell (a=b; c; α, β = 900 ; γ = 1200 : input
parameters a,c) or a rhombohedral cell (a=b=c; α = β = γ: input parameters = a, α).
See input datum IFHR.

14. It is sufficient to supply the coordinates of only one of a group of atoms equivalent under
centring translations (eg: for space group Fm3m only the parameters of the face-centred
cubic cell, and the coordinates of one of the four atoms at (0,0,0), (0, 12 , 21 ), ( 12 ,0, 12 ) and
( 21 , 12 ,0) are required).
The coordinates of only one atom among the set of atoms linked by centring translations
are printed. The vector basis is relative to the centred conventional cell. However when
Cartesian components of the direct lattice vectors are printed, they are those of the
primitive cell.

15. The conventional atomic number NAT is used to associate a given basis set with an
atom (see Basis Set input, Section 1.2, page 19). The real atomic number is given by the
remainder of the division of the conventional atomic number by 100 (Example: NAT=237,
Z=37; NAT=128, Z=28). Atoms with the same atomic number, but in non-equivalent
positions, can be associated with different basis sets, by using different conventional
atomic numbers: e.g. 6, 106, 1006 (all electron basis set for carbon atom); 206, 306 (core
pseudo-potential for carbon atom, Section 2.2, page 65).
If the remainder of the division is 0, a ”ghost” atom is identified, to which no nuclear
charge corresponds (it may have electronic charge). This option may be used for enriching
the basis set by adding bond basis function [5], or to allow build up of charge density on
a vacancy. A given atom may be transformed into a ghost after the basis set definition
(input block 2, keyword GHOSTS, page 64).

16. The keyword SLABCUT (Geometry editing input, page 55) allows the creation of a
slab (2D) of given thickness from the 3D perfect lattice. See for comparison test4-test24;
test5-test25; test6-test26; test7- test27.

17. For slabs (2D), when two settings of the origin are indicated in the International Tables
for Crystallography, setting number 2 is chosen. The setting can not be modified.

18. Conventional orientation of slabs and polymers: Polymers are oriented along the x axis.
Slabs are parallel to the xy plane.

19. The keywords MOLECULE (for molecular crystals only; page 46) and CLUSTER
(for any n-D structure; page 36) allow the creation of a non-periodic system (molecule(s)
or cluster) from a periodic one.

18
1.2 Basis set
rec variable value meaning
• ∗ NAT n conventional atomic 3 number
<200> 1000 all-electron basis set (Carbon, all electron BS: 6, 106, 1006)
>200 valence electron basis set (Carbon, ECP BS: 206, 306) . ECP
(Effective Core Pseudopotential) must be defined (page 65)
=99 end of basis set input section
NSHL n number of shells
0 end of basis set input (when NAT=99)
if NAT > 200 insert ECP input (page 65) II
NSHL sets of records - for each shell
• ∗ ITYB type of basis set to be used for the specified shell:
0 general BS, given as input
1 Pople standard STO-nG (Z=1-54)
2 Pople standard 3(6)-21G (Z=1-54(18)) Standard polarization
functions are included.
LAT shell type:
0 1 s AO (S shell)
1 1 s + 3 p AOs (SP shell)
2 3 p AOs (P shell)
3 5 d AOs (D shell)
4 7 f AOs (F shell)
NG Number of primitive Gaussian Type Functions (GTF) in the con-
traction for the basis functions (AO) in the shell
1≤NG≤10 for ITYB=0 and LAT ≤ 2
1≤NG≤6 for ITYB=0 and LAT = 3
2≤NG≤6 for ITYB=1
6 6-21G core shell
3 3-21G core shell
2 n-21G inner valence shell
1 n-21G outer valence shell
CHE formal electron charge attributed to the shell
SCAL scale factor (if ITYB=1 and SCAL=0., the standard Pople scale
factor is used for a STO-nG basis set.
if ITYB=0 (general basis set insert NG records II
• ∗ EXP exponent of the normalized primitive GTF
COE1 contraction coefficient of the normalized primitive GTF:
LAT=0,1 → s function coefficient
LAT=2 → p function coefficient
LAT=3 → d function coefficient
LAT=4 → f function coefficient
COE2 LAT=1 → p function coefficient
optional keywords terminated by END/ENDB or STOP II

The choice of basis set is the most critical step in performing ab initio calculations of periodic
systems, with Hartree-Fock or Kohn-Sham Hamiltonians. Optimization criteria are discussed in
Chapter 8.2. When an effective core pseudo-potential is used, the basis set must be optimized
with reference to that potential (Section 2.2, page 65).

1. A basis set (BS) must be given for each atom with different conventional atomic number
defined in the crystal structure input. If atoms are removed (geometry input, keyword
ATOMREMO, page 32), the corresponding basis set input can remain in the input
stream. The keyword GHOSTS (page 64) removes the atom, leaving the associated
basis set.

2. The basis set for each atom has NSHL shells, whose constituent AO basis functions

19
 are built from a linear combination (’contraction’) of individually normalized primitive
Gaussian-type functions (GTF) (Chapter 11, page 221).

3. A conventional atomic number NAT links the basis set with the atoms defined in the
crystal structure. The atomic number Z is given by the remainder of the division of the
conventional atomic number by 100 (Example: NAT=108, Z=8, all electron; NAT=228,
Z=28, ECP). See point 5 below.

4. A conventional atomic number 0 defines ghost atoms, that is points in space with an
associated basis set, but lacking a nuclear charge (vacancy). See test 28.

5. Atoms with equal conventional atomic number are associated with the same basis set.

NAT < 200>1000: all electron basis set. A maximum of two different basis sets may be
given for the same chemical species in different positions: NAT=Z,
NAT=Z+100, NAT=Z+1000.
NAT > 200: valence electron basis set. A maximum of two different BS may be
given for the same chemical species in positions not symmetry-related:
NAT=Z+200, NAT=Z+300. A core pseudo-potential must be defined.
See Section 2.2, page 65, for information on core pseudo-potentials.
Suppose we have four non-equivalent carbon atoms in the unit cell. Conventional atomic
numbers 6 106 1006 206 306 mean that carbon atoms (real atomic number 6) unrelated
by symmetry are to be associated with different basis sets: the first tree (6, 106, 1006)
all-electron, the second two (206, 306) valence only, with pseudo-potential.

6. The basis set input ends with the card:

99 0 conventional atomic number 99, 0 shell.
Optional keywords may follow.

In summary:

1. CRYSTAL can use the following all electrons basis sets:

a) general basis sets, including s, p, d, f functions (given in input);

b) standard Pople basis sets [71] (internally stored as in Gaussian 94 [53]).
STOnG, Z=1 to 54
6-21G, Z=1 to 18
3-21G, Z=1 to 54

The standard basis sets b) are stored as internal data in the CRYSTAL code. They are
all electron basis sets, and can not be combined with ECP.

2. Warning The standard scale factor is used for STO-nG basis set when the input datum
SCAL is 0.0 in basis set input. All the atoms of the same row are attributed the same
Pople STO-nG basis set when the input scale factor SCAL is 1.

3. Standard polarization functions can be added to 6(3)-21G basis sets of atoms up to Z=18,
by inserting a record describing the polarization shell (ITYB=2, LAT=2, p functions on
hydrogen, or LAT=3, d functions on 2-nd row atoms; see test 12).

H Polarization functions exponents He

1.1 1.1
__________ ______________________________
Li Be B C N O F Ne
0.8 0.8 0.8 0.8 0.8 0.8 0.8 --
___________ ______________________________
Na Mg Al Si P S Cl Ar
0.175 0.175 0.325 0.45 0.55 0.65 0.75 0.85
_____________________________________________________________________

The formal electron charge attributed to a polarization function must be zero.

20
 4. The shell types available are :

shell shell n. order of internal storage

code type AO
0 S 1 s
1 SP 4 s, x, y, z
2 P 3 x, y, z
3 D 5 2z 2 − x2 − y 2 , xz, yz, x2 − y 2 , xy
4 F 7 (2z 2 − 3x2 − 3y 2 )z, (4z 2 − x2 − y 2 )x, (4z 2 − x2 − y 2 )y,
(x2 − y 2 )z, xyz, (x2 − 3y 2 )x, (3x2 − y 2 )y

When symmetry adaptation of Bloch functions is active (default; NOSYMADA in block3

to remove it), if F functions are used, all lower order functions must be present (D, P ,
S).
The order of internal storage of the AO basis functions is an information necessary to
read certain quantities calculated by the program properties. See Chapter 8: Mul-
liken population analysis (PPAN, page 96), electrostatic multipoles (POLI, page 182),
projected density of states (DOSS,page 162) and to provide an input for some options
(EIGSHIFT, input block 3, page 82).
5. Spherical harmonics d-shells consisting of 5 AOs are used.
6. Spherical harmonics f-shells consisting of 7 AOs are used.
7. The formal shell charges CHE, the number of electrons attributed to each shell, are
assigned to the AO following the rules:
shell shell max rule to assign the shell charges
code type CHE
0 S 2. CHE for s functions
1 SP 8. if CHE>2, 2 for s and (CHE−2) for p functions,
if CHE≤2, CHE for s function
2 P 6. CHE for p functions
3 D 10. CHE for d functions
4 F 14. CHE for f functions - it may be 6= 0 in CRYSTAL09.
8. A maximum of one open shell for each of the s, p and or d atomic symmetries is allowed
in the electronic configuration defined in the input. The atomic energy expression is not
correct for all possible double open shell couplings of the form pm dn . Either m must
equal 3 or n must equal 5 for a correct energy expression in such cases. A warning
will be printed if this is the case. However, the resultant wave function (which is a
superposition of atomic densities) will usually provide a reasonable starting point for the
periodic density matrix.
9. When extended basis sets are used, all the functions corresponding to symmetries (an-
gular quantum numbers) occupied in the isolated atom are added to the atomic basis
set for atomic wave function calculations, even if the formal charge attributed to that
shell is zero. Polarization functions are not included in the atomic basis set; their input
occupation number should be zero.
10. The formal shell charges are used only to define the electronic configuration of the atoms
to compute the atomic wave function. The initial density matrix in the SCF step may
be a superposition of atomic (or ionic) density matrices (default option, GUESSPAT,
page 91). When a different guess is required ( GUESSP), the shell charges are not used,
but checked for electron neutrality when the basis set is entered.
11. F shells functions are not used to compute the “atomic” wave function, to build an atomic
density matrix SCF guess. If F shells are occupied by nf electrons, the “atomic” wave
function is computed for an ion (F electrons are removed), and the diagonal elements of
the atomic density matrix are then set to nf /7. The keyword FDOCCUP (input block
3, page 84 allows modification of f orbitals occupation.

21
 12. Each atom in the cell may have an ionic configuration, when the sum of formal shell
charges (CHE) is different from the nuclear charge. When the number of electrons in
the cell, that is the sum of the shell charges CHE of all the atoms, is different from the
sum of nuclear charges, the reference cell is non-neutral. This is not allowed for periodic
systems, and in that case the program stops. In order to remove this constraint, it is
necessary to introduce a uniform charged background of opposite sign to neutralize the
system [38]. This is obtained by entering the keyword CHARGED (page 62) after the
standard basis set input. The value of total energy must be carefully checked.

13. It may be useful to allow atoms with the same basis set to have different electronic
configurations (e.g, for an oxygen vacancy in MgO one could use the same basis set for
all the oxygens, but begin with different electronic configuration for those around the
vacancy). The formal shell charges attributed in the basis set input may be modified for
selected atoms by inserting the keyword CHEMOD (input block 2, page 62).

14. The energies given by an atomic wave function calculation with a crystalline basis set
should not be used as a reference to calculate the formation energies of crystals. The
external shells should first be re-optimized in the isolated atom by adding a low-exponent
Gaussian function, in order to provide and adequate description of the tails of the isolated
atom charge density [25] (keyword ATOMHF, input block 3, page 72).

Optimized basis sets for periodic systems used in published papers are available in:

http://www.crystal.unito.it

22
1.3 Computational parameters, hamiltonian,
SCF control
Default values are set for all computational parameters. Default choices may be modified
through keywords. Default choices:
default keyword to modify page

hamiltonian: RHF UHF (SPIN) 102

tolerances for coulomb and exchange sums : 6 6 6 6 12 TOLINTEG 102
Pole order for multipolar expansion: 4 POLEORDR 96
Max number of SCF cycles: 50 MAXCYCLE 93
Convergence on total energy: 10−6 TOLDEE 102

For periodic systems, 1D, 2D, 3D, the only mandatory input information is the shrinking
factor, IS, to generate a commensurate grid of k points in reciprocal space, according to Pack-
Monkhorst method [85]. The Hamiltonian matrix computed in direct space, Hg , is Fourier
transformed for each k value, and diagonalized, to obtain eigenvectors and eigenvalues:
X
Hk = Hg eigk
g

Hk Ak = SkAk Ek
A second shrinking factor, ISP, defines the sampling of k points, ”Gilat net” [57, 56], used
for the calculation of the density matrix and the determination of Fermi energy in the case of
conductors (bands not fully occupied).
The two shrinking factors are entered after the keyword SHRINK (page 97).
In 3D crystals, the sampling points belong to a lattice (called the Pack-Monkhorst net), with
basis vectors:
b1/is1, b2/is2, b3/is3 is1=is2=is3=IS, unless otherwise stated
where b1, b2, b3 are the reciprocal lattice vectors, and is1, is2, is3 are integers ”shrinking
factors”.
In 2D crystals, IS3 is set equal to 1; in 1D crystals both IS2 and IS3 are set equal to 1.
Only points ki of the Pack-Monkhorst net belonging to the irreducible part of the Brillouin
Zone (IBZ) are considered, with associated a geometrical weight, wi . The choice of the recip-
rocal space integration parameters to compute the Fermi energy is a delicate step for metals.
See Section 11.7, page 227.
Two parameters control the accuracy of reciprocal space integration for Fermi energy calcula-
tion and density matrix reconstruction:

IS shrinking factor of reciprocal lattice vectors. The value of IS determines the number of
k points at which the Fock/KS matrix is diagonalized.
In high symmetry systems, it is convenient to assign IS magic values such that all low
multiplicity (high symmetry) points belong to the Monkhorst lattice. Although this
choice does not correspond to maximum efficiency, it gives a safer estimate of the integral.
The k-points net is automatically made anisotropic for 1D and 2D systems.

23
 The figure presents the reciprocal lattice cell of 2D graphite (rhombus), the first
Brillouin zone (hexagon), the irreducible part of Brillouin zone (in grey), and the
coordinates of the ki points according to a Pack-Monkhorst sampling, with shrinking
factor 3 and 6.

ISP shrinking factor of reciprocal lattice vectors in the Gilat net (see [103], Chapter II.6).
ISP is used in the calculation of the Fermi energy and density matrix. Its value can be
equal to IS for insulating systems and equal to 2*IS for conducting systems.
The value assigned to ISP is irrelevant for non-conductors. However, a non-conductor
may give rise to a conducting structure at the initial stages of the SCF cycle (very often
with DFT hamiltonians), owing, for instance, to a very unbalanced initial guess of the
density matrix. The ISP parameter must therefore be defined in all cases.
Note. The value used in the calculation is ISP=IS*NINT(MAX(ISP,IS)/IS)
In the following table the number of sampling points in the IBZ and in BZ is given for a
fcc lattice (space group 225, 48 symmetry operators) and hcp lattice (space group 194, 24
symmetry operators). The CRYSTAL code allows 413 k points in the Pack-Monkhorst net,
and 2920 in the Gilat net.

IS points in IBZ points in IBZ points BZ

fcc hcp
6 16 28 112
8 29 50 260
12 72 133 868
16 145 270 2052
18 195 370 2920
24 413 793 6916
32 897 1734 16388
36 1240 2413 23332
48 2769 5425 55300

1. When an anisotropic net is user defined (IS=0), the ISP input value is taken as ISP1
(shrinking factor of Gilat net along first reciprocal lattice) and ISP2 and ISP3 are set to:
ISP2=(ISP*IS2)/IS1,
ISP3=(ISP*IS3)/IS1.
2. User defined anisotropic net is not compatible with SABF (Symmetry Adapted Bloch
Functions). See NOSYMADA, page 95.

Some tools for accelerating convergence are given through the keywords LEVSHIFT (page
92 and tests 29, 30, 31, 32, 38), FMIXING (page 87), SMEAR (page 99), BROYDEN

24
(page 73) and ANDERSON (page 71).
At each SCF cycle the total atomic charges, following a Mulliken population analysis scheme,
and the total energy are printed.
The default value of the parameters to control the exit from the SCF cycle (∆E < 10−6
hartree, maximum number of SCF cycles: 50) may be modified entering the keywords:
TOLDEE (tolerance
; on change in total energy) page 102
TOLDEP (tolerance
; on SQM in density matrix elements) page 102
MAXCYCLE (maximum
. number of cycles) page 93

Spin-polarized system
By default the orbital occupancies are controlled according to the ’Aufbau’ principle.
To obtain a spin polarized solution an open shell Hamiltonian must be defined (block3, UHF
or DFT/SPIN). A spin-polarized solution may then be computed after definition of the (α-
β) electron occupancy. This can be performed by the keywords SPINLOCK (page 101) and
BETALOCK (page 73).

25
Chapter 2

Wave function calculation -

Advanced input route

2.1 Geometry editing

The following keywords allow editing of the crystal structure, printing of extended informa-
tion, generation of input data for visualization programs. Processing of the input block 1 only
(geometry input) is allowed by the keyword TEST[GEOM].
Each keyword operates on the geometry active when the keyword is entered. For instance, when
a 2D structure is generated from a 3D one through the keyword SLABCUT, all subsequent
geometry editing operates on the 2D structure. When a dimer is extracted from a molecular
crystal through the keyword MOLECULE, all subsequent editing refers to a system without
translational symmetry.
The keywords can be entered in any order: particular attention should be paid to the action of
the keywords KEEPSYMM 2.1 and BREAKSYM 2.1, that allow maintaining or breaking
the symmetry while editing the structure.
These keywords behave as a switch, and require no further data. Under control of the
BREAKSYM keyword (the default), subsequent modifications of the geometry are allowed
to alter (reduce: the number of symmetry operators cannot be increased) the point-group sym-
metry. The new group is a subgroup of the original group and is automatically obtained by
CRYSTAL. However if a KEEPSYMM keyword is presented, the program will endeavor
to maintain the number of symmetry operators, by requiring that atoms which are symmetry
related remain so after a geometry editing (keywords: ATOMSUBS, ATOMINSE, ATOM-
DISP, ATOMREMO).
The space group of the system may be modified after editing. For 3D systems, the file FIND-
SYM.DAT may be written (keyword FINDSYM). This file is input to the program findsym
(http://physics.byu.edu/ stokesh/isotropy.html), that finds the space-group symmetry of a
crystal, given the coordinates of the atoms.

Geometry keywords
Symmetry information

ATOMSYMM printing of point symmetry at the atomic positions 34 –

MAKESAED printing of symmetry allowed elastic distortions (SAED) 45 –
PRSYMDIR printing of displacement directions allowed by symmetry. 53 –
SYMMDIR printing of symmetry allowed geom opt directions 60 –
SYMMOPS printing of point symmetry operators 60 –
TENSOR print tensor of physical properties up to order 4 60 I

26
Symmetry information and control

BREAKELASsymmetry breaking according to a general distortion 35 I

BREAKSYM allow symmetry reduction following geometry modifications 34 –
KEEPSYMM maintain symmetry following geometry modifications 44 –
MODISYMM removal of selected symmetry operators 45 I
PURIFY cleans atomic positions so that they are fully consistent with the 53 –
group
SYMMREMO removal of all symmetry operators 60 –
TRASREMO removal of symmetry operators with translational components 60 –

Modifications without reduction of symmetry

ATOMORDE reordering of atoms in molecular crystals 32 –

NOSHIFT no shift of the origin to minimize the number of symmops with 52 –
translational components before generating supercell
ORIGIN shift of the origin to minimize the number of symmetry operators 52 –
with translational components
PRIMITIV crystallographic cell forced to be the primitive cell 52 –
ROTCRY rotation of the crystal with respect to the reference system cell 54 I

Atoms and cell manipulation (possible symmetry reduction (BREAKSYMM)

ATOMDISP displacement of atoms 32 I

ATOMINSE addition of atoms 32 I
ATOMREMO removal of atoms 32 I
ATOMROT rotation of groups of atoms 33 I
ATOMSUBS substitution of atoms 34 I
ELASTIC distortion of the lattice 38 I
POINTCHG point charges input 52 I
SCELPHONO generation of supercell for phonon dispersion 58 I
SUPERCEL generation of supercell - input refers to primitive cell 56 I
SUPERCON generation of supercell - input refers to conventional cell 56 I
USESAED given symmetry allowed elastic distortions, reads δ 61 I

From crystals to slabs (3D→2D)

SLABINFO definition of a new cell, with xy k to a given plane 56 I

SLABCUT generation of a slab parallel to a given plane (3D→2D) 55 I

From slabs to nanotubes (2D→1D)

NANOTUBE building a nanotube from a slab 47 I

SWCNT building a nanotube from an hexagonal slab 50 I

From periodic structure to clusters

CLUSTER cutting of a cluster from a periodic structure (3D→0D) 36 I

HYDROSUB border atoms substituted with hydrogens (0D→0D) 44 I

Molecular crystals

MOLECULE extraction of a set of molecules from a molecular crystal 46 I

(3D→0D)
MOLEXP variation of lattice parameters at constant symmetry and molec- 46 I
ular geometry (3D→3D)
MOLSPLIT periodic structure of non interacting molecules (3D→3D) 47 –
RAYCOV modification of atomic covalent radii 53 I

BSSE correction

27
MOLEBSSE counterpoise method for molecules (molecular crystals only) 45 I
(3D→0D)
ATOMBSSE counterpoise method for atoms (3D→0D) 31 I

Auxiliary and control keywords

ANGSTROM sets input units to Ångstrom 31 –

BOHR sets input units to bohr 34 –
BOHRANGS input bohr to Å conversion factor (0.5291772083 default value) 34 I
BOHRCR98 bohr to Å conversion factor is set to 0.529177 (CRY98 value) –
END/ENDG terminate processing of geometry input –
FRACTION sets input units to fractional 44 –
NEIGHBOR number of neighbours in geometry analysis 51 I
PARAMPRT printing of parameters (dimensions of static allocation arrays) 52 –
PRINTCHG printing of point charges coordinates in geometry output 52
PRINTOUT setting of printing options by keywords 53 –
SETINF setting of inf array options 54 I
SETPRINT setting of printing options 54 I
STOP execution stops immediately 56 –
TESTGEOM stop after checking the geometry input 60 –

Output of data on external units

COORPRT coordinates of all the atoms in the cell 37 –

EXTPRT write file in CRYSTAL geometry input format 41 –
FINDSYM write file in FINDSYM input format 44 –
STRUCPRT cell parameters and coordinates of all the atoms in the cell 56 –

External electric field - modified Hamiltonian

FIELD electric field applied along a periodic direction 41 I
FIELDCON electric field applied along a non periodic direction 43 I

Geometry optimization - see index for keywords full list

28
OPTGEOM Geometry optimization input block - closed by END 105 I
Type of optimization (default: atom coordinates)
FULLOPTG full geometry optimization –
CELLONLY cell parameters optimization –
INTREDUN optimization in redundant internal coordinates 115 –
ITATOCEL iterative optimization (atom/cell) –
CVOLOPT full geometry optimization at constant volume 117 –
Initial Hessian
HESSIDEN initial guess for the Hessian - identity matrix –
HESSMOD1 initial guess for the Hessian - model 1 (default) –
HESSMOD2 initial guess for the Hessian - model 2 –
HESSNUM initial guess for the Hessian - numerical estimate –
Convergence criteria modification
TOLDEG RMS of the gradient [0.0003] I
TOLDEX RMS of the displacement [0.0012] I
TOLDEE energy difference between two steps [10−7 ] I
MAXCYCLE max number of optimization steps I
Optimization control
FRAGMENT partial geometry optimization 119 I
RESTART data from previous run –
FINALRUN Wf single point with optimized geometry I
Gradient calculation control
NUMGRATO numerical atoms first derivatives 114 –
NUMGRCEL numerical cell first derivatives 114 –
NUMGRALL numerical atoms and cell first derivatives 114 –
External stress
EXTSTRESS apply external stress 121 I
EXTPRESS apply external hydrostatic pressure 121 I
Printing options
PRINTFORCES atomic gradients –
PRINTHESS Hessian –
PRINTOPT optimization procedure –
PRINT verbose printing –

Frequencies at Γ - Harmonic calculation

29
FREQCALC Frequencies at Γ (harmonic calculation) input block - closed by END 127 I
ANALYSIS –
[NOANALYSIS] –
DIELISO I
DIELTENS I
END[FREQ] –
FRAGMENT I
INTENS –
INTLOC - IR intensities through Wannier functions 134 –
[INTPOL] - IR intensities through Berry phase [default] 134 –
ISOTOPES 129 I
[MODES] –
[NOINTENS] –
NOMODES –
NORMBORN –
NOUSESYMM –
NUMDERIV I
PRESSURE I
PRINT –
RESTART –
SCANMODE I
STEPSIZE I
TEMPERAT I
TESTFREQ –
[USESYMM] –

ANHARM Frequency at Γ (anharmonic calculationinput block - closed by END142 I

END[ANHA] –
ISOTOPES modification of atomic mass 143 I
KEEPSYMM 44 –
NOGUESS 143 –
POINTS26 143 –
PRINT –
TEST[ANHA] –

CONFCNT - Configuration counting and cluster expansion 147

CPHF - Coupled Perturbed Hartree-Fock 145

ELASTCON - Second order elastic constants 149

EOS - Equation of state 145

Geometry input optional keywords

ANGLES
This option prints the angle the AXB,
\ where X is one of the irreducible (that is, non symmetry
equivalent) atoms of the unit cell, and A and B belong to its m-th and n-th stars of neighbors.

30
 rec variable meaning
• ∗ NATIR number of X atoms to be considered; they are the first NATIR in the list of
irreducible atoms (flag ”T” printed) generated by CRYSTAL
∗ NSHEL number of stars of neighbors of X to be considered; all the angles AXB,
\
where A and B belong to the first NSHEL neighbors of X, are printed out

Though the keyword ANGLES can be entered in geometry input, full input deck must be
supplied (block 1-2-3), in order to obtain information on bond angles, when neighbors analysis
is printed.
Example. Bulk Silicon. There is 1 irreducible atom, and the first star of neighbors contain 4
atoms: (from CRYSTAL output):
COORDINATES OF THE EQUIVALENT ATOMS (FRACTIONAL UNITS)

N ATOM ATOM Z X Y Z
IRR EQUIV
1 1 1 14 SI 1.250000E-01 1.250000E-01 1.250000E-01
2 1 2 14 SI -1.250000E-01 -1.250000E-01 - 1.250000E-01
- - - - - - - - - - - - - - - - - - - - - - -
N NUMBER OF NEIGHBORS AT DISTANCE R

STAR ATOM N R/ANG R/AU NEIGHBORS (ATOM LABELS AND CELL INDICES)
1 1 SI 4 2.3469 4.4351 2 SI 0 0 0 2 SI 1 0 0 2 SI 0 1 0
2 SI 0 0 1

The number of angles having the irreducible Silicon as vertex is: (4)*(4-1)/2 = 6
ANGLES (DEGREES) ARE INDICATED AS A-X-B(I),I=1,L

at A cell at X at B cell
angle at B cell angle at B cell angle
AXB AXB AXB
2 SI( 0 0 0) 1 SI 2 SI( 1 0 0) 109.47 2 SI( 0 1 0) 109.47 2 SI( 0 0 1) 109.47
2 SI( 1 0 0) 1 SI 2 SI( 0 1 0) 109.47 2 SI( 0 0 1) 109.47
2 SI( 0 1 0) 1 SI 2 SI( 0 0 1) 109.47

If it is required to consider 6 stars of neighbors to compute all the angles having the irreducible
Silicon as vertex, the number of angles computed will be:
(4+12+12+6+12+24)*(4+12+12+6+12+24-1)/2 = 2415

ANGSTROM - unit of measure

The unit of length in geometry editing is set to Ångstrom, (default value).

ANHARM - Anharmonic calculation of frequencies of X-H (X-D) bond

stretching
See Chapter 4, page 142.

ATOMBSSE - counterpoise for closed shell atoms and ions

rec variable meaning
• ∗ IAT label of the atom in the reference cell
NSTAR maximum number of stars of neighbors included in the calculation.
RMAX maximum distance explored searching the neighbors of the atom.

A cluster is defined including the selected atom and the basis functions belonging to the NSTAR
sets of neighbors, when their distance R from the central atom is smaller than RMAX. The
atomic wave function is not computed by the atomic package, but by the standard CRYSTAL
route for 0D, 1 atom system. UHF and SPINLOCK must be used to define a reasonable
orbital occupancy. It is suggested to compute the atomic wave function using a program
properly handling the electronic configuration of open shell atoms.

31
Warning. The system is 0D. No reciprocal lattice information is required in the scf input
(Section 1.3, page 23).

ATOMDISP
rec variable meaning
• ∗ NDISP number of atoms to be displaced
insert NDISP records II
• ∗ LB label of the atom to be moved
DX,DY,DZ increments of the coordinates in the primitive cell [Å].

Selected atoms are displaced in the primitive cell. The point symmetry of the system may be
altered (default value BREAKSYM, page 34). To displace all the atoms symmetry related,
KEEPSYMM must be inserted before ATOMDISP.
Increments are in Ångstrom, unless otherwise requested (keyword BOHR, FRACTION,
page 31). See tests 17, 20, 37.

ATOMINSE
rec variable meaning
• ∗ NINS number of atoms to be added
insert NINS records II
• ∗ NA conventional atomic number
X,Y,Z coordinates [Å] of the inserted atom. Coordinates refer to the primitive cell.

New atoms are added to the primitive cell. Coordinates are in Ångstrom, unless otherwise
requested (keyword BOHR, FRACTION, page 31). Remember that the original symmetry
of the system is maintained, applying the symmetry operators to the added atoms if the
keyword KEEPSYMM (page 34) was previously entered. The default is BREAKSYM
(page 34). Attention should be paid to the neutrality of the cell (see CHARGED, page 62).
See tests 16, 35, 36.

ATOMORDE
After processing the standard geometry input, the symmetry equivalent atoms in the reference
cell are grouped. They may be reordered, following a chemical bond criterion. This simplifies
the interpretation of the output when the results of bulk molecular crystals are compared with
those of the isolated molecule. See option MOLECULE (page 46) and MOLSPLIT (page
47). No input data are required.
For molecular crystals only.

ATOMREMO
rec variable meaning
• ∗ NL number of atoms to remove
• ∗ LB(L),L=1,NL label of the atoms to remove

Selected atoms, and related basis set, are removed from the primitive cell (see test 16). A
vacancy is created in the lattice. The symmetry can be maintained (KEEPSYMM), by
removing all the atoms symmetry-related to the selected one, or reduced (BREAKSYM,
default). Attention should be paid to the neutrality of the cell (see CHARGED, page 62).
NB. The keyword GHOSTS (basis set input, page 64) allows removal of selected atoms,
leaving the related basis set.

32
ATOMROT
rec variable value meaning
• ∗ NA 0 all the atoms of the cell are rotated and/or translated
>0 only NA selected atoms are rotated and/or translated.
<0 the atom with label |NA| belongs to the molecule to be rotated. The
program selects all the atoms of the molecule on the base of the sum of
their atomic radii (Table on page 53).
if NA > 0, insert NA data II
• ∗ LB(I),I=1,NA label of the atoms to be rotated and/or translated.
• ∗ ITR >0 translation performed. The selected NA atoms are translated by -r,
where r is the position of the ITR-th atom. ITR is at the origin after
the translation.
≤0 a general translation is performed. See below.
=999 no translation.
IRO >0 a rotation around a given axis is performed. See below.
<0 a general rotation is performed. See below.
=999 no rotation.
if ITR<0 insert II
• ∗ X,Y,Z Cartesian components of the translation vector [Å]
if ITR=0 insert II
• ∗ N1,N2 label of the atoms defining the axis.
DR translation along the axis defined by the atoms N1 and N2, in the di-
rection N1 → N2 [Å].
if IRO<0 insert II
• ∗ A,B,G Euler rotation angles (degree).
IPAR defines the origin of the Cartesian system for the rotation
0 the origin is the barycentre of the NAT atoms
>0 the origin is the atom of label IPAR
if IRO>0 insert II
• ∗ N1,N2 label of the atoms that define the axis for the rotation
ALPHA 6= 0. rotation angle around the N1–N2 axis (degrees)
0. the selected atoms are rotated anti-clockwise in order to orientate the
N1–N2 axis parallel to the z axis.

This option allows to rotate and/or translate the specified atoms. When the rotation of a
molecule is required (NA < 0), the value of the atomic radii must be checked, in order to
obtain a correct definition of the molecule. It is useful to study the conformation of a molecule
in a zeolite cavity, or the interaction of a molecule (methane) with a surface (MgO).
The translation of the selected group of atoms can be defined in three different ways:
1. Cartesian components of the translation vector (ITR < 0);
2. modulus of the translation vector along an axis defined by two atoms (ITR = 0);
3. sequence number of the atom to be translated to the origin. All the selected atoms are
subjected to the same translation (ITR > 0).
The rotation can be performed in three different ways:
1. by defining the Euler rotation angles α, β, γ and the origin of the rotating system (IRO
< 0). The axes of the rotating system are parallel to the axes of the Cartesian reference
system. (The rotation is given by: Rαz Rβx Rγz , where R are the rotation matrices).
2. by defining the rotation angle α around an axis defined by two atoms A and B. The
origin is at A, the positive direction A→B.
3. by defining a z’ axis (identified by two atoms A and B). The selected atoms are rotated,
in such a way that the A–B z’ axis becomes parallel to the z Cartesian axis. The origin
is at A and the positive rotation anti clockwise (IRO>0, α=0).

33
The selected atoms are rotated according to the defined rules, the cell orientation and the
cartesian reference frame are not modified. The symmetry of the system is checked after the
rotation, as the new geometry may have a different symmetry.
See tests 15, rotation of the N H3 molecule in a zeolite cavity, and 16, rotation of the H2 O
molecule in the zeolite cavity.

ATOMSUBS
rec variable meaning
• ∗ NSOST number of atoms to be substituted
insert NSOST records II
• ∗ LB label of the atom to substitute
NA(LB) conventional atomic number of the new atom

Selected atoms are substituted in the primitive cell (see test 17, 34, 37). The symmetry can be
maintained (KEEPSYMM), by substituting all the atoms symmetry-related to the selected
one, or reduced (BREAKSYM, default). Attention should be paid to the neutrality of the
cell: a non-neutral cell will cause an error message, unless allowed by entering the keyword
CHARGED, page 62.

ATOMSYMM
The point group associated with each atomic position and the set of symmetry related atoms
are printed. No input data are required. This option is useful to find the internal coordinates
to be relaxed when the unit cell is deformed (see ELASTIC, page 38).

BOHR
The keyword BOHR sets the unit of distance to bohr. When the unit of measure is modified,
the new convention is active for all subsequent geometry editing.
The conversion factor Ångstrom/bohr is 0.5291772083 (CODATA 1998). This value can be
modified by entering the keyword BOHRANGS and the desired value in the record following.
The keyword BOHRCR98 sets the conversion factor to 0.529177, as in the program CRYS-
TAL98.
CRYSTAL88 default value was 0.529167).

BOHRANGS
rec variable meaning
• ∗ BOHR conversion factor Ångstrom/bohr

The conversion factor Ångstrom/bohr can be user-defined.

In CRYSTAL88 the default value was 0.529167.
In CRYSTAL98 the default value was 0.529177.

BOHRCR98
The conversion factor Ångstrom/bohr is set to 0.529177, as in CRYSTAL98. No input data
required.

BREAKSYM
Under control of the BREAKSYM keyword (the default), subsequent modifications of the
geometry are allowed to alter (reduce: the number of symmetry operators cannot be increased)
the point-group symmetry. The new group is a subgroup of the original group and is automat-
ically obtained by CRYSTAL.

34
The symmetry may be broken by attributing different spin (ATOMSPIN, block34, page 72)
to atoms symmetry related by geometry.
Example: When a CO molecule is vertically adsorbed on a (001) 3-layer MgO slab, (D4h
symmetry), the symmetry is reduced to C4v , if the BREAKSYM keyword is active. The
symmetry operators related to the σh plane are removed. However, if KEEPSYMM is
active, then additional atoms will be added to the underside of the slab so as to maintain the
σh plane (see page 32, keyword ATOMINSE).

BREAKELAS (for 3D systems only)

This keyword breaks the symmetry of 3D sysems according to a general distortion (3x3 adi-
mensional matrix, not necessarily symmetric):

rec variable value meaning

• ∗ D11 D12 D13 first row of the matrix.
• ∗ D21 D22 D23 second row of the matrix.
• ∗ D31 D32 D33 third row of the matrix.

BREAKELAS can be used when the symmetry must be reduced to apply an external stress
(see EXTSTRESS, OPTGEOM input block, page 121) not compatible with the present sym-
metry.
BREAKELAS reduces the symmetry according to the distortion defined in input, but does
not perform a distortion of the lattice.
Another possibility is when you compute elastic constants, and you want to fix a reference
geometry with FIXINDEX. If your reference geometry has a symmetry higher than the dis-
torted one, then you had to break the symmetry by applying e.g. a tiny elastic distortion
with ELASTIC. By using BREAKELAS you can reduce the symetry without distortion of the
lattice.
Example - Geometry optimization of MgO bulk, cubic, with an applied uniaxial stress modi-
fying the symmetry of the cell.

TEST11 - MGO BULK

CRYSTAL
0 0 0
225
4.21
2
12 0. 0. 0.
8 0.5 0.5 0.5
BREAKELAS the number of symmops is reduced, from 48 to 16
0.001 0. 0. the cell has a tetragonal symmetry now
0. 0. 0.
0. 0. 0.
OPTGEOM
FULLOPTG
EXTSTRESS
0.001 0. 0.
0. 0. 0.
0. 0. 0.
ENDOPT

When EXTSTRESS is requested, the code automatically checks if the required distortion is
possible or not (if the symmetry had not been broken properly beforehand, an error message
comes).

35
CONFCNT - Mapping of CRYSTAL calculations to model Hamilto-
nians
See Chapter 6, page 147.

CPHF - performs the Coupled Perturbed HF/KS calculation up to

the second order
See Chapter 5, page 145.

CLUSTER - a cluster (0D) from a periodic system

The CLUSTER option allows one to cut a finite molecular cluster of atoms from a periodic
lattice. The size of the cluster (which is centred on a specified ’seed point’ A) can be controlled
either by including all atoms within a sphere of a given radius centred on A, or by specifying
a maximum number of symmetry-related stars of atoms to be included.
The cluster includes the atoms B (belonging to different cells of the direct lattice) satisfying
the following criteria:

1. those which belong to one of the first N (input data) stars of neighbours of the seed point
of the cluster.
and

2. those at a distance RAB from the seed point which is smaller then RMAX (input datum).

The resulting cluster may not reproduce exactly the desired arrangement of atoms, particularly
in crystals with complex structures such as zeolites, and so it is possible to specify border
modifications to be made after definition of the core cluster.
Specification of the core cluster:

rec variable value meaning

• ∗ X, Y, Z coordinates of the centre of the cluster [Å] (the seed point)
NST maximum number of stars of neighbours explored in defining the core
cluster
RMAX radius of a sphere centred at X,Y,Z containing the atoms of the core
cluster
• ∗ NNA 6= 0 print nearest neighbour analysis of cluster atoms (according to a radius
criterion)
NCN 0 testing of coordination number during hydrogen saturation carried out
only for Si (coordination number 4), Al (4) and O(2)
N N user-defined coordination numbers are to be defined
if NNA 6= 0 insert 1 record II
• ∗ RNNA radius of sphere in which to search for neighbours of a given atom in
order to print the nearest neighbour analysis
if NCN 6= 0 insert NCN records II
• ∗ L conventional atomic number of atom
MCONN(L) coordination number of the atom with conventional atomic number L.
MCONN=0, coordination not tested

Border modification:

36
 rec variable value meaning
• ∗ NMO number of border atoms to be modified
if NMO > 0 insert NMO records II
• ∗ IPAD label of the atom to be modified (cluster sequence)
NVIC number of stars of neighbours of atom IPAD to be added to the cluster
IPAR =0 no hydrogen saturation
6= 0 cluster border saturated with hydrogen atoms
BOND bond length Hydrogen-IPAD atom (direction unchanged).
if NMO < 0 insert II
• ∗ IMIN label of the first atom to be saturated (cluster sequence)
IMAX label of the last atom to be saturated (cluster sequence)
NVIC number of stars of neighbours of each atom to be added to the cluster
IPAR =0 no hydrogen saturation
6= 0 cluster border saturated with hydrogen atoms
BOND H-cluster atom bond length (direction unchanged).

The two kinds of possible modification of the core cluster are (a) addition of further stars of
neighbours to specified border atoms, and (b) saturation of the border atoms with hydrogen.
This latter option can be essential in minimizing border electric field effects in calculations for
covalently-bonded systems.
(Substitution of atoms with hydrogen is obtained by HYDROSUB).
The hydrogen saturation procedure is carried out in the following way. First, a coordination
number for each atom is assumed (by default 4 for Si, 4 for Al and 2 for O, but these may
be modified in the input deck for any atomic number). The actual number of neighbours of
each specified border atom is then determined (according to a covalent radius criterion) and
compared with the assumed connectivity. If these two numbers differ, additional neighbours are
added. If these atoms are not neighbours of any other existing cluster atoms, they are converted
to hydrogen, otherwise further atoms are added until the connectivity allows complete hydrogen
saturation whilst maintaining correct coordination numbers.
The label of the IPAD atoms refers to the generated cluster, not to the original unit cell. The
preparation of the input thus requires two runs:

1. run using the CLUSTER option with NMO=0, in order to generate the sequence number
of the atoms in the core cluster. The keyword TESTGEOM should be inserted in the
geometry input block. Setting NNA 6= 0 in the input will print a coordination analysis of
all core cluster atoms, including all neighbours within a distance RNNA (which should
be set slightly greater than the maximum nearest neighbour bond length). This can be
useful in deciding what border modifications are necessary.

2. run using the CLUSTER option with NMO 6= 0, to perform desired border modifica-
tions.

Note that the standard CRYSTAL geometry editing options may also be used to modify the
cluster (for example by adding or deleting atoms) placing these keywords after the specification
of the CLUSTER input.
Warning. The system is 0D. No reciprocal lattice information is required in the scf input
(Section 1.3, page 23). See test 16.

COORPRT
Geometry information is printed: cell parameters, fractional coordinates of all atoms in the
reference cell, symmetry operators.
A formatted file, ”fort.33” , is written. See Appendix D, page 262. No input data are re-
quired. The file ”fort.33” has the right format for the program MOLDEN [116] which can be
downloaded from:
www.cmbi.ru.nl/molden/molden.html

37
ELASTCON - Calculation of elastic constants
See Chapter 7, page 149.

ELASTIC
An elastic deformation of the lattice may be defined in terms of the Z or  strain tensors defined
in section 11.9, page 229.

rec variablevalue meaning

• ∗ IDEF ±1 deformation through equation 11.36, Z matrix.
±2 deformation through equation 11.35:  matrix.
>0 volume conserving deformation (equation 11.37).
<0 not volume conserving (equation 11.36 or 11.35).
• ∗ D11 D12 D13 first row of the matrix.
• ∗ D21 D22 D23 second row of the matrix.
• ∗ D31 D32 D33 third row of the matrix.

2
The elastic constant is V −1 ∂∂E2 |i =0 , where V is the volume of the primitive unit cell.
i
The symmetry of the system is defined by the symmetry operators in the new crystallographic
cell. The keyword MAKESAED gives information on symmetry allowed elastic distortions.
The calculation of the elastic constants with CRYSTAL requires the following sequence of
steps:

1. select the ij matrix elements to be changed ( for example, 4 ≡ 23 + 32 ), and set the
others j to zero;

2. perform calculations with different values of the selected matrix element(s) i : 0.02, 0.01,
0.001, -0.001, -0.01, -0.02, for example, and for each value compute the total energy E;
3. perform a polynomial fit of E as a function of i .
 is adimensional, Z in Å(default) or in bohr (page 31). The suggested value for IDEF is
-2 (deformation through equation 11.35, not volume conserving). The examples refer to this
setting.
Example
Geometry input deck to compute one of the energy points used for the evaluation of the C44
(page 232) elastic constants of Li2 O [41].
CRYSTAL
000 3D code
225 3D space group number
4.5733 lattice parameter (Å)
2 2 non equivalent atoms in the primitive cell
8 0.0 0.0 0.0 Z=8, Oxygen; x, y, z
3 .25 .25 .25 Z=3, Lithium; x, y, z
ATOMSYMM printing of the point group at the atomic positions
ELASTIC
-2 deformation not volume conserving through equation 11.35
0. 0.03 0.03  matrix input by rows
0.03 0. 0.03
0.03 0.03 0.
ATOMSYMM printing of the point group at the atomic positions after the defor-
mation
. . . . . . .

A rhombohedral deformation is obtained, through the  matrix. The printout gives information
on the crystallographic and the primitive cell, before and after the deformation:

LATTICE PARAMETERS (ANGSTROMS AND DEGREES) OF

38
(1) ORIGINAL PRIMITIVE CELL
(2) ORIGINAL CRYSTALLOGRAPHIC CELL
(3) DEFORMED PRIMITIVE CELL
(4) DEFORMED CRYSTALLOGRAPHIC CELL

A B C ALPHA BETA GAMMA VOLUME

(1) 3.233811 3.233811 3.233811 60.000000 60.000000 60.000000 23.912726
(2) 4.573300 4.573300 4.573300 90.000000 90.000000 90.000000 95.650903
(3) 3.333650 3.333650 3.333650 56.130247 56.130247 56.130247 23.849453
(4) 4.577414 4.577414 4.577414 86.514808 86.514808 86.514808 95.397811

After the deformation of the lattice, the point symmetry of the Li atoms is C3v , where the C3
axis is along the (x,x,x) direction. The Li atoms can be shifted along the principal diagonal,
direction (x,x,x) of the primitive cell without altering the point symmetry, as shown by the
printing of the point group symmetry obtained by the keyword ATOMSYMM (page 34).
See test20 for complete input deck, including shift of the Li atoms.
See test38 (KCoF3 ).

END
Terminate processing of block 1, geometry definition, input. Execution continues. Subsequent
input records are processed, if required.
Processing of geometry input block stops when the first three characters of the string are
”END”. Any character can follow: ENDGEOM, ENDGINP, etc etc.

EOS - Equation of state

E vs V curves can be computed automatically by specifying the keyword EOS in input block
1 (geometry). EOS must be the last keyword in geometry input.

. . . geometry input . . . .
EOS
END[EOS] close EOS input block
END close geometry input block

EOS input block admits several options (subkeywords), and terminates with keyword END
(or END[EOS], END[—]: the first three characters only are processed).
A constant volume geometry optimization is performed for each point and results are then
fitted to most common equation of states (e.g. Birch-Murnagham 3rd-order). The constant
volume geometry optimization can be only performed in fractional coordinates.
Note that the initial geometry must be fully relaxed.
Work is in progress to extend this scheme to E vs P calculations

Keywords, options, and defaults

A default value is chosen for all computational parameters.
SCF energy convergence threshold is set to 10−8 . To modify it, see keyword TOLDEE in
input block 3, page 102.
Optional keywords are (in any order):

39
 rec variable meaning

• A smRANGE keyword to specify the range of volumes in the E(V) curve

∗ VMIN minimum (compression) variation of the initial volume [def ault =
0.92]
∗ VMAX maximum (expansion) variation of the initial volume [def ault = 1.08]
∗ NPOINTS number of points in the selected range [def ault = 10]
The interval is specified as the minimum (compression) and maximum (expansion) variation of
the volume of the initial geometry. The set of volumes is then defined according to the number
of points in the selected range. For instance, to set the default values:
...
EOS
RANGE
0.92 1.08 10
END[EOS]
...
corresponds to 10 points between 0.92*Vini and 1.08*Vini , where Vini is the volume of the
initial geometry given as input (assumed to be the fully optimized structure) or as obtained
after a preliminary geometry optimization.
Note that the initial E-V value is always included in the final data for fitting (i.e. 11 points in
the example above).

rec variable meaning

• A PRINT This option turns on a higher level of diagnostic printing and will
generally not be necessary for the typical user. LPRINT(250) is
set equal to 1 with this option. Fitting of the E(V) points is also
performed for increasing set of points from 3 to the maximum num-
ber of points. (def ault is minimal printing and no additional lines
should exist)

• A RESTART Allows for a RESTART using file EOSIN F O.DAT from a previous
run.

• A PREOPTGEOM A preliminary geometry optimization of cell and atomic positions

is performed before starting the E(V) curve calculation.

Geometry optimization is performed at convergence criteria tighter than the ones given in
OPTGEOM. Some values can be modified by inserting the following keywords:

• A TOLDEG EOS default [0.00006] - see OPTGEOM, page 110

• A TOLDEX EOS default [0.00012] - see OPTGEOM, page 110
• A TOLDEE EOS default [8] - see OPTGEOM, page 110
• A MAXCYCLE see OPTGEOM, page 113
• A NOTRUSTR see OPTGEOM, page 111
• A TRUSTRADIUS see OPTGEOM, page 111
• A MAXTRADIUS see OPTGEOM, page 111

Important remarks
• The structure given as input should be fully optimized. Note that the same tighter
convergence criteria as in the calculation of the E vs V curve must be used as well as

40
 the FINALRUN=4 option.

• Integrals classification is done on the input geometry. Therefore, initial geometry must
be the equilibrium structure to avoid slightly shifted curves with respect to the true
minimum.

EXTPRT
A formatted input deck with explicit structural/symmetry information is written in file
”fort.34”. If the keyword is entered many times, the data are overwritten. The last geom-
etry is recorded. The file may be used as crystal geometry input to CRYSTAL through the
EXTERNAL keyword.
For instance, to enter the final optimized geometry, or a geometry obtained by editing opera-
tions who modified the original space group or periodicity.
When geometry optimization is performed, the name of the file is ”optc(a)xxx”, being xxx the
number of the cycle, and it is automatically written at each cycle.
Please refer to the standard script for running CRYSTAL09 to handle input/output file names.
See Appendix D, page 258.
No input data are required.

FIELD - Electric field along a periodic direction

rec variable value meaning

• ∗ E0MAX electric field intensity E0 (in atomic units)
• ∗ DIRE(I),I=1,3 crystallographic (Miller) indices of the plane perpendicular to the elec-
tric field
• ∗ SMFACT supercell expansion factor
∗ IORTO 0 non-orthogonal supercell
1 orthogonal supercell
• ∗ MUL number of term in Fourier expansion for triangular electric potential
∗ ISYM +1 triangular potential is symmetric with respect to the z = 0 plane
−1 triangular potential is anti-symmetric with respect to the z = 0 plane

This option can be used with polymers, slabs and crystals and permits to apply an electric
field along a periodic direction of the system.
~ is taken into account according to a perturbation
The effect of a periodic electric field (E)
scheme. The Hamiltonian (Fock or Kohn-Sham) can be written as::

~
Ĥ = Ĥ0 + Ĥ1 (E) (2.1)

~ the electric potential term.

where Ĥ0 is the unperturbed Hamiltonian and Ĥ1 (E)
During the SCF procedure crystalline orbitals are relaxed under the effect of the field, leading
to a perturbed wave function and charge density.

The applied electric field has a square-wave form, that corresponds to a triangular (”sawtooth”)
electric potential.
Due to the form of the potential, a single unit cell must contain both positive and negative
part of the square wave electric field. Then, in order to maintain translational invariance of the
system a new, expanded, unit cell is automatically created by adopting a supercell approach
(see keywords SUPERCEL/SUPERCON, page 56).
This procedure consists in two automatic steps: the re-orientation of the c lattice parameter

41
along the chosen field direction and the multiplication of this lattice vector according to the
~
supercell expansion factor (C=SMFACT·~ c, see fig. 2.1). By varying this parameter is possible
to control the period of the electric potential and therefore the length of the constant region
of the electric field.

Figure 2.1: Triangular electric potential (”sawtooth”) in a supercell with SM F ACT = 4.

Then, for computational reasons, an automatic rotation of the crystal in the cartesian reference
system is performed by aligning C ~ (and therefore E)~ along the z cartesian direction (see
keyword ROTCRY, page 54). After these transformations the field is along the z direction,
~ takes the form:
and the perturbation Ĥ1 (E)

Ĥ1± (Ez ) = V (z) = −qE0 · f ± (z) (2.2)

where the f + (f − ) function is expanded as a Fourier series and is chosen according to the
symmetry of the supercell in the direction of the applied field as follows:
+∞  
2C X 1 2π(2k + 1)z
f + (z) = cos (2.3)
π2 (2k + 1)2 C
k=0

+∞
2C X (−1)k
 
2π(2k + 1)z
f − (z) = sin (2.4)
π2 (2k + 1)2 C
k=0

1. In order to evaluate the dielectric constant of a material in the direction of the applied
field it is necessary to run a PROPERTIES calculation with the keyword DIEL (see page
160). In this way the perturbed wave function is used for the calculation of , following
a macroscopic average scheme, as described in references [54], [32].

2. The field is along the z axis for 3D-crystal calculations; it is along the x for 1D-polymer
and 2D-slab calculations.

3. In calculations of the dielectric constant, more accurate results can be achieved by in-
creasing the SMFACT value. This will lead to systems characterized by a high number
of atoms with large computational costs. The option IORTO = 0 allows to consider
non-orthogonal supercells, characterized by the same dielectric properties of orthogonal
cells, but with a lower number of atoms.

4. In 3D-crystals, the electric potential takes a triangular form to maintain translational

symmetry and electric neutrality of cell. The symmetry of triangular potential has two
options:

a) ISYM=+1, triangular potential is symmetric with respect to the center of the su-
percell, along the z axis. Use this option if there is a symmetry plane orthogonal to
the z axis.
b) ISYM=−1, triangular potential is anti-symmetric. This option can be used when
the supercell does not have a symmetry plane orthogonal to z axis.

42
Figure 2.2: Left: symmetric triangular electric potential (ISIM = 1). Right anti-symmetric
triangular electric potential (ISYM=-1).

5. MUL, the number of terms in the Fourier expansion, can take values between 1 and 60.
MUL=40 is sufficient to adequately reproduce the triangular shape of the potential.

6. High E0MAX values are inconsistent with perturbation method, the choice of E0MAX
depends on the dielectric susceptibility of the system and on the gap width. For small gap
cases, use of eigenvalue level shifting technique is recommended (keyword LEVSHIFT,
page 92).

7. When an external field is applied, the system can become conducting during the SCF
procedure. In order to avoid convergence problems, it is advisable to set the shrinking
factor of the Gilat net ISP equal to 2× IS, where IS is the Monkhorst net shrinking factor
(see SCF input, page 97).

Conversion factors for electric field:

1 AU = 1.71527E+07 ESU·CM−2 = 5.72152E+01 C ·M−2 = 5.14226E+11 V·M−1

FIELDCON - Electric field along non-periodic direction

rec variable meaning

• ∗ E(I),I=N,3 field components along x,y,z directions

For a brief theoretical introduction see keyword FIELD.

This option can be used with molecules, polymers, slabs and permits to apply an electric field
along a non-periodic direction of the system.

1. For molecules (N=1) three components of the field must be supplied, as the field can be
directed along any direction.

2. For polymers (N=2) two components (y,z) of the field must be defined; the x component
of the field must be zero because the default orientation of polymers is along the x axis.

3. For slabs (N=3) just one component (z) of the field have to be defined; the x,y components
must be zero because the default orientation of slabs in is in x-y plan.

Conversion factors for electric field:

1 AU = 1.71527E+07 ESU·CM−2 = 5.72152E+01 C ·M−2 = 5.14226E+11 V·M−1
This option can evaluate the dielectric response of the molecule, polymer or slab in a direction
of non periodicity (see option FIELD for a field along a periodicity direction).

43
Consider the following expansion of the total energy of the system as a function of the applied
field:

1 ( 1
E(F0 ) = E0 − µF0 − αF02 − )3!βF03 − γF04 − · · · (2.5)
2! 1 4!
By fitting the E vs F0 data the µ, α, β and γ values can be derived. See
http://www.crystal.unito.it → tutorials → Static dielectric constants..

FINDSYM
Geometry information is written in file FINDSYM.DAT, according to the input format of the
program FINDSYM.
http://stokes.byu.edu/findsym.html
FINDSYM: Identify the space group of a crystal, given the positions of the atoms in a unit
cell. When geometry editing modifies the basic input space group, the symmetry of the system
is identified by the symmetry operators only. The program FINDSYM allows identification of
the space group.

FRACTION
The keyword FRACTION means input coordinates given as fraction of the lattice parameter
in subsequent input, along the direction of translational symmetry:

x,y,z crystals (3D)

x,y slabs (2D; z in Ångstrom or bohr)
x polymers (1D; y,z in Ångstrom or bohr)

no action for 0D. When the unit of measure is modified, the new convention is active for all
subsequent geometry editing.

FREQCALC - Harmonic frequencies at Γ

See Chapter 4, page 127.

HYDROSUB - substitution with hydrogen atoms

rec variable meaning
• ∗ NSOST number of atoms to be substituted with hydrogen
insert NSOST records II
• ∗ LA label of the atom to substitute
LB label of the atom linked to LA
BH bond length B-Hydrogen

Selected atoms are substituted with hydrogens, and the bond length is modified. To be used
after CLUSTER.

KEEPSYMM
In any subsequent editing of the geometry, the program will endeavour to maintain the number
of symmetry operators, by requiring that atoms which are symmetry related remain so after
geometry editing (keywords: ATOMSUBS, ATOMINSE, ATOMDISP, ATOMREMO)
or the basis set (keywords CHEMOD, GHOSTS).

44
Example: When a CO molecule is vertically adsorbed on a (001) 3-layer MgO slab, (D4h
symmetry) (see page 32, keyword ATOMINSE), the symmetry is reduced to C4v , if the
BREAKSYM keyword is active. The symmetry operators related to the σh plane are re-
moved. However, if KEEPSYMM is active, then additional atoms will be added to the
underside of the slab so as to maintain the σh plane.

MAKESAED
Symmetry allowed elastic distortions are printed. No input data required.

MODISYMM
rec variable meaning
• ∗ N number of atoms to be attached a flag
• ∗ LA,LF(LA),L=1,N atom labels and flags (n couples of integers in 1 record).

The point symmetry of the lattice is lowered by attributing a different ”flag” to atoms related
by geometrical symmetry. The symmetry operators linking the two atoms are removed and the
new symmetry of the system is analyzed. For instance, when studying spin-polarized systems, it
may be necessary to apply different spins to atoms which are related by geometrical symmetry.

MOLDRAW
The last version of the program MOLDRAW reads crystal standard output, and can generate
a movie from an optimization run. See:
http://www.moldraw.unito.it .

MOLEBSSE - counterpoise for molecular crystals

rec variable meaning
• ∗ NMOL number of molecules to be isolated
II insert NMOL records II
• ∗ ISEED label of one atom in the n-th molecule
J,K,L integer coordinates (direct lattice) of the primitive cell containing the ISEED
atom
• ∗ NSTAR maximum number of stars of neighbours included in the calculation
RMAX maximum distance explored searching the neighbours of the atoms belonging
to the molecule(s)

The counterpoise method [17] is applied to correct the Basis Set Superposition Error in molecu-
lar crystals. A molecular calculation is performed, with a basis set including the basis functions
of the selected molecules and the neighbouring atoms. The program automatically finds all
the atoms of the molecule(s) containing atom(s) ISEED (keyword MOLECULE, page 46).
The molecule is reconstructed on the basis of the covalent radii reported in Table on page 53.
They can be modified by running the option RAYCOV, if the reconstruction of the molecule
fails. The radius of the hydrogen atom is very critical when intermolecular hydrogen bonds
are present.
All the functions of the neighbouring atoms in the crystal are added to the basis set of the
selected molecule(s) such that both the following criteria are obeyed:

1. the atom is within a distance R lower than RMAX from at least one atom in the molecule
and

45
 2. the atom is within the NSTAR-th nearest neighbours of at least one atom in the molecule.

For molecular crystals only.

Warning Do not use with ECP
Warning. The system obtained is 0D. No reciprocal lattice information is required in the scf
input (Section 1.3, page 23). See test 19.

MOLECULE - Extraction of n molecules from a molecular crystal

rec variable meaning
• ∗ NMOL number of molecules to be isolated
II insert NMOL records II
• ∗ ISEED label of one atom in the nth molecule
J,K,L integer coordinates (direct lattice) of the primitive cell containing the
ISEED atom

The option MOLECULE isolates one (or more) molecules from a molecular crystal on the
basis of chemical connectivity, defined by the sum of the covalent radii (Table on page 53).
The covalent radii can be modified by running the option RAYCOV, if the reconstruction of
the molecule fails. The covalent radius of the hydrogen atom is very critical when intermolecular
hydrogen bonds are present.
The input order of the atoms (atoms symmetry related are grouped) is modified, according
to the chemical connectivity. The same order of the atoms in the bulk crystal is obtained by
entering the keyword ATOMORDE (see Section 2.1, page 32). The total number of electrons
attributed to the molecule is the sum of the shell charges attributed in the basis set input (input
block 2, Section 1.2, page 19) to the atoms selected for the molecule.
The keyword GAUSS98, entered in input block 2 (basis set input), writes an input deck to
run Gaussian 98 (see page 64)
For molecular crystals only.
Warning. The system is 0D. No reciprocal lattice information is required in the scf input
(Section 1.3, page 23).
Test 18 - Oxalic acid. In the 3D unit cell there are four water and two oxalic acid molecules.
The input of test 18 refers to a cluster containing a central oxalic acid molecule surrounded by
four water molecules.

MOLEXP - Variation of lattice parameters at constant symmetry

and molecular geometry
rec variable meaning
• ∗ δa,[δb],[δc], increments of the minimal set of crystallographic cell parameters:
[δα],[δβ] translation vectors length [Ångstrom],
[δγ] crystallographic angles (degrees)

The cell parameters (the minimum set, see page 16) are modified, according to the increments
given in input. The volume of the cell is then modified. The symmetry of the lattice and the
geometry (bond lengths and bond angles) of the molecules within the cell is kept. The fractional
coordinates of the barycentre of the molecules are kept constant, the cartesian coordinates
redefined according to the modification of the lattice parameters. Optimization of the geometry
with reference to the compactness of the lattice is allowed, keeping constant the geometry of
the molecules. When there are very short hydrogen bonds linking the molecules in the lattice,

46
it may be necessary a modification of the atomic radii to allow proper identification of the
molecules (see option RAYCOV, page 53)

MOLSPLIT - Periodic lattice of non-interacting molecules

In order to compare bulk and molecular properties, it can be useful to build a density ma-
trix as a superposition of the density matrices of the isolated molecules, arranged in the same
geometry as in the crystal. The keyword MOLSPLIT (no additional input required) per-
forms an expansion of the lattice, in such a way that the molecules of the crystal are at an
”infinite” distance from each other. The crystal coordinates are scaled so that the distances
inside the molecule are fixed, and the distances among the molecules are expanded by a factor
100, to avoid molecule-molecule interactions. The 3D translational symmetry is not changed.
Reciprocal lattice information is required in the scf input (Section 1.3, page 23).
A standard wave function calculation of the expanded crystal is performed. The density
matrix refers to the non-interacting subsystems. Before running properties, the lattice
is automatically contracted to the bulk situation given in input. If a charge density or
electrostatic potential map is computed (ECHG, POTM options), it corresponds to the
superposition of the charge densities of the isolated molecules in the bulk geometry.
This option must be used only for molecular crystals only (no charged fragments).
Warning: the DFT grid is not designed for the expanded lattice yet. Large memory allocation
may be necessary.

See test 21.

NANOTUBE - Building a nanotube from a slab

Nanotubes are cylindrical structures, periodic along one direction. They are therefore charac-
terised by a single lattice vector. However, in order to study their symmetry and orientation,
it is easier to start from a 2D lattice where the additional periodicity becomes the cylinder
period. This 2D periodic flat structure will be referred in the following, as the nanotube flat
lattice.
~
In this flat lattice, 3 vectors are important for the structure description: the rolling vector, R,
the longitudinal vector, L ~ and the helical vector, H.
~
~ = n1 a~1 +n2 a~2 (where a~1 and a~2 are the slab cell vectors), is sufficient
• The rolling vector, R
to completely define a nanotube. It is used in the literature to characterise the systems
(for example the (4,3) nanotube is a nanotube with a rolling vector R ~ = 4a~1 +3a~2 ). In the
~
flat nanotube lattice, R is a nanotube unit cell vector. Once the nanotube wrapped, the
rolling vector becomes a circle normal to the cylinder axis; its norm corresponds to the
cylinder perimeter. If R~ passes N times through the lattice nodes, a N -order rotational
axis will exist along the corresponding nanotube axis.
~ = l1 a~1 + l2 a~2 , is the shortest lattice vector normal to R.
• The longitudinal vector, L ~ In
the nanotube, it becomes the 1D lattice parameter and gives the 1D periodicity along
the tube axis.
~ = h1 a~1 +h2 a~2 , is a lattice vector defining with R
• The helical vector, H ~ an area which is N
times the area of the unit cell of the flat slab. It satisfies, then, the following relationship
~ H)
S(R, ~
= |n1 h2 − n2 h1 | = N (2.6)
S(a~1 , a~2 )

47
 where S(~vi , v~j ) is the surface defined by the v~i and v~j vectors. The helical vector defines a
correspondence between a translation in the flat slab and a roto-translation in the curved
~ has a rotational component along the circumference vector and a translational
surface; H
component along the lattice parameter.

The direct product between the rotational and roto-translational operations generates the full
symmetry of the nanotube.
The three vectors listed previously are represented in figure 2.3 for graphene; the example
refers to the construction of a (4,2) single-walled carbon nanotube (SWCNT).

Figure 2.3: Building the (4,2) SWCNT from graphene.

Further information about the implemented method, the computational costs and the symme-
try features of nanotubes are found in Ref. [31] and [132, 7].

rec variable meaning

• ∗ n1 , n 2 components of the roll-up-vector of the nanotube in the slab unit
cell basis. The roll-up-vector, whose length corresponds to the tube
circumference, is expressed as a linear combination of the unit cell
vectors of the slab before rolling up, n1 and n2 being integer coeffi-
cients.

The convention adopted in CRYSTAL is n1 ≥ n2 . In cases when n2 is required to be larger

than n1 , it is sufficient to exchange the x and y coordinates of the reference 2D slab.
NANOTUBE can be used with the following 2D systems:

• square and hexagonal lattices: any (n1 , n2 ) combination

• rectangular lattices: any n1 , n2 = 0

• rhombohedral lattices: n2 = ±n1

In the other cases, the existence of a lattice vector perpendicular to R~ (so the 1D periodicity
along the tube axis) is not guaranteed.
The orthogonality condition between R ~ and L ~ provides the following equation:

~ ·R
L ~ = (l1 a~1 + l2 a~2 ) · (n1 a~1 + n2 a~2 ) = n1 l1 |a~1 |2 + n2 l2 |a~2 |2 + (n1 l2 + n2 l1 )|a~1 ||a~2 |cos(γ) = 0(2.7)

48
γ being the angle between a~1 and a~2 . From Equation 2.7 one obtains:

l1 n2 |a2 |2 + n1 |a1 ||a2 |cos(γ)

=− (2.8)
l2 n1 |a1 |2 + n2 |a1 ||a2 |cos(γ)

If the right term is divided and multiplied by |a2 |2 , one obtains

l1 n2 + n1 aq cos(γ)
=− (2.9)
l2 aq (n1 aq + n2 cos(γ))
|a1 |
with aq = |a 2|
.
The above equation cannot be satisfied for any (a~1 ,a~2 ,cos(γ)) combination. This observation
is based on the fact that, as l1 and l2 are integers, ll12 is a rational number, whereas, in general,
cos(γ) and aq are real numbers. In the following the five 2D Bravais lattices are considered
separately, in order to show which conditions satisfy the periodicity along the tube axis and
which do not.

• Hexagonal lattice: a~1 = a~2 , cos(γ) = ±1/2. Equation 2.9 becomes:

l1 n2 + n1 /2 2n2 + n1
=− =− (2.10)
l2 n1 + n2 /2 2n1 + n2

Any roll-up vector is possible.

• Square lattice: a~1 = a~2 , cos(γ) = 0.

l1 n2
=− (2.11)
l2 n1
Any roll-up vector is possible.

• Rectangular lattice: a~1 6= a~2 , cos(γ) = 0.

l1 n2 |a2 |2
=− (2.12)
l2 n1 |a1 |2

In this case the right term corresponds to a rational number either if |a1 | = n|a2 |, with
n being a rational number, or if n2 =0. More generally, for rectangular lattices, the
periodicity along the tube axis is always satisfied for (n, 0) and (0, n) roll-up vectors.

• Rhombohedral (centred rectangular) lattice: a~1 = a~2 , any cos(γ).

l1 n2 + n1 cos(γ)
=− (2.13)
l2 n1 + n2 cos(γ)

The right term provides a rational number only when n1 = n2 or n1 = −n2 , so that
Equation 2.13 becomes:

l1 n1 [1 + cos(γ)] l1 n1 [1 − cos(γ)]
=− = −1 = =1 (2.14)
l2 n1 [1 + cos(γ)] l2 n1 [1 − cos(γ)]

• Oblique lattice: a~1 6= a~2 , any cos(γ). Equation 2.9 remains as such, and the right term
is always an irrational number.

When the above conditions are not satisfied, however, it is possible to manipulate the geometry
of the starting slab and force it to assume a suitable form, by building supercells or with minor
modifications to the cell parameters.

49
Note that, in cases of hexagonal lattices, the standard convention adopted in CRYSTAL con-
siders the reference flat lattice cell vectors forming a 120◦ angle. In the carbon nanotube
literature, however, the rolling indices refer to a 60◦ angle. For this reason, the same input
can be obtained with the following choices, where the (6,3) indices in the 60◦ unit cell become
(9,3) in the 120◦ option:

Example I: (6,3) SWCNT from graphene, 60 degrees

SLAB
1
2.47 2.47 60.000
2
6 0.33333333333 0.33333333333 0.000
6 -0.33333333333 -0.33333333333 0.000
NANOTUBE
6 3

Example II: (6,3) SWCNT from graphene, default choice

SLAB
77
2.47
1
6 0.333333 0.666667 0.000000
NANOTUBE
9 3

In order to avoid ambiguity in the choice of the carbon nanotubes indices, use the SWCNT
(page 50) keyword.

SWCNT - Building a nanotube from an hexagonal slab (60◦ unit cell

choice)
rec variable meaning
• ∗ n1 , n 2 Only for hexagonal cells. Same as NANOTUBE keyword but the
components of the rolling vector refer to a 60◦ hexagonal cell.
vspace0.3cm

Example: (6,3) SWCNT from graphene

SLAB
77
2.47
1
6 0.333333 0.666667 0.000000
SWCNT
6 3

See NANOTUBE (page 47) for further details.

Nanotube rebuild: build a nanotube starting from the structure of

another one, with same directions but a different radius.
Two restart keywords, NANORE and SWCNTRE, allow to build a (n1 , n2 ) nanotube by start-
ing from the structure of another one (a previously (nold1 , nold2 ) optimised one, read from file
fort.35). The ”old” nanotube is unrolled and re-rolled according to a ”new” R ~ vector, with
minor modifications to the structure. The rolling direction of the two tubes must be the same.
It is particularly helpful for the geometry optimisation of inorganic nanotubes (thick slabs,
large systems, the geometry of the tubes is very different from the one of the slab), as the
number of optimisation steps is reduced (see Ref. [33] for documentation of computational
gain).

50
NANORE
To be used for tubes built with NANOTUBE (page 47).

rec variable meaning

• ∗ nold1 , nold2 Nanotube rebuild: indices of the starting nanotube.
∗ n1 , n 2 New indices of the rolling vector.

Consider, for example, the (8,8) and the (10,10) carbon nanotubes (built with NANOTUBE,
indices refer to the 120◦ unit cell choice). We have optimised the structure of the former, and
we want to build the latter starting from its geometry. With NANORE the (8,8) nanotube is
unrolled and re-rolled as (10,10). In order to do this, the information on geometry of both the
starting slab (graphene) and the (8,8) nanotube is required. The first one is given in input (or
read with EXTERNAL from file fort.34), the second one is read with an EXTERNAL strategy
from file fort.35. The input syntax is then:

SLAB
77
2.47
1
6 0.333333 0.666667 0.000000
NANORE
8 8
10 10

For nanotubes built with SWCNT (page 50) see SWCNTRE (page 51).

SWCNTRE
It is the same as NANORE, with the same input syntax, but in for hexagonal lattices with 60◦
unit cell reference (see SWCNT, page 50).

rec variable meaning

• ∗ nold1 , nold2 Nanotube rebuild: indices of the starting nanotube.
∗ n1 , n 2 New indices of the rolling vector.

NANOJMOL - Graphical visualisation of nanotubes with Jmol

A file named NANOJMOL.DAT containing the nanotube structure is generated. This file can
be directly used by the 3D structure viewer Jmol (www.jmol.org). This keyword must appear
in the geometry block before all the keywords related to nanotube construction.

NEIGHBOR/NEIGHPRT
rec variable meaning
• ∗ INEIGH number of neighbours of each non-equivalent atom to be printed

The option is active when analyzing the crystal structure (bond lengths and bond angles) and
when printing the bond populations following Mulliken analysis. Full input deck must be given
(block 1-2-3),in order to obtain neighbors analysis of all the non-equivalent atoms
For each non-equivalent atom information on the first INEIGH neighbours is printed: number,
type, distance, position (indices of the direct lattice cell).
Warning: the neighbors analysis is performed after the symmetry analysis and the screening
of the integrals. If very soft tolerances for the integrals screening are given in input, it may
happen that the information is not given for all the neighbors requested, as their are not taken
into account when truncation criteria are applied.

51
NOSHIFT
It may be used before SUPERCEL keyword. It avoids shift of the origin in order to minimize
the number of symmetry operators with finite translation component. No input data are
required.

OPTGEOM - Full geometry optimization

See Chapter 3, page 105.

ORIGIN
The origin is moved to minimize the number of symmetry operators with finite translation
components. Suggested before cutting a slab from a 3D structure (option SLABCUT, page
55). No input data are required.

PARAMPRT - printing of parametrized dimensions

The parameters controlling the dimensions of the static allocation arrays of the program are
printed. No input data are required.

POINTCHG
rec variablemeaning
• ∗ NCH number of point charges to be added
insert NCH records II
• ∗ X,Y,Z,QC cartesian coordinates [Å], charge(au). Coordinates refer to the primitive cell.

Dummy atoms with formal atomic number 93, mass zero, nuclear charge as given in input (file
POINTCHG.INP), are added to the primitive cell. Data are read in free format.

record type of data content

1 1 integer N, number of point charges
2..2+N-1 4 real x y z charge

Coordinates are in Ångstrom, unless otherwise requested (keyword BOHR, page 31). Charges
are net charges (1 electron = -1). The symmetry of the system must be removed by the keyword
SYMMREMO.
As point charges are formally considered as ”atoms”, they must be the last addition of centres
to the system.
No electron charge should be attributed to those atoms in basis set input (no atomic wave
function calculation is possible). The default basis set defined by the program is a single s
gaussian, with exponent 100000.
Attention should be paid to the neutrality of the cell. If the absolute value of the sum of the
charges is less than 10−3 , the value of the charges is ”normalized” to obtain 0.
The data given in input are printed. To obtain printing of coordinates and neighbour analysis
of the dummy atoms in geometry output, insert the keyword PRINTCHG.
Not compatible with: OPTGEOM, FREQCALC, ANHARM, ELASTCON, EOS,
FIELD, FIELDCON, NOBIPOLA.

PRIMITIV
Some properties (XFAC, EMDL, EMDP, PROF) input the oblique coordinates of the k
points in the reciprocal lattice with reference to the conventional cell, though the computation
refers to the primitive one. This option allows entering directly the data with reference to the
primitive cell. The transformation matrix from primitive to crystallographic (Appendix A.5,
page 244) is set to the identity. No effect on the CPU time: CRYSTAL always refers to the
primitive cell. No input data are required.

52
PRINTCHG
Coordinates of the dummy atoms inserted after the keyword POINTCHG are printed in
geometry output, basis set output, neighbor analysis. No input data required.

PRINTOUT - Setting of printing environment

Extended printout can be obtained by entering selected keywords in a printing environment
beginning with the keyword PRINTOUT and ending with the keyword END. The possible
keywords are found in the fifth column of the table on page 256.
Extended printing request can be entered in any input block. Printing requests are not trans-
ferred from wave function to properties calculation.
See Appendix C, page 254.

PRSYMDIR
Printing of the so-called symmetry allowed directions adopted in the geometry optimization.
The printing is done after the neighbor analysis, before computing the wave function. Full
input must be supplied (3 blocks). Test run allowed with the keyword TESTPDIM.
No input data required.

PURIFY
This cleans up the atomic positions so that they are fully consistent with the group (to within
machine rounding error). Atomic position are automatically redefined after basic geometry
input. No input data are required.

RAYCOV - covalent radii modification

rec variable meaning
• ∗ NCOV number of atoms for which the covalent radius is redefined
insert NCOV records II
• ∗ NAT atomic number (0 ≤NAT ≤ 92)
RAY covalent radius of the atom with atomic number NAT ([Å], default,
or bohr, if the keyword BOHR precedes in the deck)

The option RAYCOV allows modification of the covalent radius default value for a given
atom.
Table of covalent radii (Angstrom)
H He
0.68 1.47
--------- -----------------------------
Li Be B C N O F Ne
1.65 1.18 0.93 0.81 0.78 0.78 0.76 1.68
--------- -----------------------------
Na Mg Al Si P S Cl Ar
2.01 1.57 1.50 1.23 1.15 1.09 1.05 1.97
-----------------------------------------------------------------------------------------
K Ca Sc Ti V Cr Mn Fe Co Ni Cu Zn Ga Ge As Se Br Kr
2.31 2.07 1.68 1.47 1.41 1.47 1.47 1.47 1.41 1.41 1.41 1.41 1.36 1.31 1.21 1.21 1.21 2.10
-----------------------------------------------------------------------------------------
Rb Sr Y Zr Ni Mo Tc Ru Rh Pd Ag Cd In Sn Sb Te I Xe
2.31 2.10 1.94 1.60 1.52 1.52 1.42 1.36 1.42 1.47 1.68 1.62 1.62 1.52 1.52 1.47 1.47 2.66
-----------------------------------------------------------------------------------------
Cs Ba La Hf Ta W Re Os Ir Pt Au Hg Tl Pb Bi Po At Rn
2.73 2.10 1.94 1.63 1.63 1.63 1.63 1.63 1.63 1.63 1.63 1.63 1.99 1.89 1.68 1.42 1.42 1.62
-----------------------------------------------------------------------------------------
The choice of the covalent radius of hydrogen may be very critical when extracting a molecule
from a hydrogen bonded molecular crystal. See test 15.

53
ROTCRY - Rotation of the crystal with respect to the reference
system - developers only

This option allows to rotate the crystal with respect to the original orthonormal Cartesian
reference system. The SCF procedure, both for HF and DFT calculations, is performed in the
rotated geometry.
The rotation can be performed in three different ways:

1. By defining the Euler rotation angles α, β, γ and the origin of the rotating system. (The
rotation is given by: Rα β γ θ
z Rx Rz , where Rt are the rotation matrices about t by angle θ).

2. By explicitly defining the rotation matrix.

3. An automatic procedure that reorient the crystal aligning ~c along z Cartesian axis.

ANGROT Rotation defined by Euler angles α, β, γ

rec variable meaning
• ∗ ALPHA,BETA,GAMMA α, β, γ rotation Euler angles (dgrees)
or

MATROT Rotation matrix by input

rec variable meaning
• ∗ R11 R12 R13 first row of the matrix.
• ∗ R21 R22 R23 second row of the matrix.
• ∗ R31 R32 R33 third row of the matrix.
or

AUTO Automatically align c along z

The rotation involves: direct and reciprocal lattice parameters, coordinates of atoms and
symmetry operators. When a DFT calculation is performed also the points of the numerical
integration grid are rotated in order to preserve numerical accuracy.
Note that this keyword is different from ATOMROT (see page 33) that rotates a group of
atoms without affecting the reference system.

SETINF - Setting of INF values

rec variable meaning
• ∗ NUM number of INF vector positions to set
• ∗ J,INF(J),I=1,NUM position in the vector and corresponding value

The keyword SETINF allows setting of a value in the INF array. It can be entered in any
input section.

SETPRINT - Setting of printing options

rec variable meaning
• ∗ NPR number of LPRINT vector positions to set
• ∗ J,LPRINT(J),I=1,NPR prtrec; position in the vector and corresponding value

The keyword SETPRINT allows setting of a value in the LPRINT array, according to the
information given in Appendix C, page 256. It can be entered in any input section.

54
SLABCUT (SLAB)
rec variable meaning
• ∗ h, k, l crystallographic (Miller) indices of the plane parallel to the surface
• ∗ ISUP label of the surface layer
NL number of atomic layers in the slab

The SLABCUT option is used to create a slab of given thickness, parallel to the given plane
of the 3D lattice.
A new Cartesian frame, with the z axis orthogonal to the (hkl) plane, is defined. A layer is
defined by a set of atoms with same z coordinate, with reference to the new Cartesian frame.
The thickness of the slab, the 2D system, is defined by the number of layers. No reference is
made to the chemical units in the slab. The neutrality of the slab is checked by the program.

1. The crystallographic (Miller) indices of the plane refer to the conventional cell (cubic and
hexagonal systems).

2. A two-sided layer group is derived from the 3D symmetry group of the original crystal
structure: the origin may be shifted to maximize the order of the layer group (keyword
ORIGIN, page 52).

3. The unit cell is selected with upper and lower surface parallel to the (hkl) plane.

4. The 2D translation vectors a1 and a2 are chosen according to the following criteria:

(a) minimal cell area;

(b) shortest translation vectors;
(c) minimum |cos(γ)|, where γ is the angle between a1 and a2 .

5. The surface layer ISUP may be found from an analysis of the information printed by the
SLABINFO (page 56) option. This information can be obtained by a test run, inserting
in the geometry input block the keyword TESTGEOM (page 60). Only the geometry
input block is processed, then the program stops.

Two separate runs are required in order to get the information to prepare the input for a full
SLABCUT option run:

1. keyword SLABINFO: Rotation of the 3D cell, to have the z axis perpendicular to the
(hkl) place, with numbering of the atomic layers in the rotated reference cell, according
to the z coordinate of the atoms (insert STOP after SLABINFO to avoid further
processing).

2. keyword SLAB: Definition of the 2D system, a slab of given thickness (NL, number of
atomic layers) parallel to the (hkl) crystallographic plane, with the ISUP-th atom on the
surface layer

The SLABCUT option, combined with ATOMINSE (page 32), ATOMDISP (page 32),
etc. can be used to create a slab of given thickness, with an atom (or group of atoms) adsorbed
at given position. This is achieved by adding new atoms to the 2D structure, obtained after
executing the SLAB option.
Test cases 5-6-7 refer to a 2D system; test cases 25-26-27 refer to the same system, but generated
from the related 3D one. See also tests 35, 36, 37.

55
SLABINFO - 3D cell with z axis orthogonal to a given plane
rec variable meaning
• ∗ h,k,l Crystallographic (Miller) indices of the basal layer of the new 3D unit cell

1. A new unit cell is defined, with two lattice vectors perpendicular to the [hkl] direction.
The indices refer to the Bravais lattice of the crystal; the hexagonal lattice is used for
the rhombohedral systems, the cubic lattice for cubic systems (non primitive).

2. A new Cartesian reference system is defined, with the xy plane parallel to the (hkl) plane.

3. The atoms in the reference cell are re-ordered according to their z coordinate, in order
to recognize the layered structure, parallel to the (hkl) plane.

4. The layers of atoms are numbered. This information is necessary for generating the input
data for the SLABCUT option.

5. After neighboring analysis, the program stops. If the keyword ROTATE was entered,
execution continues. The shape of the new cell may be very different, computational
parameters must be carefully checked.

6. the keyword ORIGIN can be used to shift the origin after the rotation of the cell, and
minimize the number of symmetry operators with translational component. Useful to
maximize the point group of the 2D system that can be generated from 3D using the
keyword SLABCUT (page 55).

STOP
Execution stops immediately. Subsequent input records are not processed.

STRUCPRT
A formatted deck with cell parameters and atoms coordinates (bohr) in cartesian reference is
written in the file STRUC.INCOOR . See appendix D, page 263.

SUPERCEL
rec variable meaning
• ∗ E expansion matrix E (IDIMxIDIM elements, input by rows: 9 reals (3D); 4 reals
(2D); 1 real (1D)

A supercell is obtained by defining the new unit cell vectors as linear combinations of the
primitive cell unit vectors (See SUPERCON for crystallographic cell vectors reference). The
point symmetry is defined by the number of symmetry operators in the new cell. It may be
reduced, not increased.
The new translation vectors b01 , b02 , b03 are defined in terms of the old vectors b1 , b2 , b3 and of
the matrix E, read in input by rows, as follows:

b01 = e11 · b1 + e12 · b2 + e13 · b3

b02 = e21 · b1 + e22 · b2 + e23 · b3
b03 = e31 · b1 + e32 · b2 + e33 · b3

The symmetry is automatically reduced to the point symmetry operators without translational
components and a further reduction of the symmetry is also possible.

56
Before building the supercell, the origin is shifted in order to minimize the number of sym-
metry operators with translational components (see page 17). To avoid this operation, insert
NOSHIFT before SUPERCEL
Atoms that are related by translational symmetry in the unit cell are considered nonequivalent
in a supercell.
The supercell option is a useful starting point for the study of defective systems, of chemisorp-
tion and anti ferromagnetism, by combining the SUPERCELoption with the options de-
scribed in this chapter: ATOMREMO (page 32), ATOMSUBS (page 34), ATOMINSE
(page 32), ATOMDISP (page 32), SLAB (page 55).
To study anti ferromagnetic (AFM) states, it may be necessary to generate a supercell, and
then attribute different spin to atoms related by translational symmetry (ATOMSPIN, input
block 3, page 72). See tests 17, 30, 31, 34, 37, 43, 47.

Example. Construction of supercells of face-centred cubic 3D system (a = 5.42 Å).

The crystallographic cell is non-primitive, the expansion matrix refers to primitive
cell vectors. The E matrix has 9 elements:
PRIMITIVE CELL
DIRECT LATTICE VECTORS COMPONENTS
X Y Z
B1 .000 2.710 2.710
B2 2.710 .000 2.710
B3 2.710 2.710 .000

2 UNITS SUPERCELL (a)

EXPANSION MATRIX DIRECT LATTICE VECTORS
E1 .000 1.000 1.000 B1 5.420 2.710 2.710
E2 1.000 .000 1.000 B2 2.710 5.420 2.710
E3 1.000 1.000 .000 B3 2.710 2.710 5.420

2 UNITS SUPERCELL (b)

EXPANSION MATRIX DIRECT LATTICE VECTORS
E1 1.000 1.000 -1.000 B1 .000 .000 5.420
E2 .000 .000 1.000 B2 2.710 2.710 .000
E3 1.000 -1.000 .000 B3 -2.710 2.710 .000

4 UNITS SUPERCELL (c) crystallographic cell

EXPANSION MATRIX DIRECT LATTICE VECTORS
E1 -1.000 1.000 1.000 B1 5.420 .000 .000
E2 1.000 -1.000 1.000 B2 .000 5.420 .000
E3 1.000 1.000 -1.000 B3 .000 .000 5.420

8 UNITS SUPERCELL
EXPANSION MATRIX DIRECT LATTICE VECTORS
E1 2.000 .000 .000 B1 .000 5.420 5.420
E2 .000 2.000 .000 B2 5.420 .000 5.420
E3 .000 .000 2.000 B3 5.420 5.420 .000

16 UNITS SUPERCELL
EXPANSION MATRIX DIRECT LATTICE VECTORS
E1 3.000 -1.000 -1.000 B1 -5.420 5.420 5.420
E2 -1.000 3.000 -1.000 B2 5.420 -5.420 5.420
E3 -1.000 -1.000 3.000 B3 5.420 5.420 -5.420

27 UNITS SUPERCELL
EXPANSION MATRIX DIRECT LATTICE VECTORS
E1 3.000 .000 .000 B1 .000 8.130 8.130
E2 .000 3.000 .000 B2 8.130 .000 8.130
E3 .000 .000 3.000 B3 8.130 8.130 .000

32 UNITS SUPERCELL
EXPANSION MATRIX DIRECT LATTICE VECTORS
E1 -2.000 2.000 2.000 B1 10.840 .000 .000
E2 2.000 -2.000 2.000 B2 .000 10.840 .000
E3 2.000 2.000 -2.000 B3 .000 .000 10.840

a), b) Different double cells

c) quadruple cell. It corresponds to the crystallographic, non-primitive cell, whose parameters

57
are given in input (page 17).

Example. Construction of supercells of hexagonal R3̄ (corundum lattice) cubic 3D system.

The crystallographic cell is non-primitive: CRYSTAL refer to the primitive cell, with volume
1/3 of the conventional one. The E matrix has 9 elements:

GEOMETRY INPUT DATA:

LATTICE PARAMETERS (ANGSTROMS AND DEGREES) - CONVENTIONAL CELL
A B C ALPHA BETA GAMMA
4.76020 4.76020 12.99330 90.00000 90.00000 120.00000

TRANSFORMATION WITHIN CRYSTAL CODE FROM CONVENTIONAL TO PRIMITIVE CELL:

LATTICE PARAMETERS (ANGSTROMS AND DEGREES) - PRIMITIVE CELL

A B C ALPHA BETA GAMMA VOLUME
5.12948 5.12948 5.12948 55.29155 55.29155 55.29155 84.99223

3 UNITS SUPERCELL crystallographic cell

EXPANSION MATRIX DIRECT LATTICE VECTORS
E1 1.000 -1.000 .000 B1 4.122 -2.380 .000
E2 .000 1.000 -1.000 B2 .000 4.760 .000
E3 1.000 1.000 1.000 B3 .000 .000 12.993

LATTICE PARAMETERS (ANGSTROM AND DEGREES)

A B C ALPHA BETA GAMMA VOLUME
4.76020 4.76020 12.99330 90.000 90.000 120.000 254.97670

SCELPHONO
rec variable meaning
• ∗ E expansion matrix E (IDIMxIDIM elements, input by rows: 9 reals (3D); 4 reals
(2D); 1 real (1D)

This keyword is specific for phonon dispersion calculations.

A supercell is constructed, as in the case of keyword SUPERCELL (see page 56), as linear
combination of the primitive cell unit vectors.
The atomic ordering of the resulting supercell is different with respect to the SUPERCELL
case. The first atoms in the list are the ones of the primitive cell, as required for phonon
dispersion calculations. Example:

TEST11 - MGO BULK

CRYSTAL
0 0 0
225
4.21
2
12 0. 0. 0.
8 0.5 0.5 0.5
SCELPHONO
2 0 0
0 2 0
0 0 2
TESTGEOM
END
======================= output - atoms in the small cell first ================
PRIMITIVE CELL - CENTRING CODE 1/0 VOLUME= 149.236922 - DENSITY 3.559 g/cm^3
A B C ALPHA BETA GAMMA
5.95383910 5.95383910 5.95383910 60.000000 60.000000 60.000000
*******************************************************************************
ATOMS IN THE ASYMMETRIC UNIT 5 - ATOMS IN THE UNIT CELL: 16
ATOM X/A Y/B Z/C
*******************************************************************************

58
 1 T 12 MG 0.000000000000E+00 0.000000000000E+00 0.000000000000E+00
2 T 8 O 2.500000000000E-01 2.500000000000E-01 2.500000000000E-01
3 T 12 MG 0.000000000000E+00 -5.000000000000E-01 0.000000000000E+00
4 F 12 MG 0.000000000000E+00 -5.000000000000E-01 -5.000000000000E-01
5 F 12 MG -5.000000000000E-01 0.000000000000E+00 0.000000000000E+00
6 F 12 MG -5.000000000000E-01 0.000000000000E+00 -5.000000000000E-01
7 F 12 MG -5.000000000000E-01 -5.000000000000E-01 0.000000000000E+00
8 T 12 MG -5.000000000000E-01 -5.000000000000E-01 -5.000000000000E-01
9 F 12 MG 0.000000000000E+00 0.000000000000E+00 -5.000000000000E-01
10 T 8 O 2.500000000000E-01 2.500000000000E-01 -2.500000000000E-01
11 F 8 O 2.500000000000E-01 -2.500000000000E-01 2.500000000000E-01
12 F 8 O 2.500000000000E-01 -2.500000000000E-01 -2.500000000000E-01
13 F 8 O -2.500000000000E-01 2.500000000000E-01 2.500000000000E-01
14 F 8 O -2.500000000000E-01 2.500000000000E-01 -2.500000000000E-01
15 F 8 O -2.500000000000E-01 -2.500000000000E-01 2.500000000000E-01
16 F 8 O -2.500000000000E-01 -2.500000000000E-01 -2.500000000000E-01

TEST11 - MGO BULK

CRYSTAL
0 0 0
225
4.21
2
12 0. 0. 0.
8 0.5 0.5 0.5
SUPERCEL
2 0 0
0 2 0
0 0 2
TESTGEOM

======================= output - atoms of same type first ================

PRIMITIVE CELL - CENTRING CODE 1/0 VOLUME= 149.236922 - DENSITY 3.559 g/cm^3
A B C ALPHA BETA GAMMA
5.95383910 5.95383910 5.95383910 60.000000 60.000000 60.000000
*******************************************************************************
ATOMS IN THE ASYMMETRIC UNIT 5 - ATOMS IN THE UNIT CELL: 16
ATOM X/A Y/B Z/C
*******************************************************************************
1 T 12 MG 0.000000000000E+00 0.000000000000E+00 0.000000000000E+00
2 T 12 MG 0.000000000000E+00 0.000000000000E+00 -5.000000000000E-01
3 F 12 MG 0.000000000000E+00 -5.000000000000E-01 0.000000000000E+00
4 F 12 MG 0.000000000000E+00 -5.000000000000E-01 -5.000000000000E-01
5 F 12 MG -5.000000000000E-01 0.000000000000E+00 0.000000000000E+00
6 F 12 MG -5.000000000000E-01 0.000000000000E+00 -5.000000000000E-01
7 F 12 MG -5.000000000000E-01 -5.000000000000E-01 0.000000000000E+00
8 T 12 MG -5.000000000000E-01 -5.000000000000E-01 -5.000000000000E-01
9 T 8 O -2.500000000000E-01 -2.500000000000E-01 2.500000000000E-01
10 T 8 O -2.500000000000E-01 -2.500000000000E-01 -2.500000000000E-01
11 F 8 O -2.500000000000E-01 2.500000000000E-01 2.500000000000E-01
12 F 8 O -2.500000000000E-01 2.500000000000E-01 -2.500000000000E-01
13 F 8 O 2.500000000000E-01 -2.500000000000E-01 2.500000000000E-01
14 F 8 O 2.500000000000E-01 -2.500000000000E-01 -2.500000000000E-01
15 F 8 O 2.500000000000E-01 2.500000000000E-01 2.500000000000E-01
16 F 8 O 2.500000000000E-01 2.500000000000E-01 -2.500000000000E-01

It can be used to generate a supercell for DISPERSION calculation only.

59
SUPERCON
The first step in geometry definition in CRYSTAL is the transformation of the cell from crystal-
lographic to primitive, according to the transformation matrices given in appendix A.5, when
the crystallographic cell is non-primitive.
From the point of view of CRYSTAL, the crystallographic cell is a supercell, in that case.
Using the keyword SUPERCON, the crystallographic cell is built from the primitive one,
before applying the expansion matrix.
See SUPERCEL, page 56 for expansion matrix input instructions.

Note - The easiest way to work on crystallographic cell is the following input:

SUPERCON
1. 0. 0.
0. 1. 0.
0. 0. 1.
\

The expansion matrix is the identity, leaving the crystallographic cell unmodified.

SYMMDIR
The symmetry allowed directions, corresponding to internal degrees of freedom are printed.
No input data are required.

SYMMOPS
Point symmetry operator matrices are printed in the Cartesian representation. No input data
are required.

SYMMREMO
All the point group symmetry operators are removed. Only the identity operator is left. The
wave function can be computed. No input data are required.
Warning: the CPU time may increase by a factor MVF (order of point-group), both in the
integral calculation and in the scf step. The size of the bielectronic integral file may increase
by a factor MVF2 .

TENSOR
rec variable meaning
• ∗ IORD order of the tensor (≤ 4)

This option evaluates and prints the non zero elements of the tensor of physical properties up
to order 4.

TESTGEOM
Execution stops after reading the geometry input block and printing the coordinates of the
atoms in the conventional cell. Neighbours analysis, as requested by the keyword NEIGH-
BOR, is not executed. The geometry input block must end with the keyword END or ENDG.
No other input blocks (basis set etc) are required.

TRASREMO
Point symmetry operators with fractional translation components are removed. It is suggested
to previously add the keyword ORIGIN (page 52), in order to minimize the number of sym-
metry operators with finite translation component. No input data are required.

60
USESAED
rec variable meaning
• ∗ δ(i),i=1,nsaed δ for each distortion

Given the symmetry allowed elastic distortion (SAED), (printed by the keyword MAKE-
SAED, page 45) δ for the allowed distortion are given in input.

61
2.2 Basis set input
Symmetry control

ATOMSYMM printing of point symmetry at the atomic positions 34 –

Basis set modification

CHEMOD modification of the electronic configuration 62 I
GHOSTS eliminates nuclei and electrons, leaving BS 64 I

Auxiliary and control keywords

CHARGED allows non-neutral cell 62 –

NOPRINT printing of basis set removed 64 –
PARAMPRT printing of parameters controlling code dimensions 52 –
PRINTOUT setting of printing options 53 I
SETINF setting of inf array options 54 I
SETPRINT setting of printing options 54 I
STOP execution stops immediately 56 –
SYMMOPS printing of point symmetry operators 60 –
END/ENDB terminate processing of basis set definition keywords –

Output of data on external units

GAUSS98 printing of an input file for the GAUSS94/98 package 64 –

Basis set input optional keywords

ATOMSYMM
See input block 1, page 34

CHARGED - charged reference cell

The unit cell of a periodic system must be neutral. This option forces the overall system to
be neutral even when the number of electrons in the reference cell is different from the sum
of nuclear charges, by adding a uniform background charge density to neutralize the charge in
the reference cell.
Warning - Do not use for total energy comparison.

CHEMOD - modification of electronic configuration

rec variable meaning
• ∗ NC number of configurations to modify
• ∗ LA label of the atom with new configuration
∗ CH(L),L=1,NS shell charges of the LA-th atom. The number NS of shells must coincide
with that defined in the basis set input.

The CHEMOD keyword allows modifications of the shell charges given in the basis set input,
which are used in the atomic wave function routines. The original geometric symmetry is
checked, taking the new electronic configuration of the atoms into account. If the number of
symmetry operators should be reduced, information on the new symmetry is printed, and the
program stops. No automatic reduction of the symmetry is allowed. Using the information
printed, the symmetry must be reduced by the keyword MODISYMM (input block 1, page
45).

62
See test 37. MgO supercell, with a Li defect. The electronic configuration of the oxygen nearest
to Li corresponds to O− , while the electronic configuration of those in bulk MgO is O2− . The
basis set of oxygen is unique, while the contribution of the two types of oxygen to the initial
density matrix is different.

END
Terminate processing of block 2, basis set, input. Execution continues. Subsequent input
records are processed, if required.

GAUSS98 - Printing of input file for GAUSS98 package

The keyword GAUSS98 writes in file GAUSSIAN.DAT an input deck to run Gaussian (Gaus-
sian 98 or Gaussian03) [52, 51]. The deck can be prepared without the calculation of the wave
function by entering the keyword TESTPDIM in input block 3 (page 101). For periodic
systems, coordinates and basis set for all the atoms in the reference cell only are written (no
information on translational symmetry).
If the keyword is entered many times, the data are overwritten. The file GAUSSIAN.DAT
contains the data corresponding to the last call.
The utility program gautocry reads basis set input in Gaussian format (as prepared by
http://www.emsl.pnl.gov/forms/basisform.html) and writes it in CRYSTAL format. No
input data required.

1. The route card specifies:

method HF
basis set GEN 5D 7F
type of job SP
geometry UNITS=AU GEOM=COORD

2. The title card is the same as in CRYSTAL input.

3. The molecule specification defines the molecular charge as the net charge in the reference
cell. If the system is not closed shell, the spin multiplicity is indicated with a string ”??”,
and must be defined by the user.

4. Input for effective core pseudopotentials is not written. In the route card PSEUDO =
CARDS is specified; the pseudopotential parameters used for the crystal calculation are
printed in the crystal output.

5. The scale factors of the exponents are all set to 1., as the exponents are already scaled.

6. the input must be edited when different basis sets are used for atoms with the same
atomic number (e.g., CO on MgO, when the Oxygen basis set is different in CO and in
MgO)

Warning: Only for 0D systems! The programs does not stop when the keyword GAUSS98
is entered for 1-2-3D systems. Coordinates and basis set of all the atoms in the primitive cell
are written, formatted, in file GAUSSIAN.DAT, following Gaussian scheme.
Warning If you run Gaussian 98 using the input generated by CRYSTAL with the keyword
GAUSS98, you do not obtain the same energy. There are 3 main differences between a standard
CRYSTAL run and a GAUSSIAN run.

1. CRYSTAL adopts by default bipolar expansion to compute coulomb integrals when the
two distributions do not overlap. To compute all 2 electron integrals exactly, insert
keyword NOBIPOLA in input block 3;

2. CRYSTAL adopts truncation criteria for Coulomb and exchange sums: to remove them,
in input block 3 insert:

63
 TOLINTEG
20 20 20 20 20

3. CRYSTAL adopts the NIST conversion factor bohr/Angstrom CODATA98:

1 Å= 0.5291772083 bohr
To modify the value, in input block 1 insert:

BOHRANGS
value_of_new_conversion_factor

GHOSTS
rec variable meaning
• ∗ NA number of atoms to be transformed into ghosts
• ∗ LA(L),L=1,NA label of the atoms to be transformed.

Selected atoms may be transformed into ghosts, by deleting the nuclear charge and the shell
electron charges, but leaving the basis set centred at the atomic position. The conventional
atomic number is set to zero., the symbol is “XX”.
If the system is forced to maintain the original symmetry (KEEPSYMM), all the atoms
symmetry related to the given one are transformed into ghosts.
Useful to create a vacancy (Test 37), leaving the variational freedom to the defective region
and to evaluate the basis set superposition error (BSSE), in a periodic system. The periodic
structure is maintained, and the energy of the isolated components computed, leaving the basis
set of the other one(s) unaltered. For instance, the energy of a mono-layer of CO molecules on
top of a MgO surface can be evaluated including the basis functions of the first layer of MgO,
or, vice-versa, the energy of the MgO slab including the CO ad-layer basis functions.
See test36 and test37.
Warning Do not use with ECP.
Warning The keyword ATOMREMO (input block 1, page 32) creates a vacancy, removing
nuclear charge, electron charge, and basis functions. The keyword GHOSTS creates a vacancy,
but leaves the basis functions at the site, so allowing better description of the electron density
in the vacancy.
Warning - Removal of nuclear and electron charge of the atoms selected is done after complete
processing of the input. They look still as ”atoms” in the printed output before that operation.

NOPRINT
Printing of basis set is removed. No input data required.

PARAMPRT - Printing of parametrized dimensions

See input block 1, page 52.

PRINTOUT - Setting of printing environment

See input block 1, page 53.

SETINF - Setting of INF values

See input block 1, page 54.

SETPRINT - Setting of printing options

See input block 1, page 54.

64
STOP
Execution stops immediately. Subsequent input records are not processed.

TEST[BS]
Basis set input is checked, then execution stops. Subsequent input records (block3) are not
processed.

SYMMOPS
See input block 1, page 60

Effective core pseudo-potentials - ECP

rec variable value meaning
• A PSN pseudo-potential keyword:
HAYWLC Hay and Wadt large core ECP.
HAYWSC Hay and Wadt small core ECP.
BARTHE Durand and Barthelat ECP.
DURAND Durand and Barthelat ECP.
INPUT free ECP - input follows.
if PSN = INPUT insert II
• ∗ ZNUC effective core charge (ZN in eq. 2.16).
M Number of terms in eq. 2.17
M0 Number of terms in eq. 2.18 for ` =0.
M1 Number of terms in eq. 2.18 for ` =1.
M2 Number of terms in eq. 2.18 for ` =2.
M3 Number of terms in eq. 2.18 for ` =3.
M4 Number of terms in eq. 2.18 for ` =4. - CRYSTAL09
insert M+M0+M1+M2+M3+M4 records II
• ∗ ALFKL Exponents of the Gaussians: αk` .
CGKL Coefficient of the Gaussians: Ck` .
NKL Exponent of the r factors: nk` .

Valence-electron only calculations can be performed with the aid of effective core pseudo-
potentials (ECP). The ECP input must be inserted into the basis set input of the atoms with
conventional atomic number > 200.

The form of pseudo-potential Wps implemented in CRYSTAL is a sum of three terms: a

Coulomb term (C), a local term (W0) and a semi-local term (SL):

Wps = C + W 0 + SL (2.15)

where:
C = −ZN /r (2.16)

M
X 2
W0 = rnk Ck e−αk r (2.17)
k=1

4 X̀M
X 2
SL = [ rnk` Ck` e−αk` r ]P` (2.18)
`=0 k=1

ZN is the effective nuclear charge, equal to total nuclear charge minus the number of electrons
represented by the ECP, P` is the projection operator related to the ` angular quantum number,
and M, nk , αk , M` , nk` , Ck` , αk` are atomic pseudo-potential parameters.

65
 1. Hay and Wadt (HW) ECP ([69, 67]) are of the general form 2.15. In this case, the NKL
value given in the tables of ref. [69, 67] must be decreased by 2 (2 → 0, 1 → -1, 0 → -2).

2. Durand and Barthelat (DB) ([8] - [45], [9], [10]), and Stuttgart-Dresden [63] ECPs contain
only the Coulomb term C and the semi-local SL term.

3. In Durand and Barthelat ECP the exponential coefficient α in SL depends only on ` (i.e.
it is the same for all the Mk terms).
3 M
X 2 X̀
SL = e−α` r [ rnk` Ck` ]P` (2.19)
`=0 k=1

The core orbitals replaced by Hay and Wadt large core and Durand-Barthelat ECPs are as
follows:

Li-Ne = [He]
Na-Ar = [Ne]
first series = [Ar]
second series = [Kr]
third series = [Xe]4f 14 .

The core orbitals replaced by Hay and Wadt small core ECPs are as follows:

K-Cu = [Ne]
Rb-Ag = [Ar] 3d10
Cs-Au = [Kr] 4d10 .

The program evaluates only those integrals for which the overlap between the charge distri-
bution ϕ0µ ϕgν (page 221) and the most diffuse Gaussian defining the pseudopotential is larger
than a given threshold Tps (the default value is 10−5 ). See also TOLPSEUD (Section 1.3).

Pseudopotential libraries
The following periodic tables show the effective core pseudo-potentials included as internal
data in the CRYSTAL code.

66
 HAY AND WADT LARGE CORE ECP. CRYSTAL92 DATA
------- ------------------
Na Mg Al Si P S Cl Ar
------------------------------------------------------
K Ca Sc Ti V Cr Mn Fe Co Ni Cu Zn Ga Ge As Se Br Kr
------------------------------------------------------
Rb Sr Y Zr Nb Mo Tc Ru Rh Pd Ag Cd In Sn Sb Te I Xe
------------------------------------------------------
Cs Ba Hf Ta W Re Os Ir Pt Au Hg Tl Pb Bi
------------------------------------------------------

HAY AND WADT SMALL CORE ECP. CRYSTAL92 DATA

-------------------------------------------------------
K Ca Sc Ti V Cr Mn Fe Co Ni Cu
-------------------------------------------------------
Rb Sr Y Zr Nb Mo Tc Ru Rh Pd Ag
-------------------------------------------------------
Cs Ba Hf Ta W Re Os Ir Pt Au
-------------------------------------------------------

DURAND AND BARTHELAT’S LARGE CORE ECP - CRYSTAL92 DATA

------ ------------------
Li Be B C N O F Ne
------ ------------------
Na Mg Al Si P S Cl Ar
-------------------------------------------------------
K Ca Sc Ti V Cr Mn Fe Co Ni Cu Zn Ga Ge As Se Br Kr
-------------------------------------------------------
Rb Y Ag In Sn Sb I
-------------------------------------------------------
Tl Pb Bi
-------------------------------------------------------

BARTHE, HAYWSC and HAYWLC pseudopotential coefficients and exponents are in-
serted as data in the CRYSTAL code. The data defining the pseudo-potentials where in-
cluded in CRYSTAL92, and never modified. The keyword INPUT allows entering updated
pseudo-potentials, when available. An a posteriori check has been possible for HAYWLC
and HAYWSC only, as the total energy of the atoms for the suggested configuration and
basis set has been published [69, 68]. Agreement with published atomic energies data is satis-
factory (checked from Na to Ba) for Hay and Wadt small core and large core pseudo-potentials,
when using the suggested basis sets. The largest difference is of the order of 10−3 hartree.
For Durand and Barthelat the atomic energies are not published, therefore no check has been
performed. The printed data should be carefully compared with those in the original papers.
The authors of the ECP should be contacted in doubtful cases.

Valence Basis set and pseudopotentials

Hay and Wadt ([69, 68]) have published basis sets suitable for use with their small and large core
pseudopotentials. and in those basis set the s and p gaussian functions with the same quantum
number have different exponent. It is common in CRYSTAL to use sp shells, where basis
functions of s and p symmetry share the same set of Gaussian exponents, with a consequent
considerable decrease in CPU time. The computational advantage of pseudopotentials over
all-electron sets may thus be considerably reduced.
Basis set equivalent to those suggested by Hay and Wadt can be optimized by using CRYSTAL
as an atomic package (page 72), or any atomic package with effective core pseudopotentials.
See Chapter 8.2 for general comments on atomic basis function optimization. Bouteiller et al

67
[16] have published a series of basis sets optimized for Durand and Barthelat ECPs.

Stuttgart-Dresden ECP (formerly STOLL and PREUSS ECP)

The most recent pseudopotential parameters, optimized basis sets, a list of references and
guidelines for the choice of the pseudopotentials can be found in:
http://www.theochem.uni-stuttgart.de/pseudopotentials/index.en.html
The clickable periodic table supplies, in CRYSTAL format, ECP to be used in CRYSTAL via
the INPUT keyword (basis set input, block2, page 65).
http://www.theochem.uni-stuttgart.de/pseudopotentials/clickpse.en.html

RCEP Stevens et al.

Conversion of Stevens et al. pseudopotentials An other important family of pseudopotentials
for the first-, second-, third-, fourth and fifth-row atoms of the periodic Table (excluding
the lanthanide series) is given by Stevens et al. [121, 122]. Analytic Relativistic Compact
Effective Potential (RCEP) are generated in order to reproduce the ”exact” pseudo-orbitals
and eigenvalues as closely as possible. The analytic RCEP expansions are given by:
X 2
r2 Vl (r) = Alk rnl,k e−Blk r
k

An example of data for Ga atom (Table 1, page 616 of the second paper) is:

Alk nlk Blk

Vd -3.87363 1 26.74302
Vs−d 4.12472 0 3.46530
260.73263 2 9.11130
-223.96003 2 7.89329
Vp−d 4.20033 0 79.99353
127.99139 2 17.39114

The corresponding Input file for the CRYSTAL program will be as follows:

INPUT
21. 1 3 2 0 0
26.74302 -3.87363 -1
3.46530 4.12472 -2
9.11130 260.73263 0
7.89329 -223.96003 0
79.99353 4.20033 -2
17.39114 127.99139 0
Note that for the r-exponent (nlk ), -2 has been subtracted to the value given in their papers,
as in the case of Hay and Wadt pseudopotentials.

68
2.3 Computational parameters, hamiltonian,
SCF control
Single particle Hamiltonian

RHF Restricted Closed Shell 96 –

UHF Unrestricted Open Shell 102 –
DFT DFT Hamiltonian 102 –
SPIN spin-polarized solution 75 –
Choice of the exchange-correlation functionals
CORRELAT correlation functional 74 I
LYP Lee-Yang-Parr [78] (GGA)
P86 Perdew 86 [88] (LDA)
PBE Perdew-Becke-Ernzerhof [90] (GGA)
PBESOL GGA. PBE functional revised for solids [96]
PWGGA Perdew-Wang 91 (GGA)
PWLSD Perdew-Wang 92 [92, 93, 91] (GGA)
PZ Perdew-Zunger [94] (LDA)
VBH von Barth-Hedin [127] (LDA)
WL GGA - Wilson-Levy [133]
VWN Vosko,-Wilk-Nusair [128] (LDA)
EXCHANGE exchange functional 75 I
BECKE Becke [13] (GGA)
LDA Dirac-Slater [34] (LDA)
PBE Perdew-Becke-Ernzerhof [90] (GGA)
PBESOL GGA. PBE functional revised for solids [96]
PWGGA Perdew-Wang 91 (GGA)
SOGGA second order GGA. [136]
VBH von Barth-Hedin [127] (LDA)
WCGGA GGA - Wu-Cohen [135]
Global keywords: exchange+correlation
B3PW B3PW parameterization 76 –
B3LYP B3LYP parameterization 76 –
PBE0 Adamo and Barone [2] 76
SOGGAXC see [136] 76
Custom made dft exchange-local exchange
HYBRID hybrid mixing 76 I
NONLOCAL local term parameterization 76 I

69
 Numerical accuracy control
ANGULAR definition of angular grid I
BATCHPNT grid point grouping for integration I
[BECKE] selection of Becke weights (default) –
[CHUNKS] max n. points in a batch for numerical int. I
LGRID ”large” predefined grid I
RADSAFE safety radius for grid point screening I
RADIAL definition of radial grid I
SAVIN selection of Savin weights –
TOLLDENS density contribution screening 6 I
TOLLGRID grid points screening 14 I
XLGRID ”extra large” predefined grid I
XXLGRID ”extra extra large” predefined grid I
Atomic parameters control
RADIUS customized atomic radius 81 I
FCHARGE customized formal atomic charge 81 I
Auxiliary
PRINT extended printing
END close DFT input block

Numerical accuracy and computational parameters control

BIPOLAR Bipolar expansion of bielectronic integrals 73 I

BIPOSIZE size of coulomb bipolar expansion buffer 73 I
EXCHSIZE size of exchange bipolar expansion buffer 73 I
INTGPACK Choice of integrals package 0 92 I
NOBIPCOU Coulomb bielectronic integrals computed exactly 95 –
NOBIPEXCH Exchange bielectronic integrals computed exactly 95 –
NOBIPOLA All bielectronic integrals computed exactly 95 –
POLEORDR Maximum order of multipolar expansion 4 96 I
TOLINTEG Truncation criteria for bielectronic integrals 6 6 6 6 12 102 I
TOLPSEUD Pseudopotential tolerance 6 102 I

Type of execution
MPP MPP execution 93

Type of run

ATOMHF Atomic wave functions 72 I

SCFDIR SCF direct (mono+biel int computed) 96 –
EIGS S(k) eigenvalues - basis set linear dependence check 81 –
FIXINDEX Reference geometry to classify integrals 86 –

Basis set - AO occupancy

FDAOSYM f and d degeneracies analysis 83 I

FDAOCCUP f and d orbital occupation guess 84 I
GUESDUAL Density matrix guess - different Basis set 89 I

Integral file distribution

BIESPLIT writing of bielectronic integrals in n files n = 1 ,max=10 73 I

MONSPLIT writing of mono-electronic integrals in n file n = 1 , max=10 93 I

Numerical accuracy control and convergence tools

70
ANDERSON Fock matrix mixing 71 I
BROYDEN Fock matrix mixing 73 I
FMIXING Fock/KS matrix (cycle i and i-1) mixing 0 87 I
LEVSHIFT level shifter no 92 I
MAXCYCLE maximum number of cycles 50 93 I
SMEAR Finite temperature smearing of the Fermi surface no 99 I
TOLDEE convergence on total energy 5 102 I
TOLDEP convergence on density matrix 16 102 I

Initial guess

EIGSHIFT alteration of orbital occupation before SCF no 82 I

EIGSHROT rotation of the reference frame no 82 I
GUESSP density matrix from a previous run 91 –
GUESSPAT superposition of atomic densities 91 –

Spin-polarized system

ATOMSPIN setting of atomic spin to compute atomic densities 72 I

BETALOCK beta electrons locking 73 I
SPINLOCK spin difference locking 101 I
SPINEDIT editing of the spin density matrix used as SCF guess 100 I

Auxiliary and control keywords

END terminate processing of block3 input –

KSYMMPRT printing of Bloch functions symmetry analysis 92 –
NEIGHBOR number of neighbours to analyse in PPAN 51 I
PARAMPRT output of parameters controlling code dimensions 52 –
PRINTOUT setting of printing options 53 I
NOSYMADA No Symmetry Adapted Bloch Functions 95 –
SYMADAPT Symmetry Adapted Bloch Functions (default) 101 –
SETINF setting of inf array options 54 I
SETPRINT setting of printing options 54 I
STOP execution stops immediately 56 –
TESTPDIM stop after symmetry analysis 101 –
TEST[RUN] stop after integrals classification and disk storage estimate 101 –

Output of data on external units

NOFMWF wave function formatted output not written in file fort.98. 95 –

SAVEWF wave function data written every two SCF cycles 97 –

Post SCF calculations

POSTSCF post-scf calculations when convergence criteria not satisfied 96 –
EXCHGENE exchange energy evaluation (spin polarized only) 83 –
GRADCAL analytical gradient of the energy 87 –
PPAN population analysis at the end of the SCF no 96

Computational parameters, hamiltonian, SCF control

optional keywords
ANDERSON
Anderson’s method [3], as proposed by Hamann [65], is applied. No input data are required.

71
See test49 dft, a metallic Lithium 5 layers slab, PWGGA Hamiltonian.

ATOMHF - Atomic wave function calculation

The Hartree-Fock atomic wave functions for the symmetry unique atoms in the cell are com-
puted by the atomic program [6]. Full input (geometry, basis set, general information, SCF) is
processed. No input data are required. The density matrix, constructed from a superposition
of atomic densities, is computed and written on Fortran unit 9, along with the wave function
information. The crystal program then stops. It is then possible to compute charge density
(ECHG) and classical electrostatic potential (CLAS) maps by running the program proper-
ties. This option is an alternative to the keyword PATO in the program properties (page
180), when the calculation of the periodic wave function is not required. The atomic wave
function, eigenvalues and eigenvectors, can be printed by setting the printing option 71.

1. The atomic basis set may include diffuse functions, as no periodic calculation is carried
out.

2. A maximum of two open shells of different symmetry (s, p, d) are allowed in the electronic
configuration. In the electronic configuration given in input the occupation number of
the shells must follow the rules given in Section 1.2.

3. For each electronic configuration, the highest multiplicity state is computed. Multiplicity
cannot be chosen by the user.

Warning: DFT wave function for isolated atoms can not be computed.

ATOMSPIN - Setting of atomic spin

rec variable meaning
• ∗ NA number of atoms to attribute a spin
• ∗ LA,LS(LA),L=1,NA atom labels and spin (1, 0, -1)

The setting of the atomic spins is used to compute the density matrix as superposition of
atomic densities (GUESSPAT must be SCF initial guess); it does not work with GUESSP).
The symmetry of the lattice may be reduced by attributing a different spin to geometrically
symmetry related atoms. In such cases a previous symmetry reduction should be performed
using the MODISYMM keyword. The program checks the symmetry taking the spin of the
atoms into account. If the spin pattern does not correspond to the symmetry, the program
prints information on the new symmetry, and then stops.
The formal spin values are given as follows:

1 atom spin is taken to be alpha;

0 atom spin is irrelevant;
-1 atom spin is taken to be beta.

In a NiO double-cell (four atoms, Ni1 Ni2 O1 O2) we might use:

atom Ni1 Ni2

spin 1 1 for starting ferromagnetic solutions: ↑ ↑
spin 1 -1 for starting anti ferromagnetic solutions: ↑ ↓

SPINLOCK forces a given nα − nβ electrons value: to obtain a correct atomic spin density
to start SCF process, the atomic spins must be set even for the ferromagnetic solution.
See test 30 and 31.

72
BETALOCK - Spin-polarized solutions
rec variable meaning
• ∗ INF97 nβ electrons
∗ INF98 number of cycles the nβ electrons is maintained

The total number of of β electrons at all k points can be locked at the input value. The number
of α electrons is locked to (N + INF95)/2, where N is the total number of electrons in the
unit cell. INF95 must be odd when the number of electrons is odd, even when the number of
electrons is even. See SPINLOCK for alternative way to define spin setting.

BIESPLIT - Splitting of large bielectronic integral files

rec variable meaning
• ∗ NFILE number of files to be used 1 (max 10)

Very compact crystalline systems, and/or very diffuse basis functions and/or very tight toler-
ances can produce billions integrals to be stored. The storage of bielectronic integrals can be
avoided by running the direct SCF code scfdir rather than the standard SCF, at the expenses
of a certain amount of CPU time.
When the standard SCF code is used, distributing the integrals on several disk files can improve
performance.

BIPOLAR - Bipolar expansion approximation control

rec variable meaning
• ∗ ITCOUL overlap threshold for Coulomb 14
∗ ITEXCH overlap threshold for exchange 10

The bipolar approximation is applied in the evaluation of the Coulomb and exchange integrals
(page 225). ITCOUL and ITEXCH can be assigned any intermediate value between the default
values (14 and 10) (see page 225) and the values switching off the bipolar expansion (20000
and 20000).

BIPOSIZE -Size of buffer for Coulomb integrals bipolar expansion

rec variable meaning
• ∗ ISIZE size of the buffer in words

Size (words) of the buffer for bipolar expansion of Coulomb integrals (default value 100000.
The size of the buffer is printed in the message:

BIPO BUFFER LENGTH (WORDS) = XXXXXXX

or
COULOMB BIPO BUFFER TOO SMALL - TO AVOID I/O SET BIPOSIZE = XXXXXX

BROYDEN
rec variable meaning
• ∗ W0 W0 parameter in Anderson’s paper [75]
∗ IMIX percent of Fock/KS matrices mixing when Broyden method is switched on
∗ ISTART SCf iteration after which Broyden method is active (minimum 2)

73
A modified Broyden [18] scheme, following the method proposed by Johnson [75], is applied
after the ISTART SCF iteration, with IMIX percent of Fock/KS matrices simple mixing. The
value of % mixing given in input after the keyword FMIXING is overridden by the new one.
Level shifter should be avoided when Broyden method is applied.
Suggested values:

FMIXING
80
BROYDEN
0.0001 50 2

See test50 dft, a metallic Lithium 5 layers slab, PWGGA Hamiltonian.

END
Terminate processing of block 3,(last input block). Execution continues. Subsequent input
records are not processed.

DFT
The Kohn-Sham [76, 74] DFT code is controlled by keywords, that must follow the general
keyword DFT, in any order. These keywords can be classified into four groups:

1 Choice of the exchange-correlation functionals

2 Integration grid and numerical accuracy control (optional)

3 DF energy gradient (optional)

4 Atomic parameters (optional)

The DFT input block ends with the keyword END or ENDDFT. Default values are sup-
plied for all computational parameters. Choice of exchange and/or correlation potential is
mandatory.

1. Choice of the exchange-correlation functionals

EXCHANGE and CORRELAT keywords, each followed by an alpha-numeric record, allow
the selection of the exchange and correlation functionals.
If the correlation potential is not set (keyword CORRELAT), an exchange-only potential is
used in the Hamiltonian. If the exchange potential is not set (keyword EXCHANGE), the
Hartree-Fock potential is used.

74
CORRELAT Correlation Potential (default: no correlation).
Insert one of the following keywords II
LYP GGA. Lee-Yang-Parr [78]
P86 GGA. Perdew 86 [88]
PBE GGA. Perdew-Burke-Ernzerhof [90]
PBESOL GGA. PBE functional revised for solids [96]
PWLSD LSD. Perdew-Wang parameterization of the Ceperley-Alder free electron
gas correlation results [91]
PWGGA GGA. Perdew-Wang [95]
PZ LSD. Perdew-Zunger parameterization of the Ceperley-Alder free electron
gas correlation results [94]
VBH LSD. von Barth-Hedin [127]
WL GGA - Wilson-Levy [133]
VWN LSD. Vosko-Wilk-Nusair parameterization of the Ceperley-Alder free elec-
tron gas correlation results [128]

EXCHANGE Exchange potential (default: Hartree-Fock exchange).

Insert one of the following keywords II
BECKE GGA. Becke [13]
LDA LSD. Dirac-Slater [34]
PBE GGA. Perdew-Becke-Ernzerhof [90]
PBESOL GGA. PBE functional revised for solids [96]
PWGGA GGA. Perdew-Wang [95]
SOGGA second order GGA. [136]
VBH LSD. von Barth-Hedin [127]
WCGGA GGA - Wu-Cohen [135]

All functionals are formulated in terms of total density and spin density. Default is total
density. To use functionals of spin density insert the keyword SPIN.

SPIN unrestricted spin DF calculation (default: restricted)

It is also possible to incorporate part of the exact Hartree-Fock exchange into the exchange
functional through the keyword HYBRID. Any mixing (0-100) of exact Hartree-Fock and
DFT exchange can be used.
NONLOCAL allows modifying the relative weight of the local and non-local part both in
the exchange and the correlation potential with respect to standard definition of GGA type
potentials.

75
HYBRID Hybrid method - 1 record follows:
• ∗A Fock exchange percentage (default 100.)

NONLOCAL setting of non-local weighting parameters - 1 record follows:

• ∗B exchange weight of the non-local part of exchange
C weight of the non-local correlation

B3PW Becke’s 3 parameter functional [14] combined with the non-local correla-
tion PWGGA [89, 92, 93, 91]

B3LYP Becke’s 3 parameter functional [14] combined with the non-local correla-
tion LYP

PBE0 Adamo and Barone [2]

SOGGAXC see [136]

B3PW, B3LYP, PBE0 and SOGGAXC are global keywords, defining hybrid exchange-
correlation functionals completely. They replace the following sequences:

B3PW B3LYP
corresponds to the sequence: corresponds to the sequence:
EXCHANGE EXCHANGE
BECKE BECKE
CORRELAT CORRELAT
PWGGA LYP
HYBRID HYBRID
20 20
NONLOCAL NONLOCAL
0.9 0.81 0.9 0.81
PBE0
corresponds to the sequence: SOGGAXC
EXCHANGE corresponds to the sequence:
PBE EXCHANGE
CORRELAT SOGGA
PBE CORRELAT
HYBRID PBE
25

B3LYP in CRYSTAL is based on the ’exact’ form of the Vosko-Wilk-Nusair correlation poten-
tial (corresponds to a fit to the Ceperley-Alder data). In the original paper [128]) it is reported
as functional V, which is used to extract the ’local’ part of the LYP correlation potential.
The Becke’s 3 parameter functional can be written as:
Exc = (1 − A) ∗ (ExLDA + B ∗ ExBECKE ) + A ∗ ExHF + (1 − C) ∗ EcV W N + C ∗ EcLY P
A, B, and C are the input data of HYBRYD and NONLOCAL.

Examples of possible selection of the correlation and exchange functionals are:

76
exchange correlation
—– PWGGA Hartree-Fock exchange, non local Perdew-Wang correlation.
LDA VWN probably the most popular LDA formulation
VBH VBH was the most popular LDA scheme in the early LDA solid state
applications (1975-1985).
PWGGA PWGGA ------
BECKE LYP ------

2. Integration grid and numerical accuracy control

No input data are required: Becke weights are chosen by default, as well as a set of safe values
for the computational parameters of integration.
The generation of grid points in CRYSTAL is based on an atomic partition method, originally
developed by Becke [11] for molecular systems and then extended to periodic systems [124].
Each atomic grid consists of a radial and an angular distribution of points. Grid points are
generated through a radial and an angular formula: Gauss-Legendre radial quadrature and
Lebedev two-dimensional angular point distribution are used.
Lebedev angular grids are classified according to progressive accuracy levels, as given in the
following table:
LEV CR98 ` Nang LEV CR98 ` Nang
1 1 9 38 16 53 974
2 2 11 50 17 59 1202
3 13 74 * 18 65 1454 Index of Lebedev accuracy levels
4 15 86 19 71 1730 LEV: Lebedev accuracy level
5 3 17 110 20 77 2030 CR98: corresponding index in CRYSTAL98
6 19 146 21 83 2354 `: maximum quantum number of spher-
7 21 170 22 89 2702 ical harmonics used in Lebedev
8 4 23 194 23 95 3074 derivation
9 25 230 * 24 101 3470 Nang : number of angular points generated
10 5 27 266 * 25 107 389 per radial point
∗
11 6 29 302 26 113 4334 : sets with negative weights, to be
12 31 350 27 119 4802 avoided
13 7 35 434 28 125 5294
14 41 590 29 131 5810
15 47 770

If one Lebedev accuracy level is associated with the whole radial range, the atomic grid is
called unpruned, or uniform. In order to reduce the grid size and maintain its effectiveness, the
atomic grids of spherical shape can be partitioned into shells, each associated with a different
angular grid. This procedure, called grid pruning, is based on the assumption that core electron
density is usually almost spherically symmetric, and surface to be sampled is small.
Also, points far from the nuclei need lower point density, as associated with relatively small
weights, so that more accurate angular grids are mostly needed within the valence region than
out of it.
The choice of a suitable grid is crucial both for numerical accuracy and need of computer
resources.
Different formulae have been proposed for the definition of grid point weights. In CRYSTAL
Becke and Savin weights are available; Becke weights are default, and provide higher accuracy.

[BECKE] Becke weights [12]. Default choice.

SAVIN Savin weights [115]

77
A default grid is available in CRYSTAL, however the user can redefine it by the following
keywords:

RADIAL Radial integration information

rec variable meaning
• ∗ NR number of intervals in the radial integration [default 1]
• ∗ RL(I),I=1,NR radial integration intervals limits in increasing sequence [default 4.0]
(last limit is set to ∞)
• ∗ IL(I),I=1,NR number of points in the radial quadrature in the I-th interval
[default 55].

ANGULAR Angular integration information

rec variable meaning
• ∗ NI number of intervals in the angular integration [default 1]
• ∗ AL(I),I=1,NI upper limits of the intervals in increasing sequence. The last limit must
be 9999.0 [default 9999.0]
• ∗ LEV(I),I=1,NI accuracy level in the angular integration over the I-th interval; positive
for Lebedev level (see Lev in page 77) [default 13]

The default grid is a pruned (55,434) grid, having 55 radial points and a maximum number of
434 angular points in regions relevant for chemical bonding. Each atomic grid is split into ten
shells with different angular grids.
This grid is good enough for either single-point energy calculations or medium-accuracy
geometry optimizations. Due to the large pruning, the cost of the calculation is modest.
Default grid - corresponds to the sequence:
RADIAL Keyword to specify the radial grid
1 Number of intervals in the radial part
4.0 Radial integration limits of the i-th interval
55 Number of radial points in the i-th interval
ANGULAR Keyword to specify the angular grid
10 Number of intervals in the angular part
0.4 0.6 0.8 0.9 1.1 2.3 2.4 2.6 2.8 9999.0 Angular integration limits of the i-th interval
1 2 5 8 11 13 11 8 5 1 Angular grid accuracy level of the i-th interval

Information on the size of the grid, grid thresholds, and radial (angular) grid is reported in the
CRYSTAL output with the following format:

SIZE OF GRID= 40728

BECKE WEIGHT FUNCTION
RADSAFE = 2.00
TOLERANCES - DENSITY:10**- 6; POTENTIAL:10**- 9; GRID WGT:10**-14

RADIAL INTEGRATION - INTERVALS (POINTS,UPPER LIMIT): 1( 55, 4.0*R)

ANGULAR INTEGRATION - INTERVALS (ACCURACY LEVEL [N. POINTS] UPPER LIMIT):

1( 1[ 38] 0.4) 2( 2[ 50] 0.6) 3( 5[ 110] 0.8) 4( 8[ 194] 0.9)
5( 11[ 302] 1.1) 6( 13[ 434] 2.3) 7( 11[ 302] 2.4) 8( 8[ 194] 2.6)
9( 5[ 110] 2.8) 10( 1[ 38]9999.0)

Two more pre-defined grids are available which can be selected to improve accuracy by inputing
the following global keywords:

LGRID Large grid

78
Global keyword to choose a larger grid than default, corresponding to the sequence:
RADIAL
1
4.0
75
ANGULAR
5
0.1667 0.5 0.9 3.05 9999.0
2 6 8 13 8

The large grid is a pruned (75,434) grid, having 75 radial points and a maximum number
of 434 angular points in the region relevant for chemical bonding. Five shells with different
angular points are adopted to span the radial range as proposed by Gill et al. [58]. Due to
a larger number of radial points and less aggressive pruning, this grid gives more accurate
results than the default grid. It is recommended for high-accuracy energy calculations and
geometry optimizations. It is also recommended for periodic systems containing second-row
and third-row atoms (transition metals).

XLGRID Extra large grid

Global keyword to choose an even larger integration grid, corresponding to the sequence:
RADIAL
1
4.0
75
ANGULAR
5
0.1667 0.5 0.9 3.5 9999.0
4 8 12 16 12

The extra-large grid is a pruned (75,974) grid, consisting of 75 radial points and 974 angular
points in the region of chemical interest. This is a very accurate grid and is recommended
when numerical derivatives of energy or related properties (i.e. spontaneous polarization) and
gradients have to be computed (e.g. bulk modulus, elastic constants, piezoelectric tensor,
ferroelectric transitions). It is also recommended for heavy atoms (fourth-row and heavier).

XXLGRID Extra extra large grid

Very large grid use for benchmark calculations. It corresponds to:

RADIAL
1
4.0
99
ANGULAR
5
0.1667 0.5 0.9 3.5 9999.0
6 10 14 18 14

DISTGRID developers only

This forces the code to distribute the DFT grid across the available processors. Only parallel
jobs are affected.

79
REPLGRID developers only

This forces the code to replicate the DFT grid on the available processors. Only parallel jobs
are affected.
(This directive at present has no effect - the defaults is a replicated grid. However I am thinking
of changing the default for MPP jobs to distgrid, and then replgrid will have a use for certain
medium sized jobs where load balancing is an issue).IJB

Unpruned grids
To switch from a pruned grid to the corresponding unpruned grid, only one shell must be
defined in the radial part and the same angular accuracy is used everywhere. The use of
unpruned grids increases the cost of the calculations by about 50-60% with respect to the
pruned grid.
For example, to transform the default grid to the corresponding unpruned grid input the
following data:
ANGULAR
1
9999.0
13

Numerical accuracy and running time are also controlled by the following keywords:

TOLLGRID
• ∗ IG DFT grid weight tolerance [default 14]
TOLLDENS
• ∗ ID DFT density tolerance [default 6]

The DFT density tolerance ID controls the level of accuracy of the integrated charge density
Nel (number of electron per cell):
Z
g+g0 0
X X
Nel = ρ(r)dr = Pµ,ν w(ri )ϕgµ (ri )ϕgν (ri )
cell µ,ν,g,l i

all contributions where |ϕµ (ri )| < 10−ID or |ϕν (ri )| < 10−ID are neglected (see Chapter 11.11
for notation). The default value of ID is 6.
Grid points with integration weights less than 10−IG are dropped. The default value of IG is
14.

RADSAFE
• ∗ RAD for developers only [default 2]

BATCHPNT
• ∗ BATCH average number of points in a batch for numerical integration [default 100]

Default value of BATCH is 100. In the calculation of the exchange-correlation contribution to

the Kohn-Sham hamiltonian matrix elements and energy gradients, the grid is partitioned into
batches of points as suggested by Ahlrichs [125]. However, in CRYSTAL the number of points
per batch is not constant, as it depends on point density, so that BATCH does not correspond
to the maximun number of points in a batch. As a consequence, in special cases, memory
requirement may become huge and cause problems in dynamic allocation at running time.
When the program runs out of memory, it stops with the following error message:

ERROR *** sub_name *** array_name ALLOCATION

80
where array_name is one of the following:
DF0 KSXC1 KSXC2 KSXC2Y KSXC2Z DFXX DFYY DFZZ DFXY DFYZ DFXZ
RHO FRHO AXJ,AYJ,AZJ,VGRID GRAZ GRAY GRAZ
In these cases it is recommended that the value of BATCH be reduced, although this may
result in some degree of inefficiency (minimum value: 1).

CHUNKS
• ∗ NCHU maximum number of points allowed in a batch for numerical integration
[default 10000000]

Does not work for MPP run.

3. DF energy gradient

[NEWTON]

The current default when computing DFT analytical gradients in CRYSTAL is to include
weight derivatives. Weight derivatives are mandatory when low quality grids are adopted.

4. Atomic parameters
The radius attributed to each atom for the integration is computed from the nuclear charge
and the net charge. It is possible to enter for selected atoms a given atomic radius or a formal
charge.
No check on symmetry requirements is performed. If the selected atoms has other atoms
symmetry related, radius (or charge) of those atoms must be defined. The keyword ATOM-
SYMM inserted in the first input block (geometry) prints the set of atoms symmetry related.

• A RADIUS
• ∗ NUMAT number of atoms selected
insert NUMAT records II
• ∗ LB label of the atom
RAD(LB) radius (Å) attributed to the atom

• A FCHARGE
• ∗ NUMAT number of atoms selected
insert NUMAT records II
• ∗ LB label of the atom
FCH(LB) formal charge attributed to the atom

EIGS - Check of basis set linear dependence

In order to check the risk of basis set linear dependence, it is possible to calculate the eigenvalues
of the overlap matrix. Full input (geometry, basis set, general information, SCF) is processed.
No input data are required. The overlap matrix in reciprocal space is computed at all the k-
points generated in the irreducible part of the Brillouin zone, and diagonalized. The eigenvalues
are printed.
The higher the numerical accuracy obtained by severe computational conditions, the closer
to 0 can be the eigenvalues without risk of numerical instabilities. Negative values indicate
numerical linear dependence. The crystal program stops after the check (even if negative
eigenvalues are not detected).
The Cholesky reduction scheme [77], adopted in the standard SCF route, requires linearly
independent basis functions.
MPP doesn’ support EIGS.

81
EIGSHIFT - Alteration of orbital occupation before SCF
rec variable meaning
• ∗ NORB number of elements to be shifted
> 0 level shift of diagonal elements only
< 0 off-diagonal level shift
insert NORB records
if NORB > 0
• ∗ IAT label of the atom
ISH sequence number of the shell in the selected atom Basis Set (as given in Basis
Set input)
IORB sequence number of the AO in the selected shell (see Section 1.2, page 21).
SHIF1 α (or total, if Restricted) Fock/KS matrix shift
[SHIF2 β Fock/KS matrix shift - spin polarized only ]
if NORB < 0
• ∗ IAT label of the atom
ISH sequence number of the shell in the selected atom Basis Set
IORB1 sequence number of the AO in the selected shell
IORB2 sequence number of the AO in the selected shell
SHIF1 α (or total, if Restricted) Fock/KS matrix shift
[SHIF2 β Fock/KS matrix shift - spin polarized only]

Selected diagonal Fock/KS matrix elements can be shifted upwards when computing the initial
guess, to force orbital occupation. This option is particularly useful in situations involving d
orbital degeneracies which are not broken by the small distortions due to the crystal field, but
which are broken by some higher-order effects (e.g. spin-orbit coupling). The EIGSHIFT
option may be used to artificially remove the degeneracy in order to drive the system to a
stable, non-metallic solution. The eigenvalue shift is removed after the first SCF cycle.
If the shift has to be applied to matrix elements of atoms symmetry related, the input data
must be repeated as many times as the atoms symmetry related.
Example: KCoF3 (test 38). In the cubic environment, two β electrons would occupy the three-
fold degenerate t2g bands. A state with lower energy is obtained if the degeneracy is removed
by a tetragonal deformation of the cell (keyword ELASTIC), and the dxy orbital (see page
21 for d orbital ordering) is shifted upwards by 0.3 hartree.
Warning EIGSHIFT acts on the atoms as specified in input. If there are atoms symmetry-
related to the chosen one, hamiltonian matrix elements shift is not applied to the others. The
programs checks the symmetry compatibility, and, if not satisfied, stops execution. The matrix
elements of all the atoms symmetry-related must be shifted, if the symmetry of the systems
must be kept
The keyword ATOMSYMM (input block 1, page 34) prints information on the atoms sym-
metry related in the cell.

EIGSHROT
Consider now the case of CoF2 . The first six neighbors of each Co2+ ion form a slightly
distorted octahedron (2 axial and 4 equatorial equivalent distances); also in this case, then, we
are interested in shifting upwards the dxy orbital, in order to drive the solution towards the
following occupation:
α: all five d orbitals
β: dxz and dyz
The principal axis of the CoF6 octahedron, however, is not aligned along the z direction, but lies
in the xy plane, at 450 from the x axis. The cartesian reference frame must then be reoriented
before the shift of the dxy orbital.
To this aim the option EIGSHROT must be used. The reoriented frame can be specified in
two ways, selected by a keyword:

82
 rec variable meaning
• MATRIX keyword - the rotation matrix R is provided
• ∗ R11 R12 R13 first row of the matrix.
• ∗ R21 R22 R23 second row of the matrix.
• ∗ R31 R32 R33 third row of the matrix.
or
• ATOMS keyword - the rotation is defined by three atoms of the crystal
• ∗ IA label of first atom in the reference cell
AL,AM,AN indices (direct lattice, input as reals) of the cell where the first atom
is located
• ∗ IB label of second atom in the reference cell
BL,BM,BN indices (direct lattice, input as reals) of the cell where the second
atom is located
• ∗ IC label of third atom in the reference cell
CL,CM,CN indices (direct lattice, input as reals) of the cell where the third atom
is located
insert EIGSHIFT input records (Section 2.3, page 82)

When the rotation is defined by three atoms, the new reference frame is defined as follows :
Z-axis from atom 2 to atom 1
X-axis in the plane defined by atoms 1-2-3
Y-axis orthogonal to Z- and X-axis

Notice that the wave function calculation is performed in the original frame: the aim of the
rotation is just to permit a shift of a particular orbital. An equivalent rotation of the eigen-
vectors can be obtained in properties by entering the keyword ROTREF, so allowing AO
projected Density of States according to the standard orientation of the octahedron.
Example:

CoF2 example

EXCHGENE - Exchange energy calculation

In RHF calculations Coulomb and exchange integrals are summed during their calculation,
and there is no way to separate the exchange contribution to the total energy. In UHF/ROHF
calculations, this option allows the independent calculation and printing of the exchange con-
tribution to the total energy. See equation 11.19, page 224.
No input data are required. See tests 29, 30, 31, 38.

EXCHSIZE - Size of buffer for exchange integrals bipolar expansion

rec variable meaning
• ∗ ISIZE size of the buffer in words

Size (words) of the buffer for bipolar expansion of exchange integrals (default value 100000).
The size of the buffer is printed in the message:

EXCH. BIPO BUFFER: WORDS USED = XXXXXXX

or
EXCH. BIPO BUFFER TOO SMALL - TO AVOID I/O SET EXCHSIZE = XXXXXX

FDAOSYM - f and d degeneracies analysis

rec variable meaning
• ∗ NA number of atoms for which the AOs mixing check is performed
• ∗ ISCAT(J),J=1,NA atomic label (output order) of atoms to be checked
This keyword performs the symmetry analysis that permits to know if AOs belonging to the
same shell are mixed or not by the symmetry operators that don’t move the atom, to which

83
they belong to. This analysis is implemented for d and f shell types only. The FDAOSYM
keyword must be inserted in the third block of the CRYSTAL input.
In the output, the subgroup of operators that do not move the atom is performed first (ATOM-
SYMM keyword). Then the AOs are listed with the indication of mixing with other AOs (if
any).
This keyword is useful for partially occupied shells (d or f ). AOs that mix will form a sym-
metry constrained degeneracy subset. If n AOs of the shell mix generating these subsets, and
m<n (open shell case) electrons are supposed to populate the shell, then Jan-Teller symmetry
breaking should be taken into account. If a guess AOs occupation is defined with FDOCCUP
keyword or forced with EIGSHIFT keyword, information obtained by FDAOSYM can indicate
which AOs will have the same occupation.
Example. Ce2 O3 bulk. Structure 4d shell is completely filled and 4f shell contains one elec-
tron. Suppose Ce atom is labeled 1. Information produced by FDAOSYM indicates which
AOs will have the same occupation. If combined with TESTRUN such information is obtained
at zero cost and then a new input with FDOCCUP or EIGSHIFT keywords can be run. The
input is

FDAOSYM
1
1
TESTRUN

Output obtained contains also d shell information, here just f part is reported.

ANALYSIS FOR SHELL TYPE F

COMPONENT MIXES WITH SYMMETRY OPERATORS
1
2 3 2 3 10 12
3 2 2 3 10 12
4 5 2 3 10 12
5 4 2 3 10 12
6
7

That is: AOs 1, 6 and 7 do not mix with any other AO, 2 and 3 mix as well as 4 and 5.
Label and characteristic polynomial for quantum angular symmetry d and f are reported in
the FDOCCUP keyword description.

FDOCCUP - f and d orbital occupation guess

rec variable meaning
• ∗ INSELOC number of records describing the occupation of d or f shells
insert INSELOC records II
IATNUM atom label, output order
ISHATO shell label, atomic list
SHTYP angular quantum label of ISHATO(I)
• ∗ IAOCALPH(J) α occupation of the 5 d (J=1,5) or 7 f (J=1,7) AOs of the shell
• ∗ IAOC BETA(J) β occupation of the 5 d (J=1,5) or 7 f (J=1,7) AOs of the shell

Label and characteristic polynomial for quantum angular symmetry d and f are reported in
the following. Spherical harmonics are used, it follows that 5 and 7 functions are reported for
d and f symmetry respectively.

84
Symmetry d, label 3 Symmetry f , label 4
Label Polynomial label Polynomial
1 (2z 2 − x2 − y 2 ) 1 (2z 2 − 3x2 − 3y 2 )
2 xz 2 x(4z 2 − x2 − y 2 ))
3 yz 3 y(4z 2 − x2 − y 2 )
2 2
4 (x − y ) 4 z(x2 − y 2 )
5 xy 5 (xyz)
6 x(x2 − 3y 2 )
7 y(3x2 − y 2 )

The option FDOCCUP permits to define the occupation of specific f or d orbitals in a given
shell in the initial guess calculation. This option can be used only for open shell cases, where
electrons belonging to partially filled shells can be assigned to selected AOs.
INSELOC specifies the number of shell occupation descriptions; for each shell, the atom IAT-
NUM (output order) and the shell label ISHATO (basis set order) is specified. As a cross
check, also the shell label type SHTYP (3 and 4 for d and f respectively) is required. Finally
the α and β occupation for the 5 (d) or 7 (f ) AOs in the shell is indicated. The occupation
numbers in IAOCALPH(J) and IAOCBETA(J) are normalized a posteriori to the number of
d or f electrons resulting from the input charges.
Note that the input information must be inserted just one time per atomic number (say 26,
Fe) and IATNUM can be any of the output order atom labels for the selected atom. See
EXAMPLE 1 for test case.
To attribute different AOs guess occupation to atoms with the same atomic number, it is just
needed to insert as many input as the number of different occupation one wants to set with
IATNUM indicating atom labels of atoms with same atomic number but not symmetry related.
If different AOs occupation for atoms symmetry related is required, the program stops and an
indication of symmetry relations between atoms is reported (as the one obtained by the use of
ATOMSYMM keyword).
FDOCCUP can be used in conjunction with EIGSHIFT, ATOMSPIN and SPINLOCK. Note
that FDOCCUP and ATOMSPIN act at CYCLE 0, EIGSHIFT at CYCLE 1 and SPINLOCK
works from CYCLE 1 for a defined number of cycles.
On the contrary FDOCCUP, as ATOMSPIN, is incompatible with SPINEDIT, where the ini-
tial guess calculation is bypassed because the initial guess is obtained by a previous calculation.

Example. Ce2 O3 . The Ce3+ ion is required to have one f electron. Suppose Ce is the first
atom in the input list and that the seventh shell in the Basis Set list is the f shell with charge
one. Suppose to distribute the f electron in two AOs with similar occupation: f(2z2 −3x2 −3y2 )z
(first f AO) and f(x2 −3y2 )y (sixth f AO). Use of FDOCCUP in this case is reported in the
following:
FDOCCUP
1
1 7 4
1 0 0 0 0 1 0
0 0 0 0 0 0 0

Output is printed in the ATOMIC WAVEFUNCTION(S) part and is reported in the following.
Note that the nuclear charge is 30 instead of 58 because pseudopotential is used.
NUCLEAR CHARGE 30.0 SYMMETRY SPECIES S P D
N. ELECTRONS 27.0 NUMBER OF PRIMITIVE GTOS 11 11 5
NUMBER OF CONTRACTED GTOS 4 4 1
NUMBER OF CLOSED SHELLS 2 2 1
OPEN SHELL OCCUPATION 0 0 0

ZNUC SCFIT TOTAL HF ENERGY KINETIC ENERGY VIRIAL THEOREM ACCURACY

30.0 7 -4.706164398E+02 1.769968771E+02 -3.658896854E+00 3.4E-06

FDOCCUP ACTIVE - ATOM 1 SHELL 6 F SYMMETRY WITH CHARGE 1.0

DIAGONAL ELEMENTS OF DENSITY MATRIX FOR SHELL 6
ALPHA+BETA 0.500 0.000 0.000 0.000 0.000 0.500 0.000
ALPHA-BETA 0.500 0.000 0.000 0.000 0.000 0.500 0.000

Note: the printed information refers to α + β and α − β, while in input α and β are inserted
separately.

85
FIXINDEX
No input data required.
When the geometrical and/or the basis set parameters of the system are changed, maintaining
the symmetry and the setting, the truncation criteria of the Coulomb and exchange series,
based on overlap (Chapter 11) can lead to the selection of different numbers of bi-electronic
integrals. This may be the origin of numerical noise in the optimization curve. When small
changes are made on the lattice parameter or on the Gaussian orbital exponents, the indices of
the integrals to be calculated can be selected for a reference geometry (or basis set), ”frozen”,
and used to compute the corresponding integrals with the modified geometry (or basis set).
This procedure is recommended only when basis set or geometry modifications are relatively
small.

• The corresponding irreducible atoms in the two geometries must be entered in the same
order, and their position in the second geometry must be slightly shifted in comparison
with the first geometry (reference);

• the reference geometry must correspond to the most compact structure;

• the reference basis set must have the lowest outer exponent.

This guards against the loss of significant contributions after, for example, expansion of the
lattice.
If estimate of resource is requested with TESTRUN, the reference geometry is used.
Two sets of input data must be given:

1. complete input (geometry, Section 1.1; basis set, Section 1.2; general information, Section
1.3; SCF, Section 1.3), defining the reference basis set and/or geometry;

2. ”restart” option input, selected by one of the following keywords (format A) to be added
after the SCF input:

restart with new geometrical parameters

GEOM
insert geometry input, page 13
or
restart with new basis set
BASE
insert basis set input, page 19
or
restart with new basis set and new geometrical parameters
GEBA
insert geometry input, page 13
insert basis set input, page 19

BASE: the only modification of the basis set allowed is the value of the orbital exponent
of the GTFs and the contraction coefficient; the number and type of shells and AOs cannot
change.
GEOM: geometry variation must keep the symmetry and the setting unchanged.

The resulting structure of the input deck is as follows:

0 Title

1 standard geometry input (reference geometry). Section 1.1

1b geometry editing keywords (optional; Section 2.1)

END

86
 2 standard basis set input (reference basis set). Section 1.2

2b basis set related keywords (optional; Section 2.2)

END
3 FIXINDEX
3 block3 keywords
END

GEOM BASE GEBA

⇓ ⇓ ⇓
geometry input(p 13) basis set input(p 19) geometry input (p 13)
(block 1, 1b) (block 2, 2b) (block 1, 1b)
END END END
basis set input(p 19)
(block 2, 2b)
END

Warning: The reference geometry and/or basis set is overwritten by the new one, after
symmetry analysis and classification of the integrals. If the reference geometry is edited through
appropriate keywords, the same editing must be performed through the second input. Same
for basis set input.
If the geometry is defined through the keyword EXTERNAL, the reference geometry data
should be in file fort.34, the wave function geometry in file fort.35.
Note In geometry optimization (OPTGEOM, page 105) FIXINDEX is automatically set,
with input geometry as reference geometry. See keyword FINALRUN, page 112 to redefine
the reference geometry.

See tests 5 and 20.

FMIXING - Fock/KS matrix mixing

rec variable meaning
• ∗ IPMIX percent of Fock/KS matrices mixing

The Fock/KS matrix at cycle i is defined as:

Fi0 = (1 − p)Fi + pFi−1

0

where p, input datum IPMIX, is the % of mixing. Too high a value of p (>50%) causes higher
number of SCF cycles and can force the stabilization of the total energy value, without a real
self consistency.

GRADCAL
No input data required.
Analytic calculation of the nuclear coordinates gradient of the HF, UHF, DFT energies after
SCF (all electrons and ECP).
If numerical gradient is requested for the geometry optimization (NUMGRALL, page 114;
NUMGRATO, page 114; NUMGRCEL, page 114;), analytical gradient is not computed.

87
GRIMME - Grimme dispersion correction for energy and gradient
The keyword GRIMME, inserted in third input block, calculates a London-type pairwise
empirical correction to the energy as proposed by Grimme [61], to include long-range dispersion
contributions to the computed ab initio total energy and gradients.
Therefore geometry optimization and vibrational frequency calculation can be carried out by
including the empirical correction.

rec variable meaning

• A GRIMME keyword
• * s6 scaling factor
* d steepness
* Rcut cutoff distance to truncate direct lattice summation
• * NATS number of atomic species
II insert NATS records II
• * NAT “conventional” atomic number (page 20).
* C6 dispersion coefficient for atomic number NAT (Jnm6 mol−1 )
* Rvdw van der Waals radius for atomic number NAT (Å)

The keyword GRIMME is followed by a set of computational parameters [i.e. scaling factor,
steepness and cutoff radius for g (direct lattice) summation], the number of atomic species and
for each atomic species the “conventional atomic number” (page 20) and the value of C6 and
Rvdw (in Jnm6 mol−1 and Å, respectively).
Note that the atomic number is defined as the “conventional atomic number”. In this way, if one
wants to use different atomic parameters for the same atomic species, a different conventional
atomic number can be used. For example, 1 and 101 for hydrogen.
The current implementation has been mainly tested and used in combination with the B3LYP
method [27, 28, 126], even if it can be applied to whatever level of theory.
The original proposal by Grimme was to augment DFT methods with an empirical London-
type correction. To that purpose, Grimme defined a general set of atomic empirical parameters
(see Table 1 of ref. [61]) and used an optimized scaling factor to adjust the dispersion correction
for each DFT method.
The total energy is given by
EDF T −D = EKS−DF T + Edisp
where Edisp , the empirical dispersion correction, is given as
at −1 X
NX Nat X C ij
6
Edisp = −s6 fdump (Rij,g )
i=1 j=i+1 g
R ij,g

The summation is over all atom pairs and g lattice vectors with the exclusion of the i=j
contribution (i.e. self interaction) for g = 0, C6ij is the dispersion coefficient for the atom pair
ij, s6 is a scaling factor that depends only on the adopted DFT method (e.g. s6 is 1.05 for
B3LYP) and Rij,g is the interatomic distance between atoms i in the reference cell and j in
the neighbouring cells at distance |g|. For a set of molecular crystals, a cutoff distance of 25.0
Å was used to truncate the summation over lattice vectors which corresponds to an estimated
error of less than 0.02 kJ/mol on computed cohesive energies with respect to larger cutoffs
[27, 28].
A damping function is used to avoid near-singularities for small interatomic distances:
fdump (Rij,g)= 1
−d(Rij,g /Rvdw −)
1+exp

i j
where Rvdw is the sum of atomic van der Waals radii (i.e. Rvdw = Rvdw + Rvdw ) and d
determines the steepness of the damping function (d = 20). Van der Waals radii and atomic

88
C6 coefficients were taken from Table 1 of ref. [61]. From the latter, the C6ij dispersion
coefficients are computed by using a geometric mean:
q
C6ij = C6i C6j

The input requires to specify all parameters that enter equations above and it looks like (urea
molecular crystal):

Urea
CRYSTAL
0 0 0
113
5.565 4.684
5
6 0.0000 0.5000 0.3260
8 0.0000 0.5000 0.5953
7 0.1459 0.6459 0.1766
1 0.2575 0.7575 0.2827
1 0.1441 0.6441-0.0380
Optional keywords
END (ENDG)
Basis set input
END
DFT
. . . . . .
END
. . . . .
GRIMME
1.05 20. 25.
1.05 20. 25. s6 (scaling factor) d (steepness) Rcut (cutoff radius)
4
1 0.14 1.001 Hydrogen Conventional Atomic number , C6 , Rvdw
6 1.75 1.452 Carbon Conventional Atomic number , C6 , Rvdw
7 1.23 1.397 Nitrogen Conventional Atomic number , C6 , Rvdw
8 0.70 1.342 Oxygen Conventional Atomic number , C6 ,‘Rvdw
SHRINK
. . . . . .
END

GUESDUAL - SCF guess

density matrix from a previous run with different basis set
rec variable value meaning
• ∗ NFR number of modification (NFR ≥ 1) in the atomic basis set given in input
IC 1 complete SCF calculation
0 stop before the first hamiltonian matrix diagonalization (to be used
in Cryscor - see http://www.cryscor.unito.it when the dual basis set
option is activated)
insert NFR records - for each shell
• ∗ NAT formal atomic number of the atom whose basis set is modified
* NSH sequence number of the reference shell in the atomic basis set, starting
from which shell(s) is(are) inserted/deleted
∗ NU number of shells inserted/deleted after the reference shell NSH

The keyword is to be inserted in the third (Hamiltonian-SCF) input block. It defines how the
basis set given in input differs from the one used to compute the density matrix used as SCF
guess. The two basis sets must be marginally different. It can be used to add polarization

89
functions, or diffuse functions (see http://www.cryscor.unito.it).
The present calculation and the one used to compute the density matrix SCF guess have same
symmetry and number of atoms, but different number of shells in the BS of some atoms. Dif-
ferent geometrical parameters and computational conditions are allowed.
The density matrix P0 (direct lattice) of the previous run, SCF guess, is read from external
unit (copy file fort.9 of the previous run in file fort.20). Density matrix is reorganized: ele-
ments corresponding to basis functions removed are removed, elements corresponding to basis
functions added are set to 0.
Warning. The efficiency of this guess can be dramatically compromised if not congruent BS
modifications, i.e. insertion or elimination of significantly populated shells, are made. Check
the normalization factor.
Example. LiH bulk. The BS of Li in the reference calculation is referred to as BS1 and reported
on the left, the modified one is referred to as BS2 and reported on the right. The GUESDUAL
option following BS2 illustrates how BS2 differs with respect to BS1.
BS1 | BS2

3 3 | 3 4
0 0 6 2.0 1.0 | 0 0 6 2.0 1.0
700. .001421 | 700. .001421
220. .003973 | 220. .003973
70. .01639 | 70. .01639
20. .089954 | 20. .089954
5. .31565 | 5. .31565
1.5 .4946 | 1.5 .4946
0 0 1 0.0 1.0 | 0 0 1 0.0 1.0
0.5 1. | 0.5 1.
0 2 1 0.0 1.0 | 0 2 1 0.0 1.0
0.6 1. | 0.6 1.
1 4 | 0 3 1 0. 1.
0 0 5 2. 1. | 0.4 1.
120. .000267 | 1 5
40. .002249 | 0 0 5 2. 1.
12. .006389 | 120. .000267
4. .032906 | 40. .002249
1.2 .095512 | 12. .006389
0 0 1 0. 1. | 4. .032906
0.45 1. | 1.2 .095512
0 0 1 0. 1. | 0 0 1 0. 1.
0.13 1. | 0.45 1.
0 2 1 0. 1. | 0 0 1 0. 1.
0.3 1. | 0.13 1.
99 0 | 0 2 1 0. 1.
END | 0.6 1.
| 0 2 1 0. 1.
| 0.3 1.
| 99 0
| END
GUESDUAL
2 0
3 3 1
1 3 1
END
In this case two modifications (NFR=2) has been introduced in BS2 with respect to BS1.
Complete SCF calculation (IC=1) is required.

90
First modification in Li (formal atomic number NAT=3) basis set, after the third shell
(NSH=3), one shell was added (NU=1).
Second modification in H (atomic number IA=1) basis set, after the third shell (NSH=3), one
shell (NU=1) was inserted.

GUESSF - Fock/KS matrix from a previous run

Obsolete - See GUESSP, page 91.

GUESSP - SCF guess - Density matrix from a previous run

The density matrix from a previous run, P0 (direct lattice), is read from disk, and used as SCF
guess. No input data are required.
The density matrix can be edited to modify the spin state. See SPINEDIT, page 100.
The density matrix used as SCF guess was written with wave function information in file fort.9
at the end of a previous SCF run, and read as file fort.20.
The two cases, the present one and that used for the restart, must have same symmetry, and
same number of atoms, basis functions and shells. Atoms and shells must be in the same order.
The program does not check the 1:1 old–new correspondence.
Different geometrical parameters, computational conditions or exponents of the Gaussian prim-
itives are allowed.
In geometry and/or basis set optimization, this technique will significantly reduce the number
of SCF cycles. The following scheme shows how to proceed.

1. First run to generate the density matrix

Program inp. block section comments

crystal 0 1 Title
1 1.1 geometry input
2 1.2 basis set input
3 1.3 computational parameters and SCF input
save wf in file fort.9 (binary) or file.98 (formatted)

2. Second run - the density matrix is read in as a guess to start SCF

copy file fort.9 to fort.20 (or convert file fort.98 and then copy)
Program inp. block section comments
crystal 0 1 Title
1 1.1 geometry input
2 1.2 basis set input
3 1.3 computational parameters and SCF input
(GUESSP)

Warning The modification of the geometry may result in a different order in the storage of
the matrix elements associated to each overlap distribution in the present and the previous
run. To avoid the mismatch it is strongly recommended to classify the integrals of the present
case using the geometry of the previous case (FIXINDEX, page 86).
Warning When wave function information are stored formatted in file fort.98, the data must
be converted to binary by the keyword RDFMWF, page 167 of the properties program).
Warning SCF guess from a density matrix computed with a different basis set is allowed by
the keyword GUESDUAL, page 89.

GUESSPAT - Superposition of atomic densities

The standard initial guess to start the SCF cycle is the superposition of atomic (or ionic)
densities. No input data are required. The electronic configuration of the atoms is entered as
a shell occupation number in the basis set input (page 19). Different electronic configurations

91
may be assigned to atoms with the same atomic number and basis set (but not symmetry
related) through the keyword CHEMOD (page 19).

INTGPACK - Choice of bielectronic integrals package

rec variable value meaning
• ∗ IPACK [0] s, sp shells→ POPLE; p, d shells → ATMOL
1 ATMOL for Coulomb integrals;
POPLE for exchange integrals
2 POPLE for Coulomb integrals;
ATMOL for exchange integrals
3 ATMOL for Coulomb integrals;
ATMOL for exchange integrals

By default the bielectronic integrals are computed using a set of routines derived from Pople’s
GAUSSIAN 70 package [70], if s and sp shells are involved, and by routines derived from
ATMOL [84] for p and d shells. The value of IPACK allows different choices. Integrals involving
p or d shells are always computed by ATMOL. The ATMOL package can compute integrals over
functions of any quantum number, but the symmetry treatment implemented in the CRYSTAL
package allows usage of s, p and d functions only. The use of sp shells (s and p orbitals sharing
the same exponent) reduces the time required to compute the integrals considerably.

KSYMMPRT
Symmetry Adapted Bloch Functions [138, 137] (page 101)are used as basis for the Fock/KS
matrix diagonalization. The results of the symmetry analysis in reciprocal space are printed.
At each k-point: number of point symmetry operators, number of active IRs, maximum IR
dimension and maximum block dimension in the Fock/KS matrix factorization. TESTRUN
stops execution after this information is printed.
No input data required.
Extended information can be obtained by setting the value N of LPRINT(47) (keyword SET-
PRINT, page 54) before KSYMMPRT.
N information
0 Basic Symmetry Information - At each k-point: list of point symmetry operators,
IR dimensions and number of Irreducible Sets.
>0 Symmetry Information - At each k-point ≤ N: class structure, character table
and IR information concerning the K-Little Group. For the rest of the k-point
the same information as -1 is printed.
< −1 Full Symmetry Information - At each k-point: the same information as N > 0,
together with the matrix representatives of the point operators.
MPP doesn’t support KSYMMPRT.

LEVSHIFT - Eigenvalue level shifting

rec variable value meaning
• ∗ ISHIFT The level shifter is set to ISHIFT *0.1 hartree.
ILOCK 0 no locking
1 causes a lock in a particular state (eg non-conducting) even if the so-
lution during the SCF cycles would normally pass through or even con-
verge to a conducting state.

The eigenvalue level shifting technique is well known in molecular studies [62, 29], and may
also be used for periodic systems. The technique involves the addition of a negative energy
shift to the diagonal Fock/KS matrix elements (in the Crystalline Orbital basis) of the occupied
orbitals and thus reducing their coupling to the “unoccupied” set. This shift may be main-
tained (ILOCK=1) or removed (ILOCK=0) after diagonalization. The former case causes a

92
lock in a particular state (eg non- conducting) even if the solution during the SCF cycles would
normally pass through or even converge to a conducting state.
This option provides an alternative damping mechanism to Fock/KS matrix mixing (FMIX-
ING, page 87). The locking is effective only if ISHIFT is large enough. If locking is used, the
Fermi energy and the eigenvalues are depressed by the value of the level shifter. Suggested
values :

1. Normal cases require no mixing of Fock/KS matrices in successive cycles to converge:

ISHIFT=0 (default).

2. When 20% to 30% mixing of Fock/KS matrices is necessary, an ISHIFT value of between
1 and 3 (giving a level shift of 0.1 to 0.3 hartree) may produce an equivalent or even
superior convergence rate.

3. If serious convergence difficulties are encountered, ISHIFT=10 will normally be adequate,

corresponding to a level shift of 1 hartree. But it may happen that the system moves
towards an excited state, and no convergence is obtained.

Warning - Hamiltonian eigenvalues are modified by the level shifter. Take into account when
looking at top and bottom bands eigenvalues printed.
See tests 29, 30, 31, 32, 38.

MAXCYCLE
rec variable meaning
• ∗ NMAX maximum number of SCF cycles [50]

The possibility to modify the maximum number of SCF cycles allows: increasing the number
of cycles in case of very slow convergence (metals, magnetic systems, DFT);
The keyword POSTSCF forces saving wave function data in file fort.9, even if SCF ends
before reaching convergence for ”too many cycles”.

MONSPLIT - Splitting of large monoelectronic integral files

rec variable meaning
• ∗ NFILE number of files to be used [1] (max 10)

Very large basis sets can produce billions monoelectronic integrals to be stored, as the number
of monoelectronic integrals scales as the square of basis set size. The multipolar expansion
technique based on the atoms reduces the disk space up to a factor 3, compared to the value
printed as estimate. The distribution of the integrals over several disk files may be necessary,
if available disk space is limited.

MPP - Massive Parallel Execution - Programmers only

No input data required.
This option is available only with MPPcrystal, the massively parallel version of CRYSTAL09,
where Massive Parallel Libraries are linked, and matrices in the reciprocal space are distributed
over the processors.
MPPcrystal is designed to compute the total energy and wavefunction of large unit cell systems
with large memory requirements. In particular, The MPP option is eligible when nk /nprocs ,
i.e. the ratio between nk , the number of k points sampled in the reciprocal space and nprocs ,
the number of processors used, is small.
In choosing nprocs , the user should consider that MPP is efficient when nAO /nprocs ≥ 50 (nAO
is the number of Atomic Orbitals per cell, or basis functions). If nprocs is too large compared to

93
the size of the system, some parts of the calculation (matrix diagonalization) become inefficient
because of increasing communications among processors.

Advantages of MPP:
• matrix diagonalization is well balanced because many processors are involved in the di-
agonalization of one matrix (use of standard linear algebra libraries for massively parallel
calculation)
• memory requirement per processor decreases with nprocs because data are distributed to
processors
• I/O is limited to reading the user’s input deck and writing the output files with the
results of a job.
Warnings.
• The following options are not available with MPP:

keyword block

CPHF CPHF 1
Frequency calculation - IR intensities FREQCALC/INTENS 1
Bloch Functions Symmetry Adapted SYMADAPT 3
Printing of eigenvalues of overlap matrix in k space EIGS 3

• use of option NOBIPOLA (page 95) is recommended for very large systems, because
data distribution has not been implemented for the bipolar approximation of two-electron
integrals (such approximation becomes less efficient with large unit-cell systems so that
switching to NOBIPOLA does generally not affect performance significantly)
• MPP calculations run with NOSYMADA (page 95) active (no use of symmetry in the
reciprocal space)
• keyword DCDIAG may activate Divide&Conquer diagonalization strategy (not default)
which improves performance; however, it must be used with caution as one failure was
reported in one case where eigenvectors and eigenvalues were not determined correctly
(with no other clear symptom)
• SCF convergence may proceed along slightly different paths with and without MPP,
particularly in sensitive cases, and SCF convergence acceleration methods may exhibit
different behaviours. A particular setting of convergence tool parameters, for example a
given mixing rate of matrices along several SCF steps, may be either effective or ineffective
when using MPP or not. However, a proper choice of the parameters will always result
into achievement of the SCF cycle convergence.
If keyword MPP is not part of the input deck, MPPcrystal will run as Pcrystal (replicated data
storage in memory, matrix diagonalization by k-point with one full matrix diagonalization per
processor, no link to external libraries).

MYBIPOLA - Bipolar expansion approximation control

rec variable meaning
• ∗ ILCOUL maximum multipole order for Coulomb 4
∗ ITCOUL overlap threshold for Coulomb 14
∗ IFCOUL reducing factor for Coulomb 90
• ∗ ILEXCH maximum multipole order for exchange 2
∗ ITEXCH overlap threshold for exchange 10
∗ IFEXCH reducing factor for exchange 70

94
The bipolar approximation is applied in the evaluation of the Coulomb and exchange integrals
(page 225). Maximum values for ILCOUL and ILEXCH are 8 and 4, respectively. ITCOUL and
ITEXCH can be assigned any intermediate value between the default values (14 and 10) (see
page 225) and the values switching off the bipolar expansion (20000 and 20000). Increasing
IFCOUL and IFEXCH the threshold is lightly modified in order to increase the number of
approximated integrals, and vice versa.
Warning - for developers only

NEIGHBOR/NEIGHPRT
See input block 1, page 51

NOBIPOLA - Bipolar expansion approximation suppression

All the bielectronic integrals, coulomb and exchange, are evaluated exactly. The overlap thresh-
old both for coulomb and exchange integrals is set to 2000000.
No input data required. The CPU time in the integrals program may increase up to a factor
3.

NOBIPCOU - Bipolar expansion approximation of coulomb integrals

suppression
Coulomb bielectronic integrals are evaluated exactly. The overlap threshold for coulomb inte-
grals is set to 2000000.
No input data required.

NOBIPEXC - Bipolar expansion approximation of exchange integrals

suppression
Exchange bielectronic integrals are evaluated exactly. The overlap threshold for exchange
integrals is set to 2000000. No input data required.

NOFMWF - Wave function formatted output

CRYSTAL writes the formatted wave function in file fort.98 at the end of SCF by default.
This keyword deletes this feature.

NOMONDIR - Monoelectronic integrals on disk

No input data required.
In the SCF step bielectronic integrals are computed at each cycle, while monoelectronic inte-
grals are computed once and read from disk at each cycle.

NOSYMADA
The Symmetry Adapted Functions are not used in the Hamiltonian matrix diagonalization. No
input data are required. This choice increases the diagonalization CPU time when the system
has symmetry operators.

PARAMPRT - - printing of parametrized dimensions

See input block 1, page 52.

95
POLEORDR - Maximum order of multipolar expansion
rec variable meaning
• ∗ IDIPO maximum order of pole [4]

Maximum order of shell multipoles in the long-range zone for the electron-electron Coulomb
interaction. Maximum value = 6. See Section 11.3, page 223.

POSTSCF
Calculation to be done after scf (gradient, population analysis) are performed even if conver-
gence is not reached. It may be useful when convergence is very slow, and scf ends for ”TOO
MANY CYCLES” very close to the convergence criteria required.
No input data are required.

PPAN/MULPOPAN - Mulliken Population Analysis

Mulliken population analysis is performed at the end of SCF process.
No input data are required.
Bond populations are analysed for the first n neighbours (n default value 3; see NEIGHBOR,
page 51, to modify the value).
Computed data:
P P g g
1. aµ = ν g Pµν Sµν orbital charges
P
2. sl = µ∈l aµ shell charges
P
3. qA = l∈A sl atomic charges

4. b(A0 , B g ) = µ∈A ν∈B Pµν g g

P P
Sµν bond populations between the non-equivalent atoms in
the unit cell (A0 ) and their first NVI neighbours (B g in cell g). The printed values must
be multiplied by 2 when B6=A to compare with standard molecular calculations.

Formatted data are written in file PPAN.DAT (opened in forrtran unit 24).
See Appendix D, page 263.

PRINTOUT - Setting of printing environment

See input block 1, page 53.

RHF [default]
A restricted closed-shell hamiltonian calculation is performed ([108, 103], Chapter 8 of ref.
[100]). Default choice.

ROHF
Obsolete. See UHF

SCFDIR
No input data required.
In the SCF step monoelectronic and bielectronic integrals are evaluated at each cycle. No
screening of the integrals is performed.

96
SAVEWF
The wave function is written in file fort.79 every two cycles. The format is the same as in file
fort.9, written at the end of SCF.
To restart SCF cycles using the density matrix written in file fort.79, it has to be copied in file
fort.20
No input data required.

SETINF - Setting of INF values

See input block 1, page 54

SETPRINT - Setting of printing options

See input block 1, page 54.

SHRINK - Pack-Monkhorst/Gilat shrinking factors

rec variable value meaning
if the system is periodic insert II
• ∗ IS Shrinking factor in reciprocal space (Section 11.7, page 227)
ISP Shrinking factor for a denser k point net (Gilat net) in the
evaluation of the Fermi energy and density matrix.
if IS = 0 insert II
• ∗ IS1,IS2,IS3 Shrinking factors along B1,B2,B3 (reciprocal lattice vectors);
to be used when the unit cell is highly anisotropic
optional keywords terminated by END or STOP II

For periodic systems, 1D, 2D, 3D, the mandatory input information is the shrinking factor, IS,
to generate a commensurate grid of k points in reciprocal space, according to Pack-Monkhorst
method. The Hamiltonian matrix computed in direct space, Hg , is Fourier transformed for
each k value, and diagonalized, to obtain eigenvectors and eigenvalues:
X
Hk = Hg eigk
g

Hk Ak = SkAk Ek
A second shrinking factor, ISP, defines the sampling of k points, ”Gilat net” [57, 56], used
for the calculation of the density matrix and the determination of Fermi energy in the case of
conductors (bands not fully occupied).
In 3D crystals, the sampling points belong to a lattice (called the Pack-Monkhorst net), with
basis vectors:
b1/is1, b2/is2, b3/is3 is1=is2=is3=IS, unless otherwise stated
where b1, b2, b3 are the reciprocal lattice vectors, and is1, is2, is3 are integers ”shrinking
factors”.
In 2D crystals, IS3 is set equal to 1; in 1D crystals both IS2 and IS3 are set equal to 1. Only
points ki of the Pack-Monkhorst net belonging to the irreducible part of the Brillouin Zone
(IBZ) are considered, with associated a geometrical weight, wi . The choice of the reciprocal
space integration parameters to compute the Fermi energy is a delicate step for metals. See
Section 11.7, page 227. Two parameters control the accuracy of reciprocal space integration
for Fermi energy calculation and density matrix reconstruction:

IS shrinking factor of reciprocal lattice vectors. The value of IS determines the number of k
points at which the Fock/KS matrix is diagonalized. Multiples of 2 or 3 should be used,
according to the point symmetry of the system (order of principal axes).

97
 In high symmetry systems, it is convenient to assign IS magic values such that all low
multiplicity (high symmetry) points belong to the Monkhorst lattice. Although this
choice does not correspond to maximum efficiency, it gives a safer estimate of the integral.
The k-points net is automatically made anisotropic for 1D and 2D systems.

The figure presents the reciprocal lattice cell of 2D graphite (rhombus), the first
Brillouin zone (hexagon), the irreducible part of Brillouin zone (in grey), and the
coordinates of the ki points according to a Pack-Monkhorst sampling, with shrinking
factor 3 and 6.

ISP shrinking factor of reciprocal lattice vectors in the Gilat net (see [103], Chapter II.6).
ISP is used in the calculation of the Fermi energy and density matrix. Its value can be
equal to IS for insulating systems and equal to 2*IS for conducting systems.
The value assigned to ISP is irrelevant for non-conductors. However, a non-conductor
may give rise to a conducting structure at the initial stages of the SCF cycle, owing, for
instance, to a very unbalanced initial guess of the density matrix. The ISP parameter
must therefore be defined in all cases.
Note. The value used in the calculation is ISP=IS*NINT(MAX(ISP,IS)/IS), a multiple
integer of IS. For instance:

input data IS ISP ISP for wf calculation

3 4 3
3 6 6
3 8 6

In the following table the number of sampling points in the IBZ and in BZ is given for a
fcc lattice (space group 225, 48 symmetry operators) and hcp lattice (space group 194, 24
symmetry operators). The CRYSTAL code allows 413 k points in the Pack-Monkhorst net,
and 2920 in the Gilat net.

98
 IS points in IBZ points in IBZ points BZ
fcc hcp
6 16 28 112
8 29 50 260
12 72 133 868
16 145 270 2052
18 195 370 2920
24 413 793 6916
32 897 1734 16388
36 1240 2413 23332
48 2769 5425 55300
1. When an anisotropic net is user defined (IS=0), the ISP input value is taken as ISP1
(shrinking factor of Gilat net along first reciprocal lattice) and ISP2 and ISP3 are set to:
ISP2=(ISP*IS2)/IS1,
ISP3=(ISP*IS3)/IS1.
2. User defined anisotropic net is not compatible with SABF (Symmetry Adapted Bloch
Functions). See NOSYMADA, page 95.
Some tools for accelerating convergence are given through the keywords LEVSHIFT (page
92 and tests 29, 30, 31, 32, 38), FMIXING (page 87), SMEAR (page 99), BROYDEN
(page 73) and ANDERSON (page 71).
At each SCF cycle the total atomic charges, following a Mulliken population analysis scheme,
and the total energy are printed.
The default value of the parameters to control the exit from the SCF cycle (∆E < 10−6 hartree,
maximum number of SCF cycles: 50) may be modified entering the keywords:
TOLDEE (tolerance on change in total energy) page 102;
TOLDEP (tolerance on SQM in density matrix elements) page 102;
MAXCYCLE (maximum number of cycles) page 93.

SMEAR
rec variable meaning
• ∗ WIDTH temperature smearing of Fermi surface

Modifies the occupancy of the eigenvalues (fj ) used in reconstructing the density matrix from
the step function, (equation 11.9, page 222) to the Fermi function;
(j −F )

fj = (1 + e kb T
)−1 (2.20)
where F is the Fermi energy and kb T is input as WIDTH in hartree.
The smearing of the Fermi surface surface may be useful when studying metallic systems in
which the sharp cut-off in occupancy at F can cause unphysical oscillations in the charge
density. It may also result in faster convergence of the total energy with respect to k-point
sampling.
In density functional theory the use of Fermi surface smearing finds a formal justification in
the finite temperature DFT approach of Mermin [83]. In this case the “free energy” of the
system may be computed as:

F = E(T ) − kb T S(T )
NX
states

= E − kb T fi ln fi + (1 − fi ) ln(1 − fi ) (2.21)
i

where S is the electronic entropy. Often we wish to compute properties for the athermal limit
(T=0). For the free electron gas the dependencies of the energy and entropy on temperature
are:

99
 E(T ) = E(0) + αT 2
S(T ) = 2αT (2.22)
and so the quantity

F (T ) + E(T )
E0 = = E(0) + O(T 3 ) (2.23)
2
may be used as an estimate of E(0).

0.72
+ Using E(T) kT=0.001H ♦
0.7 Using E(T) kT=0.02H +
Using E0 kT=0.02H 
0.68

0.66
+ + + + + + + + + + + + + + +
0.64

0.62 ♦

0.6

0.58 ♦
♦ 
♦  ♦
   ♦   
♦ 
♦  
♦
♦  
♦ ♦ ♦ ♦
♦
♦
♦
0.56 ♦
0 2 4 6 8 10 12 14 16 18

Figure 2.4: The surface energy (J/M2 ) of Li(100) for various numbers of layers in a slab model
showing the effects of WIDTH (0.02H and 0.001H) and the use of E(T) or E0

Figure 2.4 shows the effect of WIDTH on the convergence of the Li(100) surface energy. Despite
the dense k-space sampling (IS=24, ISP=48) the surface energy is rather unstable at low
temperature (0.001H). There is a significant improvement in the stability of the solution for
higher values of WIDTH (0.02H) but use of E(T) results in a surface energy of 0.643 J/M2
significantly above that obtained by extrapolating E(T) to the T=0 limit (0.573 J/M2 ). The
use of E0 at WIDTH=0.02H results in an excellent estimate of the surface energy - 0.576 J/M2 .
Note that for conducting systems analytic first derivatives are not fully implemented when the
keyword SMEAR is used. In that case, numerical first derivatives should be computed (see
page 113). For very small value of smearing (around 0.001 hartree) analytical gradients should
be reliable in geometry optimization.

SPINEDIT - Editing of the spin density matrix

rec variable meaning
• ∗ N number of atoms for which spin must be reversed
• ∗ LB, L=1,N atom labels

The spin density matrix from a previous run is edited to generate an approximate guess for a
new spin configuration. The sign of the elements of the spin density matrix of selected atoms
is reversed. The keyword SPINEDIT must be combined with UHF (input block 3, page 102)
or DFT/SPIN (input block 3, page 75) and GUESSP.
Example: the anti ferromagnetic solution for the spinel MnCr2 O4 can be obtained by calculat-
ing the ferro magnetic solution, and using as guess to start the SCF process the density matrix
of the ferromagnetic solution with reversed signs on selected atoms.

100
SPINLOCK - Spin-polarized solutions
rec variable meaning
• ∗ NSPIN nα -nβ electrons
∗ NCYC number of cycles the difference is maintained

The difference between the number of α and β electrons at all k points can be locked at the
input value. The number of α electrons is locked to (N + NSPIN)/2, where N is the total
number of electrons in the unit cell; the number of β electrons is locked to (N - NSPIN)/2.
NSPIN must be odd when the number of electrons is odd, even when the number of electrons
is even.
Example. Bulk NiO. If a anti ferromagnetic solution is required, a double cell containing 2
NiO units must be considered (test 30). The two Ni atoms, related by translational symmetry,
are considered nonequivalent. The number of electron is 72, each Ni ion is expected to have
two unpaired electrons.

INF95 type of solution corresponding to the spin setting

0 anti ferromagnetic ↑ ↓ ↑ ↓
4 ferromagnetic ↑ ↑ ↑ ↑

Warning To lock the difference between α and β electrons α and β eigenvalues are forced to
be split. Their printed value is meaningless, until locking is active.
See tests 29, 30, 32, 33, 37, 38.

STOP
Execution stops immediately. Subsequent input records are not processed.

SYMADAPT
A computational procedure for generating space-symmetry-adapted Bloch functions, when BF
are built from a basis of local functions (AO), is implemented. The method, that applies to
any space group and AOs of any quantum number, is based on the diagonalization of Dirac
characters [138, 137].
The Symmetry Adapted Functions are used in the Hamiltonian matrix diagonalization. No
input data are required. This choice reduces the diagonalization CPU time when the system
has symmetry operators. Default choice.
Not supported by MPP execution.

TESTPDIM
The program stops after processing of the full input (all four input blocks) and performing
symmetry analysis. The size of the Fock/KS and density matrices in direct space is printed.
No input data are required.
It may be useful to obtain information on the neighbourhood of the non equivalent atoms (up
to 3, default value; redefined through the keyword NEIGHBOR, input block 1, page 51).

TEST[RUN] - Integrals classification and selection

Full input (geometry, basis set, general information, SCF) is processed.
The symmetry analysis is performed, and the monoelectronic and bielectronic integrals classi-
fied and selected, according to the the truncation criteria adopted. The size of the Fock/KS
and density matrices (direct lattice) and the disk space required to store the bielectronic are
printed. The value printed as ”disk space for monoelectronic integrals” is an upper limit. The
new technique of atomic multipolar expansion reduces the required space to about 1/3 of the
printed value.
No input data required.

101
This type of run is fast, and allows an estimate of the resources to allocate for the traditional
SCF wave function calculation.

TOLDEE - SCF convergence threshold on total energy

rec variable meaning
• ∗ ITOL 10−IT OL threshold for convergence on total energy

Different default values are set for different type of calculation:

SCF single point 6

Geometry optimization OPTGEOM 7
Frequency calculation FREQCALC 10
Elastic constants ELASTCON 8
Equation of state EOS 8

TOLDEP - SCF convergence threshold on density matrix

rec variable meaning
• ∗ ITOL 10−IT OL threshold for convergence on ∆P

For developers only.

TOLINTEG - Truncation criteria for bielectronic integrals

(Coulomb and HF exchange series)
rec variable meaning
• ∗ ITOL1 overlap threshold for Coulomb integrals- page 223 6
ITOL2 penetration threshold for Coulomb integrals-page 224 6
ITOL3 overlap threshold for HF exchange integrals-page 224 6
ITOL4 pseudo-overlap (HF exchange series-page 224) 6
ITOL5 pseudo-overlap (HF exchange series-page 224) 12

The five ITOL parameters control the accuracy of the calculation of the bielectronic Coulomb
and exchange series. Selection is performed according to overlap-like criteria: when the overlap
between two Atomic Orbitals is smaller than 10−IT OL , the corresponding integral is disregarded
or evaluated in a less precise way. Criteria for choosing the five tolerances are discussed in
Chapter 11.

TOLPSEUD - Truncation criteria for integrals involving ECPs

rec variable meaning

• ∗ ITPSE overlap threshold for ECP integrals 6

The program evaluates only those integrals for which the overlap between the charge distri-
bution ϕ0µ ϕgν (page 221) and the most diffuse Gaussian defining the pseudopotential is larger
than a given threshold Tps =10−IT P SE (default value 10−6 ; it was 5 in CRYSTAL98).

UHF - Hamiltonian for Open Shell Systems

For the description of systems containing unpaired electrons (such as molecules with an odd
number of electrons, radicals, ferromagnetic and anti ferromagnetic solids) a single determinant
is not an appropriate wave-function; in order to get the correct spin eigenfunction of these
systems, it is necessary to choose a linear combination of Slater determinants (whereas, in

102
 6 ψnα

ψnββ
6 ψnβ +1
ψnαα
6 ?

ψnβ ψnββ −1
ψnαα −1
6 6
? ?

6 ψnβ −1
?
ψ3β
6 ψ3α ?

6 ψ2 ψ2β
? 6 ψ2α ?

ψ1 ψ1β
ψ1α
6 6
? ?

Figure 2.5: Molecular Orbitals diagram for the Restricted Open Shell method (ROHF, left)
and for the Unrestricted Open Shell method (UHF, right)

closed shell systems, a single determinant gives always the appropriate spin eigenfunction)
([103, 4], Chapter 6 of ref. [100]).
In the Restricted Open Shell [108] Hamiltonian, the same set of molecular (i.e. crystalline)
orbitals describes alpha and beta electrons; levels can be doubly occupied (by one alpha and
one beta electron, as in the RHF closed shell approach), singly occupied or left vacant. The
wave-function is multi-determinantal; in the special case of half-closed shell systems, where we
can define a set of orbitals occupied by paired electrons and a second set occupied by electrons
with parallel spins, the wave-function is formed by a single determinant.
Another mono-determinantal approach for the study of open-shell systems is the UHF method
[104]. In this theory, the constraint of double occupancy is absent and α electrons are allowed
to populate orbitals other than those occupied by the β electrons. Energy levels corresponding
to a ROHF and UHF description are plotted in fig. 2.5.
The double occupancy constraint allows the ROHF approach to obtain solutions that are eigen-
c2 , whereas UHF solutions are formed by a mixture of spin states.
functions of the spin operator, S
The greater variational freedom allows the UHF method to produce wave-functions that are
energetically more stable than the corresponding ROHF ones; another advantage of the UHF
method is that it allows solutions with locally negative spin density (i.e. anti ferromagnetic
systems), a feature that ROHF solutions can never exhibit.
ROHF solution is not supported by CRYSTAL any more.
Related keywords
SPINLOCK definition of (nα - nβ electrons)
BETALOCK definition of nβ electrons.

Developers only
FULLTIME - Detailed timing report
A more detailed report of the timing data is generated:

TTTTTTTTTTTTTTTTTTTTTTTTTTTTTT SHELXG TELAPSE 19.68 TCPU 18.42

WWWWWWWWWWWWWWW SHELXG MX 1.07 MN 1.07 MD 1.07
QQQQQQQQQQQQQQQ SHELXG MX 1.07 MN 0.92 MD 0.98

103
The first line is the standard data. The second line reports the minimum, maximum and mean
wall time since the last report. The last line reports the minimum, maximum and mean cpu
time since the last report. The minimum, maximum and mean operations are across processors,
and so this directive is most useful for parallel job.

DCDIAG - ”divide and conquer” diagonalization

This directive is ONLY for MPP jobs. It instructs the code to use the divide and conquer
algorithm for the diagonalization stage. This algorithm can be up to four times quicker than
the standard, but it has been found very, very, occasionally to generate incorrect results.

CMPLXFAC - Detailed timing report

This directive is ONLY for MPP jobs. For load balancing reasons the MPP code must know
how many times more expensive a calculation on a complex k point is relative to a real one.
This allows the user to specify a value for this. The default value is 2.333333.

QVRSGDIM - limiting size switch for multipole moments gradients

rec variable meaning
• ∗ NFILE limiting size of multipole moment gradients to switch from generation by
pairs to generation by shells. Default 90000000.

104
Chapter 3

Geometry optimization

Geometry optimization at constant symmetry is invoked by the keyword OPTGEOM in input

block 1 (geometry). OPTGEOM must be the last keyword in geometry input. OPTGEOM
input block admits several options (sub keywords), and terminates with keyword END (or
END[OPT], END[—]: the first three characters only are processed).
crystal allows geometry optimization of systems with any periodicity: molecules, polymers,
slabs, and crystals. Unconstrained relaxation of the structure and different optimizations with
constraints can be carried out. The full symmetry of the system is preserved.
Geometry optimization can be performed in symmetrized fractional coordinates [default] or
redundant internal coordinates (optional choice, page 115).
OPTGEOM sub keywords can be classified as follow:
1. General sub keywords:
A - Optimization type (page 109)
B - Initial Hessian (page 109)
C - Hessian updating technique (page 107)
D - Convergence criteria (page 106)
E - Step control (page 110)
F - Coordinate system related options (page 111)
G - Optimization procedure control (page 112)
H - Numerical (first) derivatives (page 113)
I - Printing options (page 114)
2. Geometry optimization in redundant coordinates (page 115).
3. Geometry optimization with constraints (page 117).
A - Constant volume optimization (page 117)
B - Fixing lattice deformations (page 118)
C - Linear constraints between atomic coordinates (page 119)
D - Partial optimization of atomic positions (page 119)
E - Fixing internal coordinates (page 120)
4. Geometry optimization with application of an external stress (page 121)
5. Searching a transition state (page 123)
Default values are supplied for all computational parameters.
By default an unconstrained geometry optimization of the atomic positions at fixed
cell is performed.
Users can find supplementary information and input examples in the CRYSTAL Tutorials
Project web page at the CRYSTAL web site (http://www.crystal.unito.it/tutorials).

105
Geometry optimization strategy
A Quasi-Newton optimization scheme is implemented. Gradients are evaluated every time the
total energy is computed; the second derivative matrix (i.e. Hessian matrix) is built from the
gradients. The default choice for the initial Hessian matrix is obtained from a model Hessian,
as proposed by Schlegel and updated by using the BFGS algorithm[19, 20, 50, 59, 120].
By default the step considered is the Newton step (direction and length) controlled by the
Trust Radius scheme (see ALLOWTRUSTR page 111). NOTRUSTR to remove trust
radius control (CRYSTAL06 default choice).
HF and DFT (pure and hybrid functionals) analytical gradients are used for insulators and
conductors, both for all-electron and ECP calculations.
Note that for conducting systems analytic first derivatives are not fully implemented when
the keyword SMEAR (page 99) is used. In that case, numerical first derivatives should be
computed (see page 113).
For very small value of smearing (around 0.001 hartree) analytical gradients can be used.
For atomic positions, geometry optimization is performed in symmetrized fractional coordi-
nates, in order to exploit the point group symmetry of the lattice. The keyword PRSYMDIR
(input block 1, page 53) may be used to print the so-called symmetry allowed directions adopted
in the geometry optimization. If there are no symmetry allowed directions, the program prints
a warning message and stops, unless FULLOPTG or CELLONLY is requested (see page
109).
To optimize the lattice parameters a set of symmetry preserving cell deformations (see Symme-
try Allowed Elastic Distortions, USESAED, page 61) related to changes of isotropic volume
and of axial ratios is defined. By default, the symmetry allowed deformations are printed in
the output file.
When a full optimization of atom positions and cell parameters is carried out, a normalized
combined set of symmetrized directions and deformations is adopted.
Optional choice (keyword INTREDUN, page 115) is the geometry optimization in redundant
internal coordinates. In such a case, atomic displacements and cell deformations are implicitly
determined by the internal coordinate system.

Default choices
- Type of optimization:
The default geometry optimization type is the relaxation of the nuclear coordinates at fixed
lattice parameters in symmetrized fractional coordinates.
Optional choices: see page 109.

- Convergence criteria
A stationary point on the potential energy surface is found when the forces acting on atoms are
numerically zero. Geometry optimization is usually completed when the gradients are below a
given threshold.
In crystal, the optimization convergence is checked on the root-mean-square (RMS) and the
absolute value of the largest component of both the gradients and the estimated displacements.
When these four conditions are all satisfied at a time, optimization is considered complete.
In some cases (see page. 123), the optimization process stops with a warning message controlled
by the threshold in the energy change between consecutive optimization steps.
Default values are set for all computational parameters, and they may be modified through
keywords. Default choices:

106
 default keyword

RMS on gradient 0.000300 a.u. TOLDEG

largest component of gradient 1.5 * 0.000300 1.5 * TOLDEG
RMS on estimated displacements 0.0012 a.u. TOLDEX
absolute value of largest displacement 1.5 * 0.0012 1.5 * TOLDEX
max number of optimization cycles 100 MAXCYCLE
energy change between optimization steps threshold 10−7 a.u. TOLDEE

Optimization convergence criteria are set to different values according to the context where
geometry optimization is performed.

RMS on gradient RMS on displacement

Standard geometry opt 0.0003 0.0012
preopt in frequency calculation 0.00003 0.00012
preopt in EOS 0.00006 0.00012
preopt in elastic constants 0.00006 0.00012

- Initial Hessian guess

The initial Hessian is generated by means of a classical model as proposed by Schlegel.
H.B. Schlegel, Theoret. Chim. Acta 66 (1984) 333
J.M. Wittbrodt and H.B. Schlegel, J. Mol. Struct. (Theochem) 398-399 (1997) 55
It adopts a simple valence force field. Empirical rules are used to estimate the diagonal force
constants for a set of redundant internal coordinates (stretches, bends and torsions). Parame-
ters are available from H to At.
Warning - To define bonds the sum of covalent radii (see page 53) is used. For ionic systems
it may be necessary to modify the default values (see RAYCOV, page 53).

- Hessian updating technique

BFGS Broyden-Fletcher-Goldfarb-Shanno scheme [19, 20, 50, 59, 120].
Optional choices:

1. Schlegel’s updating scheme [117], (OLDCG, page 110), optimization scheme as in CRYS-
TAL03
2. Powell’s updating scheme (POWELL, page 110)

- SCF convergence and guess

The default value for SCF convergence criterion on total energy is set to 10−7 (TOLDEE in
input block 3 to modify it: never reduce accuracy).
After the first step, at each SCF cycle, the density matrix is recovered from the previous
geometry optimization step (implicit GUESSP, page 91 option).
This choice may be modified by inserting the keyword NOGUESS. A superposition of atomic
densities is then adopted on each step as SCF initial guess.
If the SCF solution at a given optimization step does not correspond to real convergence,
but to an energy stabilization due to the techniques applied to help convergence (LEVSHIFT,
FMIXING, BROYDEN..), the hamiltonian eigenvalues may be unphysical, and there is no
chance to recover the SCF process. In those cases it may be better to use an atomic guess
(keyword NOGUESS).

107
- Output files
The following formatted files are written during geometry optimization, and may be saved for
further processing.

fort.33 Cartesian coordinates of the atoms in the unit cell and total energy for each geometry
optimization step are written to file fort.33 in a simple xyz format (see Appendix D, page
262). This file is suitable to be read by molecular graphics programs (e.g. Molden...) to
display the animation of a geometry optimization run.

fort.34 If optimization is successful, the last geometry in written in file fort.34 (format de-
scribed in Appendix D, page 264).
The file can be read to define the basic geometry input. See EXTERNAL, page 15
opta(c)xxx At each xxx optimization step, the geometry is written in file optaxxx (opti-
mization of atoms coordinates only), or optcxxx (optimization of cell parameters or full
optimization) in the format of ”fort.34” file (see Appendix D, page 264). The file must
be renamed ”fort.34” if used to enter geometry input (keyword EXTERNAL).
The ”history” of the optimization allows restarting from a given step with different
parameters, when the procedure did not converge.

OPTINFO.DAT contains information to restart optimization. (see keyword RESTART in

OPTGEOM input block, page 113).

HESSOPT.DAT The hessian matrix is written, and can be read to define the initial guess
for the Hessian (keyword HESSOPT) in geometry optimization of a system with same
geometry and symmetry (it may have different BS, Hamiltonian, computational param-
eters).

SCFOUT.LOG SCF and optimization process printout is routed to file SCFOUT.LOG after
the first cycle. Keyword ONELOG: full printing on standard output.

1 - General sub keywords

A number of optional keywords allow tuning of the optimization procedure.

A - Type of optimization (page 109)

B - Initial Hessian (page 109)

C - Hessian updating technique (page 110)

D - Convergence criteria (page 106)

E - Step control (page 110)

F - Coordinate system related options (page 111)

G - Optimization procedure control (page 112)

H - Numerical first derivatives (page 113)

I - Printing options (page 114)

108
A - Type of optimization
Optional choices:

FULLOPTG full optimization, atom coordinates and cell parameters. The cell volume
may change (see CVOLOPT to optimize at constant volume)

CELLONLY only cell parameters are optimized. Default: the cell volume may change
(see CVOLOPT, page 117, to optimize at constant volume)

ITATOCEL full optimization, iterative procedure optimization: atoms-cell-atoms-cell-

. . . .

INTREDUN full optimization of atomic positions and cell parameters in redundant

internal coordinates (page 115).

B - Initial Hessian
By default an estimated model Hessian is adopted. The Hessian matrix is stored in file HES-
SOPT.DAT at each optimization step. This may be useful to restart the optimization from a
previous run performed at a lower level of theory (e.g. a smaller basis set). An initial Hessian
can also be obtained as numerical first-derivative (HESSNUM), but this process can be very
expensive.

HESSFREQ initial guess for the hessian - input from file HESSFREQ.DAT obtained
from frequencies calculations (developers only)

HESSIDEN initial guess: identity matrix

HESSOPT external guess (read from file HESSOPT.DAT)

HESSMOD1 initial guess: Lindh’s model Hessian [81]

A model Hessian based on a simple 15-parameter function of the nuclear positions as proposed
by Lindh et al. is used as initial Hessian. Parameters are available for the first three rows of
the periodic table.
R. Lindh, A. Bernhardsson, G. Karlstrom and P.-A. Malmqvist, Chem. Phys. Lett. 241
(1996) 423

HESSMOD2 initial guess: Schlegel’s model Hessian [118, 134] [default]

The initial Hessian is generated by means of a classical model as proposed by Schlegel.

H.B. Schlegel, Theoret. Chim. Acta 66 (1984) 333
J.M. Wittbrodt and H.B. Schlegel, J. Mol. Struct. (Theochem) 398-399 (1997) 55
It adopts a simple valence force field. Empirical rules are used to estimate the diagonal

109
force constants for a set of redundant internal coordinates (stretches, bends and torsions).
Parameters are available from H to At.
Warning - To define bonds the sum of covalent radii (see page 53) is used. For ionic systems
it may be necessary to modify the default values (see RAYCOV, page 53).

HESSNUM initial guess: numerical estimate

C - Hessian updating technique

Different Hessian updating schemes are available for minimization:

BFGS Hessian update - Broyden-Fletcher-Goldfarb-Shanno scheme [19, 20, 50,

59, 120] - [default]

OLDCG Hessian updating - old Schlegel updating scheme[117] (CRYSTAL03)

POWELL Hessian update - symmetric Powell scheme [105]

D - Optimization convergence criteria

These options are available to modify the default values:

TOLDEE threshold on the energy change between optimization steps

• ∗ IG |∆E| < 10−IG (default: 7)

The value of IG must be larger or equal to the threshold adopted for the SCF convergence.
The value is checked when input block 3, defining the SCF convergence criteria, is processed.

TOLDEG convergence criterion on the RMS of the gradient

• ∗ TG max RMS of the gradient (default: 0.0003)

TOLDEX convergence criterion on the RMS of the displacement

• ∗ TX max RMS of the displacement (default: 0.0012)

E - Step control
To avoid the predicted step size being too large, two options are available:
Simple scaling
Simple scaling of the displacement vector. Each component is scaled by a factor that makes
the largest component of the displacement vector equal to 0.5 a.u.
Trust Radius [default]
A more sophisticated and accurate technique to control the step size is the trust radius region

110
scheme. The trust radius limits the step length of the displacement at each cycle, according
to the quadratic form of the surface in the actual region. The default maximum value for
minimization is 0.5.
To run CRYSTAL06 as CRYSTAL09 the keyword ALLOWTRUSTR must be specified along
with BFGS.
To run CRYSTAL09 as CRYSTAL06, the keyword NOTRUSTR must be specified in geom-
etry optimization input
Related keywords are discussed below:

ALLOWTRUSTR activate the trust radius technique to control the step size [0.5 for geometry
optimization; 0.1 for transition state search] [default]

The step at each cycle is computed by means of a Newton-like scheme, while the length is
determined by linear minimization along an extrapolated quadratic polynomial.

MAXTRADIUS optional
• ∗ TRMAX maximum value allowed for the trust radius - default [16]

This is useful in transition state optimizations or in minimizations along flat potential surfaces
in order to avoid too large displacements from one point to the next one. Default value:
geometry optimization: ∞ ; transition states search : 0.3

NOTRUSTR not using trust radius to limit displacement

TRUSTRADIUS
• ∗ TRADIUS set the initial value for trust radius
Set the initial value of the trust radius to [TRADIUS]. The trust radius limitates the step
length of the displacement at each cycle, according to the quadraticity of the surface in the
actual region.
Default: geometry optimization 0.5; transition state search 0.1.

Warning - When the Trust Radius technique is active, the value of the trust radius could
become too small and the geometry optimization process stops with an error message:
”TRUST RADIUS TOO SMALL”.
In this case, we suggest to restart the optimization from the last geometry, written to file
optc(a)xxx, being xxx the optimization cycle number.

F - Coordinate system related options

Geometry optimization can be performed in fractional (default) or redundant internal coor-
dinates (see INTREDUN). Default fractional coordinates are defined as symmetry allowed
directions (atomic positions) and deformations (cell). The latter are related to changes of
isotropic volume and of axial ratios.

Some options related to the choice of the coordinate systems are also available:

111
CRYDEF crystallographic-like symmetrized cell deformations, corresponding to
symmetrized strains of the unit-cell edges (consistent with symmetry).
This set of deformations is useful for fixing lattice parameters in con-
strained optimizations in combination with the keyword FIXDEF (page
118) - 3D only.

FRACTION optimization in fractional coordinates

FRACTIOO optimization in normalized fractional coordinates [default when FUL-

LOPTG is requested]

FRACTCOOR third type of symmetrized fractional coordinates (non-orthogonal; the ori-

gin on polar axes must be explicitly fixed by the FIXCOOR option [to be
used with constraints])

RENOSAED renormalize symmetry allowed deformations [default when FULLOPTG

is requested]

G - Optimization procedure control

EXPDE
• ∗ DE expected energy change used to estimate the initial step [default 10−3 Ha,
if model 1 initial hessian; 10−4 Ha, otherwise]

FINALRUN action after geometry optimization - integrals classification is based on the

last geometry. See page 122
• ∗ ICODE Action type code:
0 the program stops (default)
1 single-point energy calculation
2 single-point energy and gradient calculation
3 single-point energy and gradient calculation - if convergence criteria on
gradients are not satisfied, optimization restarts
4 step 3 is iterated until full stable optimization

FIXDELTE
• ∗ IE 10−ie hartree: threshold on the total energy change for redefining the
geometry to which integral classification is referred - see FIXINDEX,
page 86 - [default -1000, no reclassification]

FIXDELTX
• ∗ DX RMS (bohr) of the displacement for redefining the geometry to which
integral classification is referred - [default: -1, no reclassification]

112
FIXDEIND the reference geometry for integrals classification does not change during
optimization [default choice]

FITDEGR
• ∗ N degree of polynomial fitting function for linear search:
2 parabolic fit [default]
3 cubic polynomial function
4 constrained quartic fitting

HESEVLIM limits for the allowed region of hessian eigenvalues (hartree)

• ∗ VMIN lower limit [default 0.001]
VMAX upper limit [default 1000.]

ITACCONV
• ∗ DE energy difference threshold for ITATOCEL [default 0.1 * TOLDEE be-
tween 2 optimization cycles]

MAXITACE
• ∗ MAXI max number of iteration cycles in atom/cell iterative optimization [default
100]

MAXCYCLE
• ∗ MAX maximum number of optimization steps [default 100]

N.B. When optimization is restarted, the first restarted optimization cycle number is the last
of the previous run + 1. Set MAXCYCLE value accordingly.

NOGUESS SCF guess at each geometry point: superposition of atomic densities at

each SCF calculation (default choice in geometry optimization: GUESSP

NRSTEPS
• ∗ DE number of stored steps to be used in the OLDCG Hessian updating scheme
[default: number of degrees of freedom]

RESTART restart geometry optimization from a previous run.

See page 123

SORT sorting of the previous optimization steps information when the OLDCG
scheme is active [default:nosort]

H - Numerical first derivatives

The nuclear coordinate gradients of the energy can also be computed numerically. A three-point
numerical derivative formula is adopted. A finite positive (and then negative) displacement

113
is applied to the desired coordinate and a full SCF calculation is performed. The gradient is
then computed as
E∆xi − E−∆xi
gi =
2 ∆xi
where ∆xi is the finite displacement along the i-coordinate.
Such a computation is very expensive compared to analytical gradients, since the cost is 2·N ·t,
where N is the number of coordinates to be optimized and t the cost of the SCF calculation.
Numerical first-derivatives should be avoided whenever possible, but sometimes they are the
only way to obtain gradients (i.e. for metals) and therefore to optimize the atoms coordinates.
One choice only, NUMGRCEL, NUMGRATO, NUMGRALL, is allowed.

NUMGRALL geometry optimization - numerical atomic and cell gradient

NUMGRATO geometry optimization - numerical atomic gradients

NUMGRCEL geometry optimization - numerical cell gradients

STEPSIZE modify step for numerical gradient [default 0.001 au] (developers only)
• ∗ STEP new stepsize value

I - Printing options

ONELOG This causes all output to be sent to the standard log file, instead of to
SCFOUT.LOG

NOXYZ printing of cartesian coordinates at the end of optimization removed

NOSYMMOPS printing of symmetry operators at the end of optimization removed

PRINTFORCES printing atomic gradients

PRINTHESS printing Hessian information

PRINTOPT prints information on optimization process

PRINT verbose printing

114
2 - Optimization in redundant internal coordinates

INTREDUN geometry optimization in internal redundant coordinates

An optimization in redundant internal coordinates is performed when the keyword

INTREDUN is inserted in OPTGEOM input block.

Optional keywords related to geometry optimization in redundant internal coordinates must

follow.

A symmetrized set of internal coordinates (i.e. bonds, angles and torsions) is defined, which
contains more coordinates than the requisite internal degrees of freedom.

Redundant internal coordinates are generated according to a hierarchical scheme: bond

lengths are firstly identified by using covalent radii. Then, angles are determined on the basis
of the irreducible set of distances and, finally, dihedral angles are defined. Note that to define
bonds the sum of covalent radii (see page 53) is used. For ionic systems it may be necessary
to reduce the default values (see RAYCOV, page 53). In case of systems constituted by
unconnected fragments (ie some molecular crystals or adsorption complexes), fragments are
linked to each other by pseudo “bond lengths” between the closest pair of atoms belonging to
each fragment.

There has been substantial controversy in recent years concerning the optimal coordinate
system for optimizations.

For molecular systems, it is now well-established that redundant internal coordinates require
fewer optimization steps than Cartesian coordinates. However, this is not definitely demon-
strated for periodic systems. Nevertheless, the use of internal coordinates can be very useful in
several respects: for a chemical intuitive view (e.g. internal coordinates can easily be added),
for constrained geometry optimization (see below) and for searching transition states (under
development).

By default, optimization of internal redundant coordinates involves both atomic positions and
cell parameters. To avoid optimizing cell parameters the keyword FIXCELL page 116 must
be specified.

Before running a geometry optimization in redundant internal coordinates, the set of coordi-
nates generated automatically by CRYSTAL should be checked for consistency. This can be
done by specifying the keyword TESTREDU.

Optional keywords related to the geometry optimization in redundant internal coordinates are
listed below.

Managing with almost linear angles

Linear or almost linear angles (i.e. close to 1800 ) can lead to numerical instabilities in the
computation of the dihedrals. To avoid this problem a common practice is to split the angle
in two ones. The double angles are defined by the angles obtained by projection of the vectors
onto two suitable perpendicular planes, in order to avoid the indetermination around 1800 . The
threshold value, beyond which the almost linear angle is split, is controlled by the keyword
ANGTODOUBLE.

115
ANGTODOUBLE minimum value (degrees) beyond which a double angle is defined
• ∗ AL value of the angle (degrees) - default [1650 ]

The default value is set to 1650 . This means that all angles larger than 1650 are automatically
split into two.
This option can be required, for instance, when optimizing zeolitic structures where siloxane
bridges could change a lot during the geometry minimization. In that case, it is better to
reduce the default value to 1500 .

A list of angle to be converted into two can also be explicitly given by specifying

DBANGLIST list of angles chosen to be converted in double angles - advanced option

• ∗ MU number of angles to convert in double
• ∗ (IN(I), list of the angles
I=1,MU)

This keyword provides the list of angles chosen to be converted in double angles (i.e. defined
by the angles obtained by projection of the vectors onto two suitable perpendicular planes) in
order to avoid the indetermination around 180 degrees). The labels used for the angles are
those provided by a previous automatic generation of internal coordinates computed in a test
run (TESTREDU keyword).

Adding internal coordinates - bonds and angles

When some relevant internal coordinates are missing (e.g. intermolecular bonds) they can be
added by means of two keywords: DEFLNGS and DEFANGLS.

DEFLNGS definition of bond lengths

• ∗ NL number of bonds to be added
insert NL sets of 5 data to define the bond AB II
LA label of the atom A (it must be in the reference cell)
LB label of the atom B
I1, I2, I3 indices of the cell where the atom B is located

DEFANGLS definition of bond angles

• ∗ NL number of angles to be defined
insert NL sets of 9 data to define the angle ABC
\ II
LA label of the atom A (it must be in the reference cell)
LB label of the atom B
I1, I2, I3 indices of the cell where the atom B is located
LC label of the atom C
I1, I2, I3 indices of the cell where the atom C is located

Other optional keywords

FIXCELL keep cell parameters fixed in internal coordinates optimization

116
STEPBMAT step used for numerical bmat calculation (developers only)
• ∗ I integer - step = 10I (default 7: step=107 )

TESTREDU request test run for checking automatic definition of internal coordinates
.

TOLREDU tolerance used to eliminate redundancies (developers only)

• ∗ I tolerance 10−I (default: 7, 10−7 )
.

3 - Geometry optimization with constraints

Along with an unconstrained relaxation of the crystalline structure, options are available to
perform different optimizations with constraints.
In particular:
A - Constant volume optimization (page 117)
B - Fixing lattice deformations (page 118)
C - Linear constraints between atomic coordinates (page 119)
D - Partial optimization of atomic positions (page 119)
E - Fixing internal coordinates (page 120)
All constraining strategies are compatible with any choice of coordinate system adopted for the
optimization process to perform the optimization process. On the other hand, option E is only
operative together with the choice of a redundant internal coordinate system (INTREDUN
page 115).
The examples at the CRYSTAL Tutorial Project web page illustrate the use of the available
keywords for constrained geometry optimizations.

A - Constant volume optimization

CVOLOPT constant volume optimization.

Only active with CELLONLY (cell parameters only optimization) or FULLOPTG (atom
coordinates and cell parameters optimization).
The volume is kept fixed at the value corresponding to the input unit cell; all cell angles and
ratios between cell edges unconstrained by the point-group symmetry are optimized.
Examples: in the tetragonal symmetry, only the c/a ratio, and in the monoclinic symmetry
the a/b and b/c ratios and the beta angle, respectively, are optimized.

This option is useful for computing E vs V curves point-by-point by relaxing the crystalline
structure at different values of the cell volume. In this case, the keyword FIXINDEX must be
used to obtain a smooth curve. The reference geometry must correspond either to the smallest
volume to be explored, or to the equilibrium structure obtained from a prior optimization run
(FULLOPTG).
Warning: if large changes of the individual unit-cell parameters occur in the optimization
process, the linear strain approximation may not be strictly obeyed and very small volume
variations (of the order of 0.01%) may ensue.

117
B - Fixing lattice deformations
Linear constraints between unit cell deformations can be set up during optimization by means
of the keyword FIXDEF:

FIXDEF optimization with constrained symmetrized cell deformation - 3D only

• ∗ NFIXC number of constraints relating pairs of cell deformations
insert NFIXC records II
• ∗ LA,LB integer sequence number of the two constrained symmetrized cell defor-
mations.
CA,CB real coefficients multiplying the two cell deformations in the linear com-
bination constraint. If LA=0, the cell deformation denoted by the second
integer (LB) is kept fixed during the optimization (the coefficients in this
case can take any value).

FIXDEF can also be combined with the keyword CRYDEF, that sets crystallographic-like
cell deformations (i.e. a, b, c, α, β, γ) to fix lattice parameters. Integer sequence number given
as input refer to the minimal set of lattice parameters:

1 2 3 4 5 6
cubic a
hexagonal a, c
rhombohedral hexagonal cell a, c
rhombohedral cell a, α
tetragonal a, c
orthorhombic a, b, c
monoclinic a, b, c, β
a, b, c, γ
a, b, c, α
triclinic a, b, c, α, β, γ

Note that the labels of the symmetry allowed deformations must correspond to the ones
printed in the output file.

As an example, a constrained optimization of the crystalline structure of α-quartz (hexagonal)

with the c unit cell edge kept fixed follows
QUARTZ ALFA STO-3G
CRYSTAL
0 0 2
154
0 0 16
4.916 5.4054
2
14 0.4697 0. 0.
8 0.4135 0.2669 0.1191
OPTGEOM
FULLOPTG
CRYDEF
FIXDEF
1
0 2 0.0 0.0 : the second lattice parameter, c, is kept fixed
ENDOPT
END

118
C - Linear constraints between atomic coordinates
Linear constraints between atomic coordinates can be set up during optimization by using the
keyword FIXCOOR.

FIXCOOR optimization with constrained symmetrized coordinates

• ∗ NFIX number of constraints relating pairs of coordinates
insert NFIX records II
• ∗ LA,LB integer sequence number of the two constrained symmetrized coordinates
(sequence numbers are read from the output of PRSYMDIR)
CA,CB real coefficients multiplying the two coordinates in the linear combination
constraint. If LA=0, the coordinate denoted by the second integer (LB)
is kept fixed during the optimization (the coefficients in this case can take
any value).

Note that the labels of the symmetry allowed directions must correspond to the one printed
in the output file (PRSYMDIR keyword for coordinates).

In the following example on α-quartz, two constraints are set up on coordinates

QUARTZ ALFA - Linear constraints between atomic coordinates

CRYSTAL
0 0 2
154
0 0 16
4.916 5.4054
2
14 0.4697 0. 0.
8 0.4135 0.2669 0.1191
OPTGEOM
FULLOPTG
FRACTCOOR
FIXCOOR
2
2 3 1.0 1.0
0 4 0.0 0.0
ENDOPT
END

1. The x and y fractional coordinates of Oxygen are forced to change by the same amount,
so that their difference remains constant.

2. The z coordinate of Oxygen is kept fixed.

In general, any of the structural parameters can be kept fixed in the optimization process by
the combined use of FIXCOOR and FIXDEF keywords.

D - Partial optimization of atomic positions

FRAGMENT Partial geometry optimization (default: global optimization)

• ∗ NL number of atoms ”free”
• ∗ LB(L),L=1,NL label of the atoms to move

119
Optimization is limited to an atomic fragment. Symmetrized cartesian coordinates are gen-
erated according to the list of atoms allowed to move. Note that no advantage is taken in
the gradient calculation to reduce the number of atoms, i.e. gradients are calculated on the
whole system. The symmetrized forces are then computed by using the new set of symmetrized
coordinates. See example in section 9.4, page 206.

E - Fixing internal coordinates

Constraints on internal coordinates can be easily imposed during geometry optimization.
The following two options allow users to both define and freeze one or more bond lengths or
angles:

LNGSFROZEN explicitly freezes bond lengths

• ∗ MU number of bond lengths to freeze
insert NL sets of 5 data to define the bond AB II
LA label of the atom A (it must be in the reference cell)
LB label of the atom B
I1, I2, I3 indices of the cell where the atom B is located

ANGSFROZEN definition of bond angles to be frozen

• ∗ NL number of angles to be frozen
insert NL sets of 9 data to define the angle ABC
\ II
LA label of the atom A (it must be in the reference cell)
LB label of the atom B
I1, I2, I3 indices of the cell where the atom B is located
LC label of the atom C
I1, I2, I3 indices of the cell where the atom C is located

According to the list of redundant internal coordinates automatically generated by the code,
bond lengths or angles can also be frozen by means of the FREEZINT option:

FREEZINT freeze internal coordinates (active with INTREDUN only):

• ∗ NB first NB bond length are frozen
NA first NA bond angles are frozen
ND first ND dihedral angles are frozen (not active)

The list of redundant coordinates can be obtained from a prior run, by inserting the keyword
TESTREDU (the program stops after printing internal coordinates).

Note that for a better control over the selected frozen internal coordinates we suggest using
the keywords LNGSFROZEN and ANGSFROZEN.

Constraint optimization combining internal coordinates and fractional coordinates can also be
performed.

For instance, one can keep fixed a bond angle together with the constraint that the x and y
fractional coordinates of a given atom change by the same amount. Such a combination of
constraining strategies must be used with caution, as it may lead to undesired behavior in the
optimization process.

The constraining of internal coordinates is performed with numerical techniques (particu-

larly in the back-transformation from redundant internal to cartesian coordinate systems)

120
and the fixed values may be affected by some small changes (in general of the order of 10−4 au).

The following example corresponds to a rigid tetrahedral geometry optimization of α-quartz:

QUARTZ ALFA fixing internal coordinates

CRYSTAL
0 0 2
154
0 0 16
4.916 5.4054
2
14 0.4697 0. 0.
8 0.4135 0.2669 0.1191
OPTGEOM
INTREDUN
LGNSFROZEN
2
4 1 0 0 0
5 1 -1 0 0
ANGSFROZEN
4
4 1 0 0 0 7 0 0 0
4 1 0 0 0 5 1 0 0
4 1 0 0 0 8 1 0 0
5 1 -1 0 0 8 0 0 0
ENDOPT
END

The two independent Si-O bond lengths and then the four O-Si-O angles of the SiO4 tetrahe-
dron are frozen in order to relax just the Si-O-Si bridges and the dihedral angles.

FREEZDIH freeze a list of dihedral (active with INTREDUN only):

• ∗ NDH number of dihedrals to be frozen
• ∗ IFR(I), list of dihedrals to be frozen
I=1,NDH)

The list of dihedrals, to choose the ones to be frozen, is obtained performing a previous run
with the keyword TESTREDU into OPTGEOM input block (the program stops after the
printing of the internal coordinates, see page 117).
After using FREEZINT (page 120) it turns out that the order of the dihedral angles in the
output changes: the frozen parameters appear at the beginning of the list. All of them are
now labeled as frozen (”FRZ=T”) in the printed list of redundant coordinates.

4 - Geometry optimization with application of an external

stress
Geometry optimization can be done under an hydrostatic pressure (EXTPRESS) or a general
external stress. The choice of the external stress is controlled by keywords.

121
EXTPRESS to apply external, hydrostatic pressure
• ∗ pres pressure in hartree/bohr3

EXTSTRESS to apply a general external stress - stress matrix follows

• ∗ D11 D12 D13 first row of the stress matrix.

• ∗ D21 D22 D23 second row of the stress matrix.
• ∗ D31 D32 D33 third row of the stress matrix.

The stress matrix, input by row - must be a symmetric 3x3 matrix, all elements in
hartree/bohr3 )

EXTSTRESS may require to reduce the symmetry, e.g. when you apply uniaxial pressure to
a cubic system, then you have to break the symmetry. The keyword BREAKELAS (geometry
input, input block 1, page 35) breaks the symmetry according to a general distortion, but does
not perform the distortion.‘

EXTPRESS
0.001 hydrostatic pressure of hartree/bohr^3 s applied

EXTSTRESS the distortion matrix allows for instance a

0.001 0. 0. cubic lattice to become tetragonal‘
0. 0. 0.
0. 0. 0.

Notes on geometry optimization

On the integrals classification during a geometry optimization
Truncation of infinite Coulomb and exchange series, based on the overlap between two atomic
functions (see chapter 11.11), depends on the geometry of a crystal. With default thresholds,
different selection of integrals are evaluated with different geometries. This introduces small
discontinuities in the PES, producing artificial noise in the optimization process. To avoid
noise in interpolation of PES, the FIXINDEX option is always active during optimization.
The adopted selection pattern refers to the starting geometry.
If equilibrium geometry is significantly different from the starting point, reference truncation
pattern may be inappropriate and the use of proper truncation becomes mandatory.
Since both total energy and gradients are affected by the integrals classification, a single-point
energy calculation ought to be run always with the final structure, and integrals classified
according to the new final geometry, to calculate correct total energy and gradients.
If during the final run the convergence test on the forces is not satisfied, optimization has to be
restarted, keeping the integrals classification based on the new geometry. The FINALRUN
option has been implemented to this aim.
The four different options of FINALRUN allow the following actions, after classification of
integrals:

1. single-point energy calculation (correct total energy),

2. single-point energy and gradient calculation (correct total energy and gradients),

3. single-point energy and gradient computation, followed by a new optimization process,

starting from the final geometry of the previous one, (used to classify the integrals), if
the convergence test is not satisfied.

122
 4. step 3 is iterated until full stable optimization
If the starting and final geometry are close, the energy and gradient calculated from the final
geometry, with integral classification based on the initial geometry, are not very different from
the values obtained with correct classification of integrals. In some cases (e.g. optimization
of the geometry of a surface, with reconstruction) the two geometries are very different, and
a second optimization cycle is almost mandatory (ICODE=3 in FINALRUN input). This is
strongly recommended.

Optimization of flat surfaces

Flat regions of surfaces often behave as non quadratic. This may give rise to erratic opti-
mization paths when using the linear minimization to control the step length. In these cases
it is recommendable using the trust radius strategy set by the keyword ALLOWTRUSTR.
Under this scheme the step is controlled so as to never go out from the quadratically behaved
regions of the surface (the trust regions). Additionally, one can set the maximum trust radius
to a given value MAXTRADIUS [def ∞], in order to avoid too large displacements from one
point to the next one.
Additional combined test on gradient and energy are adopted for treating special cases:
1. If the gradient criteria are satisfied (but not the displacement criteria) and the energy
difference between two steps is below a given threshold (see TOLDEE), the optimization
stops with a warning message;
2. If both the gradient and displacements criteria are not satisfied, but the energy does not
change (TOLDEE parameter) for four subsequent steps, the optimization stops with a
warning message.

Restart optimization
Restart of geometry optimization is possible for a job which is abruptly terminated (e.g. number
of steps exceeded, available cpu time exceeded,...).
The optimization restarts from the last complete step of the previous run.
The geometry at each step is written to file optc number of step, and can be read by EXTER-
NAL (see page 15).
If optimization ended successfully, the optimized geometry is written to file fort.34 (EXTER-
NAL format).
When restarting an optimization, information on previous optimization steps is read from file
OPTINFO.DAT. Optimization then proceeds, saving information in the same file at each step.
The SCF guess, read from file fort.20, is the density matrix written in file fort.9 at the end of
SCF of the last successful step of the optimization process.
The same input deck as for the initial geometry optimization must be used when
the RESTART keyword is added.

Visualizing the optimization process

CRYSTAL output is read by the software MOLDRAW:
http://www.moldraw.unito.it to visualize the optimization process.

File fort.33 contains the geometry at each step, in xyz format.

5 - Searching a transition state

TSOPT transition state search requested [default: false]

123
Transition state optimization is invoked by the keyword TSOPT in the OPTGEOM input
block.
The reference to be quoted is:
C. M. Zicovich-Wilson, M. L. San-Romn, A. Ramrez-Sols,
Mechanism of F− Elimination from Zeolitic D4R Units: A Periodic B3LYP Study on the
Octadecasil Zeolite
J. Phys. Chem. C 114 (2010) 2989-2995.
An example of transition state search is presented in http://www.crystal.unito.it/tutorials
By default the eigenvector that corresponds to the lowest eigenvalue is followed uphill according
to the scheme proposed by Simmons and Nichols [Simmons, J., and Nichols, J.: , Int. J.
Quantum Chem. Symp. 24, volume 24, 263, (1990)].
To adopt other choices for directions to be followed uphill see keywords: MODEFOLLOW,
PATHFOLLOW, FITTOPATH and CHNGTSFOL.

Transition state control keywords

MODEFOLLOW
• ∗ MODEFOL mode to follow

ABS(MODEFOL) is the label of the eigenvector to be followed uphill initially, DIR(0).

If MODEFOL < 0, the initial uphill direction, DIR(0), is the opposite to that of the eigenvector
of label ABS(MODEFOL)
In a general optimization step, NSTEP, the current uphill direction DIR(NSTEP) is chosen
as the hessian eigenvector of maximum overlap with the direction chosen in the previous step,
DIR(NSTEP-1). In this scheme the uphill direction is allowed to smoothly change along
the optimization. Some problems might appear when there are quasi-degeneracies between
the eigenvalue of the uphill direction and other that corresponds to a direction to be followed
downhill. In such a case the optimization might go in troubles. (Improvements still in progress)

PATHFOLLOW only with redundant internal coords

• ∗ NPATHFOL max coord to choose the mode to follow

Only valid together with INTREDUN.

The uphill direction is the eigenvector that has maximal absolute contribution of the internal
valence coordinate labeled ABS(NPATHFOL), which is thus supposed to dominate the reaction
coordinate.
If NPATHFOL < 0, the uphill search is such that the value of coordinate ABS(NPATHFOL)
decreases along the reaction coordinate, otherwise the opposite direction is chosen.
At variance with the MODEFOLLOW case, where the reference direction changes from step
to step, here the same strategy is employed in every step of the optimization. This prevents
troubles when near-degeneracies occur (see keyword MODEFOLLOW).

FITTOPATH only with redundant internal coords

• ∗ NPATHFOL2 integer
NPATHWEIGHT integer

Only valid together with INTREDUN and PATHFOLLOW

ABS(NPATHFOL2) is the label of a second valence coordinate, coordinate (II), that together
with the one labeled ABS(NPATHFOL), coordinate (I), mostly contributes to the reaction

124
coordinate.
Once the eigenvector with maximum contribution of coordinate (I), namely XMAXCONTR,
is obtained (see PATHFOLLOW keyword), the eigenvectors are once more scanned and those
having a contribution of this coordinate larger in absolute value than XMAXCONTR*(100-
NPATHWEIGHT)/100 selected. If NPATHFOL2 > 0 the previously selected eigenvector with
maximum contribution of coordinate (II) with the same sign of XMAXCNTR is chosen as uphill
direction. Otherwise, the one with maximum contribution having opposite sign is considered.

CHNGTSFOL only with redundant internal coords

Valid together with INTREDUN, PATHFOLLOW (FITTOPATH).

The optimization follows uphill the path according to the PATHFOLLOW (+FITTOPATH)
scheme while the chosen eigenvector is not the first one in the list ordered by increasing eigenval-
ues, i.e. it has the lowest eigenvalue. Once this occurs, the scheme changes to MODEFOLLOW
in the following steps and the uphill direction is chosen according to the criterion of maximum
overlap with the previous uphill directionr.

Scan keywords

SCANATOM
• ∗ NATSCAN (integer): label of the atom to be scanned
TARGET (real array dim 3): last position of the atom in the scan
MSCAN number of steps in which the previous displacement is carried out

Only for P1 structures.

Perform a series of optimizations in which one atom is kept fixed at different contiguous posi-
tions and the remainder of the structure fully or partially relaxed.
Compatible with the optimization of atomic positions (default) and atoms+cell (FUL-
LOPTG).
Not compatible with INTREDUN.
This directive must be used associated with FRAGMENT. The atom to be scanned together
with an other atom that belongs to other moiety of the system must be kept fixed using
FRAGMENT (the remainder atoms can be relaxed).

SCANREDU To be used with INTREDUN only

• ∗ IREDSCA (integer): type of valence coordinate to be scanned (1, bond length; 2
angle; 3 dihedral)
ENDSCA (real): last value taken by the chosen coordinate along the scan The initial
value is the current one with the geometry defined for the structure.
MAXSCA number of points considered in the scan

Perform a series of optimizations (scan) in which one (or two) redundant valence internal
coordinate(s) are kept fixed at different values while the remainder are fully relaxed.
To be used only with INTREDUN.
The directive SCANREDU must be accompanied with freezing the redundant valence internal
coordinate(s) one wants to scan (see keywords FREEZINT (page 120), FREEZDIH (page
121), ANGSFROZEN (page 120), LNGSFROZEN (page 120).
According to the order of the coordinates given in the output (see INTREDUN, TESTREDU)
the scan is performed on the last frozen coordinate (indicated in the output by a ”T”) of type
IREDSCA. If SCANREDU is requested twice with the same IREDSCA, the second time it
refers to the last but one frozen coordinate of type IREDSCA.

125
SCANREDU may be requested at maximum twice so as to carry out a bidimensional scan.

126
Chapter 4

Vibration calculations at Γ point

FREQCALC - Frequency harmonic calculation

Harmonic frequency calculation at Γ is invoked by the keyword FREQCALC in input block
1 (geometry). FREQCALC must be the last keyword in geometry input. FREQCALC
input block admits several options (subkeywords), and terminates with keyword END (or
END[FREQ], END[—]: the first three characters only are processed).
References to be quoted when using this module:
F. Pascale, C.M. Zicovich-Wilson, F. Lopez, B. Civalleri, R. Orlando, R. Dovesi
The calculation of the vibration frequencies of crystalline compounds and its implementation
in the CRYSTAL code., J. Comput. Chem. 25 (2004) 888-897
C.M. Zicovich-Wilson, F. Pascale, C. Roetti, V.R. Saunders, R. Orlando, R. Dovesi
The calculation of the vibration frequencies of alpha-quartz: the effect of Hamiltonian and
basis set., J. Comput. Chem. 25 (2004) 1873-1881

Besides harmonic frequency calculation at Γ [default] it allows:

1. Calculation of IR intensities (keyword (INTENS, page 133).

2. Scanning of geometry along selected normal modes (SCANMODE, page 136).

3. Calculation of the IR reflectance spectrum (REFLECTANCE, page 140).

4. Harmonic calculation of phonon dispersion (DISPERSION, page 144.

The second derivatives of the energy are computed numerically by using the analytical first
derivatives. Frequencies are obtained by diagonalizing the mass-weighted Hessian in cartesian
coordinates.

Symmetry exploitation
The point group symmetry of the lattice is used to reduce the number of SCF+gradient cal-
culation to be performed. At each point, the residual symmetry is exploited for the SCF
calculation.
Second derivatives calculations are done on the irreducible atoms only.
The full hessian matrix is then generated by applying the point group symmetry to the irre-
ducible part.
The mass-weighted hessian matrix is diagonalized to obtain eigenvalues, which are converted
in frequencies (cm−1 ), and eigenvectors, i.e. the normal modes.

127
Geometry
The first step to compute frequencies is the calculation of the wave function at the equilibrium
geometry. The geometry of the system must correspond to a stationary point on the potential
energy surface.
Geometry optimization can be controlled by two keywords:

NOOPTGEOM do not perform previous optimization in FREQCALC job [default]

PREOPTGEOM perform optimization before starting the Vibrational modes calcula-

tion.
insert OPTGEOM keywords (close with END) II

keyword END II

An input block is open that must finish with keyword END . The numerical conditions for
the optimization are controlled by means of the same keywords as documented in page 105
(keyword OPTGEOM).
The conditions adopted by default in geometry optimization before frequency calculation are
different than those considered for normal optimizations, to obtain much more accurate min-
ima numerical second derivatives. This ensures a good accuracy in the computation of the
frequencies and modes. The defaults are:

TOLDEG 0.00003
TOLDEX 0.00012
FINALRUN 4
MAXTRADIUS 0.25
TRUSTRADIUS .TRUE.

If frequency calculation is restarted (keyword RESTART, page 133) the input geometry must
be the final optimized geometry found by PREOPTGEOM.

Default choices
SCF guess for wave function calculation for all subsequent geometries defined to compute
numerical second derivatives is the density matrix obtained at equilibrium geometry.

The default value for SCF convergence criterion on total energy is set to 10−9 (TOLDEE in
block3 input to modify it).
The default choice for DFT grid, when DFT Hamiltonian is used, corresponds to XLGRID
(page 79). It was LGRID in CRYSTAL06.

The calculation of longitudinal optical (LO) frequencies and IR intensities is not performed by
default. If the INTENS (page 133) keyword is used, intensities are evaluated.
The frequency input block must be closed by the keyword END (or ENDFREQ). All the
keywords are optional.

Output files
Files written during frequency calculation, to be saved to restart a calculation.

SCFOUT.LOG The output from the wave function and gradient calculation is printed in
standard output for the reference geometry only. The output is then written in file
SCFOUT.LOG.

FREQINFO.DAT Formatted. Contains information on the hessian. Updated at each point,

it is necessary to restart a frequency calculation.

128
HESSFREQ.DAT Formatted. Contains the hessian in cartesian coordinates to be read by
HESSFREQ in geometry optimization input block..
fort.9 Binary. Wave function computed at the equilibrium geometry. Full symmetry exploited
by default. When those data are used to restart, file fort.9 is read as file fort.20 (SCF
guess).
fort.13 Binary. Reducible density matrix at central point. To exploit maximum symmetry in
numerical second derivatives calculations.
fort.28 Binary. Data for restart of IR intensities calculation through Berry phase approach.
fort.80 Binary. Localized Wannier functions, computed only if IR intensities are computed
through Wannier functions.

Optional keywords

• A ANALYSIS Analysis of the vibrational modes

• A DIELISO to be used if INTENS is active. Calculation of the LO/TO shifts by

using the dielectric tensor. The isotropic dielectric tensor (dielectric
constant) should be calculated previously with options CPHF (page
145) or (SUPERCELL/FIELD) and (DIEL) applied for each axis of
the system.
• ∗ DIEL dielectric constant

• A DIELTENS to be used if INTENS is active. Calculation of the LO/TO shifts by

using the dielectric tensor.
• ∗ TENS(1:9) Dielectric tensor matrix TENS (3x3 elements, input by rows: 9 reals
(3D).
The dielectric tensor elements can be obtained from the literature
or computed with CRYSTAL using CPHF (page 145) or SUPER-
CEL/FIELD (page 41).
It must be computed in order to obtain LO modes. See keyword
INTENS (page 133).

• A FRAGMENT Frequency calculation on a moiety of the system

• ∗ NL number of atoms active for frequencies
• ∗ LB(L),L=1,NL label of the active atoms

Frequency calculation can be limited to an atomic fragment, instead of the whole system.
Symmetry is removed. If a fragment contains symmetry related atoms, they must be explicitly
defined. A reduced hessian is computed, according to the list of atoms belonging to the
fragment. A chemically sound moiety of the system must be considered to avoid random
results.

• A ISOTOPES atomic masses modified

• ∗ NL number of atoms whose atomic mass must be modified
II insert NL records II
• ∗ LB,AMASS label and new atomic mass (amu) of the atom.
II II

129
When the isotopic mass of one atom symmetry related to others is modified, the symmetry of
the electronic wave function is not modified, as the mass of the atoms is not present in the
single particle electronic Hamiltonian. For instance, if in a methane molecule (point group Td )
we want to substitute H with D, we can redefine the mass of the 1, 2, 3, 4 Hydrogen atoms; if
C is the first atom, the corresponding input are:
1 H atom 2 H atoms 3 H atoms 4 H atoms

ISOTOPES ISOTOPES ISOTOPES ISOTOPES

1 2 3 4
2 2.000 2 2.000 2 2.000 2 2.000
3 2.000 3 2.000 3 2.000
4 2.000 4 2.000
5 2.000

If a single D is inserted, the symmetry is reduced, (point group C3v ), the three-fold degeneracy
becomes two-fold. When all the four Hydrogens are substituted, the three-fold degeneracy is
restored.
If a frequency calculation was performed with standard atomic masses, new frequencies values
with different atomic masses for selected atoms can be computed from the hessian already
computed, at low computational cost, by inserting the keyword RESTART in FREQCALC
input block, and supplying the file FREQINFO.DAT written by the previous run.

130
• A MODES Printing of eigenvectors [default]

• A NEGLEFRE Reads the number of lowest vibration frequencies to be neglected

in the computation of the thermodynamical properties. By default,
the (rotational)+translational degrees of freedom are automatically
neglected. This keyword allows to neglect further soft vibrations with
low frequencies which may carry numerical issues on the computation
of thermodynamical properties.
• ∗ N Number of frequencies to be neglected

• A NOANALYSIS No analysis of the vibrational modes [default]

• A NOINTENS No calculation of the IR intensities [default choice].

• A NOMODES No printing of eigenvectors

• A NORMBORN Normalize Born tensor to fulfill sum rule

• A NOUSESYMM Symmetry is removed, the space group is set to P1

• A NUMDERIV specifies the technique to compute the numerical first-derivatives

h(x)=dg(x)/dx of the gradient g(x)=dE(x)/dx
• ∗ N 1 different quotient formula: h(x)=(g(x+t)-g(x))/t t=0.001 Å
(one displacement for each atom along each cartesian direction)
2 Central-difference formula: h(x)=(g(x+t)-g(x-t))/2t t=0.001 Å
(two displacements for each atom along each cartesian direction)

• A PRESSURE Pressure range for thermodynamic analysis

• ∗ NP,P1,P2 3 reals, NP is the number of intervals in the pressure range P1 to P2
(GPa) [1,0.101325,0.101325]

• A PRINT Extended printing active (hessian and hessian eigenvectors)

131
• A RESTART Restart frequency calculation from a previous run. See page 133.

• A STEPSIZE Modify the step size of displacement along each cartesian axis
• ∗ STEP step (Å) for numerical derivatives [0.003]

• A TEMPERAT Temperature range for thermodynamic analysis

• ∗ NT,T1,T2 3 reals, NT is the number of interval in the range T1 to T2 tempera-
ture (K) [1,298.0,298.0]

• A TEST[FREQ] Frequency test run

• A USESYMM Maximum space group symmetry used to compute wave function at

each point [default]

Partition of the modes into Building Unit contributions

The external portion of the motion of the b-th Building Unit (BU) in mode i is quantified by
means of
2
P(b) ei
bi = , (4.1)
Nbi
(b)
where matrix P(b) ≡ PAα,Bβ , with atoms A, B belonging to the b-th BU, is the projector
onto the
Proto-translational
P3 degrees of freedom of the unit taken as an isolated fragment and
2
Nbi = A∈b e
α=1 i,Aα is a normalization factor. Accordingly, the corresponding internal
contribution is given by 1 − bi .
The contribution of the b-th BU to mode i is computed as
3 X 2
X ei,Aα
ξbi = , (4.2)
α=1
mA
A∈b

where mA is the number P of BUs to which atom A belongs. If the mode vector is normalized
Eq. (4.2) ensures that b ξbi gives the portion of the mode covered by the partition (1 if a full
partition is considered). By adopting such partitions, the internal and external contributions
(n) (x)
(per cent) of the b-th BU to mode i are given by Γbi = (1 − bi ) × ξbi × 100 and Γbi =
bi × ξbi × 100, respectively. In some cases one may be interested in considering the overall
external contribution under a given BU partition. This is given by
P (x)
Γ
Ξi = Pb bi , (4.3)
b ξbi
P
where i Ξi is not 100%, but the percentage of the structure covered by the BUs considered
in the partition.
Though this analysis may be somehow arbitrary, the resulting indices provide a systematic
and clear description of most of the significant features of the vibrational modes of the system
under study.

The keyword BUNITSDECO performs a building unit decomposition of the vibrational modes.
The vibrational modes are decomposed in terms of internal and external motions of some units
defined by input. The latter coorrespond to rotations and translations of the units behaving
like rigid, while the former to the relative motions of the constitutive atoms.

132
BUNITSDECO perform a building unit decomposition of the vibrational modes.

• ∗ NBDNGUNIT number of building units irreducible by symmetry considered (the

units symmetry-equivalent are automatically generated)
• ∗ MBDNGUNIT(I), number of atoms of each unit. The sum defines NATOMS, the
I=1,NBDNGUNIT total number of atoms considered
• ∗ (LBDNGUNIT(JA,IU), identification of the atoms: for each atom, the sequence number
JA=1,4),IU=1,NATOMSand three cell indexes, in the order given in MBDNGUNIT

Restart a calculation
A frequency calculation for a job abruptly terminated (e.g. machine crash, exceeded the
available cpu time,....). can be restarted exactly from the last point of the previous run.
The same input deck used for the incomplete calculation, with the keyword RESTART in
the FREQCALC input block is submitted. The following files, written by the previous job,
must be present:

FREQINFO.DAT formatted - information on the part of the hessian already computed.

fort.20 binary - wave function at the equilibrium geometry, with no symmetry, as guess for
SCF process (fort.9 saved at the end of single point calculation).

fort.28 (binary) Data for restart of IR intensities calculation through Berry phase approach.

fort.80 (binary) localized Wannier functions (if IR intensities through Wannier cunctions are
computed).

fort.13 binary - Reducible density matrix at central point. To exploit maximum symmetry
in numerical second derivatives calculations.

IR intensities calculation must be present in the first frequency calculation, it can not be
inserted in restart only.
The restart option can be used to modify the algorithm used to compute gradients (switch from
different quotient formula to Central-difference formula, keyword NUMDERIV). In this case
the new points only are calculated. The same input deck as for the initial frequency calculation
must be used.
Restart can be used to evaluate frequencies for a system with different isotopes of selected
atoms (keyword RESTART followed by ISOTOPES 129).

1 - IR intensities
Calculation of IR intensities is invoked by the keyword INTENS in FREQCALC input block.
Two different techniques can be adopted:

• IR intensities through Berry phase - keyword INTPOL [default choice in CRYSTAL09]

• IR intensities through Wannier functions - keyword INTLOC [default choice in CRYS-

TAL06]

In order to obtain the LO modes, the high frequency dielectric tensor must be provided. See
keyword DIELTENS, page 129.
The dielectric tensor elements can be obtained from the literature or computed with CRYSTAL
using CPHF (page 145) or SUPERCEL/FIELD (page 41).

133
- IR intensities through Berry phase [default CRYSTAL09]
Calculation of IR intensities through Berry Phase approach, keyword INTPOL, is the default
choice.
This is possible for 3D, 2D, 1D and 0D systems, but only for insulating system.
Born charges are the key quantities for the calculation of the IR intensities, the longitudinal
optical (LO) frequencies and the static dielectric tensor. The Berry phase approach consists
in evaluating the Born charges, that is the derivative of the dipole moment with respect to
the atomic displacements, as polarization differences between the central and the distorted
geometries: the polarization difference is then equal to the time-integrated transient macro-
scopic current that flows through the insulating sample during the vibrations.
The scheme operates on the crystalline-orbital eigenfunctions in the reciprocal space. As a
consequence of that, the accuracy of IR intensities might be sensitive to the density of the
Monkhorst net.
There are no additional keywords related to this method.

- IR intensities through Wannier functions [default CRYSTAL06]

Calculation of IR intensities through Wannier functions (default of CRYSTAL06) is invoked
by the keyword INTLOC following INTENS in FREQCALC input block.
Many keywords are related to the Wannier functions calculation, and should be used by devel-
opers and very experienced users only.
If keyword INTLOC is activated in FREQCALC input block, IR intensities, Born charges
and LO-TO split are evaluated through the Wannier functions, obtained by localizing the
Crystalline Orbitals. This is possible for insulators only.
IR intensities calculation through localization is very demanding, in terms of memory alloca-
tion. NOINTENS, default choice, avoids intensity calculation, when not necessary.
As regards the computation of the IR intensities, they are obtained by means of the Wan-
nier Function (WnF) approach, in which those functions span the occupied manifold and are
explicitly constructed in real space. They are at time obtained from the eigenvectors of the one-
electron Hamiltonian (Bloch Functions) by numerical integration in reciprocal space through
the definition of a Pack-Monkhorst net. The system must be an insulator. By default the
dipole moment in the non central points are computed with Wannier Functions that are the
projection onto the occupied space of the current point of those obtained by localization at the
central point. If RELOCAL is requested these WFs are relocalized at each point
This procedure leads not to real WnFs, but to an approximation contained into a cyclic space.
In the mapping (unfolding) that permits to convert cyclic to real WnFs, CRYSTAL exploits
the classification of the lattice vectors made at the very beginning of the SCF calculation that,
obviously, does not involve the infinite space, but just a cluster of a finite number of cells,
ordered by increasing length (i.e. it covers a close to spherical region of the real space).
In all the tested cases, this classification provides sufficient room to represent the matrices
needed in the SCF part within the required accuracy. This is also so in what concerns the
(post-SCF) computation of the WnFs, apart from very particular cases in which the primitive
cell is oblong and the corresponding unfolded cyclic cluster associated to the Monkhorst-Pack
net (also very elongated in one direction) does not fit into the real cluster (always close to
spherical shape).

A set of keywords can be used to modify the localisation process (see properties input, keyword
LOCALWF, page 169) They are entered after the DIPOMOME keyword. Modification of
default choices is not recommended, it should be restricted to developers only.
The keyword DIPOMOME defines an input block (closed by END) with keywords allowing

134
modification of the localization process.
To be modified by developers only.
• A DIPOMOME Calculation of the dipole moment - see Localisation part (properties,
keyword LOCALWF, page 169
To be modified by developers only.
• A END end of the DIPOMOME block.
all keywords are optional II
• A DMACCURA (Optional) Change the final dipole moment tolerance
• ∗ NTOL Value of the new tolerance as TOLWDM=0.1−N T OL

• A RELOCAL (Optional) Relocalize all points in frequency calculations

• A BOYSCTRL see LOCALWF, page 171
• A CAPTURE see LOCALWF, page 172
• A WANDM see LOCALWF, page 175
• A FULLBOYS see LOCALWF, page 175
• A MAXCYCLE see LOCALWF, page 171
• A CYCTOL see LOCALWF, page 170
• A ORTHNDIR see LOCALWF, page 175
• A CLUSPLUS see LOCALWF, page 174
• A PHASETOL see LOCALWF, page 170
• A RESTART see LOCALWF, page 170
• A IGSSBNDS see LOCALWF, page 172
• A IGSSVCTS see LOCALWF, page 171
• A IGSSCTRL see LOCALWF, page 171

135
 2 - Scanning of geometry along selected normal modes
Scanning of geometry along selected normal modes is invoked by the keyword SCANMODE
in FREQCALC input block. Preliminary frequency calculation is required to single out the
selected mode.
rec variable meaning

• ∗ NMO |N M O| number of modes to be scanned.

> 0 SCF calculation at each point along the path - energy is computed
< 0 only the geometry along the path is computed (no SCF calculation)
INI Initial point for the scan
IFI Final point for the scan
STEP Step given as a fraction of the maximum classical displacement, that
corresponds to the 1.0 value
• ∗ N(I),I=1,NMO sequence number of the modes selected.

Let |r0 > be the equilibrium configuration; then the following configurations are explored:
|ri >= |r0 > +i∆|u >, where |u > is the eigenvector of the selected mode, i is a positive or
negative integer, running from IN I to IF I, and ∆ is the step. IF I − IN I + 1 is the number
of points that will be considered in the IN I ∗ ST EP − IF I ∗ ST EP interval. If the STEP
variable is set to 1.0, the maximum classical displacement is computed. This displacement
corresponds to the point where the potential energy in the harmonic approximation is equal
to the energy of the fundamental vibrational state as follows:

V = E0vib
1 2 1
kx = ~ω
2 2
Where x=|rmax i − |r0 i and the force constant k is given by:

k = ω2 µ

The final expression of the maximum classical displacement is therefore:

s
~
x=
ωµ

This option can be useful in two different situations.

Let us consider νi as the frequency of the Qi normal mode:
νi > 0 we want to explore the energy curve along Qi normal mode and check the deviation of
the energy from the harmonic behaviour. See example 1;

νi < 0 the system is in a transition state. We want to explore the Qi normal mode in order to
find a total energy minimum; usually Qi is not total-symmetric, the symmetry of the
structure needs to be reduced. CRYSTAL determines automatically the subgroup of the
original group to which the symmetry of the mode belongs. See example 2.

At each point, the geometry is written in file ”SCANmode number frequencyvalue DISP i∆”
(see below), in a format suitable to be read by the keyword EXTERNAL (geometry input,
page 15).
The geometry of the system then has to be re-optimized in this new subgroup using as a
starting geometry one of those external files (better the one corresponding to the minimum).
Frequencies can then be evaluated in the new minimum and the new set of frequencies should
contain only positive values (apart from the three referring to translations).

136
Example 1 - Methane molecule
First run: optimization of the geometry (full input at page 212).
Second run: calculation of the vibrational frequencies of CH4 in the optimized geometry.
The optimized geometry corresponds to a minimum, as all frequencies are positive (modes 1-3,
translational mode; modes 4-6, rotational modes).

MODES EIGV FREQUENCIES IRREP IR RAMAN

(HARTREE**2) (CM**-1) (THZ)
1- 3 -0.1863E-11 -0.2995 -0.0090 (F2 ) A A
4- 6 0.7530E-07 60.2270 1.8056 (F1 ) I I
7- 9 0.4821E-04 1523.8308 45.6833 (F2 ) A A
10- 11 0.6302E-04 1742.3056 52.2330 (E ) I A
12- 12 0.2099E-03 3179.3763 95.3153 (A ) I A
13- 15 0.2223E-03 3272.4193 98.1047 (F2 ) A A

Third run: Scanning of a selected mode.

To explore the 12th normal mode, corresponding to C-H symmetric stretching, the following
lines must be inserted before the end of geometry input (RESTART to read from external file
vibrational modes, computed in 2nd run):

FREQCALC
RESTART
SCANMODE
1 -10 10 0.2
12
END

The potential energy function as well as its harmonic approximation is computed are repre-
sented in the figure. The anharmonicity of C–H stretching is evident.

Figure 4.1: Scanning of the energy along normal mode 12, ν=3179.3763 cm−1 , corresponding to
C–H symmetric stretching

137
Example 2 - PbCO3
The space group of this carbonate, as it can be found in the literature [ICSD database], is
Pmcn (orthorhombic lattice).
First run: full optimization of the geometry in Pmcn space group (full input at page 212).
Second run: frequency calculation. The output would look as follows:

MODES EIGV FREQUENCIES IRREP IR RAMAN

(HARTREE**2) (CM**-1) (THZ)
1- 1 -0.3212E-07 -39.3362 -1.1793 (AU ) I I
2- 2 -0.1388E-09 -2.5858 -0.0775 (B3U) A I
3- 3 -0.6924E-10 -1.8262 -0.0547 (B2U) A I
4- 4 -0.2405E-11 -0.3404 -0.0102 (B1U) A I
5- 5 0.4141E-07 44.6637 1.3390 (AG ) I A
6- 6 0.4569E-07 46.9137 1.4064 (B3G) I A
7- 7 0.5304E-07 50.5476 1.5154 (B1G) I A
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
53- 53 0.4245E-04 1429.9950 42.8702 (AU ) I I
54- 54 0.4338E-04 1445.5993 43.3380 (B1G) I A
55- 55 0.4340E-04 1445.8649 43.3459 (AG ) I A
56- 56 0.4401E-04 1455.9714 43.6489 (B1U) A I
57- 57 0.4408E-04 1457.1539 43.6844 (B3G) I A
58- 58 0.4417E-04 1458.5583 43.7265 (B3U) A I
59- 59 0.4475E-04 1468.2070 44.0157 (B2U) A I
60- 60 0.5007E-04 1553.0286 46.5586 (B2G) I A

Four negative frequencies are present. Modes 2, 3 and 4 are translations, as results from their
small values (< 2 cm−1 ) and from a visual analysis (program MOLDRAW [87]); mode 1,
frequency -39.3362 cm−1 , corresponds to a maximum along the Q1 normal coordinate.
Third run: scanning of the first normal mode. The input lines for the frequency calculation

Figure 4.2: Scanning of the energy along normal mode 1, corresponding to a frequency of -39.3362
cm−1 (L. Valenzano, unpuplished results)

block are now the following:

138
FREQCALC
RESTART
SCANMODE
1 -10 10 0.4 scanning of 1 mode, initial point -10, final +10, step 0.4
1
END

where we are asking to perform the scan of 1 mode (mode 1), computing energy in 21 points
in the interval -10/+10 with a step equal to 0.4. Figure 2 shows the energy computed, and the
energy in the harmonic approximation.
The optimized geometry of PbCO3 in Pmcn space group corresponds to a transition state.
Fourth run:
We need to fully re-optimize the geometry of the system with symmetry as a subgroup (P21 21 2,
space group number 19) of the original space group (Pmcn). The geometry, with correct
reduced symmetry, is read (EXTERNAL, page 15) from one of the files written during the scan,
copied as file fort.34. For instance SCAN1 -39.3361 DISP -2.400 (scan of mode 1, frequency
-39.3361 cm−1 , displacement -2.4 the classical amplitude).
Please refer to the standard script for running CRYSTAL09 to handle input/output file names.
Fifth run: After full geometry optimization, we are ready to run a new frequency calculation.
The new frequency output looks like (just the first four lines are given):

MODES EIGV FREQUENCIES IRREP IR INTENS RAMAN

(HARTREE**2) (CM**-1) (THZ) (KM/MOL)
1- 1 -0.1504E-09 -2.6917 -0.0807 (B1 ) A ( 0.00) A
2- 2 -0.1414E-09 -2.6097 -0.0782 (B3 ) A ( 0.00) A
3- 3 -0.1690E-11 -0.2853 -0.0086 (B2 ) A ( 0.00) A
4- 4 0.4363E-07 45.8409 1.3743 (A ) I ( 0.00) A
[. . . . . . . . . . . . .]

Only the three expected negative (translational) modes are present, the fourth negative fre-
quency is not present any more. The PbCO3 structure corresponds now to a minimum in the
potential energy surface.

139
3 - Calculation of the IR reflectance spectrum
Keyword REFLECTANCE, inserted in the FREQCALC input block, activates the calcu-
lation of the IR reflectance spectrum. Prior calculation of IR intensities is required (keyword
INTENS, page 133) and definition of the dielectric tensor matrix (keyword DIELTENS,
page 129) or of the isotropic dielectric constant (keyword DIELISO, page
For each desired direction 3 quantities are calculated (for a given frequency range):

1. Reflectance (R)

2. Imaginary part of the inverse dielectric function (Im(1/)). The maxima provide the
Longitudinal Optical frequencies (LO)

3. Imaginary part of the dielectric function (Im())

The keyword REFLECTANCE defines an input block (closed by END).

rec variable meaning

• REFLDIR keyword to define directions

• ∗ NDIR Number of directions along which the spectrum is computed [1]
• ∗ IDIR(I),I=1,NDIR Directions are labeled from 1 to 6 as in Voigt notation (see Chap-
ter 11.11, page 230
END

REFLECTANCE is at the moment limited to:

• 3D systems

• directions 1,2 and 3 only (11 , 22 , 33 , according to Voigt notation, page 230))

• dielectric tensor in diagonal form (all crystal systems, apart from monoclinic and triclinic,
can be oriented in order to have diagonal dielectric tensor)

Generalization is in progress.
REFLECTANCE data are written in file REFLECTANCE.DAT and can be directly plotted
with gnuplot (http://www.gnuplot.info, appendix D, page 266).
Once the dielectric tensor and the vibrational spectrum are calculated, the reflectance spectrum
can be calculated at almost zero computational time with a RESTART in the FREQCALC
input block (the FREQINFO.DAT file is required):

FREQCALC
RESTART
INTENS
[options for INTENS]
DIELTENS or DIELISO
...
[optional keywords]
REFLECTANCE
REFLDIR
...
...
[other optional keywords]
END [end REFLECTANCE]
END [end FREQCALC]

To be modified by developers only.

140
There are several keywords which allow to modify the default values.

rec variable meaning

• A REFLNSTEP .
• ∗ N Number of points used for the reflectance spectrum [1000].
• A REFLSTEPS
• ∗ X Step size used for the reflectance spectrum [no default]. To be used
as an alternative to REFLNSTEP
• A REFLRANGE
• ∗ X1,X2 Frequency range used for the reflectance spectrum Default: [0., 1.3
νmax ], where νmax is the frequency of the highest calculated mode .

• A REFLDAMP
• ∗ GAMMA Damping factor (related to peak width) used for the reflectance spec-
trum [8.0]

• A REFLANGLE
• ∗ ALPH Angle of incidence used for the reflectance spectrum (degrees) [10.0]
.

141
4 - Anharmonic calculation of frequencies of X-H (X-D)
bond stretching

Anharmonic calculation of frequencies of X-H (X-D) bond stretching is invoked by the keyword
ANHARM in FREQCALC input block.

rec variable meaning

• ∗ LB label of the atom to be displaced (it must have atomic number 1, Hydrogen
or Deuterium. The first neighbour (NA) of the LB atom is identified. LB
moves along the (NA-LB) direction.
• A END End of ANHARM input block

This keyword allows the calculation of the anharmonic X-Y stretching. The selected X-Y
bond is considered as an independent oscillator. This condition is fulfilled when H or D are
involved. It can be used for X-H (or X-D) only.

S. Tosoni, F. Pascale, P. Ugliengo, R. Orlando, V.R. Saunders and R. Dovesi,

”Vibrational spectrum of brucite, Mg(OH)(2): a periodic ab initio quantum mechanical
calculation including OH anharmonicity”
Chem. Phys. Lett. 396, 308-315 (2004)].

Frequencies are calculated as follows:

i) the X-H distance is varied around the equilibrium value, d0 [default: d0 + (-0.2, -0.16, -0.06,
0.00, 0.16, 0.24, 0.3 Å)], all other geometrical features being constant (only H moves);
ii) the total potential energy is calculated for each value of the X-H distance [default 7 points];
iii) a polynomial curve of sixth degree is used to best fit the energy points; the root mean
square error is well below 10−6 hartree;
iv) the corresponding nuclear Schrödinger equation is solved numerically following the method
proposed in reference [80]. See P. Ugliengo, ”ANHARM, a program to solve the mono
dimensional nuclear Schrödinger equation”, Torino, 1989.

The anharmonicity constant and the harmonic XH stretching frequency are computed from
the first vibrational transitions ω01 and ω02 , as:

ωe xe = (2ω01 − ω02 ) /2

ωe = ω01 + 2ωe xe
Stretching of the X-H bond may reduce the symmetry (default). If keyword KEEPSYMM is
inserted, all equivalent X-H bonds will be stretched, to maintain the symmetry. For example,
in CH4 (point group Td ), KEEPSYMM forces the four CH bonds to stretch in phase; otherwise
only the selected C-H bond is stretched, and the symmetry reduced (point group C3v ).

142
Optional keywords of ANHARM input block

ISOTOPES atomic mass of selected atoms modified

• ∗ NL number of selected atoms
II insert NL records II
• ∗ LB,AMASS label and new atomic mass (amu) of the atom.
II II

KEEPSYMM all atoms symmetry equivalent to the selected one are displaced

NOGUESS scf guess at each geometry point: superposition of atomic densities at each
scf calculation

POINTS26 26 points: dX−H range: d0 -0.2 ÷ d0 +0.3 with a step of 0.02 Å.

PRINT extended printing

PRINTALL printing for programmers

TEST[ANHA] Preliminary test to check if the neighbour(s) of the selected atom is cor-
rectly identified and the X-Y direction properly set. No energy calculations
is performed.

It has been verified that calculations with 7 points provides very similar results to the ones
obtained with 26 points. In the following table, results for POINTS=7 and 26 are reported for
three systems. All values are in cm−1 .
system NPOINTS 26 NPOINTS 7
W01 4358.6 4359.0
HF (molecule) W02 8607.3 8608.1
We 4468.6 4468.8
We Xe 55.0 54.9
W01 3325.3 3325.8
Be(OH)2 (bulk) W02 6406.3 6407.4
We 3569.5 3569.9
We Xe 122.1 122.1
W01 3637.2 3637.5
Ca(OH)2 (bulk) W02 7111.4 7111.9
We 3800.3 3800.7
We Xe 81.5 81.6

143
5 - Harmonic calculation of phonon dispersion
Keyword DISPERSION, inserted in the FREQCALC input block, activates the calculation
of phonon dispersion on the reciprocal points, compatible with the supercell given in input.
The supercell must be generated using the keyword SCELPHONO (page 58). in order to
perform a dispersion calculation.
DISPERSION can not be run with INTENS.
The program finds the symmetry irreducible set of points in the first Brillouin zone correspond-
ing to the supercell. The frequencies and symmetry adapted modes are computed for these
points, allowing a better description of thermodynamic and other related properties.

NOKSYMDISP
Active if DISPERSION is active. Do not factorize the Dinamic Matrices in reciprocal space
according to the irreducible representations of the Space Groups. The symmetry properties of
the phonons are not recognized.

144
Chapter 5

Coupled Perturbed HF/KS

calculation

CPHF - Coupled Perturbed HF/KS calculation up to the

second order
This keyword performs Coupled Perturbed HF/KS calculation up to the second order. It
must be the last keyword in geometry input block.
Users of this module are requested to cite the following papers:
M. Ferrero, M. Rérat, R. Orlando and R. Dovesi
Coupled perturbed Hartree-Fock for periodic systems: the role of symmetry and related
computational aspects
J. Chem. Phys. 128, Art.N. 014100 (2008)
M. Ferrero, M. Rérat, R. Orlando and R. Dovesi
The calculation of static polarizabilities in 1-3D periodic compounds. The implementation in
the CRYSTAL code.
J. Comput. Chem. 29, 1450–1459 (2008)

rec variable meaning

• A END[CPHF] end of CPHF input block
• A FMIXING permits to mix the Fock/KS matrix derivatives
between CPHF-SCF1 cycles n and n − 1
∗ IPMIX percentage of cycle n − 1 [IPMIX=0, no mixing]
• A TOLALPHA threshold on energy first derivative change between CPHF-SCF1 cy-
cles
∗ ITOL |∆E| < 10−IT OL [default: 4]
∗ ITOL |∆E| < 10−IT OL [default: 4]
• A MAXCYCLE permits to modify the max number of CPHF-SCF1 cycles [default:
100]
∗ NMAX maximum number of SCF cycles
• A TOLUDIK maximum allowed difference between degenerate eigenvalues
• ∗ ITOL |∆E| < 10−IT OL [default: 6]

CPHF-SCF1 is the self consistent field that calculates the U [1] matrix, the polarizability tensor
(α) and, through the 2n + 1 formula, the first hyperpolarizability tensor (β).
CPHF-SCF2 is the self consistent field that calculates the U [2] matrix, the first hyperpolariz-
ability tensor (β).
The keyword TOLUDIK may be useful for calculations extended above the first order (CPHF-
SCF1). In this case, in fact, the non − canonical approach leads to the problem of calculating

145
 [ ]
the difference between eigenvalues to obtain the ∂Uij b /∂ka matrix derivative. The latter is
defined as a function of Qkijα , the derivative of the density matrix with respect to the kα point

ka ~ ka ~
Kij (k) − Ej Rij (k)
Qkija (~k) = (5.1)
Ej − Ei
ka ~ ka ~
where Kij (k) and Rij (k) are, respectively, the Fock matrix and the overlap matrix derivatives
with respect to ka . Qkija can be conveniently redefined in terms of the equivalent hermitian
matrix
1 ka
Q̃kija = Qkija + Rij
2
that, as can be shown, leads to
Q̃kiia = 0
and to
ka ~ ka ~
Kij (k) − 12 Rij (k)(Ej + Ei )
Qkija (~k) = i 6= j
Ej − Ei
Thus, it becomes necessary to define a threshold value (10−IT OL ) as the maximum difference
allowed between degenerate eigenvalues, in order to separate these two cases.
RESTART
A CPHF/CPKS run can be started from the results of an analogous previous run (even an
incomplete run). Every CPHF/CPKS run dumps the necessary information for a restart to file
fort.31. This file must be copied to file fort.32 before running a calculation with the RESTART
directive. GUESSP (SCF guess from density matrix of a previous run, input block 3, page 91)
is not applied by default, but its use is recommended.
For computational detail on the implementation of CPHF method see [48]. For applications
see [47, 49, 106].

146
Chapter 6

Mapping of CRYSTAL
calculations to model
Hamiltonians

CONFCNT - Configuration counting and cluster expan-

sion
The CONFCNT environment can be used to study possible configurations within a given cell
if one or more than one sites are activated, i. e. atoms lying in this site can be substituted
or its spin flipped. If DUMMYFIT keyword is inserted a subsequent mapping to two body
Hamiltonian is performed. This method is usually indicated as cluster expansion method and
can be applied to solid solutions (elemental substitution within the given cell) or magnetic
configurations (Ising model to map different anti ferromagnetic and ferrimagnetic ordering).
In this manual section the allowed keywords are described, for a longer explanation of the use
of this part of the code, please refer to the tutorial page:
http://www.crystal.unito.it → tutorials → Combinatorial mapping of solid solutions and
magnetic configurations..
For the use and implementation of this scheme in the CRYSTAL code see:
A. Meyer, R. Orlando, R. Dovesi
Andradite - grossular solid solution: an hybrid functional all electron calculation and cluster
expansion scheme approach.
For a discussion concerning the cluster expansion method see:
M. H. F. Sluiter, Y. Kawazoe
Cluster expansion method for adsorption: Application to hydrogen chemisorption on graphene.,
Phys. Rev. B 68 (2003) 085410

Default choices

CONFCNT
• ∗ N number of active sites
ST number of analyzed stars of neighbors of N sites
• ∗ NA(L),L=1,N label of the active atoms
CONFCNT as such is an information keyword. A symmetry and neighbors analysis of N
irreducible atoms is performed. ST stars of neighbors of the N activated atoms are printed.
Note that only atoms symmetry equivalent to the activated ones are reported as neighbors.

147
Atom pair neighbors characterized by the same distance but symmetry in-equivalent are split
in subsets. Each one of the N atoms is specified with the label NA (CRYSTAL output).
Atoms symmetry related to all NA atoms are printed. Let us indicate the sum of all atoms
equivalent to the activated sites as IALL. The number of in-equivalent configurations that can
be obtained by substituting from one to IALL/2 (or IALL/2 plus one if IALL is odd) is printed
increasing the number of substitutions that take place. Symmetry analysis is performed and
in-equivalent configurations only are printed. The number of in-equivalent configurations for a
fixed number of substitutions as well as the total number of configurations including all possible
number of substitutions is reported. After the CONFCNT analysis the program stops except
if DUMMYFIT keyword is inserted.
Example

CONFCNT
1 2
13
END

One site (N=1) is analyzed, two (ST=2) stars of neighbors of the activated site are printed.
The atom label (CRYSTAL output) of the activated atom is 13.

Optional keywords

END indicates that CONFCNT environment input is concluded and no more

optional keywords will be inserted.

JDIST stars of neighbors of atom pairs characterized by the same distance are
not split even thought they are not symmetry equivalent

PRTCONF all configurations are printed, equivalent configuration subsets are distin-
guished

SETCOMP
• ∗ IX the whole configurational analysis is performed for a subset of possible
configurations, the ones referring to IX substitutions (or spin flipping).
IX must be higher than zero and lower equal to IALL/2

148
Chapter 7

Calculation of elastic constants

ELASTCON - second-order elastic constants

A fully-automated procedure for calculating the second-order elastic constants for an arbitrary
crystal is activated by specifying the keyword ELASTCON in input block 1 (geometry).
ELASTCON must be the last keyword in geometry input.

. . . geometry input . . . .
ELASTCON
END[ELAS] close ELASTCON input block
END close geometry input block

ELASTCON input block admits several options (optional sub keywords), and terminates
with keyword END (or END[ELAS], END[—]: the first three characters only are processed).

The procedure for calculating the second-order elastic constants takes maximum advantage of
the capabilities of the CRYSTAL09 according to the following steps:

1. Determine the minimum number of deformations required to calculate the entire set of
elastic constants.

2. Make a deformation followed by full optimization of the geometry.

3. When optimization has been achieved for a given deformation, the 3x3 force matrix is
calculated using the analytic first derivative.

4. Steps 2 and 3 are repeated in order to obtain numerical second derivatives of the total
energy, which give the second-order elastic constants.

Note that the user must be confident that the input structure is already well optimized. At
the beginning of the run, the forces at the central point are calculated and if they exceed 10−4 ,
then a warning is printed suggesting re-optimization.
Users of this module are requested to cite the following papers:
W.F. Perger, J. Criswell, B. Civalleri and R. Dovesi
Ab-initio calculation of elastic constants of crystalline systems with the CRYSTAL code
Comp. Phys. Comm. 180, 1753-1759 (2009)

Keywords, options, and defaults

A default value is chosen for all computational parameters.
SCF energy convergence threshold is set to 10−8 . To modify it, see keyword TOLDEE in
input block 3, page 102.
Default choices can be modified by optional keywords (in any order):

149
 rec variable meaning

• A NUMDERIV This sets the number of points for the numerical second derivative
∗ INUM number of points including the central (zero displacement) one
[def ault = 3]
• A STEPSIZE This gives the size of the displacement to be used for the calculation
of the numerical second derivative.
∗ STEP size of the strain step along a given deformation [def ault = 0.01]
For example, if N U M DERIV = 5 and ST EP SIZE = 0.002, then the displacements will be
-0.004, -0.002, 0.002, and 0.004.
rec variable meaning
• A PRINT This option turns on a higher level of diagnostic printing and will
generally not be necessary for the typical user. [def ault minimal
printing]
• A RESTART Allows RESTART using file ELASIN F O.DAT from a previous
run.
• A PREOPTGEOM A preliminary geometry optimization of cell and atomic positions
is performed before starting elastic constant calculation.

Geometry optimization is performed at convergence criteria tighter than the ones given in
OPTGEOM. Some values can be modified by inserting the following keywords:

• A TOLDEG ELASTCON default [0.00006] - see OPTGEOM, page 110

• A TOLDEX ELASTCON default [0.00012] - see OPTGEOM, page 110
• A TOLDEE ELASTCON default [8] - see OPTGEOM, page 110
• A MAXCYCLE see OPTGEOM, page 113
• A NOTRUSTR see OPTGEOM, page 111
• A TRUSTRADIUS see OPTGEOM, page 111
• A MAXTRADIUS see OPTGEOM, page 111
Example (the following lines are inserting into the ‘geometry’ section):
...
ELASTCON
NUMDERIV modify default choice of n. points [3]
5
STEPSIZE modify step size for numerical gradient [0.001]
0.002
PRINT extended printing - Warning: may be very large!!!!
END[ELAST] ELASTCON input
END geometry input

Internal Diagnostics
In order to monitor the quality of the calculation as it proceeds, the total energy after opti-
mization is stored. The recommended use of the ‘ELASTCON’ option assumes that the user
supplies an input file from a previously optimized geometry (and not experimental lattice con-
stants and atomic positions, for example). Therefore, in principle, as the various deformations
are made, the optimized total energy for each of the deformed geometries should be higher
than the energy at the undeformed, equilibrium, geometry. The code monitors each optimized
total energy for each deformed geometry and if any deformation lowers the total energy from
the equilibrium value, a warning is printed for the user to verify that the input file was really
from a previously optimized geometry.
As a second diagnostic, the optimized total energy for each deformation is tested to see if it is
within T OLDEE ∗ 100 of the equilibrium value. If it is, a warning is printed for the user to
try decreasing T OLDEE or increasing ST EP SIZE.

150
New in CRYSTAL09 v2.0.1
The calculation of elastic constants has been deeply revised in th new version. Thanks to Elisa
Albanese and Alessandro Erba for helping in revising the code.
In particular, the restart option has been modified to make it more stable and flexible. Cal-
culations can be restared from any deformation and the number of points can be increased as
well. For instance, a 5-point calculation can start from a previous 3-point calculation
WARNING: With v2.0.1, the compatibility between ELASINFO.DAT v1.0.1 and ELAS-
INFO.DAT v2.0.1 is broken. Therefore, calculations cannot restart from previous data files.

151
Chapter 8

Properties

One-electron properties and wave function analysis can be computed from the SCF wave func-
tion by running properties. At the end of the SCF process, data on the crystalline system
and its wave function are stored as unformatted sequential data in file fort.9, and as formatted
data in file fort.98 . The wave function data can be transferred formatted from one platform
to another (see keyword RDFMWF, page 167).
The data in file fort.9 (or fort.98) are read when running properties, and cannot be modified.
The data include:

1. Crystal structure, geometry and symmetry operators.

2. Basis set.

3. Reciprocal lattice k-points sampling information.

4. Irreducible Fock/KS matrix in direct space (Unrestricted: Fα , Fβ ).

5. Irreducible density matrix in direct space (Unrestricted: Pα+β Pα−β ).

The properties input deck is terminated by the keyword END. See Appendix C, page 256,
for information on printing.

8.1 Preliminary calculations

In order to compute the one-electron properties it is necessary to access wave function data
as binary data set: if binary data are not available in file fort.9, the keyword RDFMWF,
entered as 1st record, will read formatted data from file fort.98 and write them unformatted
in file fort.9.
Full information on the system is generated: :

a. symmetry analysis information stored in COMMON areas and modules

b. reducible Fock/KS matrix stored on Fortran unit 11
c. reducible density matrix
c.1 all electron stored on Fortran unit 13 (1st record)
c.2 core electron stored on Fortran unit 13 (2nd record)
c.3 valence electron stored on Fortran unit 13 (3rd record)
d. reducible overlap matrix stored on Fortran unit 3
e. Fock/KS eigenvectors stored on Fortran unit 10
1. a, b, c1, d, are automatically computed and stored any time you run the properties
program.

2. in unrestricted calculations, the total electron density matrix (α+β) and the spin density
matrix (α − β) are written as a unique record in fortran unit 13.

152
 3. The core and valence electron density matrices (c.2, c.3) are computed only by the
NEWK option when IFE=1. They are stored as sequential data set on Fortran unit
13, after the all electron density matrix. Calculation of Compton profiles and related
quantities requires such information.

4. Properties can be calculated using a new density matrix, projected into a selected range
of bands (keyword PBAN, PGEOMW), range of energy (keyword PDIDE), or con-
structed as a superposition of the atomic density matrices relative to the atoms (or ions)
of the lattice (keyword PATO). In the latter case a new basis set can be used.
When a specific density matrix is calculated [band projected (PBAN), energy projected
(PDIDE), atomic superposition (PATO)], all subsequent properties are calculated using
that matrix.

The option PSCF restores the SCF density matrix.

The keyword PMP2 (see page 181) reads the MP2 correction to the valence density
matrix. Properties can then be computed from a MP2 corrected density matrix.

8.2 Properties keywords

RDFMWF wave function data conversion formatted-binary (fort.98 → fort.9)
Preliminary calculations

NEWK Eigenvectors calculation 179 I

NOSYMADA No symmetry Adapted Bloch Functions 95 –
PATO Density matrix as superposition of atomic (ionic) densities 180 I
PBAN Band(s) projected density matrix (preliminary NEWK) 180 I
PGEOMW Density matrix from geometrical weights (preliminary NEWK) 181 I
PDIDE Energy range projected density matrix (preliminary NEWK) 181 I
PSCF Restore SCF density matrix 187 –

Properties computed from the density matrix

ADFT Atomic density functional correlation energy 155 I

BAND Band structure 157 I
BIDIERD Reciprocal form factors 158 I
CLAS Electrostatic potential maps (point multipoles approximation) 159 I
ECHG Charge density and charge density gradient - 2D grid 164 I
ECH3 Charge density - 3D grid 163 I
EDFT Density functional correlation energy (HF wave function only) 164 I
EMDLPG Electron momentum distribution from density matrix 166 I
PMP2 MP2 correction to the Valence Density Matrix 181
POLI Atom and shell multipoles evaluation 182 I
POTM Electrostatic potential - 2D grid 185 I
POT3 Electrostatic potential - 3D grid 183 I
POTC Electrostatic properties 184 I
PPAN Mulliken population analysis 96
XFAC X-ray structure factors 188 I

Properties computed from the density matrix (spin-polarized systems)

ANISOTRO Hyperfine electron-nuclear spin tensor 156 I

ISOTROPIC Hyperfine electron-nuclear spin interaction - Fermi contact 168 I
POLSPIN Atomic spin density multipoles 182 I

Properties computed from eigenvectors (after keyword NEWK)

153
ANBD Printing of principal AO component of selected CO 155 I
BWIDTH Printing of bandwidth 159 I
DOSS Density of states 162 I
EMDL Electron momentum distribution - line 165 I
EMDP Electron momentum distribution - plane maps 166 I
PROF Compton profiles and related quantities 186 I

New properties

POLARI Berry phase calculations 190 I

SPOLBP Spontaneous polarization (Berry phase approach) 192 –
SPOLWF Spontaneous polarization (localized CO approach) 193 –
PIEZOBP Piezoelectricity (Berry phase approach) preliminary 189 –
PIEZOWF Piezoelectricity (localized CO approach) - preliminary 189 –
LOCALWF Localization of Wannier functions 169 I
DIEL Optical dielectric constant 160 I

Auxiliary and control keywords

ANGSTROM Set input unit of measure to Ångstrom 31 –

BASISSET Printing of basis set, Fock/KS, overlap and density matrices 157 –
BOHR Set input unit of measure to bohr 34 –
CHARGED Non-neutral cell allowed (PATO) 62 –
END Terminate processing of properties input keywords –
FRACTION Set input unit of measure to fractional 44 –
MAPNET Generation of coordinates of grid points on a plane 177 I
NEIGHBOR Number of neighbours to analyse in PPAN 51 I
PRINTOUT Setting of printing options 53 I
RAYCOV Modification of atomic covalent radii 53 I
SETINF Setting of inf array options 54 I
SETPRINT Setting of printing options 54 I
STOP Execution stops immediately 56 –
SYMMOPS Printing of point symmetry operators 60 –

Info - Output of data on external units

ATOMIRR Coordinates of the irreducible atoms in the cell 156 –

ATOMSYMM Printing of point symmetry at the atomic positions 34 –
COORPRT Coordinates of all the atoms in the cell 37 –
CRYAPI OUT geometry, BS, direct lattice information 160 –
EXTPRT Explicit structural/symmetry information 41 –
FMWF Wave function formatted outputi in file fort.98. Section 8.2 167 –
INFOGUI Generation of file with wf information for visualization 168 –
CRYAPI OUT Reciprocal lattice information + eigenvalues 160 –

154
 ANBD - Principal AO components of selected eigenvectors
rec variable value meaning
• ∗ NK n Number of k points considered.
0 All the k points are considered.
NB n Number of bands to analyse
0 All the valence bands + 4 virtual are analysed.
TOL Threshold to discriminate the important eigenvector coefficients. The
square modulus of each coefficient is compared with TOL.
if NK > 0 insert II
• ∗ IK(J),J=1,NK Sequence number of the k points chosen (printed at the top of NEWK
output)
if NB > 0 insert II
• ∗ IB(J),J=1,NB Sequence number of the bands chosen

The largest components of the selected eigenvectors are printed, along with the corresponding
AO type and centre.

ADFT/ACOR - A posteriori Density Functional atomic

correlation energy
The correlation energy of all the atoms not related by symmetry is computed. The charge
density is always computed using an Hartree-Fock Hamiltonian (even when the wave function
is obtained with a Kohn-Shamm Hamiltonian).
The input block ends with the keyword END. Default values are supplied for all the compu-
tational parameters.
A new atomic basis set can be entered. It must be defined for all the atoms labelled with a
different conventional atomic number (not the ones with modified basis set only).

BECKE Becke weights [default] [12]

or
SAVIN Savin weights [115]

RADIAL Radial integration information

rec variable meaning
• ∗ NR number of intervals in the radial integration [1]
• ∗ RL(I),I=1,NR radial integration intervals limits in increasing sequence [4.]
• ∗ IL(I),I=1,NR number of points in the radial quadrature in the I-th interval [55].

ANGULAR Angular integration information

rec variable meaning
• ∗ NI number of intervals in the angular integration [default 10]
• ∗ AL(I),I=1,NI angular intervals limits in increasing sequence. Last limit is set to 9999.
[default values 0.4 0.6 0.8 0.9 1.1 2.3 2.4 2.6 2.8]
• ∗ IA(I),I=1,NI accuracy level in the angular Lebedev integration over the I-th interval
[default values 1 2 3 4 6 7 6 4 3 1].
PRINT printing of intermediate information - no input
PRINTOUT printing environment (see page 53)
TOLLDENS
• ∗ ID DFT density tolerance [default 9]
TOLLGRID
• ∗ IG DFT grid weight tolerance [default 18]
NEWBASIS a new atomic basis set is input
insert complete basis set input, Section 1.2

155
ANGSTROM - unit of measure
Unit of measure of coordinates (ECHG, POTM, CLAS) See input block 1, page 31.

ANISOTRO - anisotropic tensor

rec variable meaning
• A keyword enter one of the following keywords:
• A3 ALL
The anisotropic tensor is evaluated for all the atoms in the cell
or
• A6 UNIQUE (alias NOTEQUIV) The anisotropic tensor is evaluated for all the non-
equivalent atoms in the cell
or
• A6 SELECT The anisotropic tensor is evaluated for selected atoms
• ∗ N number of atoms where to evaluate the tensor
• ∗ IA(I),I=1,N label of the atoms
• A PRINT
extended printing

The anisotropic hyperfine interaction tensor is evaluated. This quantity is given in bohr−3
and is transformed into the hyperfine coupling tensor through the relationship [131]
1000 1
T[mT] = µ0 βN gN T = 3.4066697gN T
(0.529177 · 10−10 )3 4π

(see ISOTROPIC for the units and conversion factors). The elements of the T tensor at
nucleus A are defined as follows:
 2 
rA δij − 3rAi rAj
XX Z
A spin
Tij = Pµνg ϕµ (r) 5 ϕgν (r)dr
µν g
r A

where r A = |r − A| and rAi = (r − A)i (ith component of the vector).

For extended printing (tensor in original cartesian axes and in principal axis system) insert,
before the keyword ANISOTRO:
SETPRINT
1
18 1

See tests 29, 31, 32, 33.

ATOMIRR - coordinates of irreducible atoms

Cartesian and fractional coordinates of the irreducible atoms are printed. No input data
required.

ATOMSYMM
See input block 1, page 34

156
 BAND - Band structure
rec variable value meaning
• A TITLE any string (max 72 characters).
• ∗ NLINE > 0 number of lines in reciprocal space to be explored (max 20)).
ISS shrinking factor in terms of which the coordinates of the extremes of
the segments are expressed.
NSUB total number of k points along the path.
INZB first band
IFNB last band
IPLO 0 eigenvalues are not stored on disk.
=1 formatted output for plotting; see Appendix D, page 260
LPR66 6= 0 printing of eigenvalues
add NLINE records
• ∗ coordinates of the line extremes in units of |bi |/ISS
I1,I2,I3 first point coordinates.
J1,J2,J3 last point coordinates.

The band structure along a given path n the Brillouin zone is computed.
The data are printed in standard output and (if IPLO = 1) written in file fort.25 (for-
matted data processed by Crgra2006) and in file BAND.DAT (processed by DLV; see
http://www.cse.clrc.ac.uk/cmg/DLV). See Appendix D, page 260).

1. Warning : does not run for molecules!! (0D)

2. For a correct interpretation of HF band-structure and DOS’s, it must be stressed that

the HF eigenvalues are not a good approximation to the optical excitation spectrum of
the crystal. However, as discussed in section III.2 of reference [103] and in Chapter 2 of
reference [100], the band structures, in conjunction with total and projected DOS’s, can
be extremely useful in characterizing the system from a chemical point of view.
3. Note on band extremes coordinates: in two-(one-) dimensional cases I3, J3 (I2,I3,J2,J3)
are formally input as zero (0). See test 3 and 6.

4. The only purpose of ISS is to express the extremes of the segments in integer units (see
tests 8-9). It does not determine the density of k points along the lines, which depends
only on NSUB. The number of k points for each line is computed by the program. The
step is constant along each line. The step is taken as close as possible to a constant along
different lines.

5. If symmetry adapted Bloch functions are used (default option), BAND generates sym-
metry information in k points different from the ones defined by the Monkhosrt net.
Eigenvectors computed by NEWK in k points corresponding to the Monkhosrt net are
not readable any more. To compute density of states and bands, the sequence must be:
BAND - NEWK - DOSS.

See tests 3, 4, 6, 7, 8, 9, 11, 12 and 30.

BASISSET - Printing of basis set and data from SCF

rec variable value meaning
• ∗ NPR number of printing options.
if NPR 6= 0 insert prtrec (see page 54) II

This option allows printing of the basis set and the computational parameters, and, on re-
quest (keyword PRINTOUT before BASISSET), of the Fock/KS matrix (FGRED), the
overlap matrix (OVERLAP), and the reducible density matrix (PGRED), in direct lattice
representation.

157
Warning: the contraction coefficients of the primitive gaussians are different from the ones
given in input. See “Normalization coefficients”, Appendix E.
Printing options:
59 (Density matrix); 60 (Overlap matrix); 64 (Fock/KS matrix).

BIDIERD - Reciprocal form factors

This option evaluates the reciprocal form factors (RFF) (also called auto-correlation function)
for any direction directly from the direct space density matrix.
Compton Profiles (CPs) can be computed by Fourier transforming the RFF. The starting
auto-correlation function must be of good quality in order to get good CPs. Both RFFs and
CPs can be convoluted in order to be compared with the experiments (affected by the finite
resolution of the spectrometer); this procedure is performed by multiplying the RFF by the
gaussian function g(r)σr :
2
√
− r 1 2 2log2
g(r)σr = e 2σr2 where σr = =
σp F W HMp

where F W HMp , the convolution parameter, has to be defined in input by the user (in atomic
units); the r and p pedices identify quantities in coordinates and momentum space, respectively.
The anisotropies of the RFFs and the CPs can be evaluated via the DIFF keyword. This block
must be ended by ENDB. It works also for open-shell systems.
By using the PROF keyword, followed by the BR keyword, it is possible to obtain the same
quantity by Fourier transforming the Compton profiles. As the latter implies numerical inte-
gration, the BIDIERD keyword is expected to provide more accurate results.

rec variable value meaning

• ∗ NDIR number of directions along which the RFF are evaluated
NPU number of sampling points along each direction
STEP step along each direction
IMODO 0: the direction is defined by the Cartesian coordinates (bohr) of a point
6= 0: the direction is defined by the atom label and indices of the cell where
the atom is located
ICASO 1: the total density matrix is used
2: the core density matrix is used
3: the valence density matrix is used
• A4 CONV Convolution of the B(r) previously computed
• ∗ F W HMp convolution parameter (a.u.)
• A4 PROF Compton Profiles computed as Fourier Transforms of the B(r)
• ∗ STEPC step along each direction (a.u.)
• A4 DIFF B(r) and CPs anisotropies are computed
• A3 DIR The directions are specified
if IMODO=0, insert NDIR records
•∗ XYZ the explored direction is defined by the straight line going from the
origin to (X,Y,Z)
if IMODO6=0, insert NDIR records
• ∗ I XG YG label of the atom and indices of the cell where the atom is located. The
ZG explored direction is defined by the straight line going from the origin
to the atom position
• A4 END End block autocorrelation functions

Notes:
The explored interval is (NPU−1)×STEP long; X,Y,Z or I,XG,YG,ZG data are just used for
defining the direction, NOT the length of the explored interval.

BOHR - unit of measure

Unit of measure of coordinates (ECHG, POTM, CLAS) See input block 1, page 34.

158
 BWIDTH - Printing of band width
rec variable meaning
• ∗ INZB first band considered
0 analysis from first valence band
IFNB last band considered
0 analysis up to first 4 virtual bands

The Fock/KS eigenvalues are ordered in bands following their values. Band crossing is not
recognized.

CHARGED - charged reference cell

See input block 2, page 62.
To be used before PATO, when new basis set and/or electron configuration of the atoms result
in a charged cell.

CLAS - Point charge electrostatic potential maps

rec variable value meaning
• ∗ IDER 0 potential evaluation
1 calculation of potential and its first derivatives
IFOR 0 point multipoles have to be evaluated by POLI option
1 point formal charges given as input
if IFOR 6= 0 insert II
• ∗ Q(I),I=1,NAF formal net charge for all the NAF atoms in the unit cell (equivalent
and non equivalent, following the sequence printed at the top of the
properties printout)
insert MAPNET input records (page 177)

1. When IDER=0, the electrostatic potential is calculated at the nodes of a 2-dimensional

net in a parallelogram-shaped domain defined by the segments AB and BC (see keyword
MAPNET, page 177). The potential values are written formatted in file fort.25. (see
Appendix D, page 259).

2. When IDER 6= 0, the electrostatic potential gradient is computed at the nodes of the
same grid. The x, y and z components are printed on the standard output.

3. The potential is generated by an array of point multipoles up to a maximum order IDIPO

defined in the POLI option input, or by atomic point charges given in input (IFOR=1;
IDIPO = 0 is set in that case).

4. The multipoles must be previously computed by running the option POLI when IFOR
is equal to zero.

COORPRT
See input block 1, page 37.

DENSMAT - First order density matrix ρ(r, r0 ) - developers only

First order density matrix ρ(r, r0 ) along a given path is computed.
The variable r0 explores the same interval as r.
For UHF cases two matrices are generated, one corresponding to the total and the other to the
spin density matrix.

159
rec variable value meaning
• ∗ NKN number of knots in the path (=number of segments+1)
NPU number of sampling points along the full path
IPLOT 0: data are not saved for plot
= 1: data are saved in file fort.25
IMODO 0: knot coordinates (x, y, z) in a. u.
6= 0: knots are defined through atom labels
LPR 6= 0: print the ρ(r, r0 ) matrix in integer form (values are multiplied by 10000)
if IMODO=0, insert NKN records
•∗ XYZ Cartesian coordinates (bohr) of the i-th knot
if IMODO6=0, insert:
• ∗ DX DY displacement (bohr) applied to all atoms defining the path
DZ
insert NKN records
• ∗ I XG YG label of the atom and indices of the cell where the atom is located
ZG

• A NPU×NPU square matrix is generated.

• The step between contiguous sampling points belonging to different segments is the same.

• Meaning of the displacement: suppose you want the density matrix corresponding to the
π structure of benzene. Define, for example, the path H–C–C–C–H through the atom
labels and then displace it along z (if the molecule is in the x−y plane) by an appropriate
amount.

CRYAPI OUT - Geometry, BS, and full wave function information

Geometry, local function Basis Set, overlap, hamiltonian, density matrices n direct lattice are
written formatted in file GRED.DAT
Wannier functions (if file fort.80 is present; see keyword LOCALWF, page 169) are appended
to file GRED.DAT
k points coordinates (Monkhorst sampling net) and eigenvectors (if computed by NEWK page
179) in the full Brillouin zone are written formatted in file KRED.DAT.
The scripts runcry06/runprop06 save files GRED.DAT and KRED.DAT (if present) as
inpfilename.GRED and inpfilename.KRED
The utility program cryapi inp reads and prints the data. The organization of data can be
understood from the output of cryapi inp and from its source.
See Appendix D, page 267.

DIEL/DIELECT - Optical dielectric constant

rec variable meaning
• A END/ENDDIELend of DIEL input block
optional keywords II
• A PRINT extended output

The electron density must be obtained by applying an electric field (keyword FIELD, page
41). The dielectric constant is calculated by using the concept of macroscopic average of the
total charge density (see for example Fu et al. [54]) and Poisson’s equation. The charge density
is first averaged with respect to the (infinite) plane orthogonal to the field
Z
1
ρ(z) = ρ(z) dA (8.1)
A
A

where A = |~a ×~b|, and ~a and ~b are the lattice parameters of the supercell orthogonal to the field
direction. When a Fourier representation of the charge density is used, the previous equation

160
becomes:
+∞
1 X 2π`z
ρ(z) = F00` e−ı C (8.2)
V
`=−∞

F00` are structure factors (note that the two first indices are always zero) calculated analytically
from the SCF crystalline orbitals depending now on the applied field. The quantity ρ is then
averaged with respect to the z coordinate
z+∆z/2
Z
1
ρ(z) = ρ(z 0 ) dz 0 (8.3)
∆z
z−∆z/2

that is
+∞  
1 X ∆z 2π`z
ρ(z) = F00` sinc `π e−ı C (8.4)
V C
`=−∞

sin(u)
where the sinc function is the cardinal sinus (sinc(u) = u ) and ∆z has been chosen equal
to c; we can now apply Poisson’s equation to ρ(z):

∂ 2 V (z)
= −4πρ(z) (8.5)
∂z 2
or
∂E(z)
= 4πρ(z) (8.6)
∂z
because

∂V (z)
= −E(z) (8.7)
∂z
V (z), F (z) and ρ(z) are the mean values of the macroscopic electric potential, electric field
and electron density at z position along the electric field direction.
Structure factors can be separated in a real and an imaginary part:
< =
F00` = F00` + ıF00` (8.8)

Exploiting the following properties of the structure factors:

F000 = Ne (number of electrons in the supercell) (8.9)

< <
F00` = F00−`
= =
F00` = −F00−`

the real and imaginary parts of ρ take the following form:

+∞       
  Ne 2 X < 2π`z = 2π`z ∆z
< ρ(z) = + F00` cos + F00` sin sinc `π (8.10)
V V C C C
`=1
 
= ρ(z) = 0 (8.11)
As expected, the imaginary part is null. The Ne /V term can be disregarded, as it is exactly
canceled by the nuclear charges in the supercell.
According to equation 8.7, the local macroscopic field corresponds to minus the slope of V (z),
it has opposite sign with respect to the imposed outer field, according to the Lenz law, and is
obtained from ρ(z)(eq. 8.6):
+∞
"
#
2π`z 2π`z

 
8π X < sin C = cos C ∆z
E(z) = F00` 2π`

− F 00` 2π`

sinc `π (8.12)
V C C
C
`=1

161
The corresponding macroscopic electric potential can be written as follows:
+∞
"
#
2π`z 2π`z

 
−8π X < cos C = sin C ∆z
V (z) = F00` + F00` sinc `π (8.13)
V 2π` 2 2π` 2 C



`=1 C C

Since −E and E0 have opposite sign, the ratio E0 /(E0 +E) is larger than one, and characterizes
the relative dielectric constant of the material along the direction of the applied field:
E0
= (8.14)
E0 + E
The number of structure factors computed for a Fourier representation of the perturbed charge
density by default is equal to 300, the structure factors from F001 to F00 300 .
The data computed are written in file DIEL.DAT in append mode. See Appendix D, page 260.

DOSS - Density of states

rec variable value meaning
• ∗ NPRO 0 only total DOS is calculated
> 0 total DOS and NPRO projected densities are calculated. The maximum
number of projections is 15.
NPT number of uniformly spaced energy values (≤ LIM019) where DOSs are
calculated, from bottom of band INZB to top of band IFNB.
INZB first band considered in DOS calculation
IFNB last band considered in DOS calculation
IPLO 0 DOSs are not stored on disk
1 formatted output to file fort.25 for plotting (Appendix D, page 260).
2 formatted output to file DOSS.DAT for plotting (Appendix D, page 261).
NPOL number of Legendre polynomials used to expand DOSS (≤ 25)
NPR number of printing options to switch on
if INZB and IFNB < 0 insert II
• ∗ BMI,BMA Minimum and maximum energy (hartree) values to span for DOSS. They
must be in a band gap
if NPRO 6= 0, insert NPRO records II
• ∗ N > 0 DOS projected onto a set of N AOs
< 0 DOS projected onto the set of all AOs of the N atoms.
NDM(J),J=1,N vector NDM identifies the AOs (N>0) or the atoms (N<0) by their sequence
number (basis set order)
if NPR 6= 0, insert prtrec (see page 54) II

Following a Mulliken analysis, the orbital (ρµ ), atom (ρA ) and total (ρtot ) density of states can
be defined for a closed shell system as follows:
XXXZ
ρµ () = 2/VB dkSµν (k)aµj (k)a∗νj (k) eik·g δ[ − j (k)] (8.15)
j ν g BZ

X
ρA () = ρµ () (8.16)
µ∈A
X
ρtot () = ρA () (8.17)
A

where the last sum extends to all the atoms in the unit cell.
Bond population density of states are not computed.

1. Warning: do not run for molecules!

162
 2. The NEWK option must be executed (to compute Hartree-Fock/KS eigenvectors and
eigenvalues) before running DOSS. The values of the input parameters IS and ISP of
NEWK have a consequent effect on the accuracy of the DOSS calculation. Suggested
values for IS: from 4 to 12 for 3-D systems, from 6 to 18 for 2-D and 1-D systems (Section
11.7, page 227). ISP must be equal or greater than 2*IS; low values of the ratio ISP/IS
can lead to numerical instabilities when high values of NPOL are used.
If BAND is called between NEWK and DOSS, and symmetry adapted Bloch functions
are used (default option), the information generated by NEWK is destroyed. To compute
density of states and bands, the sequence must be: BAND - NEWK - DOSS.

3. DOSS are calculated according to the Fourier-Legendre technique described in Chapter

II.6 of reference 1, and in C. Pisani et al, ([101, 102]). Three computational parameters
must be defined: NPOL, IS, ISP. IS and ISP are entered in the NEWK option input.

4. NPOL is the number of Legendre polynomials used for the expansion of the DOS. The
value of NPOL is related to the values of IS and ISP, first and second input data of
NEWK option.
Suggested values for NPOL: 10 to 18.

5. Warning NEWK with IFE=1 must be run when spin-polarized solutions (SPIN-
LOCK, page 101) or level shifter (LEVSHIFT, page 92) were requested in SCF, to
obtain the correct Fermi energy and eigenvalues spectra.

6. Unit of measure: energy: hartree; DOSS: state/hartree/cell.

Computed data are written in file fort.25 (in Crgra2006 format), and in file DOSS.DAT
Printing options: 105 (density of states and integrated density of states); 107 (symmetrized
plane waves).
See tests 3, 4, 5, 6, 7, 8, 9, 11 and 30.

ECH3 - Electronic charge (spin) density on a 3D grid

rec variable meaning

•∗ NP Number of points along the first direction

if non-3D system
keyword to choose the type of grid on the non-periodic direction(s):

SCALE RANGE
length scales for non-periodic dimensions boundary for non-periodic dimensions (au)
if 2D system
• ∗ ZSCALE • ∗ ZMIN
• ∗ ZMAX
if 1D system
• ∗ YSCALE,ZSCALE • ∗ YMIN,ZMIN
• ∗ YMAX,ZMAX
if 0D system
• ∗ XSCALE,YSCALE,ZSCALE • ∗ XMIN,YMIN,ZMIN
• ∗ XMAX,YMAX,ZMAX

The electronic charge [and spin density] (electron/bohr3 )is computed at a regular 3-dimensional
grid of points. The grid is defined by the lattice vectors of the primitive unit cell and user
defined extents in non-periodic directions. NP is the number of points along the first lattice
vector (or XMAX-XMIN for a molecule). Equally spacing is used along the other vectors.
Non-periodic extents may be specified as an explicit range (RANGE) or by scaling the extent
defined by the atomic coordinates (SCALE).
Formatted data are written to file fort.31 (function value at the grid points) in the format
required by the visualization program DLV.

163
See Appendix D, page 265, for description of the format.
Function data computed at 3D grid points are written according to GAUSSIAN CUBE format
in files:
DENS CUBE.DAT charge density
SPIN CUBE.DAT spin density
PS. The sum of the density values divided by the number of points and multiplied by the
cell volume (in bohr, as printed in the output) gives a very rough estimate of the number of
electrons in the cell.

ECHG - Electronic charge density maps and charge density gradient

rec variable value meaning
• ∗ IDER n order of the derivative - < 2
insert MAPNET input records (Section 8.2, page 177)

1. IDER=0
The electron charge density (and in sequence the spin density, for unrestricted wave func-
tions) is calculated at the nodes of a 2-dimensional net in a parallelogram-shaped domain
defined by the segments AB and BC (see keyword MAPNET, page 177). The electron
density values (electron bohr−3 ) are written formatted in file fort.25 (see Appendix D,
page 259).

2. IDER=1
electron charge density, x, y, z component of first derivative, and modulus of the deriva-
tive, are written. The string in the header is always ”MAPN”.

3. When the three points define a segment (A≡B or B≡C), function data are written in file
RHOLINE.DAT. (see Appendix D, page 259)

4. When IDER 6= 0, the charge density gradient is computed at the nodes of the same grid.
The x, y and z components are printed on the standard output and written formatted in
file fort.25 (see Appendix D, page 259).

5. The electron charge density is computed from the density matrix stored in file fort.31.
The density matrix computed at the last cycle of SCF is the default.

6. Band projected (keyword PBAN), energy projected (keyword PDIDE) or atomic su-
perposition (keyword PATO) density matrices can be used to compute the charge den-
sity. The sequence of keywords must be: (NEWK-PBAN-ECHG), (NEWK-PDIDE-
ECHG) or (PATO-ECHG).

EDFT/ENECOR -A posteriori Density Functional correlation energy

Estimates a posteriori the correlation energy via a HF density. It is controlled by keywords.
The input block ends with the keyword END. All the keywords are optional, as default values
for all the integration parameters are supplied by the program, to obtain reasonably accurate
integration of the charge density. Please check the integration error printed on the output.

164
 BECKE Becke weights [default] [12]
or
SAVIN Savin weights [115]

RADIAL Radial integration information

rec variable meaning
• ∗ NR number of intervals in the radial integration [1]
• ∗ RL(I),I=1,NR radial integration intervals limits in increasing sequence [4.]
• ∗ IL(I),I=1,NR number of points in the radial quadrature in the I-th interval [55].

ANGULAR Angular integration information

rec variable meaning
• ∗ NI number of intervals in the angular integration [default 10]
• ∗ AL(I),I=1,NI angular intervals limits in increasing sequence. Last limit is set to 9999.
[default values 0.4 0.6 0.8 0.9 1.1 2.3 2.4 2.6 2.8]
• ∗ IA(I),I=1,NI accuracy level in the angular Lebedev integration over the I-th interval
[default values 1 2 3 4 6 7 6 4 3 1].
PRINT printing of intermediate information - no input
PRINTOUT printing environment (see page 53)
TOLLDENS
• ∗ ID DFT density tolerance [default 9]
TOLLGRID
• ∗ IG DFT grid weight tolerance [default 18]

EMDL - Electron Momentum Density - line maps

rec variable value meaning
• ∗ N number of directions (≤ 10)
PMAX maximum momentum value (a.u.) for which the EMD is to be calcu-
lated
STEP interpolation step for the EMD
IPLO 0 no data stored on disk
1 formatted output to file fort.25 for plotting (Appendix D, page 261).
2 formatted output to file LINEA.DAT for plotting (Appendix D,
page261).
LPR113 6= 0 printing of EMD before interpolation
• ∗ (K(I,J), directions in oblique coordinates
I=1,3),J=1,N
• ∗ NPO number of orbital projections (≤ 10)
NPB number of band projections(≤ 10)
if NPO 6= 0 insert NPO sets of records II
• ∗ NO number of A.O.’s in the I-th projection
• ∗ IQ(I),I=1,NO sequence number of the A.O.’s in the I-th projection - basis set se-
quence.
if NPB 6= 0 insert NPB sets of records II
• ∗ NB number of bands in the I-th projection
• ∗ IB(I),I=1,NB sequence number of the bands in the I-th projection

Warning The calculation of the Fermi energy is necessary (NEWK keyword with IFE =
1).
The Electron Momentum Density is calculated along given directions (equation 11.22, page
228). It can be computed also for open-shell systems. The electron momentum distribution,
EMD, is a non-periodic function; it falls rapidly to zero outside the first Brillouin zone. ρ(0)
gives the number of electrons at rest. The oblique coordinates directions given in input refer

165
to the conventional cell, not to the primitive cell, for 3D systems.
Example: in a FCC system the input directions refer to the orthogonal unit cell frame (sides
of the cube) not to the primitive non-orthogonal unit cell frame.

EMDLPG - Electron Momentum Density from Density Matrix - line

path
rec variable value meaning
• ∗ N number of directions (≤ 10)
PMAX maximum momentum value (a.u.) for which the EMD is to be calcu-
lated
STEP interpolation step for the EMD
IPLO 0 no data stored on disk
1 formatted output to file fort.25 for plotting (Appendix D, page 261).
2 formatted output to file LINEA.DAT for plotting (Appendix D,
page261).
LPR113 6= 0 printing of EMD before interpolation
• ∗ (K(I,J), directions in oblique coordinates
I=1,3),J=1,N
• ∗ ICASO 1: the total density matrix is used
2: the core density matrix is used
3: the valence density matrix is used

The Electron Momentum Density is calculated along given directions (equation 11.23, page
228) directly from the Density Matrix. The Electron Momentum Density, EMD, is a non-
periodic function; it falls rapidly to zero outside the first Brillouin zone. ρ(0) gives the number
of electrons at rest. For Open-Shell systems the α + β and the α − β EMD are computed. The
oblique coordinates directions given in input refer to the conventional cell, not to the primitive
cell, for 3D systems.

EMDP - Electron Momentum Density - plane maps

rec variable value meaning
• ∗ NP number of planes (< 5)
IS shrinking factor.
IPLO 0 no data stored on disk.
1 formatted output on Fortran unit 25 for plotting
LPR115 printing of the EMD function in output
insert NP set of records
• ∗ (L1(J),J=1,3), fractional coordinates of the reciprocal lattice vectors that identify the
(L2(J),J=1,3) plane
• ∗ PMX1 maximum p value along the first direction
PMX2 maximum p value along the second direction
• ∗ NPO number of orbital projections (≤ 10)
NPB number of band projections(≤ 10)
if NPO 6= 0 insert NPO set of records II
• ∗ NO number of A.O.’s in the I-th projection
• ∗ IQ(I),I=1,NO sequence number of the A.O.’s in the I-th projection - basis set order
if NPB 6= 0 insert NPB set of records II
• ∗ NB number of bands in the I-th projection
• ∗ IB(I),I=1,NB sequence number of the bands in the I-th projection

Warning The Fermi energy must be computed (NEWK keyword with IFE = 1)
Calculation of electron momentum density on definite planes (equation 11.22, page 228). It
works also for open-shell systems.
If LPR115 6= 0 the EMD function is printed in output (recommended).

166
The fractional coordinates of the reciprocal lattice vectors given in input refer to the conven-
tional cell, not to the primitive cell, for 3D systems.
Example: in a FCC system the input directions refer to the orthogonal unit cell frame (sides
of the cube) not to the primitive non-orthogonal unit cell frame.

END
Terminate processing of properties input. Normal end of the program properties. Subse-
quent input records are not processed.

EXTPRT
See input block 1, page 41

FMWF - Wave function formatted output

The keyword FMWF, entered in properties input, writes formatted wave function data (same
data are written in file fort.9, unformatted, at the end of SCF) in file fort.98 (LRECL=80).
The formatted data can then be transferred to another platform. No input data required.
The resources requested to compute the wave function for a large system (CPU time, disk
storage) may require a mainframe or a powerful workstation, while running properties is not
so demanding, at least in terms of disk space. It may be convenient computing the wave
function on a given platform, and the properties on a different one.
The keyword RDFMWF, entered in the first record of the properties input deck reads
formatted data from file fort.98, and writes unformatted data in file fort.9. The key dimensions
of the program computing the wave function and the one computing the properties are checked.
If the dimensions of the arrays are not compatible, the program stops, after printing the
PARAMETER statement used to define the dimension of the arrays in the program which
computed the wave function. The sequence of the operations, when transferring data from one
platform to another is the following:

platform program input action

1 properties FMWF wave function formatted to file fort.98
ftp file fort.98 from platform 1 to platform 2
2 properties RDFMWF wf read from file fort.98 (formatted) and written in
file fort.9 (unformatted)

FRACTION - unit of measure

Unit of measure of coordinates in the periodic direction (ECHG, POTM, CLAS) See input
block 1, page 44.

GRID3D - Selected property computed on a 3D grid

rec variable meaning

•∗ NP Number of points along the first direction
•A CHARGE electronic charge selected - see ECH3 input
or
•∗ POTENTIAL electronic charge selected - see POT3 input

The propertty to be computed at the grid points is chosen by a keyword. Input as required by
the selected property follows.
Computed data are written, formatted, to file fort.31. See Appendix D, page 265, for descrip-
tion of the format.

167
INFOGUI/INFO - output for visualization
No input data required.
Information on the system and the computational parameters are written formatted in fortran
unit 32, in a format suitable for visualization programs.
See Appendix D, page 265, for description of the format.

ISOTROPIC - Fermi contact - Hyperfine electron-nuclear

spin interaction isotropic component
rec variable meaning
• A keyword enter one of the following keywords:
ALL Fermi contact is evaluated for all the atoms in the cell
or
UNIQUE Fermi contact is evaluated for all the non-equivalent atoms in the cell
or
SELECT Fermi contact is evaluated for selected atoms
• ∗ N number of atoms where to evaluate Fermi contact
• ∗ IA(I),I=1,N label of the atoms

The spin density at the nuclei (hρspin (r N )i) is evaluated. This quantity is given in bohr−3 and
is transformed into the hyperfine coupling constant aN [mT] through the relationship [131]
1000 2
aN [mT] = µ0 βN gN hρspin (r N )i = 28.539649 gN hρspin (r N )i
(0.529177 · 10−10 )3 3

where

µ0 = 4π · 10−7 = 12.566370614 · 10−7 [T2 J−1 m3 ] (permeability of vacuum)

βN = 5.0507866 · 10−17 [JT−1 ] (nuclear magneton)

the nuclear gN factors for most of the nuclei of interest are available in the code and are taken
from [131]. Conversion factors:

aN [mT]ge βe
aN [MHz] = = 28.02.6 · aN [mT]
109 h[Js]
aN [MHz]108
aN [cm−1 ] = = 0.33356410 · 10−4 · aN [MHz]
c[ms−1 ]
aN [J] = ge βe 10−3 aN [mT] = 1.856954 · 10−26 aN [mT]
where:

βe = 9.2740154 · 10−24 [JT−1 ] (bohr magneton)

ge = 2.002319304386 (free-electron g factor)
c = 2.99792458 · 108 [ms−1 ] (speed of light in vacuum)
−34
h = 6.6260755 · 10 [Js] (Planck constant)

For extended printing (tensor in original cartesian axes and in principal axis system) insert,
before the keyword ISOTROPIC:

SETPRINT
1
18 1

See tests 29, 31, 32, 33.

168
KNETOUT - Reciprocal lattice information - Fock/KS eigenvalues
Obsolete. See CRYAPI OUT, page 160.

LOCALWF - Localization of Wannier Functions (WnF)

Wannier functions are computed fromBloch functions, and then localized, following the method
described in [140] and [139] The method applies to non-conductor systems only.
The localization of Wannier Functions (WnF) is controlled by parameters. Default values are
supplied for all parameters.
Optional keywords allow modification of the default choices, recommended to developers only.
The LOCALWF block is closed by the END keyword.
For UHF calculations two set of blocks must be inserted for the α and β electrons, each one
ending with the keyword END.
1. The NEWK option must be executed before running LOCALWF, to compute theBloch
functions.
2. The number of k points required for a good localization depends on the characteristics of
the bands chosen. For core electrons or valence bands in non-conducting materials, an IS
slightly larger than that used in the SCF part is enough to provide well localized WnFs.
For valence bands in semiconductors or conduction bands the k-point net is required to
be denser, but there are no recipes to determine a priori the optimum IS value. The
IS value chosen determines a Born-von Karman supercell (or cyclic cluster) from which
the program a priori estimates the memory space that should be enough to contain all
WnF coefficients lower than the threshold 10−IT DP (see the meaning of IT DP in what
follows) in real space. The size of this crystal domain in terms of unit cells is provided in
output before the localization procedure. If the crystal domain is too small usually the
localization fails and IS must be increased. On the other hand, if it is too large (very
large IS) the memory space reserved for the WnF coefficients becomes overestimated and
the calculation may stop because of a lack of memory for array allocation.
3. The efficiency of the localization can be controlled using the CYCTOL parameters. In
most cases, increasing ITDP and/or ICONV leads to larger and more accurate localiza-
tion of the WnFs.
4. The RESTART option admits MAXCYCLE = 0, then the program just reconstructs
all the information about the WnFs given in file fort.81 but does not continue the lo-
calization. This two options together with a IS=1 in NEWK is useful to perform the
analysis of the WnFs after localization by means of the PRINTPLO option.

Definition of the set of bands considered in the localization process

VALENCE
Valence bands are chosen for localization.
OCCUPIED
All the occupied bands are chosen for localization [default].
INIFIBND
rec variable value meaning
• ∗ IBAN initial band considered for localization
LBAN last band considered for localization

BANDLIST
rec variable meaning
• ∗ NB number of bands considered
• ∗ LB(I),I=1,NB labels of the bands.

169
Tolerances for short and large cycles
A short cycle is a sequence of wannierization and Boys localization steps. The accuracies in
both, the calculation of the Dipole Moments (DM) and the definition of the phases assigned
to each periodically irreducible atom, are controlled so that they increase as the localization
process evolves. This results in a significant saving of computational cost. Therefore, each
time a given criterion is fulfilled, the accuracy in the DM evaluation increases and a new large
cycle starts.
CYCTOL
rec variable value meaning
• ∗ ITDP0 >0 Initial tolerance used to calculate the DM matrix elements:
10−ITDP0 2
ITDP >0 Final tolerance used to calculate the DM matrix elements:
10−ITDP 5
ICONV >0 Convergence criterion to finish a large cycle: ABS(ADI(N) -
ADI(N-1)) < 10−ICONV , where ADI is the atomic delocaliza-
tion index and N is the short cycle number 5

PHASETOL
rec variable value meaning
• ∗ ITPH0 >0 10−ITPH0 is the initial tolerance on the atomic charge popula-
tion to attribute the phase to atoms in the wannierization step
2
ITPH >0 10−ITPH is the final tolerance used to attribute this phase 3
ICHTOL >0 DM tolerance of the cycle where ITPH0 changes to ITPH.
ITDP0+1

Restart Keywords
With this option the WnFs of a previous run are read from file fort.81 (copy of file fort.80,
written by the previous run) and projected onto the current occupied subspace. Along with
the projection the WnFs are re-orthonormalized within the Born-von Karman cyclic cluster.
Tolerances and active bands must be the same as in the previous run.
Three variants are possible, corresponding to three different keywords.

RESTART: can be used to start a new localization or to improve a previous one. After
the projection, cycles of wannierization and localization are performed until convergence is
attained.
RESTORTH: similar to the RESTART keyword but here the WnF are orthonormalized in di-
rect space (where the tails feature different topological properties than in the Born-von Karman
cyclic cluster) immediately after their reading. Next the localization properties in direct space
of the resulting WnFs may optionally be improved using the appropriate keywords (WANDM
and FULLBOYS).
FIXWF: this keyword can be used only if file fort.80 of the previous run has been gener-
ated with the keyword SYMMWF. After projection, no further Boys localization step is
performed. Both the original symmetry and the WnF labeling are kept with a negligible loss
in the localization indices.
The last option is recommended when a sequence of CRYSCOR calculations are to be per-
formed corresponding each time to small geometrical changes. Accordingly, to ensure a smooth
evolution of the energy and wave function, all indices concerning the symmetry and the label-
ing of the WnFs are kept to be the same along the sequence. This is required in particular
in geometry optimizations and scanning along geometrical parameters (for instance: in the
calculation of molecular physisorption energy curves – see the CRYSCOR Users’s Manual and
the CRYSCOR Tutorials for a detailed description of the procedure). As for the FIXINDEX
option (page 86), the reference calculation should be the one with the most compact structure.

170
General Keywords
MAXCYCLE
rec variable value meaning
• ∗ NCYC >0 maximal number of short cycles for the iterative process 30

BOYSCTRL

Parameters that control the Boys localization step. Convergence of the process is achieved
when the orbital-stability conditions: Bst = 0; Ast > 0, (see Pipek and Mezey 1989 [99]) are
fulfilled for all pairs st of WnFs. Additionally, in order to avoid nearly free rotations (for
instance in core or lone-pair WnFs) those pairs st with Ast close to 0 are not mixed (frozen).

rec variable value meaning

• ∗ IBTOL 10−IBTOL is the threshold used for the stability condition on
Bst . 4
IBFRZ If for a pair of WnFs st, |Ast | ≤ 10−IBFRZ , then the corre-
sponding WnFs are not mixed. 4
MXBCYC Maximum number of cycles allowed in the Boys localization
process 500

Initial guess options

The iterative localization process of the WnFs needs to start from a reasonable initial guess.
By default the starting functions are obtained automatically from the Hamiltonian eigenvectors
at the Γ point. When required (pure covalent bonds that link atoms in different unit cells), a
pre-localization is performed using a scheme similar to that suggested by Magnasco and Perico
(1967) [82].

IGSSCTRL

Parameters used to control the pre-localization of the Γ point eigenvectors.

rec variable value meaning

• ∗ CAPTURE The capture distance between atoms I and J is given by
CAPTURE ∗ (RAYCOV(I) + RAYCOV(J)) (RAYCOV, cova-
lent radius (default value table page 53). An inter-atomic dis-
tance lower than the capture indicates that I and J can be
covalently bonded. Default value [ 2.0 ].
MPMAXIT Maximum number of iterations in the pre-localization
process 200
ICNVMP 10−ICNVMP is the convergence threshold for the Magnasco-
Perico pre-localization 8
IOVPOP Just those pairs of atoms whose overlap population are greater
than 10−IOVPOP are considered covalently bonded 4
The initial guess can be given as input in two mutually exclusive ways, controlled by the
keywords IGSSVCTS and IGSSBNDS:

IGSSVCTS

The eigenvectors and the phases are given explicitly after the LOCALWF block (and before
the plot parameters if required), in the following format.

171
rec variable value meaning
• ∗ NGUES Number of bands whose phase is pre-assigned such that the
involved atoms are to be located in a given cell.
insert 2 × NGUES records
• ∗ IB
• ∗ IGAT(I,IB),I=1,NAF
Index of the direct lattice vector corresponding to the cell
where atom I is expected to have the largest charge population
in Wannier IB (NAF is the number of atoms per cell)
insert:
GUESSV(I),I=NDF*NOCC
where NDF is the basis set dimension and NOCC the num-
ber of bands considered. GUESSV is a matrix containing the
initial guess vectors for the iterative Wannier-Boys procedure
(GUESSV is written in free format as a one-dimensional ar-
ray).

IGSSBNDS

Use this option to explicitly indicate the WnFs that are to be assigned to covalent bonds.

rec variable value meaning

• ∗ NBOND Number of covalent bonds given as input.
insert NBOND records
• ∗ NAT1 Label of the first atom of the bond, it is assumed to be located
in the reference cell.
NAT2 Label of the second atom of the covalent bond
IC1,IC2,IC3 Indices of the cell where atom NAT2 is located
NBNDORD Bond Order

CAPTURE

The value of the CAPTURE parameter (see IGSSCTRL can be redefined.

rec variable value meaning

• ∗ CAPTURE The capture distance between atoms I and J is given by
CAPTURE ∗ (RAYCOV(I) + RAYCOV(J)) (RAYCOV, cova-
lent radius (default value table page 53). An inter-atomic dis-
tance lower than the capture indicates that I and J can be
covalently bonded. Default value [ 2.0 ].

Symmetry adaptation of Wannier Functions

SYMMWF

The WnFs a-posteriori symmetrization procedure [22], activated by the SYMMWF keyword,
is mandatory in the case of a subsequent CRYSCOR calculations. A brief outline of the
procedure can help to orient in the particular nomenclature adopted to define the symmetrized
WnFs and their symmetry relations.

1. after the localization step, the WnFs are centered at different Wickoff sites of the reference
cell which are invariant with respect to point-symmetry subgroup H of the space group
G of the crystal;
2. for each site, a coset decomposition of G by H is performed, thus leading to the definition
|G|
of NF = |H| symmetry operators (coset representatives) which rotate the reference site
into equivalent ones;
3. among each set of equivalent sites, a reference site is chosen;

172
 4. the WnFs of the reference site are symmetrized according to the corresponding subgroup
H: each of these symmetrized WnFs is basis of an irreducible representations (IRREP) of
the subgroup H and hereafter each component of the irreducible basis set will be referred
to as petal;

5. the collection of petals belonging to the same IRREP of H ,the whole irreducible basis
set, constitutes a so-called flower. It is worth noting that bi/three-dimensional IRREPs
gives rise to flower made up of two/three petals, respectively;

6. the rotation of the reference flower performed by means of the corresponding NF -1 coset
representatives (identity excluded) yields the creation of others NF -1 symmetry related
flowers; the set of such equivalent flowers constitutes a bunch;

7. in general, more than one bunch could be associated to the same reference site;

8. as a result of this procedure, each WnF is fully classified by four indices |b,f,p,gi (b
= bunch, f=flower, p= petal, g= crystal cell) such that a general symmetry operator of
the system R b |b,f,p,gi = P 0 [A(R)]pp0 |b,fR ,p’,gR i.
b ∈ G, applied to a WnF gives: R
p

In addition to the general keywords of the localization step, a set of optional keywords, recom-
mended to developers only, can be used to modify some default settings.
The SYMMWF input block must be closed by the keyword END.

rec variable value meaning

• A TOLBOND redefinition of the tolerance to classify WFs as bond or
atomic - default value [ 0.2 ]
• F TOLB if |[pi1 − pi2 ]| < tolb then the WnF ωi is a bond one (pix
is the atomic population of the two atoms which most
contribute to the WnF i)
• F TOLSYM a WnF ωi is classified as already symmetric if
hωj |R|ωi i > T OLSY M - default value [ 0.99 ]
• A NOSYMAP an alternative algorithm for the unpacking of WnFs (from
the reciprocal to the direct lattice) is followed
• A PRINT default value [ 0 ]

WnFs’s Quality. Get WnFs of good quality, in terms of norm and symmetry, is a necessary
prerequisites for obtaining reliable energy at the MP2 level. The output file contains some
useful information that can be used to check the quality of the solution: as a general and
safety rule, the final value for ”ERR PER WF” should be less than 1∗10−5 and the precision
on the scalar products between WnFs (printed by setting the PRINT option equals 2) should
not be less than 1∗10−5 .
The solution can be eventually improved by tuning some computational parameters; in partic-
ular the user can:
• set tighter tolerances for the evaluation of two-electron integrals in the HF reference solution
(TOLINTEG): despite the increase of computational time, it turns out that the localization
procedure is particularly sensitive to the the first threshold;
• increase the number of k points (NEWK) according to the suggestions reported in the LO-
CALWF introduction paragraph;
• use more severe values for the short and large cycles tolerances (CYCTOL).
Finally, in some case, the activation of the FULLBOYS option can be decisive.

173
Printing Options and Plot of the WnFs
PRINTPLO
rec variable value meaning
• ∗ IPRT 0 Does not print Wannier coefficients [default]
>0 Prints Wannier coefficients at each cycle up to the IPRT-th
star of direct lattice vectors 0
IPRP 0 Prints population analysis only at the end of the localization.
6=0 Prints analysis at each W-B cycle 0
ITPOP Only atomic population larger than 10−ITPOP are printed 2
IPLOT 0 WnFs are not computed for plot
6= 0 WnFs are computed in a grid of points, IPLOT being the
number of stars of direct lattice vectors taken into account for
WnF coefficients. Data are written in file fort.25 0

If IPLOT 6= 0 insert after the LOCALWF keyword block (defining the localization procedure
computational parameters, and terminated by END) the following data:

rec variable value meaning

• ∗ NWF number of WnF to plot
insert NWF blocks of data
• ∗ NUMBWF sequence number (output order) of the WnF to plot
MAPNET input data (Section 8.2, page 177)
Each block defines the index number of WnF to be computed in a grid of points,
followed by data defining the frame inside which the value of localized WnF has
to be computed in a grid of points (see MAPNET, 177. The package Crgra2006
(http://www.crystal.unito.it/Crgra2006.html) allows plotting the function as contour lines.
The WnFs and the WnFs densities (in this order) within the selected regions are given in file
fort.25.

New keywords - developers only

CLUSPLUS

Upon transformation from Bloch Functions to Wannier Function, the latter are defined within
a region with cyclic boundary condition imposed. We call it the ”cyclic cluster”. The volume
of this region depends on the shrinking factor used in the previous NEWK. For instance, if
IS=4, then the cyclic cluster in a 3D system will be 4**3 times larger than the primitive cell.
For the localization part to work the WnFs are required to be described in the real space,
hence the cyclic conditions and the WnFs are mapped onto a cluster in direct space. The size
of this cluster where the localization is performed is defined as follows:

1. We define a small cluster, as a spherical region that contains the minimum number of
G-vectors that fully map the cyclic cluster. Let’ s call RO its radius.
2. As the centroid of some WnFs may be at the border of the reference cell we should
consider some additional space in the direct cluster so as to allow the tails to be fully
included in the region. This additional distance R1 is calculated as the maximum G-
vector modulus of the set of cells at the neighbours of the reference one.
3. The radius of the resulting direct cluster will read: R = R0 + IPLUSCLUS*R1, where
IPLUSCLUS is given in input. By default IPLUSCLUS is 5.

rec variable value meaning

• ∗ iplusclus factor to define the radius of direct cluster

ORTHNDIR

174
After the WANNIER-BOYS localization the WnFs are not fully orthonormal in direct space
(they are just orthonormal within the cyclic cluster). To perform a true localization in direct
space (see FULLBOYS) a previous re-orthonormalization in direct space is required. This
is carried out by constructing the first order approximation of the Lowdin transformation and
applying it to the WnFs. This process is performed iteratively up to fulfill a given criterion.
ORTHNDIR sets the parameters that control this process.

rec variable value meaning

• ∗ ISTORTH >0 number of stars of G-vectors that contains the transformation
matrix.
=0 the number of stars is computed so as to contain the reference
cell and all its neighbors [default].
ITOLORTH >0 the overlap matrix elements are computed just between WnF
components gt 10**-ITOLORTH in absolute value [default 5].
NREORTHN ≥0 maximum number of iterations [default 10 in properties, 0 in
crystal.
<0 the iterative procedure is performed up to the mean normal-
ization error of the WnFs is < 10**NREORTHN in absolute
value (Default -7)

WANDM

WANDM controls the computation of the DM matrix elements between WnFs assigned to the
reference and the neighboring cells (translational images of the former).

rec variable value meaning

• ∗ INEIGH controls the extent of the DM matrix by limiting the neigh-
boring cell around the origin considered in the computation of
the matrix elements:
>0 number of stars of neighboring cells considered for the matrix
elements of DM
<0 the DM matrix is computed up to star of neighbor IS-
TAR with the condition that ABS(ALOCLEN(ISTAR)-
ALOCLEN(ISTAR-1))¡10**(-ABS(INEIGH)), where ALO-
CLEN(ISTAR) means ”localization length computed up to
star ISTAR”
• ∗ TOLDM tolerance in the WnF coefficients used to calculate the DM
matrix elements (see CYCTOL)

FULLBOYS
rec variable value meaning
• ∗ ITOLWPG >0 TOLWPG 10**(-ITOLWPG) tolerance on the DM matrix el-
ements

Request of Foster-Boys localization in direct space. The set of WnFs considered in the calcula-
tion of the DM matrix (see WANDM) are orthogonally transformed so as to obtain maximally
localized WnFs under the Boys criterion. The resulting functions keep both, orthonormality
and translational equivalence.
TOLWPG 10**(-TOLWPG) tolerance on the DM matrix elements to keep and use it in com-
pact form. A small TOLWPG means that only a few DM matrix elements are considered in
the localization process, then the calculation is quite fast and not very demanding in mem-
ory. A very large value would bring about very accurate LWFs with high computational cost.
Recommended values: 4-6.

175
Bibliography

N. Ashcroft, D. Mermin ”Solid State Physics”, Holt Rinehart and Winston: New York, 1976.
P.-O. Löwdin (Editor) ”Quantum Theory of Atoms, Molecules and the Solid State”, Academic:
New York, 1966.
S. F. Boys Rev. Mod. Phys 32 (1960) 296.
J. M. Foster and S. F. Boys Rev. Mod. Phys 32 (1960) 300.
J. Pipek and P. G. Mezey J. Chem. Phys 90 (1989) 4916.
V. Magnasco and A. Perico, J. Chem. Phys. 47 (1967) 971.

176
 MAPNET - coordinates of grid points on a plane
This is a dummy keyword, to explain the way is generated the grid of points in which is
evaluated a given function F: charge density and spin density (ECHG), electrostatic potential
(CLAS, POTM). The graphic representation of the resulting 2D function is made by external
software.
rec variable meaning
• ∗ NPY number of points on the B-A segment.
• A keyword enter a keyword to choose the type of coordinate:
• COORDINA
• ∗ XA,YA,ZA cartesian coordinates of point A
• ∗ XB,YB,ZB cartesian coordinates of point B
• ∗ XC,YC,ZC cartesian coordinates of point C
or
• ATOMS
• ∗ IA label of the atom at point A
indices (direct lattice, input as reals) of the cell where the atom is located
AL,AM,AN
• ∗ IB label of the atom at point B
indices (direct lattice, input as reals) of the cell where the atom is located
BL,BM,BN
• ∗ IC label of the atom at point C
indices (direct lattice, input as reals) of the cell where the atom is located
CL,CM,CN
optional keyword II
• RECTANGU definition of a new A’B’C’D’ rectangular window, with B’C’ on BC, A’D’
on AD and diagonals A’C’=B’D’=max(AC,BD) (see Fig 8.1)
optional keyword II
• MARGINS definition of a new A”,B”,C”,D” window including ABCD (or A’B’C’D’)
(see Fig 8.2)
• ∗ ABM margins along AB
CDM margins along CD
ADM margins along AD
BCM margins along BC
optional keyword II
• PRINT printing of the values of the function in the net
• ANGSTROM cartesian coordinates in Ångstrom (default)
• BOHR cartesian coordinates in bohr
• FRACTION cartesian coordinates in fractional units
• END end of MAPNET input block

1. Function F is mapped in a ABCD parallelogram-shaped domain defined by the sides AB

and BC of any ABC
\ angle. F is calculated at the nAB * nBC nodes of a commensurate
net (nAB and nBC integers).

2. If C ≡ B, F is calculated along the line AB. Data are written in file RHOLINE.DAT D.

3. nBC is set by the program such that all points in the net are as equally spaced as possible
( δAB ≈ δBC ).

4. formatted output is written in file fort.25 (processed by Crgra2006; see Appendix D, page
259.

5. The position of the three points A, B and C can be specified in two alternative ways:
COORDINA the cartesian coordinates of the three points are given in bohr /
Ångstrom / fractional units (default Ångstrom; see Section 2.1,
page 31)
ATOMS A,B,C correspond to the position of 3 nuclei, identified by their
sequence number in the reference cell, and the crystallographic in-
dices of the cell in which they are located (input as real numbers).

177
 A’ A D≡D’
























B≡B C C’

Figure 8.1: Definition of the window where the function F is mapped Effect of optional keyword
RECTANGU.

A’ D’

6





ADM


A ? D




















ABM -
 CDM

-



























B 6
BCM C


?

B’ C’

Figure 8.2: Definition of frame around the original window where the function F is mapped.
Effect of optional keyword MARGINS.

6. The symmetry is used to restrict the calculation of the function to the irreducible part of
the parallelogram chosen. To maximize the use of symmetry, the points of the net should
include the low multiplicity positions in the selected plane. For example, B=(0,0,0),
A=(a,0,0), C=(0,b,0) (a,b lattice vectors). Choose NPY=4n+1 for (100) faces of cubic
crystals, or NPY = 6n+1 for (0001) faces of hexagonal crystals.

NEIGHBOR/NEIGHPRT
See input block 1, page 51

178
 NEWK - Hamiltonian eigenvectors
rec variable value meaning
if system is periodic, insert II
• ∗ IS Shrinking factor for reciprocal space net (Monkhorst net). The num-
ber NKF of k points, where the Fock/KS matrix is diagonalized, is
roughly proportional to IS IDIM /M V F where IDIM denotes the pe-
riodic dimensionality of the system, and MVF denotes the number of
point symmetry operators (see page 23).
ISP Shrinking factor of the secondary reciprocal space net (Gilat net) for
the evaluation of the Fermi energy and density matrix.
if system is periodic and IS=0, insert II
• ∗ Shrinking factors of reciprocal lattice vectors
IS1 Shrinking factor along B1
IS2 Shrinking factor along B2
IS3 Shrinking factor along B3.
• ∗ IFE 0 no Fermi energy calculation is performed;
1 Fermi energy is computed, by performing integration on the new k
points net. Total, valence and core density matrices are written on
Fortran unit 13
NPR number of printing options to switch on
if NPR 6= 0 insert prtrec (see page 54) II

The Fock/KS eigenvectors are computed at a number of k points in reciprocal space, defined
by the shrinking factor IS, and written unformatted in file fort.10 (in the basis of symmetry
adapted Bloch functions) and in file fort.8 (in the basis of AO). Eigenvalues and related infor-
mation (coordinates of k points in reciprocal lattice, weights etc) are written in file KIBZ.DAT
by inserting the keyword CRYAPI OUT (page 160).
1. The Fock/KS matrix in direct space is always the SCF step final one. If the SCF con-
vergence was poor, and convergence tools were used, eigenvalues and eigenvectors may
be different from the ones that could be obtained after one more cycle without any
convergence trick.
2. The shrinking factors IS and ISP (Section 11.7, page 227) can be redefined with respect
to the ones used in the SCF process. If this value is smaller than the one used in the scf
step, numerical inaccuracy may occur in the Fourier transform of the Fock/KS matrix,
Fg → Fk (Chapter 11, equation 11.5).
3. A Fermi energy calculation must be performed (IFE=1) to run PROF the Compton
profiles option, PBAN and PDIDE in order to compute the weight associated to each
eigenvalue.
4. Warning NEWK with IFE=1 must be run to obtain the correct Fermi energy and eigen-
values spectra when a shift of eigenvalues was requested in SCF (LEVSHIFT, page 92;
SPINLOCK, page 101; BETALOCK, 73.
A new density matrix is computed. If the convergence of scf was poor, and convergence
tools were used (FMIXING, LEVSHIFT, ..), the density matrix obtained from the eigen-
vectors computed by NEWK may be different from the matrix that could be calculated
with one more scf cycle. Properties depending on the density matrix may be different if
computed before or after NEWK.
5. if BAND is called after NEWK, and symmetry adapted Bloch functions are used
(default option), the information generated by NEWK is destroyed. For instance, to
compute density of states and bands, the sequence must be: BAND - NEWK - DOSS.
The sequence NEWK BAND DOSS will give the error message:
NEWK MUST BE CALLED BEFORE DOSS
Printing options: 59 (Density matrix - direct lattice); 66 (Hamiltonian eigenvalues); 67 (Hamil-
tonian eigenvectors).

179
NOSYMADA
See input block 3, page 95

PARAMPRT - - printing of parametrized dimensions

See input block 1, page 52.

PATO - Density matrix as superposition of atomic densities

rec variable value meaning
• ∗ IBN 0 density matrix computed with the same basis set as in the crystal cal-
culation.
6= 0 new basis set and/or new electron configuration is given
NPR 6= 0 printing of the density matrix for the first NPR direct lattice vectors
if IBN 6= 0 insert basis set input (page 19) II

1. The PATO option is used for calculating crystal properties, such as charge density
(ECHG), structure factors (XFAC) with a periodic density matrix obtained as a su-
perposition of atomic solutions (periodic array of non interacting atoms). The reducible
density matrix is written to file fort.13.

2. The atomic wave function is computed by the atomic program [6], using HF hamiltonian,
s, p, d orbitals basis set, properly handling the open shell electronic configuration.

3. If the basis set used for the crystalline calculation (given as input of the integral part)
is not suitable for describing a free- atom or free-ion situation, a new basis set can be
supplied (see Section 1.2). When this option is used (IBN.NE.0) the basis set of all the
atoms with different conventional atomic number has to be provided.

4. The electronic configuration of selected atoms may be modified (CHEMOD in basis set
input). This allows calculation of the density matrix as superposition of atomic densities
or ionic densities, for the same crystal structure.

5. The wave function data stored in file fort.9 at the end of the SCF cycles are not modified.
Only the data stored on the temporary data set (reducible density matrix in fortran unit
13 and overlap matrix in fortran unit 3) are modified. The keyword PSCF restores the
scf density matrix and all the original information (including geometry and basis set).

6. See also ATOMHF, input block 3, page 72, and CHARGED, input block 2, page 62.

PBAN/PDIBAN - Band(s) projected density matrix

rec variable meaning
• ∗ NB number of bands to consider.
NPR printing of the density matrix for the first NPR direct lattice cells.
• ∗ N(I),I=1,NB sequence number of the bands summed up for the projected density ma-
trix.

A density matrix projected onto a given range of bands is computed and stored in fortran unit
13. The properties will subsequently be computed using such a matrix.
For spin polarized systems, two records are written:
first record, total density matrix (N=nα + nβ electrons);
second record, spin density matrix (Ns=nα - nβ electrons).

To be combined only with ECHG and PPAN. Fock/Kohn-Sham eigenvectors and band
weights must be precomputed by running NEWK and setting IFE=1.

180
 PGEOMW - Density matrix from geometrical weights
A density matrix projected onto the range of bands defined in input (see PBAN input instruc-
tions) is computed, using the geometrical weights of the k points in the reciprocal lattice. The
properties will subsequently be computed using such a matrix. All the bands are attributed an
occupation number 1., independently of the position of the Fermi energy. The density matrix
does not have any physical meaning, but the trick allows analysis of the virtual eigenvectors.
For spin polarized systems, two records are written:
first record, total density matrix (N=nα + nβ electrons);
second record, spin density matrix (Ns=nα - nβ electrons).
To be combined only with ECHG and PPAN.
Fock/Kohn-Sham eigenvectors and band weights must be computed by running NEWK and
setting IFE=1. Symmetry adaptation of Bloch functions is not allowed, the keyword NOSY-
MADA must be inserted before NEWK.

PDIDE - Density matrix energy projected

rec variable meaning
• ∗ EMI,EMAX lower and upper energy bound (hartree)

A density matrix projected onto a given energy range is computed and stored in file fort.13.
The properties will subsequently be computed using such a matrix. To be combined only with
ECHG and PPAN. Fock/Kohn-Sham eigenvectors and band weights must be computed by
running NEWK and setting IFE=1.
The charge density maps obtained from the density matrix projected onto a given energy range
give the STM topography [79] within the Tersoff-Haman approximation [123].

PMP2 - Reads the MP2 correction to the Valence Density Matrix

rec variable value meaning
• ∗ IMP2 0 The HF+MP2 Density Matrix is used
1 The MP2 Density Matrix is used

1. The MP2 correction to the Valence Density Matrix, as computed by the CRYSCOR
program, is read from the external unformatted file fort.63, which has to be provided by
the user.

2. The PMP2 keyword must be preceded by NEWK (with IFE=1) since the Fermi energy
has to be computed in order to split the total SCF DM into core and valence density
matrices. The MP2 correction to the DM is summed to the valence SCF DM.

3. By default the PROPERTIES program adopts the SCF density matrix (DM) but if one
adds the PMP2 keyword in the input file then any property whose corresponding key-
word is inserted after PMP2 is computed using the HF+MP2 density matrix (IMP2=0)
or the pure MP2 one (IMP2=1), where the former choice is the recommended one and
can be used in order to compute any DM-related quantity, while the latter can be used
just for plotting electron charge density maps (ECHG keyword).

4. For instance, in order to evaluate the MP2 correction to the X-rays Structure Factors Fhkl ,
the following sequence of keywords has to be used: NEWK-XFAC-PMP2-XFAC-
END; in the output file will be printed the HF Fhkl followed by the HF+MP2 Fhkl .

We report in the following the list of the properties for which the MP2 correction can be
evaluated, along with some advices one has to take in mind:

• Directional Compton Profiles (PROF) The ICORE variable must be 1 or 3 and the IVIA
variable must be 1

181
 • Auto-correlation Function (BIDIERD) The ICASO variable must be 1 or 3
• Electron Momentum Density (EMDLPG) The ICASO variable must be 1 or 3
• Electron Charge Density (ECHG) This is the only keyword that can be used if IMP2=1
• Mulliken’s Populations (PPAN)
• Structure Factors (XFAC)

POLI - Spherical harmonics multipole moments

rec variable value meaning
• ∗ IDIPO multipole order (maximum order `=6)
∗ ITENS 1 the quadrupole cartesian tensor is diagonalized
0 no action
LPR68 maximum pole order for printing:
<0 atom multipoles up to pole IDIPO
≥0 atom and shell multipoles up to pole IDIPO

The multipoles of the shells and atoms in the primitive cell are computed according to a
Mulliken partition of the charge density, up to quantum number IDIPO (0 ≤ IDIPO ≤ 6). The
first nine terms, corresponding to `=0,1,2 (for the definition of higher terms, see Appendix A1,
page 170 in reference [103]) are defined as follow:
` m
0 0 s
1 0 z
1 1 x
1 -1 y
2 0 z 2 − x2 /2 − y 2 /2
2 1 3xz
2 -1 3yz
2 2 3(x2 − y 2 )
2 -2 6xy
3 0 (2z 2 − 3x2 − 3y 2 )z
3 1 (4z 2 − x2 − y 2 )x
3 -1 (4z 2 − x2 − y 2 )y
3 2 (x2 − y 2 )z
3 -2 xyz
3 3 (x2 − 3y 2 )x
3 -3 (3x2 − y 2 )y

If ITENS=1, the cartesian quadrupole tensor is computed, and its eigenvalues and eigen-
vectors are printed after diagonalization. The components of the cartesian tensor are:
x2 , y 2 , z 2 , xy, xz, yz
Warning: the shell multipoles are not printed by default. On request (keyword POLIPRT),
they are printed in atomic units (electron charge = +1).

POLSPIN - Spin multipole moments

rec variable value meaning
• ∗ IDIPO multipole order (maximum order `=6)
∗ ITENS 1 the quadrupole cartesian tensor is diagonalized
0 no action
LPR68 maximum pole order for printing:
<0 atom multipoles up to pole IDIPO
≥0 atom and shell multipoles up to pole IDIPO

182
The electron spin density is partitioned in atomic contributions according to the Mulliken
scheme, and the spherical harmonic atomic multipoles up to the IDIPO angular quantum
number are evaluated (see theR POLI keyword for definition of the multipoles and references).
The Cartesian tensor Tij = xi xj ρspin (r) dr is computed and diagonalized, and its eigenvalues
and eigenvectors are printed. This option may be useful in the analysis of the size, shape and
orientation of localized electron holes.

POT3 - Electrostatic potential on a 3D grid

rec variable meaning

•∗ NP Number of points along the first direction
•∗ ITOL penetration tolerance (suggested value: 5) (see POTM, page 185)

if non-3D system
keyword to choose the type of grid on the non-periodic direction(s):

SCALE RANGE
length scales for non-periodic dimensions boundary for non-periodic dimensions (au)
if 2D system
• ∗ ZSCALE • ∗ ZMIN
• ∗ ZMAX
if 1D system
• ∗ YSCALE,ZSCALE • ∗ YMIN,ZMIN
• ∗ YMAX,ZMAX
if 0D system
• ∗ XSCALE,YSCALE,ZSCALE • ∗ XMIN,YMIN,ZMIN
• ∗ XMAX,YMAX,ZMAX

The electrostatic potential is computed at a regular 3-dimensional grid of points. The grid
is defined by the lattice vectors of the primitive unit cell and user defined extents in non-
periodic directions. NP is the number of points along the first lattice vector (or XMAX-XMIN
for a molecule). Equally spacing is used along the other vectors. Non-periodic extents may
be specified as an explicit range (RANGE) or by scaling the extent defined by the atomic
coordinates (SCALE).
Formatted data are written in fortran unit 31 (function value at the grid points), in the format
required by the visualization program DLV.
See Appendix D, page 265, for description of the format.
Function data computed at 3D grid points are written according to GAUSSIAN CUBE format
in file POT CUBE.DAT

183
 POTC - Electrostatic potential and its derivatives
rec variable meaning
• ∗ ICA 0 calculation of potential (V ), its first derivative (E) and second derivatives (E 0 )
in one or more points
1 not implemented
2 calculation of V (z), E(z), E 0 (z) and ρ(z) averaged in the plane at z position
(2D only)
3 calculation of V (z), E(z), E 0 (z) and ρ(z) averaged in the volume between z–ZD
and z+ZD (2D only)
NPU n number of points at which these properties are computed
0 these properties are computed at the atomic positions defined by IPA value
IPA 0 calculations are performed at each atomic positions in the cell
1 calculations are performed just for non equivalent atomic positions in the cell
if ICA = 0 then
if NPU > 0 insert NPU records II
• ∗ X,Y,Z point coordinates (cartesian, bohr)
if NPU < 0 data are read from file POTC.INP II
if ICA = 2 insert II
• ∗ ZM,ZP properties are averaged over NPU planes orthogonal to the z axis from z = ZP
to z = ZM by step of (ZP–ZM)/(NPU–1) (bohr)
if ICA = 3 insert II
• ∗ ZM,ZP properties are averaged over NPU volumes centered on planes orthogonal to
the z axis, same as ICA = 2
ZD half thickness of the volume (bohr)

The exact electrostatic potential V , its derivatives E (electric field) and E 0 (electric field
gradient) are evaluated for molecules (0D), slabs (2D) and crystals (3D). Plane and volume
averaged properties can be computed for slabs (2D) only. The plane is orthogonal to the z
axis.
For ICA = 3, the volume average is performed around a middle plane at z position, from z–ZD
to z+ZD, giving a thickness of 2∗ZD.
According to Poisson’s law, the charge density ρ(z) is defined as
1 d2 V (z) −E 0 (z)
ρ(z) = − =
4π dz 2 4π
If an electric field of intensity E0 is present (keyword FIELD, see page 2.1, only for slabs),
the total potential Vf ield (z) is calculated:
Vf ield (z) = V (z) − E0 z
where V (z) is the potential of the slab itself and −E0 z is the perturbation applied.

• ICA = 0 ; NPU > 0; 2D or 3D system

It is possible to enter the cartesian coordinates (bohr) of the points where the exact
value of the properties must be computed. It is useful when applying fitting procedure
to obtain formal point charges.
• ICA = 0 ; NPU < 0; 2D or 3D system
coordinates in bohr are read (free format) from file POTC.INP

record type of data content

1 1 integer N, number of points
2..2+N-1 4 real x y z

• ICA 6= 0; NPU 6= 0; 2D or 3D system

The data computed are written in file POTC.DAT. See Appendix D, page 262.

184
 POTM - Electrostatic potential maps and electric field
rec variable value meaning
• ∗ IDER 0 the electrostatic potential is evaluated
1 the potential and its first derivatives are evaluated
ITOL penetration tolerance (suggested value: 5)
insert MAPNET input records (page 177)

1. When IDER=0, the electrostatic potential is calculated at the nodes of a 2-dimensional

net in a parallelogram-shaped domain defined by the segments AB and BC (see keyword
MAPNET, page 177). The electrostatic potential values are written formatted in file
fort.25 (see Appendix D, page 259).

2. When IDER 6= 0, the electrostatic potential gradient is computed at the nodes of the
same grid. The x, y and z components are printed in the standard output, and written
formatted in file fort.25 (see Appendix D, page 259).

3. The electrostatic potential at r is evaluated [114] by partitioning the periodic charge

density ρ(r) in shell contributions ρhλ :
XX
ρ(r) = ρλ (r − h)
h λ

(h translation vector).

4. The long range contributions are evaluated through a multipolar expansion of ρλ (r − h)

[129]. The short range contributions are calculated exactly.

5. The separation between long and short range is controlled by ITOL: ρλ (r−h) is attributed
to the short range (exact) region if
2
e−αλ (r−sλ −h) > 10−IT OL

where: αλ = exponent of the adjoined gaussian of shell λ; sλ = internal coordinates of

shell λ in cell at h.
The difference between the exact and the approximated potential is smaller than 1%
when ITOL=5 (input datum to POTM), and IDIPO=4 (input datum to POLI), and
smaller than 0.01% when ITOL=15 and IDIPO=6 [114, 129].

6. The multipoles of shell charges are computed if POLI option was not run before POTM.

PPAN/MULPOPAN - Mulliken Population Analysis

See input block 3, page 96.

PRINTOUT - Setting of printing environment

See input block 1, page 53.

185
 PROF - Compton Profiles
rec variable value meaning
• ∗ ICORE 1 core plus valence calculation.
2 core only calculation.
3 valence only calculation.
IVIA 0 valence contribution is computed by numerical integration.
1 valence contribution is computed analytically.
NPR number of printing options to switch on.
IPLO 0 CP related data are not stored on disk
1 formatted CP data stored in file fort.25 (Appendix D, page 261)
2 formatted CP data stored in Fortran unit 24 (Appendix D, page 261)
if NPR 6= 0 insert prtrec (see page 54) II
• A2 CP calculation of Compton profiles (J(q)) along selected directions (eq.
11.27).
• ∗ ND number of directions (≤ 6).
REST maximum value of q for J(q) calculation (bohr−1 ).
RINT internal sphere radius (bohr−1 ).
IRAP shrinking factor ratio.
• ∗ (KD(J,N), J=1,3), directions in oblique coordinates; see note 9
N=1,ND
• ∗ STPJ interpolation step (in interpolated Compton profiles calculation).
• A4 DIFF CP difference between all computed directional CPs.

• A2 BR
auto-correlation function B(r) calculation (eq. 11.30).
• ∗ RMAX maximum r value (bohr) at which B(r) is computed
STBR step in computation of B(r).
• A4 CONV convolution of the data previously computed (CP, DIFF, BR) (eq.
11.29)
• ∗ FWHM convolution
p parameter (a.u.) full width half maximum;
σ = (F W HM )2 /(8 · 2log2).
• A4 ENDP End of input records for CP data

The keyword PROF starts the calculation of Compton profiles (J(q)) along selected directions
(eq. 11.27). The specific keywords DIFF BR CONV allow the calculation of the related
quantities. The card with the keyword ENDP ends the Compton profiles input section.

1. The input of the options must be given in the order in which they appear in the above
description. To enter this property, the CP option must always be selected after PROF,
while the others are optional.

2. Core and valence contributions are computed by using different algorithms. Core con-
tribution to CP’s is always computed analytically via the Pg matrix (direct lattice sum-
mation, equation 11.25); the valence contribution is computed numerically (IVIA=0) by
integrating the EMD (equation 11.23). Valence contribution can be evaluated analyti-
cally, setting IVIA=1.

3. The numerical integration is extended to a sphere (radius RINT) where EMD is sampled
at the points of a commensurate net characterized by a shrinking factor IS (in the IBZ)
and at all the points (with modulus less then RINT) obtained from these by applying
reciprocal lattice translations.
It is possible to define a second sphere (with radius REST); in the volume between
the two spheres a second net is employed with shrinking factor IS1 greater then IS.
IRAP=IS1/IS is given in INPUT card 2; a reasonable value is IRAP=2. The outer
contribution is supposed to be the same for different CP’s, and is obtained by integrating
the average EMD.

186
 4. If ICORE 6= 2 (valence electron CP’s are required) the NEWK option, with IFE=1,
must be run before the PROF option, in order to generate the eigenvectors required for
the EMD calculation, as well as the weights associated with each k point.

5. If ICORE 6= 2 and IVIA = 0 the CPs are evaluated at points resulting from the IS
partition of the reciprocal lattice translators. The interpolation is performed at STPJ
intervals (STPJ is given in input).
If ICORE = 2 or IVIA = 1 the CPs are, in any case, evaluated at points resulting from
STPJ intervals.
IVIA=0 (numerical integration) produces much more accurate results;
IVIA=1 (analytical integration) is to be used only for molecular calculations or for non
conducting polymers.

6. Total CP’s are always obtained by summing core and valence contributions.

7. Reasonable values of the integration parameters depend on the system under investi-
gation. The normalization integral of the CP’s is a good check of the accuracy of the
calculation. For instance, in the case of the valence electron of beryllium (test 9), good
values of RINT and IS are 10. a.u. and 4 respectively. In the case of silicon (test 10),
good values of the same variables are 8. a.u. and 8 respectively. Much greater RINT
values are required in order to include all the core electrons (70. a.u. in the case of
silicon, and 25. a.u. in the case of beryllium).

8. BR (auto-correlation function or reciprocal space form factor) should be calculated only

for valence electrons. All electron BR are reliable when the normalization integral, after
the analytical integration for core electrons contribution, is equal to the number of core
electrons.

9. The oblique coordinates directions given in input refer to the conventional cell, not to
the primitive cell for 3D systems.
Example: in a FCC system the input directions refer to the orthogonal unit cell frame
(sides of the cube) not to the primitive non-orthogonal unit cell frame.

Printing options: 116 (Compton profiles before interpolation); 117 (average EMD before inter-
polation); 118 (printing of core, valence etc. contribution). The LPRINT(118) option should
be used only if ICORE=1, that is, if core plus valence calculation are chosen.

PSCF - Restore SCF density matrix

The wave function data computed at the last SCF cycle are restored in common areas and
fortran units 3 (overlap matrix), 11 (Fock/KS matrix), 13 (density matrix). The basis set
defined in input block 2 is restored. Any modification in the default settings introduced in
properties is overwritten. No input data required.

RAYCOV - covalent radii modification

See input block 1, page 53

ROTREF - Rotation of eigenvectors and density matrix

This option permits the rotation of the cartesian reference frame before the calculation of the
properties.
It is useful, for example, in the population analysis or in the AO projected density of states of
systems containing transition metal atoms with partial d occupation.
Consider for example a d7 occupation as in CoF2 , where the main axis of the (slightly distorted)
CoF6 octahedron in the rutile structure makes a 450 angle with the x axis, and lies in the xy
plane, so that the three empty β states are a combination of the 5 d orbitals. Re-orienting the
octahedron permits to assign integer β occupations to dxz and dyz .

187
Input for the rotation as for EIGSHROT (page 82)

SETINF - Setting of INF values

See input block 1, page 54

SETPRINT - Setting of printing options

See input block 1, page 54.

STOP
Execution stops immediately. Subsequent input records are not processed.

SYMADAPT
See input block 3, page 101

XFAC - X-ray structure factors

rec variable value meaning
• ∗ ISS >0 number of reflections whose theoretical structure factors are calculated.
<0 a set of non-equivalent reflections with indices h,k,l < |ISS| is gener-
ated
if ISS > 0 insert ISS records II
• ∗ H,K,L Miller indices of the reflection (conventional cell) .

The Fourier transform of the ground state charge density of a crystalline system provides the
static structure factors of the crystal, which can be determined experimentally, after taking
into account a number of corrective terms, in particular those related to thermal and zero point
motion of nuclei:
Z
Fk = ρ(r) eik·r dr

where k ≡ h b1 + k b2 + l b3 . The Miller indices refer to the conventional cell. The structure
factors are integrated over the primitive cell volume.

188
8.3 Spontaneous polarization and piezoelectricity
Y. Noel, September 2002 - not fully updated to CRYSTAL09

PIEZOBP - Piezoelectricity (Berry phase approach)

The calculation the piezoelectric constants of a system, can be decompose in few steps. A
preliminary run must be performed for the undistorted system (λ = 0) with the keyword
POLARI. Then, for a first distorted system (λ = 1), a second preliminary run (with the
keyword POLARI) must be performed, followed by third run with the keyword PIEZOBP
that calculates a approximated value of the piezoelectric constants. The evaluation of the slope
dϕα
djk is computed with a single point. For more accuracy, other runs must be done for other
distortions (one run with the keyword POLARI of the new system, followed by a second run
with the keyword PIEZOBP with the undistorted and the new distorted systems). Then the
mean value of the obtained piezoelectric constants must be performed.

1. First run: preliminary calculation related to system λ = 0 (undistorted)

Program Keyword comments
crystal see deck 1 for input blocks 1 and 1b
properties NEWK additional keywords allowed
POLARI see above
save Fortran unit 27 as undistord.f27

2. Second run: preliminary calculation related to system λ = 1 (distorted)

Program Keyword comments
crystal see deck 1 for input blocks 1 and 1b
properties NEWK same input as in first run
POLARI
save Fortran unit 27 as distord1.f27

3. Third run: merging of previous data.

copy undistord.f27 to Fortran unit 28
copy distord1.f27 to fortran unit 29
Program Keyword comments
properties PIEZOBP

4. Refine the computed value

Repeat 2. and 3. for several distortions.
Then compute the mean value of the piezoelectric constants obtained in each case.

PIEZOWF - Piezoelectricity (localized CO approach)

The calculation the piezoelectric constants of a system, can be decompose in few steps. A
preliminary run must be performed for the undistorted system (λ = 0) with the keyword
LOCALWF. Then, for a first distorted system (λ = 1), a second preliminary run (with the
keyword LOCALWF) must be performed, followed by third run with the keyword PIEZOWF
that calculates a approximated value of the piezoelectric constants. The evaluation of the slope
dϕα
djk is computed with a single point. For more accuracy, other runs must be done for other
distortions (one run with the keyword LOCALWF of the new system, followed by a second
run with the keyword PIEZOWF with the undistorted and the new distorted systems). Then
the mean value of the obtained piezoelectric constants must be performed.

189
 1. First run: preliminary calculation related to system λ = 0 (undistorted)
Program Keyword comments
crystal see deck 1 for input blocks 1 and 1b
properties NEWK additional keywords allowed
LOCALWF see above
save Fortran unit 37 as undistord.f37

2. Second run: preliminary calculation related to system λ = 1 (distorted)

Program Keyword comments
crystal see deck 1 for input blocks 1 and 1b
properties NEWK same input as in first run
LOCALWF
save Fortran unit 37 as distord1.f37

3. Third run: merging of previous data.

copy undistord.f37 to Fortran unit 38
copy distord1.f37 to Fortran unit 39
Program Keyword comments
properties PIEZOWF

4. Refine the computed value

Repeat 2. and 3. for several distortions.
Then compute the mean value of the piezoelectric constants obtained in each case.

POLARI - Spontaneous polarization (steps 1 and 2)

PHASE - Spontaneous polarization (step 3)
The ferroelectric phases of a ferroelectric material exhibit two possible enantiomorphic non
centrosymmetric structures, which can be labelled by the geometric parameters λ=+1 and
λ=−1. An external electric field can force the system to change from one structure to the
other, passing through a small energy maximum. The centrosymmetric unstable structure
which sits in the middle of the λ=+1 and λ=−1 structures can be labelled by the geometric
parameters λ=0.
The spontaneous polarization in ferroelectric materials is then evaluated through the Berry
phase approach [107, 30] as the polarization difference between one of the two enantiomorphic
structures (λ=+1 or λ=−1) and the intermediate geometric structure (λ=0).
The fortran unit 70 is defined as direct access:

OPEN(UNIT=IO70,ACCESS=’DIRECT’,RECL=LREC)

Three subsequent runs are required.

1. First run: preliminary calculation related to λ=0 structure

Program Keyword comments

crystal see deck
properties NEWK see page 179
POLARI no input data required
move fortran unit 27 to zero.f27

2. Second run: preliminary calculation related to λ=+1 (or λ=−1) structure

190
Program Keyword comments
crystal see deck 2
properties NEWK same input as in first run
POLARI no input data required
move fortran unit 27 to one.f27

3. Third run: merging of previous data

fortran unit 9 of previous calculation must be present

copy zero.f27 to fortran unit 28
copy one.f27 to fortran unit 29
Program Keyword comments
properties PHASE no input data required
END terminate processing of polari keywords

Deck 1
Potassium niobate - KNbO3
CRYSTAL 3D system
000 IFLAG IFHR IFSO
123 space group, P4/mmm
3.997 4.063 lattice parameters
4 4 non equivalent atoms (5 atoms in the primitive cell)
19 0.5 0.5 0.5 Z=19, Potassium; x, y, z (multiplicity 1)
8 0.0 0.0 0.5 Z=8, Oxygen I; x, y, z (multiplicity 1)
8 0.5 0.0 0.0 Z=8, Oxygen II; x, y, z (multiplicity 2)
41 0.0 0.0 0.0 Z=41, Niobium; x, y, z (multiplicity 1)
END end of geometry input records

Deck 2
Potassium niobate - KNbO3
CRYSTAL 3D system
000 IFLAG IFHR IFSO
123 space group, P4/mmm
3.997 4.063 lattice parameters
4 4 non equivalent atoms (5 atoms in the primitive cell)
19 0.5 0.5 0.5 Z=19, Potassium; x, y, z (multiplicity 1)
8 0.0 0.0 0.5 Z=8, Oxygen I; x, y, z (multiplicity 1)
8 0.5 0.0 0.0 Z=8, Oxygen II; x, y, z (multiplicity 2)
41 0.0 0.0 0.0 Z=41, Niobium; x, y, z (multiplicity 1)
FRACTION fractional coordinates
ATOMDISP displacement of atoms
4 four atoms to be displaced
1 0.0 0.0 -0.023 displacement of atom no. 1 (Potassium)
2 0.0 0.0 -0.042 displacement of atom no. 2 (Oxygen II)
3 0.0 0.0 -0.042 displacement of atom no. 3 (Oxygen II)
4 0.0 0.0 -0.040 displacement of atom no. 4 (Oxygen I)
END end of geometry input records

1. The unit-cell has to contain an even number of electrons.

2. Cell parameters have to be the same for whatever value of the geometric parameter
λ. The difference between the λ=+1, λ=0, and λ=−1 structures is only in the atomic
positions.

3. Numerical accuracy and computational parameters in input block 3 (such as TOLIN-

TEG, POLEORDR, etc.) should be the same for the first and the second run.

4. See page 179 for the NEWK input, which has to be the same for the first and the second
run. The shrinking factor IS should be at least equal to 4. Fermi energy calculation is
not necessary, then set IFE=0.

191
 5. Data evaluated with the keyword POLARI in the first two runs do not have any physical
meaning if considered independently. Only the output produced choosing the keyword
PHASE in the third run is significant.

6. When the λ=−1 geometric structure is chosen in the second run, the spontaneous polar-
ization vector obtained at the end will have the same modulus and direction but opposite
versus with respect to the vector obtained by choosing the λ=+1 structure.

7. The spontaneous polarization is obtained through the Berry phase approach. Since a
phase is defined only in the interval −π to +π, each component of the spontaneous
polarization vector is defined to within an integer number (positive or negative) of the
correspondent component of the ”quantum of polarization” vector, which is automatically
shown in the output of the third run.
Usually there is not need to take into account the quantum of polarization vector, unless
the ferroelectric material shows a large value of the spontaneous polarization.
In case of doubt whether the quantum of polarization vector has to be considered or
not, it is possible to evaluate the spontaneous polarization by setting in the second run a
geometric structure corresponding to an intermediate geometric parameter, e.g. λ=0.25,
and then to extrapolate linearly the result to the λ=1 structure.

SPOLBP - Spontaneous polarization (Berry phase approach)

To calculate the spontaneous polarization, a preliminary with the keyword POLARI run is needed
for each of the two systems λ = 1 and λ = 0. Then a third run with the keyword SPOLBP
gives the difference of polarization between both systems.

1. First run: preliminary calculation related to system λ = 0

Program Keyword comments
crystal see deck 1 for input blocks 1 and 1b
properties NEWK additional keywords allowed
POLARI see above
save Fortran unit 27 as sys0.f27

2. Second run: preliminary calculation related to system λ = 1

Program Keyword comments
crystal see deck 1 for input blocks 1 and 1b
properties NEWK same input as in first run
POLARI
save Fortran unit 27 as sys1.f27

3. Third run: merging of previous data.

copy sys0.f27 to Fortran unit 28
copy sys1.f27 to Fortran unit 29
Program Keyword comments
properties SPOLBP

192
SPOLWF - The spontaneous polarization (localized CO approach)
To calculate the spontaneous polarization, two preliminary runs with the keyword LOCALI
is needed for each of the two systems λ = 1 and λ = 0. Then a third run with the keyword
SPOLBP computes the difference of polarization between both systems.

1. First run: preliminary calculation related to system λ = 0

Program Keyword comments
crystal see deck 1 for input blocks 1 and 1b
properties NEWK additional keywords allowed
LOCALI see above
save Fortran unit 37 as sys0.f23

2. Second run: preliminary calculation related to system λ = 1

Program Keyword comments
crystal see deck 1 for input blocks 1 and 1b
properties NEWK same input as in first run
LOCALI
save Fortran unit 37 as sys1.f37

3. Third run: merging of previous data.

copy sys0.f37 to Fortran unit 38
copy sys1.f37 to Fortran unit 39
Program Keyword comments
properties SPOLWF

193
Chapter 9

Input examples

9.1 Standard geometry input

3D - Crystalline compounds - 1st input record keyword:
CRYSTAL
Atom coordinates: fractional units of the crystallographic lattice vectors.
Sodium Chloride - NaCl (Rock Salt Structure)
CRYSTAL
000 IFLAG IFHR IFSO
225 space group, Fm3m, cubic
5.64 a (Å)
2 2 non equivalent atoms
11 .5 .5 .5 Z=11, Sodium, 1/2, 1/2, 1/2
17 .0 .0 .0 Z=17, Chlorine

Diamond - C (2nd Setting - 48 symmops - 36 with translational component)

CRYSTAL
000 IFLAG IFHR IFSO
227 space group, Fd3m, cubic
3.57 a (Å)
1 1 non equivalent atom
6 .125 .125 .125 Z=6, Carbon, 1/8, 1/8, 1/8 (multiplicity 2)

Diamond - C (1st Setting - 48 symmops - 24 with translational component)

CRYSTAL
001 IFLAG IFHR IFSO
227 space group 227, Fd3m, cubic
3.57 a (Å)
1 1 non equivalent atom
6 .0 .0 .0 Z=6, Carbon (multiplicity 2)

Zinc Blend - ZnS

CRYSTAL
000 IFLAG IFHR IFSO
216 space group 216, F4̄3m, cubic
5.42 a (Å)
2 2 non equivalent atoms
30 .25 .25 .25 Z=30, Zinc, (1/4, 1/4, 1/4)
16 .0 .0 .0 Z=16, Sulphur

Wurtzite - ZnS
CRYSTAL
000 IFLAG IFHR IFSO
186 space group 186, P63 mc, hexagonal
3.81 6.23 a,c (Å)
2 2 non equivalent atoms
30 .6666666667 .3333333333 .0 Zinc, (2/3, 1/3, 0.)
16 .6666666667 .3333333333 .375 Sulphur, (2/3, 1/3, 3/8)

194
Cuprite - Cu2 O
CRYSTAL
000 IFLAG IFHR IFSO
208 space group 208, P42 32, cubic
4.27 a (Å)
2 2 non equivalent atoms
8 .0 .0 .0 Z=8, Oxygen
29 .25 .25 .25 Z=29, Copper (1/4, 1/4, 1/4)

Aragonite - CaCO3
CRYSTAL
100 IFLAG (1, SPGR symbol) IFHR IFSO
PMCN space group Pmcn, orthorhombic
4.9616 7.9705 5.7394 a,b,c (Å)
4 4 non equivalent atoms
20 .25 .4151 .2103 Z=20, Calcium
6 .25 .7627 .085 Z=6, Carbon
8 .25 .9231 .0952 Z=8, Oxygen
8 .4729 .6801 .087 Z=8, Oxygen

Fluorite - CaF2
CRYSTAL
000 IFLAG IFHR IFSO
225 space group 225, Fm3m, cubic
5.46 a (Å)
2 2 non equivalent atoms
9 .25 .25 .25 Fluorine
20 .0 .0 .0 Calcium

Cesium chloride - CsCl

CRYSTAL
000 IFLAG IFHR IFSO
221 space group 221, Pm3m, cubic
4.12 a (Å)
2 2 non equivalent atoms
55 .5 .5 .5 Cesium
17 .0 .0 .0 Chlorine

Rutile - TiO2
CRYSTAL
000 IFLAG IFHR IFSO
136 space group 136, P42 /mnm, tetragonal
4.59 2.96 a, c (Å)
2 2 non equivalent atoms
22 .0 .0 .0 Titanium
8 .305 .305 .0 Oxygen

Graphite - C (Hexagonal)
CRYSTAL
000 IFLAG IFHR IFSO
194 space group 194, P63 /mmc, hexagonal
2.46 6.70 a,c (Å)
2 2 non equivalent atoms
6 .0 .0 .25 Carbon, 0, 0, 1/4
6 .33333333333 .66666666667 .25 Carbon, 1/3, 2/3, 1/4

Pyrite - FeS2
CRYSTAL
000 IFLAG IFHR IFSO
205 space group 205, Pa3, cubic
5.40 a (Å)
2 2 non equivalent atoms
26 .0 .0 .0 Iron
16 .386 .386 .386 Sulphur

195
Calcite - CaCO3
CRYSTAL
010 IFLAG IFHR (=1, rhombohedral representation) IFSO
167 space group 167, R3̄c, hexagonal
6.36 46.833 a (Å), α◦
3 3 non equivalent atoms
20 .0 .0 .0 Calcium
6 .25 .25 .25 Carbon
8 .007 .493 .25 Oxygen

Corundum - Al2 O3 (hexagonal representation)

CRYSTAL
000 IFLAG IFHR IFSO
167 space group 167, R3̄c, hexagonal
4.7602 12.9933 a,c (Å)
2 2 non equivalent atoms
13 0. 0. 0.35216 Aluminium
8 0.30621 0. 0.25 Oxygen

Corundum - Al2 O3 (rhombohedral representation)

CRYSTAL
010 IFLAG IFHR (=1, rhombohedral cell) IFSO
167 space group 167, R3̄c, hexagonal
5.12948 55.29155 a (Å), α◦
2 2 non equivalent atoms
13 0.35216 0.35216 0.35216 Aluminium
8 0.94376 0.25 0.55621 Oxygen

Zirconia - ZrO2 - monoclinic structure

CRYSTAL
001 IFLAG IFHR IFSO (=1, standard shift of origin)
14 space group 14, P21 /c, monoclinic
5.03177 5.03177 5.258 90.0 a,b,c (Å), β ◦
3 3 non equivalent atoms
240 0.2500 0.0000 0.25000 Zirconium, Pseudopotential (Z’ > 200)
208 0.0000 0.2500 0.07600 Oxygen, Pseudopotential
208 -0.500 -0.250 0.07600 Oxygen, Pseudopotential

Zirconia - ZrO2 - tetragonal structure

CRYSTAL
001 IFLAG IFHR IFSO (=1, standard shift of origin)
137 space group 137, P42 /nmc, tetragonal
3.558 5.258 a,c (Å)
3 3 non equivalent atoms
240 0.0 0.0 0.0 Zirconium, Pseudopotential (Z’ > 200)
208 0.0 -0.5 0.174 Oxygen, Pseudopotential
208 0.5 0.0 0.326 Oxygen, Pseudopotential

Zirconia - ZrO2 - cubic structure

CRYSTAL
001 IFLAG IFHR IFSO (=1, standard shift of origin)
225 space group 225, Fm3m, cubic
5.10 a (Å)
3 3 non equivalent atoms
240 0.00 0.00 0.00 Z=40 Zirconium, Pseudopotential (Z’ > 200)
208 0.25 0.25 0.25 Oxygen, Pseudopotential
208 -0.25 -0.25 -0.25 Oxygen, Pseudopotential

196
SiO2 , Chabazite
CRYSTAL
010 IFLAG IFHR (=1,rhombohedral representation) IFSO
166 space group 166 R3̄m, hexagonal
9.42 94.47 a (Å), β ◦
5 5 non equivalent atoms (36 atoms in the primitive cell)
14 .1045 .334 .8755 Silicon (multiplicity 12)
8 .262 -.262 .0 Oxygen (multiplicity 6)
8 .1580 -.1580 .5000 Oxygen (multiplicity 6)
8 .2520 .2520 .8970 Oxygen (multiplicity 6)
8 .0250 .0250 .3210 Oxygen (multiplicity 6)

SiO2 , Siliceous Faujasite

CRYSTAL
000 IFLAG IFHR IFSO
227 space group 227, Fd3m, cubic
21.53 a (Å)
5 5 non equivalent atoms (144 atoms in the primitive cell)
14 .1265 -.0536 .0370 Silicon (multiplicity 48)
8 .1059 -.1059 .0 Oxygen (multiplicity 24)
8 -.0023 -.0023 .1410 Oxygen (multiplicity 24)
8 .1746 .1746 -.0378 Oxygen (multiplicity 24)
8 .1785 .1785 .3222 Oxygen (multiplicity 24)

SiO2 , Siliceous Edingtonite

CRYSTAL
000 IFLAG IFHR IFSO
115 space group 115, P4̄m2, tetragonal
6.955 6.474 a, c (Å)
5 5 non equivalent atoms (15 atoms in the primitive cell)
14 .0 .0 .5000 Silicon (multiplicity 1)
14 .0 .2697 .1200 Silicon (multiplicity 4)
8 .0 .189 .3543 Oxygen (multiplicity 4)
8 .50000 .0 .8779 Oxygen (multiplicity 2)
8 .189 .189 .0 Oxygen (multiplicity 4)

SiO2 , Siliceous Sodalite

CRYSTAL
000 IFLAG IFHR IFSO
218 space group 218, P4̄3n, cubic
8.950675 a (Å)
3 3 non equivalent atoms (36 atoms in the primitive cell)
14 .25000 .50000 .0 Silicon (multiplicity 6)
14 .25000 .0 .50000 Silicon (multiplicity 6)
8 .14687 .14687 .50000 Oxygen (multiplicity 24)

197
2D - Slabs (surfaces) - 1st input record keyword: SLAB
A 2D structure can either be created by entering directly the 2D cell parameters and
irreducible atoms coordinates to obtain a slab of given thickness (keyword SLAB in the first
record of the geometry input), or it can be derived from the 3D structure through the keyword
SLABCUT (page 55), entered in the geometry editing section of 3D structure input. In that
case the layer group is automatically identified by the program. The input tests 4-24, 5-25,
6-26 and 7-27 show the two different ways to obtain the same 2D structure.
Atom coordinates: z in Ångstrom, x, y in fractional units of the crystallographic cell translation
vectors.

Test05 - graphite 2D (see test 25)

SLAB
77 layer group (hexagonal)
2.47 lattice vector length (Å)
1 1 non equivalent atom
6 -0.33333333333 0.33333333333 0. Z=6; Carbon; x,y,z

Beryllium - 3 layers slab

SLAB
78 layer group (hexagonal)
2.29 lattice vector length (Å)
2 2 non equivalent atoms
4 0.333333333333 0.666666666667 0. Z=4, Beryllium; 1/3, 2/3, z
4 0.666666666667 0.333333333333 1.795 Z=4, Beryllium; 2/3, 1/3,z

Test06 - beryllium - 4 layers slab (see test 26)

SLAB
72 layer group (hexagonal)
2.29 lattice vector length (Å)
2 2 non equivalent atoms
4 0.333333333333 0.666666666667 0.897499 Z=4, Beryllium;x,y,z
4 0.666666666667 0.333333333333 2.692499 Z=4, Beryllium;x,y,z

Test04 - Corundum 001 (0001) 2 layers slab (see test 24)

SLAB
66 layer group (hexagonal)
4.7602 lattice vector length (Å)
3 3 non equivalent atoms
13 0. 0. 1.9209 Z=13, Aluminum; x,y,z
8 0.333333333 -0.027093 1.0828 Z=8, Oxygen; x,y,z
13 -0.333333333 0.333333333 0.2446 Z=13, Aluminum; x,y,z

Test07 - Corundum 110 (1010) slab (see test 27)

SLAB
7 layer group (Oblique)
5.129482 6.997933 95.8395 a,b (Å) α (degrees)
6 6 non equivalent atoms
8 -0.25 0.5 2.1124 Z=8, Oxygen; x,y,z
8 0.403120 0.153120 1.9189 Z=8, Oxygen; x,y,z
8 0.096880 0.346880 0.4612 Z=8, Oxygen; x,y,z
8 -0.25 0.00 0.2677 Z=8, Oxygen; x,y,z
13 0.454320 0.397840 1.19 Z=13, Aluminum; x,y,z
13 0.045680 0.102160 1.19 Z=13, Aluminum; x,y,z

MgO (110) 2 layers slab

SLAB
40 layer group
4.21 2.97692 lattice vectors length (Å)
2 2 non equivalent atoms
12 0.25 0.25 0.74423 Z=12, Magnesium; x,y,z
8 0.75 0.25 0.74423 Z=8, Oxygen; x,y,z

198
MgO (110) 3 layers slab
SLAB
37
4.21 2.97692 lattice vectors length (Å)
4 4 non equivalent atoms
12 0. 0. 1.48846 Z=12, Magnesium; x,y,z
8 0.5 0. 1.48846 Z=8, Oxygen; x,y,z
12 0.5 0.5 0. Z=12, Magnesium; x,y,z
8 0. 0.5 0. Z=8, Oxygen; x,y,z

CO on MgO (001) two layers slab - one-side adsorption

SLAB
55 √
2.97692 lattice vector length [4.21/ 2] (Å)
6 6 non equivalent atoms
108 0. 0. 4.5625 Z=8, Oxygen; x,y,z
6 0. 0. 3.4125 Z=6, Carbon; x,y,z
12 0. 0. 1.0525 Z=12, Magnesium; x,y,z
8 0.5 0.5 1.0525 Z=8, Oxygen; x,y,z
12 0. 0. -1.0525 Z=12, Magnesium; x,y,z
8 0.5 0.5 -1.0525 Z=8, Oxygen; x,y,z
Two different conventional atomic numbers (8 and 108) are attributed to the Oxygen in CO and to the Oxygen
in MgO. Two different basis sets will be associated to the two type of atoms (see Basis Set input, page 19, and
test 36).

CO on MgO (001) two layers slab - two-side adsorption

SLAB
64
2.97692 lattice vector length (Å)
4 4 non equivalent atoms
108 0.25 0.25 4.5625 Z=8, Oxygen; x,y,z
6 0.25 0.25 3.4125 Z=6, Carbon; x,y,z
12 0.25 0.25 1.0525 Z=12, Magnesium; x,y,z
8 0.75 0.75 1.0525 Z=8, Oxygen; x,y,z
Two different conventional atomic numbers (8 and 108) are attributed to the Oxygen in CO and to the Oxygen
in MgO.

Diamond slab parallel to (100) face - nine layers slab

SLAB
59
2.52437 lattice vector length (Å)
5 5 non equivalent atoms
6 0. 0. 0. Z=6, Carbon; x,y,z
6 0.5 0. 0.8925 Z=6, Carbon; x,y,z
6 0.5 0.5 1.785 Z=6, Carbon; x,y,z
6 0. 0.5 2.6775 Z=6, Carbon; x,y,z
6 0. 0. 3.57 Z=6, Carbon; x,y,z

Diamond slab parallel to (100) face - ten layers slab

SLAB
39 layer group
2.52437 2.52437 lattice vectors length (Å)
5 5 non equivalent atoms
6 0.25 0. 0.44625 Z=6, Carbon; x,y,z
6 0.25 0.5 1.33875 Z=6, Carbon; x,y,z
6 0.75 0.5 2.23125 Z=6, Carbon; x,y,z
6 0.75 0 3.12375 Z=6, Carbon; x,y,z
6 0.25 0. 4.01625 Z=6, Carbon; x,y,z

199
1D - Polymers - 1st input record keyword: POLYMER
Atom coordinates: y,z in Ångstrom, x in fractional units of the crystallographic cell translation
vector.

Test03 - (SN)x polymer

POLYMER
4 rod group
4.431 lattice vector length (Å)
2 2 non equivalent atoms
16 0.0 -0.844969 0.0 Z=16, Sulphur; x, y, z
7 0.141600540 0.667077 -0.00093 Z=7, Nitrogen; x, y, z

Water polymer
POLYMER
1
4.965635 lattice vector length (Å)
6 6 non equivalent atoms
8 0. 0. 0. Z=8, Oxygen; x, y, z
1 0.032558 0.836088 -0.400375 Z=1, Hydrogen; x, y, z
1 0.168195 -0.461051 0. Z=1, Hydrogen; x, y, z
8 0.5 -1.370589 0. Z=8, Oxygen; x, y, z
1 0.532558 -2.206677 0.400375 Z=1, Hydrogen; x, y, z
1 0.668195 -0.909537 0. Z=1, Hydrogen; x, y, z

Formamide chain - test40 DFT

POLYMER
4 rod group
8.774 lattice vector length (Å)
6 6 non equivalent atoms
8 -7.548E-2 5.302E-3 0.7665 Z=8, Oxygen; x, y, z
7 0.1590 -0.8838 0.3073 Z=7, Nitrogen; x, y, z
6 5.627E-2 7.051E-2 0.2558 Z=6, Oxygen; x, y, z
1 0.2677 -0.6952 -9.1548E-2 Z=1, Hydrogen; x, y, z
1 0.1310 -1.8019 0.7544 Z=1, Hydrogen; x, y, z
1 9.244E-2 0.9973 -0.2795 Z=1, Hydrogen; x, y, z

200
0D - Molecules - 1st input record keyword: MOLECULE
Atom coordinates: x,y,z in Ångstrom.
Test00 - CO molecule
MOLECULE
1 point group
2 2 non equivalent atoms
6 0. 0. 0. Z=6, Carbon; x, y, z
8 0.8 0.5 0.4 Z=8, Oxygen; x, y, z

Test01 - CH4 Methane molecule

MOLECULE
44 point group
2 2 non equivalent atoms
6 0. 0. 0. Z=6, Carbon; x, y, z
1 0.629 0.629 0.629 Z=1, Hydrogen; x, y, z

Test02 - CO(NH2 )2 Urea molecule

MOLECULE
15 point group
5 5 non equivalent atoms
6 0. 0. 0. Z=6, Carbon; x, y, z
8 0. 0. 1.261401 Z=8, Oxygen; x, y, z
7 0. 1.14824666034 -0.69979 Z=7, Nitrogen; x, y, z
1 0. 2.0265496501 -0.202817 Z=1, Hydrogen; x, y, z
1 0. 1.13408048308 -1.704975 Z=1, Hydrogen; x, y, z

9.2 Basis set input

Optimized basis sets for periodic systems used in published papers are available on WWW:

http://www.crystal.unito.it

All electron Basis sets for Silicon atom

STO-3G
14 3 Z=14, Silicon; 3 shells
1 0 3 2. 0. Pople BS; s shell; 3G; CHE=2; standard scale factor
1 1 3 8. 0. Pople BS; sp shell; 3G; CHE=8; standard scale factor
1 1 3 4. 0. Pople BS; sp shell; 3G; CHE=4; standard scale factor

6-21G
14 4 Z=14, Silicon; 4 shells
2 0 6 2. 1. Pople 6-21 BS; s shell; 6G; CHE=2; scale factor 1 (core AO).
2 1 6 8. 1. Pople 6-21 BS; sp shell; 6G; CHE=8; scale factor 1 (core AOs).
2 1 2 4. 1. Pople 6-21 BS; sp shell; 2G; CHE=4; scale factor 1 (inner valence).
2 1 1 0. 1. Pople 6-21 BS; sp shell; 1G; CHE=0; scale factor 1 (outer valence).

NB. The 4th shell has electron charge 0. The basis functions of that shell are included in the basis set to
compute the atomic wave functions, as they correspond to symmetries (angular quantum numbers) occupied
in the ground state of the atom. The atomic basis set is: 4s, 3p.

6-21G modified
14 4 Z=14, Silicon; 4 shells
2 0 6 2. 1. Pople 6-21 BS; s shell; 6G; CHE=2; scale factor 1.
2 1 6 8. 1. Pople 6-21 BS; sp shell; 6G; CHE=8; scale factor 1.
2 1 2 4. 1. Pople 6-21 BS; sp shell; 2G; CHE=4; scale factor 1.
0 1 1 0. 1. free BS; sp shell; 1G; CHE=0; scale factor 1.
0.16 1. 1. gaussian exponent; s coefficient; p coefficient

201
3-21G
14 4 Z=14, Silicon; 4 shells
2 0 3 2. 1. Pople 3-21 BS; s shell; 3G; CHE=2; scale factor 1.
2 1 3 8. 1. Pople 3-21 BS; sp shell; 3G; CHE=8; scale factor 1.
2 1 2 4. 1. Pople 3-21 BS; sp shell; 2G; CHE=4; scale factor 1.
2 1 1 0. 1. Pople 3-21 BS; sp shell; 1G; CHE=0; scale factor 1.

3-21G*
14 5 Z=14, Silicon; 5 shells
2 0 3 2. 1. Pople 3-21 BS; s shell; 3G; CHE=2; scale factor 1.
2 1 3 8. 1. Pople 3-21 BS; sp shell; 3G; CHE=8; scale factor 1.
2 1 2 4. 1. Pople 3-21 BS; sp shell; 2G; CHE=4; scale factor 1.
2 1 1 0. 1. Pople 3-21 BS; sp shell; 1G; CHE=0; scale factor 1.
2 3 1 0. 1. Pople 3-21 BS; d shell; 1G; CHE=0; scale factor 1.

NB. The basis functions of the 5th shell, d symmetry, unoccupied in the ground state of Silicon atom, is not
included in the atomic wave function calculation.

3-21G modified+polarization
14 5 Z=14, Silicon; 5 shells
2 0 3 2. 1. Pople 3-21 BS; s shell; 3G; CHE=2; scale factor 1.
2 1 3 8. 1. Pople 3-21 BS; sp shell; 3G; CHE=8; scale factor 1.
2 1 2 4. 1. Pople 3-21 BS; sp shell; 2G; CHE=4; scale factor 1.
0 1 1 0. 1. free BS; sp shell; 1G; CHE=0; scale factor 1.
0.16 1. 1. gaussian exponent; s contraction coefficient; p contr. coeff.
0 3 1 0. 1. free BS; d shell; 1G; CHE=0; scale factor 1.
0.5 1. gaussian exponent; d contraction coefficient.

free basis set

14 4 Z=14, Silicon; 4 shells
0 0 6 2. 1. free BS; s shell; 6 GTF; CHE=2; scale factor 1.
16115.9 0.00195948 1st gaussian exponent; s contraction coefficient
2425.58 0.0149288 2nd gaussian exponent; s contraction coefficient
553.867 0.0728478 3rd gaussian exponent; s contraction coefficient
156.340 0.24613 4th gaussian exponent; s contraction coefficient
50.0683 0.485914 5th gaussian exponent; s contraction coefficient
17.0178 0.325002 6th gaussian exponent; s contraction coefficient
0 1 6 8. 1. free BS; sp shell; 6 GTF; CHE=8; scale factor 1.
292.718 -0.00278094 0.00443826 1st gaussian exp.; s contr. coeff.; p contr. coeff.
69.8731 -0.0357146 0.0326679 2nd gaussian exp.; s contr. coeff.; p contr. coeff.
22.3363 -0.114985 0.134721 3rd gaussian exp.; s contr. coeff.; p contr. coeff.
8.15039 0.0935634 0.328678 4th gaussian exp.; s contr. coeff.; p contr. coeff.
3.13458 0.603017 0.449640 5th gaussian exp.; s contr. coeff.; p contr. coeff.
1.22543 0.418959 0.261372 6th gaussian exp.; s contr. coeff.; p contr. coeff.
0 1 2 4. 1. free BS; sp shell; 2 GTF; CHE=4; scale factor 1
1.07913 -0.376108 0.0671030 1st gaussian exp.; s contr. coeff.; p contr. coeff.
0.302422 1.25165 0.956883 2nd gaussian exp.; s contr. coeff.; p contr. coeff.
0 1 1 0. 1. free BS; sp shell; 1 GTF; CHE=0; scale factor 1.
0.123 1. 1. gaussian exp.; s contr. coeff.; p contr. coeff.

Examples of ECP and valence only basis set input

Nickel atom. Electronic configuration: [Ar] 4s(2) 3d(8)

202
Durand & Barthelat large core
228 4 Z=28,Nickel; 4 shells valence basis set
BARTHE keyword; Durand-Barthelat ECP
0 1 2 2. 1. free BS;sp shell;2 GTF;CHE=2;scale factor 1
1.55 .24985 1. 1st GTF exponent;s coefficient;p coefficient
1.24 -.41636 1. 2nd GTF exponent;s coefficient;p coefficient
0 1 1 0. 1. free BS; sp shell; 1 GTF; CHE=0; scale factor 1
0.0818 1.0 1. GTF exponent;s coefficient;p coefficient
0 3 4 8. 1. free BS; d shell; 4 GTF; CHE=8; scale factor 1
4.3842E+01 .03337 1st GTF exponent; d coefficient
1.2069E+01 .17443 2nd GTF exponent; d coefficient
3.9173E+00 .42273 3rd GTF exponent; d coefficient
1.1997E+00 .48809 4th GTF exponent; d coefficient
0 3 1 0. 1. free BS; d shell; 1 GTF; CHE=0; scale factor 1
0.333 1. GTF exponent; d coefficient

Hay & Wadt Large Core - [Ar] 4s(2) 3d(8)

228 4 Z=28,Nickel; 4 shells valence basis set
HAYWLC keyword; Hay-Wadt large core ECP
0 1 2 2. 1. free BS; sp shell; 2 GTF; CHE=2; scale factor 1
1.257 1.1300E-01 2.6760E-02 exponent,s coefficient,p coefficient
1.052 -1.7420E-01 -1.9610E-02
0 1 1 0. 1. second shell,sp type,1 GTF
0.0790 1.0 1.
0 3 4 8. 1. third shell,d type,4 primitive GTF
4.3580E+01 .03204
1.1997E+01 .17577
3.8938E+00 .41461
1.271 .46122
0 3 1 0. 1. fourth shell,d type,1 GTF
0.385 1.

Hay & Wadt Small Core - [Ne] 3s(2) 3p(6) 4s(2) 3d(8)
228 6 nickel basis set - 6 shells
HAYWSC keyword; Hay-Wadt small core ECP
0 1 3 8. 1. first shell,sp type,3 primitive GTF -
2.5240E+01 -3.7000E-03 -4.0440E-02 exponent,s coefficient,p coefficient
7.2019E+00 -5.3681E-01 -7.6560E-02
3.7803E+00 4.2965E-01 4.8348E-01
0 1 2 2. 1. second shell,sp type,2 primitive GTF
1.40 .84111 .55922
0.504 .13936 .12528
0 1 1 0. 1. third shell,sp type,1 GTF
0.0803 1.0 1.
0 3 3 8. 1. fourth shell,d type,4 primitive GTF
4.1703E+01 3.5300E-02
1.1481E+01 1.8419E-01
3.7262E+00 4.1696E-01
0 3 1 0. 1. fifth shell,d type,1 GTF
1.212 1.
0 3 1 0. 1. sixth shell,d type,1 GTF
0.365 1.0

203
Free input
228 5 Z=28, nickel basis set - 5 shells (valence only)
INPUT keyword: free ECP (Large Core)- input follows
10. 545200 nuclear charge; number of terms in eq. 2.17 and 2.18
344.84100 -18.00000 -1 eq. 2.17, 5 records:
64.82281 -117.95937 0 α, C, n
14.28477 -29.43970 0
3.82101 -10.38626 0
1.16976 -0.89249 0
18.64238 3.00000 -2 eq. 2.18, 4 records ` = 0
4.89161 19.24490 -1
1.16606 23.93060 0
0.95239 -9.35414 0
30.60070 5.00000 -2 eq. 2.18, 5 records ` = 1
14.30081 19.81155 -1
15.03304 54.33856 0
4.64601 54.08782 0
0.98106 7.31027 0
4.56008 0.26292 0 eq. 2.18, 2 records ` = 2
0.67647 -0.43862 0 basis set input follows - valence only
0 1 1 2. 1. 1st shell: sp type; 1 GTF; CHE=2; scale fact.=1
1.257 1. 1. exponent, s coefficient, p coefficient
0 1 1 0. 1. 2nd shell: sp type; 1 GTF; CHE=0; scale fact.=1
1.052 1. 1.
0 1 1 0. 1. 3rd shell: sp type; 1 GTF; CHE=0; scale fact.=1
0.0790 1.0 1.
0 3 4 8. 1. 4th shell; d type; 4 GTF; CHE=8; scale fact.=1
4.3580E+01 .03204
1.1997E+01 .17577
3.8938E+00 .41461
1.271 .46122
0 3 1 0. 1. 5th shell; d type; 1 GTF; CHE=0; scale fact.=1
0.385 1.

9.3 SCF options

Example of how to edit the density matrix obtained for a given magnetic solution to define a
scf guess with a different magnetic solution.

Deck 1 - ferromagnetic solution

Spinel MnCr2O4
CRYSTAL
000
227 space group number
8.5985 lattice parameter
3 3 non equivalent atoms (14 atoms in the primitive cell)
24 0.500 0.500 0.500 Chromium - x, y, z - multiplicity 4
25 0.125 0.125 0.125 Manganese - x, y, z - multiplicity 2
8 0.2656 0.2656 0.2656 Oxygen - x, y, z - multiplicity 8
END end of geometry input records - block 1
basis set input terminated by END
UHF Unrestricted Hartree Fock
TOLINTEG the default value of the truncation tolerances is modified
7 7 7 7 14 new values for ITOL1-ITOl2-ITOL3-ITOL4-ITOL5
END end of input block 3
404 reciprocal lattice sampling (page 23)
SPINLOCK nα - nβ is locked to be 22 for 50 cycles.
22 50 All the d electrons are forced to be parallel
LEVSHIFT a level shifter of 0.3 hartree, maintained after diagonalization,
31 causes a lock in a non-conducting solution
MAXCYCLE the maximum number of SCF cycles is set to 50
50
PPAN Mulliken population analysis at the end of SCF cycles
END

Deck 2 (SCF input only)

204
404
GUESSP initial guess: density matrix from a previous run
SPINEDIT elements of the density matrix are modified
2 the diagonal elements corresponding to 2 atoms
56 label of the 2 atoms (6 is equivalent to 5)
LEVSHIFT a level shifter of 0.3 hartree, maintained after diagonalization,
31 causes a lock in a non-conducting solution
PPAN Mulliken population analysis at the end of SCF cycles
END

=====================================================================
First run - geometry output
=====================================================================

COORDINATES OF THE EQUIVALENT ATOMS (FRACTIONAL UNITS)

N. ATOM EQUIVALENT AT. NUMBER X Y Z

1 1 1 24 CR -5.000E-01 -5.000E-01 -5.000E-01

2 1 2 24 CR -5.000E-01 -5.000E-01 0.000E+00
3 1 3 24 CR 0.000E+00 -5.000E-01 -5.000E-01
4 1 4 24 CR -5.000E-01 0.000E+00 -5.000E-01

5 2 1 25 MN 1.250E-01 1.250E-01 1.250E-01

6 2 2 25 MN -1.250E-01 -1.250E-01 -1.250E-01

7 3 1 8 O 2.656E-01 2.656E-01 2.656E-01

8 3 2 8 O 2.656E-01 2.656E-01 -2.968E-01
9 3 3 8 O -2.968E-01 2.656E-01 2.656E-01
10 3 4 8 O 2.656E-01 -2.968E-01 2.656E-01
11 3 5 8 O -2.656E-01 -2.656E-01 -2.656E-01
12 3 6 8 O -2.656E-01 -2.656E-01 2.968E-01
13 3 7 8 O -2.656E-01 2.968E-01 -2.656E-01
14 3 8 8 O 2.968E-01 -2.656E-01 -2.656E-01

=====================================================================
Ferromagnetic solution: all unpaired electrons with the same spin
=====================================================================
SPIN POLARIZATION - ALPHA-BETA = 22 FOR 50 CYCLES
=====================================================================
Convergence on total energy reached in 33 cycles (level shifter active)
=====================================================================
CYCLE 33 ETOT(AU) -7.072805900367E+03 DETOT -8.168E-07 DE(K) 9.487E+00

=====================================================================
Population analysis - ferromagnetic solution
=====================================================================
MULLIKEN POPULATION ANALYSIS
ALPHA+BETA ELECTRONS - NO. OF ELECTRONS 210.000000

ATOM Z CHARGE SHELL POPULATION

s sp sp sp sp d d
1 CR 24 21.884 2.000 8.047 2.251 4.487 1.331 3.078 .690
5 MN 25 23.147 2.000 8.081 2.170 4.299 1.489 4.478 .629
7 O 8 9.521 1.996 2.644 2.467 2.414

MULLIKEN POPULATION ANALYSIS

ALPHA-BETA ELECTRONS - NO. OF ELECTRONS 22.000000

ATOM Z CHARGE
SHELL POPULATION
s sp sp sp sp d d
1 CR 24 3.057 .000 -.002 .011 .027 -.011 2.790 .242
5 MN 25 4.925 .000 -.003 .019 .055 -.052 4.408 .498
7 O 8 -.010 .000 .003 -.014 .002
================================================================================
================================================================================
Second run - Anti ferromagnetic solution:
Integrals calculation not affected by the spin state

Cr (atoms 1-2-3-4) unpaired electrons spin alpha;

Mn (atoms 5 and 6) unpaired electrons spin beta
================================================================================

RESTART FROM A PREVIOUS RUN DENSITY MATRIX

SPIN INVERSION IN SPIN DENSITY MATRIX FOR ATOMS: 5 6

=====================================================================
Convergence on total energy reached in 15 cycles

205
=====================================================================
CYCLE 15 ETOT(AU) -7.072808080821E+03 DETOT -4.930E-07 DE(K) 6.694E-06

======================================uuuu============================
Population analysis - anti ferromagnetic solution
=====================================================================
MULLIKEN POPULATION ANALYSIS
ALPHA+BETA ELECTRONS - NO. OF ELECTRONS 210.000000

ATOM Z CHARGE SHELL POPULATION

s sp sp sp sp d d
1 CR 24 21.884 2.000 8.047 2.251 4.487 1.331 3.078 .690
5 MN 25 23.149 2.000 8.081 2.170 4.299 1.489 4.479 .631
7 O 8 9.521 1.997 2.644 2.467 2.414

MULLIKEN POPULATION ANALYSIS

ALPHA-BETA ELECTRONS - NO. OF ELECTRONS 2.000000
ATOM Z CHARGE SHELL POPULATION
s sp sp sp sp d d
1 CR 24 3.049 .000 -.002 .011 .027 -.012 2.785 .240
5 MN 25 -4.917 .000 .003 -.018 -.055 .054 -4.406 -.495
7 O 8 -.045 .000 -.024 -.013 -.008
================================================================================

9.4 Geometry optimization

Crystal geometry input section (block1) for the geometry optimization of the urea molecule:

. Example

Urea Molecule Title

MOLECULE Dimension of the system
15 Point group (C2v )
5 Number of non equivalent atoms
6 0. 0. 0. Atomic number and cartesian coordinates
8 0. 0. 1.261401
7 0. 1.148247 -0.699790
1 0. 2.026550 -0.202817
1 0. 1.134080 -1.704975
OPTGEOM Keyword to perform a geometry optimization
ENDOPT End of geometry optimization input block
END end og geometry input
Basis set input As in test 12
END End of basis set input section
END block3 input - Molecule - no information on sampling in K space

Crystal output contains additional information on the optimization run after the initial part
of the geometry output:

. . . . . . . . .. . . . . . . . . . . . . . . . . .

BERNY OPTIMIZATION CONTROL

MAXIMUM GRADIENT COMPONENT 0.00045 MAXIMUM DISPLACEMENT COMPONENT 0.00030

R.M.S. OF GRADIENT COMPONENT 0.00180 R.M.S. OF DISPLACEMENT COMPONENTS 0.00120
THRESHOLD ON ENERGY CHANGE 0.100E-06 EXTRAPOLATING POLYNOMIAL ORDER 2
MAXIMUM ALLOWED NUMBER OF STEPS 100 SORTING OF ENERGY POINTS: NO
ANALYTICAL GRADIENTS
. . . . . . . . .. . . . . . . . . . . . . . . . . .

SYMMETRY ALLOWED INTERNAL DEGREE(S) OF FREEDOM: 7

. . . . . . . . .. . . . . . . . . . . . . . . . . .

206
At the first step of the optimization, the Crystal standard output contains both energy
(complete SCF cycle) and gradient parts. At the end of the first step, a convergence check is
performed on the initial forces and the optimization stops if the criteria are already satisfied.
For the subsequent steps, only few lines on the optimization process are reported in standard
output, namely: current geometry, total energy and gradients, and convergence tests (SCF
output is routed to file SCFOUT.LOG).
At each optimization step, xxx, the geometry is written in file optcxxx (in a format suitable
to be read with EXTERNAL keyword). Optimization step can be restarted from any step
geometry, by renaming optcxxx as fort.34.
The standard output for the urea molecule geometry optimization looks as follows:

OPTOPTOPTOPTOPTOPTOPTOPTOPTOPTOPTOPTOPTOPTOPTOPTOPTOPTOPTOPTOPTOPTOPTOPTOPTOPT
*******************************************************************************
GEOMETRY OPTIMIZATION - POINT 2
*******************************************************************************
ATOMS IN THE ASYMMETRIC UNIT 5 - ATOMS IN THE UNIT CELL: 8
ATOM X(ANGSTROM) Y(ANGSTROM) Z(ANGSTROM)
*******************************************************************************
1 T 6 C 0.000000000000E+00 0.000000000000E+00 2.645266012706E-02
2 T 8 O 0.000000000000E+00 0.000000000000E+00 1.241474126876E+00
3 T 7 N 0.000000000000E+00 1.150483100972E+00 -7.044307566681E-01
4 F 7 N 0.000000000000E+00 -1.150483100972E+00 -7.044307566681E-01
5 T 1 H 0.000000000000E+00 2.022583078191E+00 -2.043778206895E-01
6 F 1 H 0.000000000000E+00 -2.022583078191E+00 -2.043778206895E-01
7 T 1 H 0.000000000000E+00 1.135517317174E+00 -1.702036316144E+00
8 F 1 H 0.000000000000E+00 -1.135517317174E+00 -1.702036316144E+00

T = ATOM BELONGING TO THE ASYMMETRIC UNIT

INTRACELL NUCLEAR REPULSION (A.U.) 1.2463005288098E+02

TOTAL ENERGY(HF)(AU)( 11) -2.2379435865343E+02 DE-4.8E-08 DP 7.2E-06

SYMMETRY ALLOWED FORCES (ANALYTICAL) (DIRECTION, FORCE)

1 4.0854048E-02 2 -2.8460660E-02 3 1.4184257E-03 4 -3.0361419E-03

5 -1.7599295E-02 6 -1.3809310E-02 7 6.7962224E-03

GRADIENT NORM 0.055108 GRADIENT THRESHOLD 0.500000

MAX GRADIENT 0.040854 THRESHOLD 0.000450 CONVERGED NO

RMS GRADIENT 0.020829 THRESHOLD 0.000300 CONVERGED NO
MAX DISPLAC. 0.024990 THRESHOLD 0.001800 CONVERGED NO
RMS DISPLAC. 0.015649 THRESHOLD 0.001200 CONVERGED NO

When all four convergence tests are satisfied, optimization is completed. The final energy and
the optimized structure are printed after the final convergence tests.

******************************************************************
* OPT END - CONVERGED * E(AU): -2.237958289701E+02 POINTS 14 *
******************************************************************
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . .

FINAL OPTIMIZED GEOMETRY - DIMENSIONALITY OF THE SYSTEM 0

(NON PERIODIC DIRECTION: LATTICE PARAMETER FORMALLY SET TO 500)
*******************************************************************************
ATOMS IN THE ASYMMETRIC UNIT 5 - ATOMS IN THE UNIT CELL: 8
ATOM X(ANGSTROM) Y(ANGSTROM) Z(ANGSTROM)
*******************************************************************************
1 T 6 C 0.000000000000E+00 0.000000000000E+00 3.468988750953E-02
2 T 8 O 0.000000000000E+00 0.000000000000E+00 1.230143233209E+00
3 T 7 N 0.000000000000E+00 1.143750090534E+00 -7.056136525307E-01
4 F 7 N 0.000000000000E+00 -1.143750090534E+00 -7.056136525307E-01
5 T 1 H 0.000000000000E+00 2.001317638364E+00 -2.076003454226E-01
6 F 1 H 0.000000000000E+00 -2.001317638364E+00 -2.076003454226E-01
7 T 1 H 0.000000000000E+00 1.157946292824E+00 -1.696084062406E+00

207
 8 F 1 H 0.000000000000E+00 -1.157946292824E+00 -1.696084062406E+00

T = ATOM BELONGING TO THE ASYMMETRIC UNIT

INTRACELL NUCLEAR REPULSION (A.U.) 1.2541002823701E+02

**** 4 SYMMOPS - TRANSLATORS IN FRACTIONA LUNITS

V INV ROTATION MATRICES TRANSLATOR
1 1 1.00 0.00 0.00 0.00 1.00 0.00 0.00 0.00 1.00 0.00 0.00 0.00
2 2 -1.00 0.00 0.00 0.00 -1.00 0.00 0.00 0.00 1.00 0.00 0.00 0.00
3 3 -1.00 0.00 0.00 0.00 1.00 0.00 0.00 0.00 1.00 0.00 0.00 0.00
4 4 1.00 0.00 0.00 0.00 -1.00 0.00 0.00 0.00 1.00 0.00 0.00 0.00

The final geometry is both printed in the Crystal output and writ-
ten in file fort.34. The following input defines the geometry, read-
ing data from file fort.34 (keyword EXTERNAL, input block1, page 15).

. Example

Urea Molecule Title

EXTERNAL Geometry read from file fort.34
optional keywords
END End of the geometry input section

Optimization can restart, by adding the keyword RESTART in the same

input deck used for the first optimization run. Information to restart
are read from file OPTINFO.DAT, updated after each optimization cycle.

. Example

Initial input Restart input

Urea Molecule Urea Molecule

MOLECULE MOLECULE
15 15
5 5
6 0. 0. 0. 6 0. 0. 0.
8 0. 0. 1.261401 8 0. 0. 1.261401
7 0. 1.148247 -0.699790 7 0. 1.148247 -0.699790
1 0. 2.026550 -0.202817 1 0. 2.026550 -0.202817
1 0. 1.134080 -1.704975 1 0. 1.134080 -1.704975
OPTGEOM OPTGEOM
ENDOPT RESTART
END ENDOPT
END

Partial optimization
In order to optimize the coordinates of the hydrogens in urea molecule, the following input
must be entered:

208
. Example

Urea Molecule Title

MOLECULE Dimension of the system
15 Point group (C2v )
5 Number of non equivalent atoms
6 0. 0. 0. Atomic number and cartesian coordinates
8 0. 0. 1.261401
7 0. 1.148247 -0.699790
1 0. 2.026550 -0.202817
1 0. 1.134080 -1.704975
KEEPSYMM maintain symmetry in subsequent operations
OPTGEOM Keyword to perform a geometry optimization
FRAGMENT Keyword for a partial optimization
2 Number of atoms to be optimized
5 7 Label of the atoms to be optimized
ENDOPT End of the geometry optimization input block
END End of the geometry input section

The atoms allowed to move are the two hydrogens irreducible, 5 and 7. The symmetry is
maintained, atoms 6 and 7 are allowed to move.
*******************************************************************************
* PARTIAL OPTIMIZATION - ATOMS FREE TO MOVE 2 INPUT LIST :

5( 1) 7( 1)

SYMMETRY IS KEPT

ATOM 5 AND 6 ARE LINKED BY SYMMOP 2

ATOM 7 AND 8 ARE LINKED BY SYMMOP 2
*******************************************************************************

If the symmetry is not maintained (default, no KEEPSYMM before OPTGEOM) the symmetry
operators linking atoms 5 and 6, and atoms 7 and 8, are removed.
*******************************************************************************
* PARTIAL OPTIMIZATION - ATOMS FREE TO MOVE 2 INPUT LIST :

5( 1) 7( 1)

SYMMETRY MAY BE BROKEN

THE NUMBER OF SYMMETRY OPERATORS HAS BEEN REDUCED FROM 4 TO 2

******************************************************************************
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . .
SYMMETRY ALLOWED INTERNAL DEGREE(S) OF FREEDOM: 4
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . .

Final run
During optimization process, the classification of the integrals is done with reference to the
input geometry, and used for all subsequent wave function calculations.
In some cases, when the optimized geometry is far from the original one, the series truncation
defined with reference to the starting geometry may be inhomogeneous if applied to the final
geometry (see keyword FIXINDEX for explanation). In those cases, the total energy com-
puted for one geometry, with integrals selected according to a different geometry, may be not
correct.
A single point calculation, with the final optimized geometry, allows to check if that is the
case.

209
A safe procedure to check if that geometry corresponds to a real energy minimum, is to per-
form a second optimization process, with same truncation criteria, starting from the geometry
obtained in the first optimization (read from file fort.34, keyword EXTERNAL, page 15).
The keyword FINALRUN starts the process automatically. (it does not work with Pcrystal)
A typical example is the geometry optimization of a surface, described with a slab
model. The optimization process may lead to a structure significantly different
from the one cut from the bulk, when there is surface relaxation. As an exam-
ple, the geometry optimization of the surface (001) of the α -Al2 O3 is reported.

. Example - Optimization of surface

α -Al2 O3 - (001) surface title

CRYSTAL dimension of the system
0 0 0
167 space group
4.7602 12.9933 lattice parameters
2 number of irreducible atoms
13 0. 0. 0.35216 fractional coordinates of first atom
8 0.30624 0. 0.25 fractional coordinates of second atom
SLABCUT 3D→ 2D
0 0 1 (h, k, l) Miller indices of the surface
1 6 number of layers, starting from the first classified
OPTGEOM Keyword to perform a geometry optimization
FINALRUN keyword to check gradients vs true series truncation
3 new optimization if convergence criteria are not satisfied
ENDOPT end of the geometry optimization input block
END end of the geometry input section

Neighbors analysis on the initial geometry obtained with SLABCUT

N = NUMBER OF NEIGHBORS AT DISTANCE R 11 cycles

ATOM N R/ANG R/AU NEIGHBORS (ATOM LABELS AND CELL INDICES)
1 AL 3 1.8551 3.5057 2 O 0 0 0 3 O 0 0 0 4 O 0 1 0
1 AL 3 3.2192 6.0834 5 AL 0 0 0 5 AL 1 1 0 5 AL 0 1 0
1 AL 3 3.2219 6.0885 2 O -1 0 0 3 O 1 1 0 4 O 0 0 0
1 AL 3 3.4295 6.4808 7 O 0 1 0 8 O 0 1 0 9 O 0 0 0
1 AL 3 3.4990 6.6121 6 AL 0 0 0 6 AL -1 0 0 6 AL 0 1 0
1 AL 1 3.8419 7.2601 10 AL 0 0 0

Total energy E = -1399.7999027 hartree

Series truncation is defined with reference to that geometry. Optimization begins. After 11
cycles convergence on gradients and displacements is satisfied.
Neighbors analysis on the optimized geometry:

N = NUMBER OF NEIGHBORS AT DISTANCE Ra 6 cycles

ATOM N R/ANG R/AU NEIGHBORS (ATOM LABELS AND CELL INDICES)
1 AL 3 1.6886 3.1911 2 O 0 0 0 3 O 0 0 0 4 O 0 1 0
1 AL 1 2.6116 4.9351 10 AL 0 0 0
1 AL 3 2.8198 5.3286 7 O 0 1 0 8 O 0 1 0 9 O 0 0 0
1 AL 3 3.0425 5.7494 5 AL 0 0 0 5 AL 1 1 0 5 AL 0 1 0

210
 1 AL 3 3.0430 5.7504 6 AL 0 0 0 6 AL -1 0 0 6 AL 0 1 0
1 AL 3 3.1214 5.8987 2 O -1 0 0 3 O 1 1 0 4 O 0 0 0

Total energy E = -1400.1148194 hartree

A large geometrical relaxation occurred during the optimization: the aluminium atoms move
toward the core of the slab. In this case both the total energy and gradients should be recal-
culate using truncation series which refer to the final relaxed geometry.
This crucial step is automatically performed if the keyword FINALRUN is present in the input
file. If this is the case, CRYSTAL03 checks for the numerical consistency and it will find
that the gradients do not match the requested convergence criteria. At the end of the new
optimization the total energy is:

Total energy E = -1400.1193593 hartree

In this case, FINALRUN was followed by the keyword ICODE=3. This means that the geom-
etry optimization restarts from the very last step of the previous geometry optimization with
truncation series criteria defined relative to that geometry. After 6 new optimization cycles,
convergence criteria are satisfied.
Neighbors analysis on the final run optimized geometry

N = NUMBER OF NEIGHBORS AT DISTANCE R

ATOM N R/ANG R/AU NEIGHBORS (ATOM LABELS AND CELL INDICES)
1 AL 3 1.6863 3.1867 2 O 0 0 0 3 O 0 0 0 4 O 0 1 0
1 AL 1 2.5917 4.8976 10 AL 0 0 0
1 AL 3 2.8095 5.3092 7 O 0 1 0 8 O 0 1 0 9 O 0 0 0
1 AL 3 3.0382 5.7414 5 AL 0 0 0 5 AL 1 1 0 5 AL 0 1 0
1 AL 3 3.0387 5.7424 6 AL 0 0 0 6 AL -1 0 0 6 AL 0 1 0
1 AL 3 3.1215 5.8987 2 O -1 0 0 3 O 1 1 0 4 O 0 0 0

Total energy E = -1400.1194545 hartree

The final geometry is printed, and written in file fort.34.

A final check on total energy can be done with the following input:

alpha-Al2O3 (corundum) 001 2 LAYERS (3D-->2D)

EXTERNAL
OPTGEOM
ENDOPT
END

The keyword EXTERNAL routes the basic geometry input stream to file fort.34, written at
the end of the optimization run.
No optimization starts, convergence criteria are already satisfied.

Total energy E = -1400.1194544 hartree

/sectionScanning of vibrational modes - SCANMODE

211
Methane molecule
MOLECULE
44
2
6 0.000000000000E+00 0.000000000000E+00 0.000000000000E+00
1 6.252140000000E-01 6.252140000000E-01 6.252140000000E-01
FREQCALC
SCANMODE
-1 -1 0 0.1
12
ENDfreq
END
6 3
0 0 3 2. 1.
172.256000 0.617669000E-01
25.9109000 0.358794000
5.53335000 0.700713000
0 1 2 4. 1.
3.66498000 -0.395897000 0.236460000
0.770545000 1.21584000 0.860619000
0 1 1 0. 1.
0.195857000 1.00000000 1.00000000
1 2
0 0 2 1. 1.
5.44717800 0.156285000
0.824547000 0.904691000
0 0 1 0. 1.
0.183192000 1.00000000
99 0
END
TOLINTEG
20 20 20 20 20
END
TOLDEE
11
END

PbCO3

PBCO3 - frequency calculation

CRYSTAL
1 0 0
P m c n
5.20471446 8.45344758 6.16074145
4
282 2.500000000000E-01 4.175726169487E-01 -2.463557995068E-01
6 2.500000000000E-01 -2.363341497085E-01 -8.558132726424E-02
8 2.500000000000E-01 -8.360585350428E-02 -9.431628799197E-02
8 4.648370653436E-01 -3.129222129903E-01 -8.842925698155E-02
FREQCALC
RESTART
SCANMODE
1 -40 40 0.1
1
END

212
END
282 4
HAYWLC
0 1 2 4. 1.
1.335104 -0.1448789 -0.1070612
0.7516086 1.0 1.0
0 1 1 0. 1.
0.5536686 1.0 1.0
0 1 1 0. 1.
0.1420315 1.0 1.0
0 3 1 0. 1.
0.1933887 1.0
6 4
0 0 6 2.0 1.0
3048.0 0.001826
456.4 0.01406
103.7 0.06876
29.23 0.2304
9.349 0.4685
3.189 0.3628
0 1 2 4.0 1.0
3.665 -0.3959 0.2365
0.7705 1.216 0.8606
0 1 1 0.0 1.0
0.26 1.0 1.0
0 3 1 0.0 1.0
0.8 1.0
8 4
0 0 6 2.0 1.0
.5484671660D+04 .1831074430D-02
.8252349460D+03 .1395017220D-01
.1880469580D+03 .6844507810D-01
.5296450000D+02 .2327143360D+00
.1689757040D+02 .4701928980D+00
.5799635340D+01 .3585208530D+00
0 1 3 6.0 1.0
.1553961625D+02 -.1107775490D+00 .7087426820D-01
.3599933586D+01 -.1480262620D+00 .3397528390D+00
.1013761750D+01 .1130767010D+01 .7271585770D+00
0 1 1 0.0 1.0
.2700058226D+00 .1000000000D+01 .1000000000D+01
0 3 1 0.0 1.0
.8000000000D+00 .1000000000D+01
99 0
ENDBS
SCFDIR
DFT
B3LYP
RADIAL
1
4.0
99
ANGULAR
5
0.1667 0.5 0.9 3.5 9999.0
6 10 14 18 14
END

213
SHRINK
6 6
LEVSHIFT
5 0
TOLDEE
10
MAXCYCLE
200
ENDSCF

214
Chapter 10

Basis set

The most common source of problems with CRYSTAL is probably connected with the basis set.
It should never be forgotten that ultimately the basis functions are Bloch functions, modulated
over the infinite lattice: any attempt to use large uncontracted molecular or atomic basis sets,
with very diffuse functions can result in the wasting of computational resources. The densely
packed nature of many crystalline structures gives rise to a large overlap between the basis
functions, and a quasi-linear dependence can occur, due to numerical limitations.
The choice of the basis set (BS) is one of the critical points, due to the large variety of
chemical bonding that can be found in a periodic system. For example, carbon can be involved
in covalent bonds (polyacetylene, diamond) as well as in strongly ionic situations (Be2 C, where
the Mulliken charge of carbon is close to -4).
Many basis sets for lighter elements and the first row transition metal ions have been developed
for use in periodic systems. A selection of these which have been used in published work are
available on WWW:

http://www.crystal.unito.it

We summarize here some general considerations which can be useful in the construction of a
BS for periodic systems.
It is always useful to refer to some standard basis set; Pople’s STO-nG, 3-21G and 6-21G have
proved to be good starting points. A molecular minimal basis set can in some cases be used
as it is; larger basis sets must be re-optimized specifically for the chemical periodic structure
under study.
Let us explore the adequacy of the molecular BS for crystalline compounds and add some
considerations which can be useful when a molecular BS must be modified or when an ex novo
crystalline BS is defined.

10.1 Molecular BSs performance in periodic systems

Two sets of all electron basis sets are included in CRYSTAL (see Chapter 1.2):
1. Minimal STO-nG basis set of Pople and co-workers
obtained by fitting Slater type orbitals with n contracted GTFs (n from 2 to 6, atomic
number from 1 to 54) [73, 72, 98, 97].
The above BSs are still widely used in spite of the poor quality of the resulting wave func-
tion, because they are well documented and as a rule provide quite reasonable optimized
geometries (due to fortuitous cancellation of errors) at low cost.
2. ”Split valence” 3-21 and 6-21 BSs.
The core shells are described as a linear combination of 3 (up to atomic number 54)
or 6 (up to atomic number 18) gaussians; the two valence shells contain two and one
gaussians, respectively [15, 60]. Exponents (s and p functions of the same shell share the
same exponent) and contraction coefficients have been optimized variationally for the
isolated atoms.

215
A single set of polarization functions (p,d) can be added without causing numerical problems.
Standard molecular polarization functions are usually also adequate for periodic compounds.

When free basis sets are chosen, two points should be taken into account:

1. From the point of view of CPU time, basis sets with sp shells (s and p functions sharing
the same set of exponents) can give a saving factor as large as 4, in comparison with
basis sets where s and p have different exponents.

2. As a rule, extended atomic BSs, or ’triple zeta’ type BSs should be avoided. Many of
the high quality molecular BSs (Roos, Dunning, Huzinaga) cannot be used in CRYSTAL
without modification, because the outer functions are too diffuse. One should not forget
that the real basis functions are Bloch functions.

Let us consider in more detail the possibility of using molecular BS for periodic systems. We
can refer to five different situations:

Core functions
Valence functions: molecular crystals
covalent crystals
ionic crystals
metals.

10.2 Core functions

In this case standard (contracted) molecular BSs can be adopted without modification, be-
cause even when very strong crystal field effects are present, the deformation of inner states
is small, and can be correctly described through the linear variational parameters in SCF cal-
culation. An adequate description of the core states is important in order to avoid large basis
set superposition errors.

10.3 Valence functions

Molecular crystals
Molecular BSs, minimal and split-valence, are perfectly adequate. Tests have been performed
on bulk urea [36] and oxalic acid, where the molecules are at relatively small distances, with
STO-3G, 6-21, 6-21* and 6-21** BSs presenting no problem.

Covalent crystals.
Standard minimal and split valence BSs are usually adequate. In the split valence case the
best exponent of the most diffuse shell is always slightly higher than the one proposed for
molecules; in general it is advisable to re-optimize the exponent of this shell. This produces a
slightly improved basis, while reducing the cost of the calculation. Let us consider for example
the 6-21 basis set for carbon (in diamond) and silicon (bulk).
At an atomic level, the best exponent of the outer shell is 0.196 and 0.093 for C and Si, respec-
tively. Optimization of the valence shell has been repeated in the two crystalline compounds.
The innermost valence shell is essentially unaltered with respect to the atomic solution; for the
outer single-gaussian shell the best exponent is around 0.22 and 0.11 bohr−2 for carbon and
silicon, as shown in Table 10.1. The last entry of Table 10.1 refers to ”catastrophic” behaviour:
the low value of the exponent generates unphysical states.
A set of 5 polarization single-gaussian d functions can be added to the 6-21G basis (6-21G*
BS); the best exponents for the solid are very close to those resulting from the optimization in
molecular contexts: 0.8 for diamond [66] and 0.45 for silicon.

Basis sets for III-V and IV-IV semiconductors (all electron and valence electron (to be associ-
ated with effective core pseudopotentials) are given in references [86, 26].

216
Table 10.1: Total energy per cell and number of computed bielectronic integrals in 106 units
(N), as a function of the exponent α (bohr−2 ) of the most diffuse shell for carbon and silicon.
____________________________________________________________________
Diamond Silicon
------------------------ --------------------------
a N Et a N Et
___________________________________________________________
0.296 58 -75.6633 0.168 46 -577.8099
0.276 74 -75.6728 0.153 53 -577.8181
0.256 83 -75.6779 0.138 72 -577.8231
0.236 109 -75.6800 0.123 104 -577.8268
0.216 148 -75.6802 0.108 151 -577.8276
0.196 241 -75.6783 0.093 250 -577.8266
0.176 349 catastrophe 0.078 462 catastrophe
____________________________________________________________________

Ionic crystals.
Cations
The classification of covalent or ionic crystals is highly conventional, many systems being
midway. Let us first consider totally ionic compounds, such as LiH, MgO, or similar. For these
systems the cation valence shell is completely empty. Therefore, for cations it is convenient
to use a basis set containing the core functions plus an additional sp shell with a relatively
high exponent. For example, we used for Mg in MgO and for Li in LiH ( Li2 O and Li3 N) a
’valence’ sp shell with exponent 0.4-0.3 and 0.5-0.6, respectively [41, 25].
The crystalline total energies obtained by using only core functions for Li or Mg and by adding
a valence shell to the cation differ by 0.1 eV/atom, or less. This figure is essentially the same
for a relatively large range of exponents of the valence shell (say 0.5-0.2 for Mg) [25].
It can be difficult (or impossible) to optimize the exponents of nearly empty shells: the en-
ergy decreases almost linearly with the exponent. Very low exponent values can give rise to
numerical instabilities, or require the calculation of an enormous number of integrals (selected
on the basis of overlap criteria). In the latter cases, when the energy gain is small (∆E ≤ 1 m
hartree for ∆α = 0.2 bohr−2 ), it is convenient to use a relatively large exponent.

Anions
Reference to isolated ion solutions is only possible for halides, because in such cases the ions
are stable even at the HF level. For other anions, which are stabilized by the crystalline
field (H − , O2− , N 3− and also C 4− ), the basis set must be re-designed with reference to the
crystalline environment. For example, let us consider the optimization of the O2− BS in Li2 O
[41]. Preliminary tests indicated the fully ionic nature of the compound; the point was then to
allow the valence distribution to relax in the presence of the two extra electrons. We started
from a standard STO-6G BS. Two more gaussians were introduced in the 1s contraction, in
order to improve the virial coefficient and total energy, as a check of wave function quality. The
6 valence gaussians were contracted according to a 411 scheme; the exponents of the two outer
independent gaussians and the coefficients of the four contracted ones were optimized. Whereas
the two most diffuse gaussians are more diffuse than in the neutral isolated atom (α=0.45 and
0.15 to be compared with α=0.54 and 0.24 respectively), the rest of the O2− valence shell is
unchanged with respect to the atomic situation. The introduction of d functions in the oxygen
basis-set causes only a minor improvement in energy (1 10−4 hartree/cell, with a population
of 0.02 electrons/atom in the cell). Ionic BSs for H and N can be found in reference 1.
For anions, re-optimization of the most diffuse valence shell is mandatory; when starting from
a standard basis set, the most diffuse (or the two most diffuse) gaussians must be allowed to
relax.

217
From covalent to ionics
Intermediate situations must be considered individually, and a certain number of tests must
be performed in order to verify the adequacy of the selected BSs.
Let us consider for example α-quartz (SiO2 ) and corundum (Al2 O3 ). The exponent of the
outer shell for the 2 cations in the 6-21G BS is 0.093 (Si) and 0.064 (Al), respectively; in both
cases this function is too diffuse (in particular in the Al case it causes numerical catastrophes).
For quartz, re-optimization in the bulk gives α=0.15 bohr−2 for Si (the dependence of total
energy per Si atom on α is much smaller than the one resulting from Table 10.1; note too that
the cost at α=0.15 is only 50% of the one at α=0.09). On the contrary, the best molecular
and crystalline exponent (α=0.37) for oxygen coincide. Corundum is more ionic than quartz,
and about 2 valence electrons are transferred to oxygen. In this case it is better to eliminate
the most diffuse valence shell of Al, and to use as independent functions the two gaussians of
the inner valence shells (α=0.94 and 0.20 bohr−2 , respectively [109]).

Metals
Very diffuse gaussians are required to reproduce the nearly uniform density characterizing
simple metallic systems, such as lithium and beryllium. This is the worse situation, where a
full optimization of the atomic basis set is probably impossible. Functions which are too diffuse
can create numerical problems, as will be discussed below.
The optimization procedure can start from 6-21 BS; the most diffuse valence shell (exponent
0.028 for Li and 0.077 for Be) can be dropped and the innermost valence shell (exponents 0.54
and 0.10 for Li, and 1.29 and 0.268 for Be) can be split.

Table 10.2: Example of BS for metallic lithium and beryllium derived from the standard 6-21G
BS
.

_______________________________________________________________
Lithium Beryllium
shell Exp. Coeff. shell Exp. Coeff.
_______________________________________________________________
s 642.418 0.00215096 s 1264.50 0.00194336
96.5164 0.0162677 189.930 0.0148251
22.0174 0.0776383 43.1275 0.0720662
6.1764 0.246495 12.0889 0.237022
1.93511 0.467506 3.80790 0.468789
sp 0.640 1. 1. 1.282 1. 1.
sp 0.10 1. 1. 0.27 1. 1.
_______________________________________________________________

At this point the outer gaussian of the 6G core contraction, with very similar exponents (0.64
and 1.28) to those of the innermost valence shell (0.54 and 1.29), can be used as an independent
(sp) function, and the innermost valence shell can be eliminated.
The resulting (reasonable) BS, derived from the split valence standard one, is reported in Table
10.2. Finally, the most diffuse gaussian can be optimized; in the two cases the minimum has
not been found owing to numerical instabilities.
See [37] for a more extensive discussion of the metallic lithium case.

10.4 Hints on crystalline basis set optimization

In the definition of a valence shell BS, each exponent can be varied in a relatively narrow range:
in the direction of higher exponents, large overlaps with the innermost functions may occur
(the rule of thumb is: exponents must be in a ratio not too far from 3; ratios smaller than

218
2 can give linear dependence problems); proceeding towards lower exponents, one must avoid
large overlaps with a high number of neighbours (remember: the basis functions are Bloch
functions).
Diffuse gaussian orbitals play a critical role in HF-LCAO calculations of crystals, especially
the three-dimensional ones; they are expensive, not always useful, in some cases dangerous.

• Cost.

The number of integrals to be calculated increases dramatically with decreasing exponents;

this effect is almost absent in molecular calculations. Table 10.1 shows that the cost of the
calculation (number of bielectronic integrals) for silicon (diamond) can increase by a factor
10 (6) simply by changing the exponent of the most diffuse single-gaussian from 0.168 to
0.078 (0.296 to 0.176). The cost is largely dominated by this shell, despite the fact that large
contractions are used for the 1s, 2sp and the innermost valence shell.
A high number of contracted primitives tremendously increases the integrals computation time.

• Usefulness.

In atoms and molecules a large part of the additional variational freedom provided by diffuse
functions is used to describe the tails of the wave function, which are poorly represented by the
2
e−αr decay of the gaussian function. On the contrary, in crystalline compounds (in particular
3D non-metallic systems), low exponent functions do not contribute appreciably to the wave
function, due to the large overlap between neighbours in all directions. A small split valence
BS such as the 6-21G one, is nearer to the variational limit in crystals than in molecules.

• Numerical accuracy and catastrophic behaviour.

In some conditions, during the SCF (periodic) calculation, the system ’falls’ into non-physical
states, characterized by very low single particle and total energies (see for example the last
entry in Table 10.1 and the above discussion on metals).
This behaviour, generically interpreted in our early papers as due to ’linear dependence’, is
actually due to poor accuracy in the treatment of the Coulomb and exchange series. The
exchange series is much more delicate, for two reasons: first, long range contributions are not
taken into account (whereas the long range Coulomb contributions are included, although in an
approximate way); second, the ”pseudoverlap” criteria associated with the two computational
parameters ITOL4 and ITOL5 mimic only in an approximate way the real behaviour of the
density matrix.
The risks of ”numerical catastrophes” increase rapidly with a decreasing exponent; higher
precision is required in order to obtain physical solutions.
For non-metallic systems, and split-valence type BSs, the default computational conditions
given in section 1.3 are adequate for the optimization of the exponents of the valence shell and
for systematic studies of the energy versus volume curves.
For metallic systems, the optimization of the energy versus exponent curve could require ex-
tremely severe conditions for the exchange series and, as a consequence, for the reciprocal
space net. Reasonable values of the valence shell exponent (say 0.23 for beryllium and 0.10
for lithium, see Table 10.2), though not corresponding to a variational minimum, are reason-
ably adequate for the study of the structural and electronic properties of metallic systems (see
reference 1).

10.5 Check on basis-set quasi-linear-dependence

In order to check the risk of linear dependence of Bloch functions, it is possible to calculate
the eigenvalues of the overlap matrix in reciprocal space by running integrals and entering
the keyword EIGS (input block 3, page 81). Full input (general information, geometry, basis
set, SCF) is to be entered.

219
The overlap matrix in direct space is Fourier transformed at all the k points generated in the
irreducible part of the Brillouin zone, and diagonalized. The eigenvalues are printed.
The higher the numerical accuracy obtained by severe computational conditions, the closer
to 0 can be the eigenvalues without risk of numerical instabilities. Negative values indicate
numerical linear dependence. The program stops after the check (even if negative eigenvalues
are not detected).
The Cholesky reduction scheme [77] requires basis functions linearly independent. A symptom
of numerical dependence may produce an error message in RHOLSK or CHOLSK while running
scf.

220
Chapter 11

Theoretical framework

11.1 Basic equations

CRYSTAL is an ab initio Hartree-Fock LCAO program for the treatment of periodic systems.
LCAO, in the present case, means that each Crystalline Orbital , ψi (r; k), is a linear combina-
tion of Bloch functions (BF), φµ (r; k), defined in terms of local functions, ϕµ (r) (here referred
to as Atomic Orbitals, AOs).
X
ψi (r; k) = aµ,i (k)φµ (r; k) (11.1)
µ

X
φµ (r; k) = ϕµ (r − Aµ − g) eik·g (11.2)
g

Aµ denotes
P the coordinate of the nucleus in the zero reference cell on which ϕµ is centred, and
the g is extended to the set of all lattice vectors g.
The local functions are expressed as linear combinations of a certain number, nG , of individually
normalized (basis set) Gaussian type functions (GTF) characterized by the same centre, with
fixed coefficients, dj and exponents, αj , defined in the input:
nG
X
ϕµ (r − Aµ − g) = dj G(αj ; r − Aµ − g) (11.3)
j

The AOs belonging to a given atom are grouped into shells, λ. The shell can contain all AOs
with the same quantum numbers, n and `, (for instance 3s, 2p, 3d shells), or all the AOs with the
same principal quantum number, n, if the number of GTFs and the corresponding exponents
are the same for all of them (mainly sp shells; this is known as the sp shells constraint). These
groupings permit a reduction in the number of auxiliary functions that need to be calculated
in the evaluation of electron integrals and therefore increase the speed of calculation.
A single, normalized, s-type GTF, Gλ , is associated with each shell (the adjoined Gaussian of
shell λ). The α exponent is the smallest of the αj exponents of the Gaussians in the contraction.
The adjoined Gaussian is used to estimate the AO overlap and select the level of approximation
to be adopted for the evaluation of the integrals.
The expansion coefficients of the Bloch functions, aµ,i (k), are calculated by solving the matrix
equation for each reciprocal lattice vector, k:

F(k)A(k) = S(k)A(k)E(k) (11.4)

in which S(k) is the overlap matrix over the Bloch functions, E(k) is the diagonal energy
matrix and F(k) is the Fock matrix in reciprocal space:
X
F(k) = Fg eik·g (11.5)
g

221
The matrix elements of Fg , the Fock matrix in direct space, can be written as a sum of
one-electron and two-electron contributions in the basis set of the AO:
g g g
F12 = H12 + B12 (11.6)

The one electron contribution is the sum of the kinetic and nuclear attraction terms:
g g g
H12 = T12 + Z12 = hϕ01 | Tb | ϕg2 i + hϕ01 | Z
b | ϕg i
2 (11.7)

In core pseudopotential calculations, Ẑ includes the sum of the atomic pseudopotentials.

The two electron term is the sum of the Coulomb and exchange contributions:
g g g
B12 = C12 + X12 =

XX X 1
n
P3,4 [(ϕ01 ϕg2 | ϕh h+n
3 ϕ4 ) − (ϕ01 ϕh g h+n
3 | ϕ2 ϕ4 )] (11.8)
3,4 n
2
h

The Coulomb interactions, that is, those of electron-nucleus, electron-electron and nucleus-
nucleus, are individually divergent, due to the infinite size of the system. The grouping of
corresponding terms is necessary in order to eliminate this divergence.
The Pn density matrix elements in the AOs basis set are computed by integration over the
volume of the Brillouin zone (BZ),
Z X
n
P3,4 = 2 dkeik·n a∗3j (k)a4j (k)θ(F − j (k)) (11.9)
BZ j

where ain denotes the i-th component of the n-th eigenvector, θ is the step function, F , the
Fermi energy and n , the n-th eigenvalue. The total electronic energy per unit cell is given
by:
1 XX g
E elec = P (H g + F12g
) (11.10)
2 1,2 g 12 12

A discussion of the different contributions to the total energy is presented in [111, 114] and in
Chapter 11 of reference [100].
1 XX g XX n X 0 g h+n
E coul = P P [(ϕ1 ϕ2 | ϕh
3 ϕ4 )] (11.11)
2 1,2 g 12 3,4 n 3,4
h

1 XX g XX n X 0 h
E exch = − P P [(ϕ1 ϕ3 | ϕg2 ϕh+n )] (11.12)
4 1,2 g 12 34 n 34 4
h

11.2 Remarks on the evaluation of the integrals

The approach adopted for the treatment of the Coulomb and exchange series is based on a few
simple ideas and on a few general tools, which can be summarized as follows:
1. Where possible, terms of the Coulomb series are aggregated so as to reduce the number
of integrals to be evaluated;
2. Exchange integrals which will combine with small density matrix elements are disre-
garded;
3. Integrals between non-overlapping distributions are approximated;
4. Approximations for large integrals must be very accurate; for small integrals large per-
centage errors can be accepted;
5. Selection must be very efficient, because a large number of possible terms must be checked
(adjoined Gaussians are very useful from this point of view).

222
11.3 Treatment of the Coulomb series
For the evaluation of the Coulomb contributions to the total energy and Fock matrix, correct
coupling of electron-nucleus and electron-electron interactions is essential. The computational
technique for doing so was presented by Dovesi et al [40] and by Saunders et al. [114]. It may
be summarized as follows.
g
Consider the Coulomb bielectronic contribution to the Fock matrix (C12 ) and to the total
energy :
coul 1 XX g XX n X 0 g h+n
Eee = P P [(ϕ1 ϕ2 | ϕh3 ϕ4 ) (11.13)
2 1,2 g 12 3,4 n 3,4
h

Seven indices are involved in equation 11.13; four of them (1, 2, 3 and 4) refer to the AOs
of the unit cell; in principle, the other three (g, n and h) span the infinite set of translation
vectors: for example, ϕg2 (r) is AO number 2 in cell g. P is the density matrix; the usual
notation is used for the bielectronic integrals. Due to the localized nature of the basis set, the
total charges, q1 and q2 , associated with the two pseudo-overlap distributions: {G10 G2g } and
{G3h G4h+n }, decay exponentially to zero with increasing |g| and |n| (for example, G1 is the
adjoined gaussian of the shell to which ϕ1 belongs).
A Coulomb overlap parameter, Sc , can be defined in such a way that when either q1 or q2 are
smaller than Sc , the bielectronic integral is disregarded, and the sum over g or n truncated.
The ITOL1 input parameter is defined as ITOL1=−log10 Sc . The same parameter value is
used for selecting overlap, kinetic, and multipole integrals.
The problem of the h summation in equation 11.13 is more delicate, h being related to the
distance between the two interacting distributions. The multipolar expansion scheme illus-
trated below is particularly effective when large unit cell or low dimensionality systems are
g g
considered. The electron-electron and electron-nuclei series (C12 and Z12 ) can be rearranged
as follows:
1. Mulliken shell net charge distributions are defined as :
XX
ρλ (r − h) ≡ {λ}0 ≡ {λ} − Zλ = n
P34 ϕ3 (r − h) ϕ4 (r − h − n) − Zλ (11.14)
3∈λ 4n

where Zλ is the fraction of nuclear charge formally attributed to shell λ, and {λ} is the
electron charge distribution of shell λ.
2. Z and C contributions are reordered:
g g
XXZ
C12 + Z12 = dr dr0 ϕ01 (r) ϕg2 (r) |r − r0 − h|−1 ρλ (r0 − h) (11.15)
λ h

3. For a given shell λ, there is a finite set Bλ of h vectors for which the two interacting
distributions overlap; in this Bλ zone (bielectronic zone), all the bielectronic integrals are
evaluated explicitly. In the outer, infinite region which we define as Mλ , complementary
to Bλ (the mono-electronic zone), ρλ can be expanded in multipoles and the series can be
evaluated to infinity analytically, using Ewald’s method combined with recursion formulae
[114].
The resulting expression for the Coulomb contribution to the Fock matrix is:
g g P PBλ P P P n 0 g h+n
C12 + Z12 = λ{ [ P34 (ϕ1 ϕ2 | ϕh 3 ϕ4 )+
P h m3∈λ 4 n m
− γ (A λ ; {λ})Φ (12g; A λ + h)] + (11.16)
P ` m
P `,m `
0 m
+ h γ
`,m ` (A λ ; {λ} )Φ ` (12g; Aλ + h)}

where:
Z
γ`m (Aλ ; {λ}) = dr ρλ (r − Aλ )N`m X`m (r − Aλ ) (11.17)
Z
Φm
` (12g; Aλ + h) = drϕ01 (r)ϕg2 (r)X`m (r − Aλ − h) |r − Aλ − h|−2`−1 (11.18)

223
The Ewald term in eq. 11.16 includes zones Bλ + Mλ . The contribution from Bλ is subtracted.
The X`m functions entering in the definition of the multipoles and field terms are real, solid
harmonics, and N`m , the corresponding normalization coefficients.
The advantage of using equation 11.16 is that many four-centre (long-range) integrals can be
replaced by fewer three-centre integrals.
The attribution of the interaction between ρ1 = {10, 2g} and ρλ to the exact, short-range or
to the approximate, long-range zone is performed by comparing the penetration between ρ1
and ρλ with the ITOL2 input parameter (if ITOL2> − log S1λ , then ρλ is attributed to the
exact Bλ zone).
The multipolar expansion in the approximate zone is truncated at L = `max . The default value
of L is 4; the maximum possible value is 6, the minimum suggested value, 2 (defined via the
input keyword POLEORDR, input block 3, page 96).

11.4 The exchange series

The exchange series does not require particular manipulations of the kind discussed in the
previous section for the Coulomb series, but needs a careful selection of the terms contributing
appreciably to the Fock operator and to the total energy [24]. The exchange contribution to
the total energy can be written as follows:
1 XX g 1 XX n X 0 h
E ex = P12 [− P (ϕ1 ϕ3 | ϕg2 ϕh+n )] (11.19)
2 12 g 2 34 n 34 4
h

where the term in square brackets is the exchange contribution to the 12g element of the direct
space Fock matrix. E ex has no counterpart of opposite sign as the Coulomb term has; hence,
it must converge by itself.
The h summation can be truncated after a few terms, since the {ϕ01 ϕh 3 } overlap distribution
decays exponentially as h increases. Similar considerations apply to the second charge distri-
bution. In CRYSTAL, the h summation is, therefore, truncated when the charge associated
with either {G1 0 G3 h} or {G2 g G4 h + n} is smaller than 10−ITOL3 .
The situation is more complicated when g and n summations are analysed. Let us consider
the leading terms at large distance, corresponding to ϕ1 =ϕ3 , ϕ2 =ϕ4 , h = 0 and n = g:

eg12 = −1/4(P12
g 2
) (10 10|2g 2g) = −(pg )2 /(4|g|) (11.20)

(Here pg indicates the dominant P matrix element at long range). Since the number of terms
per unit distance of this kind increases as |g|d−1 , where d is the dimensionality of the system,
it is clear that the convergence of the series depends critically on the long range behaviour of
the bond order matrix.
Cancellation effects, associated in particular with the oscillatory behaviour of the density ma-
trix in metallic systems, are not predominant at long range. Even if the actual behaviour of
the P matrix elements cannot be predicted because it depends in a complicated way on the
physical nature of the compound [102], on orthogonality constraints and on basis set quality,
the different range of valence and core elements can be exploitedRby adopting a pseudoverlap
criterion. This consists in truncating g summations when the drϕ01 ϕg2 overlap is smaller
g g
than a given threshold,
R defined as Pex (where ITOL4 = -log10 (Pex )) and also truncating the
0 n n
n summation when drϕ3 ϕ4 overlap is smaller than the threshold, Pex (ITOL5 = -log10
n
(Pex )).
Despite its partially arbitrary nature, this criterion presents some advantages with respect to
other more elaborate schemes: it is similar to the other truncation schemes (ITOL1, ITOL2,
ITOL3), and so the same classification tables can be used; it is, in addition, reasonably efficient
in terms of space occupation and computer time.
This truncation scheme is symmetric with respect to the g and n summations. However, if
account is not taken of the different role of the two summations in the SC (Self Consistent)
stage, distortions may be generated in the exchange field as felt by charge distributions ϕ1 ϕT2 ,
where T labels the largest (in modulus) g vector taken into account according to ITOL4. This
distortion may be variationally exploited, and unphysically large density matrix elements build

224
up progressively along the SC stage, eventually leading to catastrophic behaviour (see Chapter
II.5 of reference [103] for a discussion of this point). In order to overcome this problem, the
n
threshold, Pex (ITOL5) for n summation must be more severe thanR that for g summation
(ITOL4). In this way, all the integrals whose second pseudo charge drϕ03 ϕn 4 is larger than
n
Pex are taken into account. A difference in the two thresholds ranging from three to eight
orders of magnitude is sufficient to stabilize the SC behaviour in most cases.

11.5 Bipolar expansion approximation of Coulomb and

exchange integrals
We may now return to the partition of the h summation in the Coulomb series shown in
equation 11.13. Consider one contribution to the charge distribution of electron 1, centred in
the reference cell: ρ0 = ϕ01 ϕg2 ; now consider the charge distribution ρλ (h) of shell λ centred
in cell h (equation 11.14). For small |h| values, ρλ and ρ0 overlap, so that all the related
bielectronic integrals must be evaluated exactly, one by one; for larger values of |h|, ρλ is
external to ρ0 , so that all the related bielectronic integrals are grouped and evaluated in an
approximate way through the multipolar expansion of ρλ .
h+n
However, in many instances, although ρλ is not external to ρ0 , the two-centre ϕh 3 ϕ4 con-
0 0 g
tributions to ρλ are external to ρ = ϕ1 ϕ2 ; in this case, instead of exactly evaluating the
bielectronic integral, a two-centre truncated bipolar expansion can be used (see Chapter II.4.c
in reference [103] and references therein).
In order to decide to which zone a shell may be ascribed, we proceed as follows: when, for a
given pair of shells λ01 λg2 , shell λh
3 is attributed to the B (bielectronic) zone, the penetration
between the products of adjoined Gaussians G01 Gg2 and Gh 3 G4
h+n
is estimated: the default
value of the penetration parameter is 14, and the block of bielectronic integrals is attributed
accordingly to the be (exact) or to the bb (bipolar) zone. The set of h vectors defining the B
zone of ρ0 = {12g} and ρλ ≡ {λ3 } is then split into two subsets, which are specific for each
partner λl4 of λ3 .
A similar scheme is adopted for the selected exchange integrals (see previous section) whose
pseudo charges do not overlap appreciably. The default value of the penetration parameter is
10.
The total energy change due to the bipolar expansion approximation should not be greater
than 10−4 hartree/atom; exact evaluation of all the bielectronic integrals (obtained by setting
the penetration parameter value > 20000) increases the computational cost by a factor of
between 1.3 and 3. Multipolar expansion is very efficient, because the following two conditions
are fulfilled:

1. A general algorithm is available for reaching high ` values easily and economically [40,
114]. The maximum allowed value is `=6.

2. The multipolar series converges rapidly, either because the interacting distributions are
nearly spherical (shell expansion), or because their functional expression is such that
their multipoles are zero above a certain (low) ` value.

11.6 Exploitation of symmetry

Translational symmetry allows the factorization of the eigenvalue problem in periodic calcula-
tions, because the Bloch functions are a basis for irreducible representations of the translational
group.
In periodic calculations, point symmetry is exploited to reduce the number of points for which
the matrix equations are to be solved. Point symmetry is also explicitly used in the reconstruc-
tion of the Hamiltonian, which is totally symmetric with respect to the point group operators
of the system.
In the HF-CO-LCAO scheme, the very extensive use of point symmetry allows us to evaluate
bielectronic and mono-electronic integrals with saving factors as large as h in the number of
bielectronic integrals to be computed or h2 in the number of those to be stored for the SCF part

225
of the calculation, where h is the order of the point group. The main steps of the procedure
[35] can be summarized as follows:
• The set of Coulomb and exchange integrals whose 3,4 indices identify translationally
equivalent pairs of AOs, so that the associated element of the density matrix P34 is the
same, are summed together to give D1234 elements:
X
D1,2T ;3,4Q = [(ϕ01 ϕg2 | ϕh h+n
3 ϕ4 ) − 1/2(ϕ01 ϕh g h+n
3 | ϕ2 ϕ4 )] (11.21)
n

• The products of AOs ϕ1 ϕ2 ( and ϕ3 ϕ4 ) are classified in symmetry-related sets; using the
fact that the Fock matrix is totally symmetric, only those quantities are evaluated whose
indices 1, 2 refer to the first member of a symmetry set. The corresponding saving factor
is as large as h.
• Using the symmetry properties of the density matrix, D quantities referring to 3, 4, cou-
ples belonging to the same symmetry set (and with the same 1, 2g index) can be combined
after multiplication by appropriate symmetry matrices, so that a single quantity for each
3, 4 symmetry set is to be stored, with a saving factor in storage of the order of h.
−n
• The symmetry Pn
34 = P43 is exploited.

• The symmetry Fg12 = F−g

21 is exploited.

Symmetry-adapted Crystalline Orbitals

A computational procedure for generating space-symmetry-adapted Bloch functions, when BF
are built from a basis of local functions (AO), is implemented in the CRYSTAL98 code. The
method, that applies to any space group and AOs of any quantum number, is based on the
diagonalization of Dirac characters. For its implementation, it does not require character tables
or related data as an input, since the information is automatically generated starting from the
space group symbol and the AO basis set. Formal aspects of the method, not available in
textbooks, are discussed in:
C. M. Zicovich-Wilson and R. Dovesi
On the use of Symmetry Adapted Crystalline Orbitals in SCF-LCAO periodic calculations. I.
The construction of the Symmetrized Orbitals
Int. J. Quantum Chem. 67, 299–310 (1998)
C. M. Zicovich-Wilson and R. Dovesi
On the use of Symmetry Adapted Crystalline Orbitals in SCF-LCAO periodic calculations. II.
Implementation of the Self-Consistent-Field scheme and examples
Int. J. Quantum Chem. 67, 311–320 (1998).
The following table presents the performance obtained with the new method. In all cases
convergence is reached in ten cycles.

System Chabazite Pyrope Faujasite

Space Group R3̄m Ia3d F d3m

N. of atoms 36 80 144
N. of AOs 432 1200 1728
N. symmetry operators 12 6 3 48 48
CPU time (sec) on IBM RISC-6000/365
integrals 447 900 1945 4286 815
Atomic BF(ABF) scf (total) 1380 2162 4613 24143 50975
Atomic BF scf (diagonalization) 898 898 898 19833 44970
Symmetry Adapted BF (SABF) scf (total) 526 1391 4335 3394 2729
Symmetry Adapted BF scf (diagonalization) 42 97 570 312 523
ABF/SABF scf time 2.62 1.55 1.06 7.11 18.7

226
11.7 Reciprocal space integration
The integration in reciprocal space is an important aspect of ab initio calculations for periodic
structures. The problem arises at each stage of the self-consistent procedure, when determining
the Fermi energy, F , when reconstructing the one-electron density matrix, and, after self-
consistency is reached, when calculating the density of states (DOS) and a number of observable
quantities. The P matrix in direct space is computed following equation 11.9. The technique
adopted to compute F and the P matrix in the SCF step is described in reference [46].
The Fourier-Legendre technique presented in Chapter II.6 of reference [103] is adopted in the
calculation of total and projected DOS. The Fermi energy and the integral in equation 11.9 are
evaluated starting from the knowledge of the eigenvalues, n (k) and the eigenvectors, aµn (k),
at a certain set of sampling points, {κ}. In 3D crystals, the sampling points belong to a lattice
(called the Monkhorst net, [85] ) with basis vectors b1 /s1 , b2 /s2 , b3 /s3 , where b1 , b2 and b3
are the ordinary reciprocal lattice vectors; s1 , s2 and s3 (input as IS1, IS2 and IS3) are integer
shrinking factors. Unless otherwise specified, IS1=IS2=IS3=IS. In 2D crystals, IS3 is set equal
to 1; in 1D crystals both IS2 and IS3 are set equal to 1. Only points of the Monkhorst net
belonging to the irreducible part of the Brillouin Zone (BZ) are considered, with associated
geometrical weights, wi .
In the selection of the κ points for non-centrosymmetric crystal, time-reversal symmetry is
exploited (n (κ) = n (−κ)).
The number of inequivalent sampling points, κi , is asymptotically given by the product of the
shrinking factors divided by the order of the point group. In high symmetry systems and with
small si values, it may be considerably larger because many points lie on symmetry planes or
axes.
Two completely different situations (which are automatically identified by the code) must now
be considered, depending on whether the system is an insulator (or zero gap semiconductor), or
a conductor. In the former case, all bands are either fully occupied or vacant. The identification
of F is elementary, and the Fourier transform expressed by equation 11.9 is reduced to a
weighted sum of the integrand function over the set {κi } with weights wi , the sum over n being
limited to occupied bands.
The case of conductors is more complicated; an additional parameter, ISP, enter into play.
ISP (or ISP1, ISP2, ISP3) are Gilat shrinking factors which define a net Gilat net [57, 56]
completely analogous to the Monkhorst net. The value of ISP is larger than IS (by up to a
factor of 2), giving a denser net.
In high symmetry systems, it is convenient to assign IS magic values such that all low multi-
plicity (high symmetry) points belong to the Monkhorst lattice. Although this choice does not
correspond to maximum efficiency, it gives a safer estimate of the integral.
The value assigned to ISP is irrelevant for non-conductors. However, a non-conductor may
give rise to a conducting structure at the initial stages of the SCF cycle, owing, for instance,
to a very unbalanced initial guess of the density matrix. The ISP parameter must therefore be
defined in all cases.

11.8 Electron momentum density

and related quantities
Three functions may be computed which have the same information content but different use
in the discussion of theoretical and experimental results; the momentum density itself, ρ(p) or
EMD; the Compton profile function, J(p) or CP; the autocorrelation function, or reciprocal
space form factor, B(r) or BR.
With reference to a Crystalline-Orbital (CO)-LCAO wave function, the EMD can be expressed
as the sum of the squared moduli of the occupied COs in a momentum representation, or equiv-
alently, as the diagonal element of the six-dimensional Fourier transform of the one electron
density matrix from configuration to momentum space:

227
 XZ
ρ(p) = 1/VBZ dk |ψj (k, p)|2 θ[F − j (k)] = (11.22)
j BZ
X X
= e−ip·(sµ −sν ) aµj (p0 )a∗νj (p0 )χµ (p)χ∗ν (p) θ[F − j (p0 )] (11.23)
j µν

Z
0
ρ(p) = N −1 drdr0 e−ip·(r −r) ρ(r − r0 ) (11.24)
XX g
= Pµν e−ip·(g+sµ −sν ) χµ (p)χ∗ν (p) (11.25)
µν g

In the above equations p0 is the value of momentum in the Brillouin zone (BZ), which is related
to p by a reciprocal lattice vector K, sµ is the fractional coordinate of the χµ centre, and χµ (p)
is the Fourier transform of χµ (r), calculated analytically:
Z
χµ (p) = drχµ (r) eip·r (11.26)

Equation 11.25 is used by default to compute the core band contribution, and equation 11.23
the valence band contribution.
The CPs are obtained by 2D integration of the EMD over a plane through p and perpendicular
to the p direction. After indicating with p⊥ the general vector perpendicular to p, we have:
Z
J(p) = dp0⊥ ρ(p + p0⊥ ) (11.27)

It is customary to make reference to CPs as functions of a single variable p, with reference to

a particular direction < hkl > identified by a vector

e = (ha1 + ka2 + la3 )/|(ha1 + ka2 + la3 )|

We have:
J<hkl> (p) = J(p e) (11.28)
The function J<hkl> (p) will be referred to as directional CPs.
The weighted average of the directional CPs over all directions is the average CP.

In the so called impulse approximation, J<hkl> (p) may be related to the experimental CPs,
after correction for the effect of limited resolution as a convolution of the ”infinite resolution”
0
results, J<hkl> (p), with a normalized function characterized by a given standard deviation σ:
Z +∞
σ
J<hkl> (p) = dp0 J<hkl>
0
(p0 )gσ (p − p0 ) (11.29)
−∞

In CRYSTAL gσ is a gaussian function with standard deviation σ. Once the directional CPs are
available, the numerical evaluation of the corresponding autocorrelation function, or reciprocal
space form factor, B(r) is given by the 1D Fourier Transform:
Z +∞
1
B<hkl> (r) = dpJ<hkl> (p)ei pr (11.30)
2π −∞
The average Compton profile can be evaluated from the average EMD:
Z ∞
J(q) = 2π ρ(p)pdp (11.31)
q

and can be used for the evaluation of the kinetic energy:

Z ∞
KE = p2 J(p)dp (11.32)
0

228
11.9 Elastic Moduli of Periodic Systems
The elastic constants are second derivatives of the energy density with respect to strain com-
ponents:
∂2E
Cij = 1/V · (11.33)
∂i ∂j
where V is the volume of the cell. The energy derivatives must be evaluated numerically.
Particular care is required in the selection of the computational parameters and of the points
where the energy is evaluated, in order to avoid large numerical errors in the fitting procedure
(FIXINDEX, page 86; OPTGEOM, page 105).
See http://www.crystal.unito.it ⇒ tutorials ⇒ Elastic and piezoelectric tensors
When the unit cell is deformed, the point group is reduced to a subgroup of the original
point group (see examples below). The new point group is automatically selected by the code.
Off-diagonal (partial derivatives) elastic constants can be computed as linear combinations
of single-variable energy curves. For example, for a cubic system, C12 can be obtained from
B=(C11 + 2C12 )/3 and (C11 − C12 ) (see examples below). Following the deformation of the
unit cell, internal relaxation of the atoms may be necessary (depending on the space group
symmetry) See test 20, referring to Li2 O.
The analysis of the point group at the atomic positions (printed by the ATOMSYMM option,
page 34) is useful in finding the atomic coordinates to be relaxed. Examples of deformation
strategies are discussed in references [41, 23].
In a crystalline system a point r is usually defined in terms of its fractional components:

r = h Lp
where :    
l1 l1x l1y l1z
Lp =  l2  =  l2x l2y l2z  (11.34)
l3 l3x l3y l3z
V = det(Lp )
l1 , l2 , l3 are the fundamental vectors of the primitive cell, h is the fractional vector and V the
cell volume.
Lp can be computed from the six cell parameters a, b, c, α, β, γ. For instance, the matrix Lp
for a face centred cubic lattice with lattice parameter a has the form:
 
0 a/2 a/2
Lp =  a/2 0 a/2 
a/2 a/2 0
Under an elastic strain, any particle at r migrates microscopically to r0 according to the relation:

r0 = r (I + )

where  is the symmetric Lagrangian elastic tensor.

In the deformed crystalline system:

r0 = h L0p
L0p = (I + )Lp (11.35)
or:
L0p = Lp + Z (11.36)
where
Z =  Lp
V 0 = det(L0p )

229
The deformation may be constrained to be volume-conserving, in which case the lattice vectors
of the distorted cell must be scaled as follows:
˙ /V 0 )1/3
Lp ” = L0p (V (11.37)

If a non-symmetric Lagrangian elastic tensor, η, is used, instead of , the deformation is the

sum of a strain () and a rotation (ω) of the crystal:

 = (η + η + )/2

ω = (η − η + )/2
The total energy of the crystal is invariant to a pure rotation, which allows non-symmetric η
matrices to be employed. However, a non-symmetric deformation will lower the symmetry of
the system, and therefore increase the complexity of the calculation, since the cost required is
roughly inversely proportional to the order of the point group.
The elastic constants of a crystal are defined as the second derivatives of the energy with
respect to the elements of the infinitesimal Lagrangian strain tensor .
Let us define, according to the Voigt convention:

1 = 11 4 = 32 + 23

2 = 22 5 = 13 + 31
3 = 33 6 = 12 + 21

A Taylor expansion of the energy of the unit cell to second order in the strain components
yields:
6 6
X ∂E X ∂2E
E() = E(0) + i + 1/2 i j (11.38)
i
∂i i,j
∂i ∂j

If E(0) refers to the equilibrium configuration the first derivative is zero, since there is no
force on any atom in equilibrium. The elastic constants of the system can be obtained by
evaluating the energy as a function of deformations of the unit cell parameters. The indices
of the non-zero element(s) (in the Voigt convention) of the  matrix give the corresponding
elastic constants.

Examples of  matrices for cubic systems

Consider a face-centred cubic system, for example Li2 O, with the Fm3m space group.
For cubic systems there are only three independent elastic constants (C11 , C12 and C44 ), as the
symmetry analysis shows that:

C11 = C22 = C33 ;

C44 = C55 = C66 ;
C12 = C13 = C23 ;
Cij =0 for i = 1, 6, j = 4, 6 and i 6= j.

Calculation of C11
The  matrix for the calculation of C11 is
 
δ 0 0
= 0 0 0 
0 0 0
The energy expression is:
∂2E 2
E(δ) = E(0) + 1/2 δ + · · · = a + bδ 2 + cδ 3 · · ·
∂21

230
where a, b, c are the coefficients of a polynomial fit of E versus δ, usually truncated to fourth
order (see examples below). Then

∂2E 2b
C11 = 1/V 2 =
∂1 V

The above distortion reduces the number of point symmetry operators to 12 (tetragonal dis-
tortion).

Calculation of C11 − C12

The  matrix for the calculation of the C11 − C12 combination has the form:
 
δ 0 0
 =  0 −δ 0 
0 0 0
The energy expression is:

∂2E 2 ∂2E 2 ∂2E 2

E(1 , 2 ) = E(0, 0) + 1/2 δ + 1/2 δ − δ + ··· =
∂21 ∂22 ∂1 ∂2
= E(0, 0) + V (C11 − C12 )δ 2 + · · · = a + bδ 2 + · · ·

Then C11 − C12 = b/V

With the previous form of the  matrix the number of point symmetry operators is reduced to
8, whereas the following  matrix reduces the number of point symmetry operators to 16:
 
δ 0 0
= 0 δ 0 
0 0 −2δ

E(1 , 2 , 3 ) = E(0, 0, 0) + 3V (C11 − C12 )δ 2 + · · · = a + bδ 2 + · · ·

and (C11 − C12 ) = b/3V

Calculation of C44
Monoclinic deformation, 4 point symmetry operators.

The  matrix has the form:

 
0 0 0
= 0 0 x 
0 x 0
The energy expression is (δ = 2x) (see Voigt convention and equation 11.38)

∂2E 2 ∂2E 2
E(4 ) = E(0) + 1/2 δ + · · · = E(0) + 2 x + · · · = a + bx2 + · · ·
∂24 ∂24
so that C44 = b/2V .

231
Calculation of C44
Rhombohedral deformation, 12 point symmetry operators.

The  matrix has the form:

 
0 x x
= x 0 x 
x x 0
The energy expression is (δ = 2x, C45 = C46 = C56 = 0)

∂2E 2 ∂2E
E(4 , 5 , 6 ) = E(0) + 3/2 2 δ + · · · = E(0) + 6 2 x2 + · · · = a + bx2 + · · ·
∂4 ∂4
so that C44 = b/6V .

Bulk modulus
The bulk modulus can be evaluated simply by varying the lattice constant, (1 in cubic systems)
without the use of the  matrix, and fitting the curve E(V ).
If the  matrix is used, the relation between B and Cij (cubic systems) must be taken into
account:

B = (C11 + 2C12 )/3

The  matrix has the form:
 
δ 0 0
= 0 δ 0 
0 0 δ
and the energy:

∂2E 2 ∂2E 2
E() = E(0) + 3/2 2 δ +3 δ = (11.39)
∂1 ∂1 ∂2
3V
= E(0) + [C11 + 2C12 ]δ 2 (11.40)
2
2
so that B = 9V b

N.B. Conversion factors:

1 hartree Å−3 = 4359.74812 GPa
1 GPa = 1 GN m−2 = 1 GJ m−3 = 1010 dyne cm−2 = 10−2 Mbar.

232
11.10 Spontaneous polarization through the Berry phase
approach
(λ,1)
The electronic phase of a system λ in the direction 1, ϕel , can be written as:

(λ,1) 1 XX (λ,1)
ϕel = ∆ϕj1 ,j2 ,j3 (k) (11.41)
s2s3 j2,j3 j
1

The electronic contribution to the polarization of a system λ can be written as :

(λ) 1  (λ) −1 (λ)

Pel = B ϕel (11.42)
Ω(λ)
(λ)
Where (B (λ) )−1 is the reciprocal lattice vectors components inverse matrix and ϕel the elec-
(λ,i)
tronic phase difference vector of a system λ (which components are ϕel ). The nuclear con-
(λ)
tribution to the polarization of a system λ, Pnuc can also be written as:
1 X (λ)
P(λ)
nuc = R · ZA (11.43)
Ω(λ) A A

(λ)
where RA and ZA are the position vector and the nuclear charge of the atom A respectively of
the system λ. The total polarization is the sum of these two contributions and can be written
as
(λ) (λ)
Ptot = P(λ)
nuc + Pel (11.44)
The spontaneous polarization is the difference between the systems λ = 1 and λ = 0
(λ) (λ)
P = Ptot − Ptot (11.45)

Spontaneous polarization through the localized crystalline

orbitals approach
(λ)
The electronic contribution to the polarization of a system λ, Pel , can be written as
(λ) e X
Pel = hrµ i (11.46)
Ω(λ) µ

Where hrµ i is the centroid of the Wannier function µ.

(λ)
The nuclear contribution to the polarization of a system λ, Pnuc can also be written as

(λ) 1 X
Pnuc = RA · ZA (11.47)
Ω(λ) A

where RA and ZA are the position vector and the nuclear charge of the atom A respectively.
The total polarization is the sum of these two contributions and can be written as
(λ) (λ)
Ptot = P(λ)
nuc + Pel (11.48)

The spontaneous polarization is the difference between the both systems λ = 1 and λ = 0:
(1) (2)
P = Ptot − Ptot (11.49)

To calculate the spontaneous polarization, a preliminary run is needed for each of the two
systems λ = 1 and λ = 0. Then a third run with the keyword SPOLWF gives the difference of
polarization between systems λ = 1 and λ = 0.

233
11.11 Piezoelectricity through the Berry phase approach
The electronic phase vector of a system λ, is given by (2.1). The nuclear phase vector of a
(λ)
system λ, ϕnuc , can be written as
ϕ(λ)
nuc = Ω
(λ)
B (λ) Pnuc
(λ)
(11.50)
Where B (λ) reciprocal lattice vectors components matrix.The last equation can be simplified
thanks to (11.43): X (λ)
ϕ(λ)
nuc = B
(λ)
RA · ZA (11.51)
A
(λ)
So the phase vector of a system λ, ϕ is:
(λ)
ϕ(λ) = ϕ(λ)
nuc + ϕel (11.52)
The proper piezoelectric constants can be obtained by:
1 1 X dϕα
ẽijk = − aα,i (11.53)
2π Ω α djk

Where ϕα is projection of the phase ϕ along the α direction and aα,i is the component of a
lattice vector aα along the cartesian axis i . To obtain the improper piezoelectric constants,
the following correction must done:
eijk = ẽijk + δij Pk − δjk Pi (11.54)
dϕα
In the piezoelectric constants calculations the djk
term is evaluated numerically. The calcu-
lated term is:
(1) (0)
dϕα ∆ϕα ϕα − ϕα
' = (1) (0)
(11.55)
djk ∆jk jk − jk

Piezoelectricity through the localized crystalline orbitals

approach
The electronic phase vector of a system λ, is given by:
(λ) (λ)
ϕel = Ω(λ) B (λ) Pel (11.56)
Where B (λ) reciprocal lattice vectors components matrix. The nuclear phase vector of a system
(λ)
λ, ϕnuc , can be written as
ϕ(λ)
nuc = Ω
(λ)
B (λ) Pnuc
(λ)
(11.57)
The last equation can be simplified thanks to 11.43:
X (λ)
(λ)
ϕnuc = B (λ) RA · ZA (11.58)
A

So the phase vector of a system λ, ϕ(λ) is:

(λ)
ϕ(λ) = ϕnuc
(λ)
+ ϕel (11.59)
The proper piezoelectric constants can be obtained by:
1 1 X dϕα
ẽijk = − aα,i (11.60)
2π Ω α djk

Where ϕα is projection of the phase ϕ along the α direction and aα,i is the component of a
lattice vector aα along the cartesian axis i . To obtain the improper piezoelectric constants,
the following correction must done:
eijk = ẽijk + δij Pk − δjk Pi (11.61)

234
 dϕα
In the piezoelectric constants calculations the djk
term is evaluated numerically. The calcu-
lated term is:
(1) (0)
dϕα ∆ϕα ϕα − ϕα
' = (1) (0)
(11.62)
djk ∆jk jk − jk

235
Appendix A

Symmetry groups

A.1 Labels and symbols of the space groups

The labels are according to the International Tables for Crystallography [64]. The symbols are
derived by the standard SHORT symbols, as shown in the following examples:

Symbol Input to CRYSTAL

P 6̄ 2 m → P -6 2 M ;
P 63 m → P 63 M.

For the groups 221-230 the symbols are according to the 1952 edition of the International
Tables, not to the 1982 edition. The difference involves the 3 axis: 3 (1952 edition); 3̄ (1982
edition) (Example group 221: 1952 ed. → P m 3 m ; 1982 ed. → P m 3̄ m)

236
IGR symbol IGR symbol IGR symbol
Triclinic lattices 37 Ccc2 Tetragonal lattices
1 P1 38 Amm2 75 P4
2 P 1̄ 39 Abm2 76 P 41
Monoclinic lattices 40 Ama2 77 P 42
3 P2 41 Aba2 78 P 43
4 P 21 42 F mm2 79 I4
5 C2 43 F dd2 80 I41
6 Pm 44 Imm2 81 P 4̄
7 Pc 45 Iba2 82 I 4̄
8 Cm 46 Ima2 83 P 4/m
9 Cc 47 P mmm 84 P 42 /m
10 P 2/m 48 P nnn 85 P 4/n
11 P 21 /m 49 P ccm 86 P 42 /n
12 C2/m 50 P ban 87 I4/m
13 P 2/c 51 P mma 88 I41 /a
14 P 21 /c 52 P nna 89 P 422
15 C2/c 53 P mna 90 P 421 2
Orthorhombic lattices 54 P cca 91 P 41 22
16 P 222 55 P bam 92 P 41 21 2
17 P 2221 56 P ccn 93 P 42 22
18 P 21 21 2 57 P bcm 94 P 42 21 2
19 P 21 21 21 58 P nnm 95 P 43 22
20 C2221 59 P mmn 96 P 43 21 2
21 C222 60 P bcn 97 I422
22 F 222 61 P bca 98 I41 22
23 I222 62 P nma 99 P 4mm
24 I21 21 21 63 Cmcm 100 P 4bm
25 P mm2 64 Cmca 101 P 42 cm
26 P mc21 65 Cmmm 102 P 42 nm
27 P cc2 66 Cccm 103 P 4cc
28 P ma2 67 Cmma 104 P 4nc
29 P ca21 68 Ccca 105 P 42 mc
30 P nc2 69 F mmm 106 P 42 bc
31 P mn21 70 F ddd 107 I4mm
32 P ba2 71 Immm 108 I4cm
33 P na21 72 Ibam 109 I41 md
34 P nn2 73 Ibca 110 I41 cd
35 Cmm2 74 Imma 111 P 4̄2m
36 Cmc21 112 P 4̄2c

237
 IGR symbol
IGR symbol
IGR symbol Cubic lattices
113 P 4̄21 m
155 R32 195 P 23
114 P 4̄21 c
156 P 3m1 196 F 23
115 P 4̄m2
157 P 31m 197 I23
116 P 4̄c2
158 P 3c1 198 P 21 3
117 P 4̄b2
159 P 31c 199 I21 3
118 P 4̄n2
160 R3m 200 P m3̄
119 I 4̄m2
161 R3c 201 P n3̄
120 I 4̄c2
162 P 3̄1m 202 F m3̄
121 I 4̄2m
163 P 3̄1c 203 F d3̄
122 I 4̄2d
164 P 3̄m1 204 Im3̄
123 P 4/mmm
165 P 3̄c1 205 P a3̄
124 P 4/mcc
166 R3̄m 206 Ia3̄
125 P 4/nbm
167 R3̄c 207 P 432
126 P 4/nnc
Hexagonal lattices 208 P 42 32
127 P 4/mbm
168 P6 209 F 432
128 P 4/mnc
169 P 61 210 F 41 32
129 P 4/nmm
170 P 65 211 I432
130 P 4/ncc
171 P 62 212 P 43 32
131 P 42 /mmc
172 P 64 213 P 41 32
132 P 42 /mcm
173 P 63 214 I41 32
133 P 42 /nbc
174 P 6̄ 215 P 4̄3m
134 P 42 /nnm
175 P 6/m 216 F 4̄3m
135 P 42 /mbc
176 P 63 /m 217 I 4̄3m
136 P 42 /mnm
177 P 622 218 P 4̄3n
137 P 42 /nmc
178 P 61 22 219 F 4̄3c
138 P 42 /ncm
179 P 65 22 220 I 4̄3d
139 I4/mmm
180 P 62 22 221 P m3̄m
140 I4/mcm
181 P 64 22 222 P n3̄n
141 I41 /amd
182 P 63 22 223 P m3̄n
142 I41 /acd
183 P 6mm 224 P n3̄m
Trigonal lattices
184 P 6cc 225 F m3̄m
143 P3
185 P 63 cm 226 F m3̄c
144 P 31
186 P 63 mc 227 F d3̄m
145 P 32
187 P 6̄m2 228 F d3̄c
146 R3
188 P 6̄c2 229 Im3̄m
147 P 3̄
189 P 6̄2m 230 Ia3̄d
148 R3̄
190 P 6̄2c
149 P 312
191 P 6/mmm
150 P 321
192 P 6/mcc
151 P 31 12
193 P 63 /mcm
152 P 31 21
194 P 63 /mmc
153 P 32 12
154 P 32 21

238
A.2 Labels of the layer groups (slabs)
The available layer groups belong to a subset of the 230 space groups. Therefore they can be
identified by the corresponding space group.
The first column gives the label to be used in the input card (IGR variable).
The second column gives the Hermann-Mauguin symbol of the corresponding space group
(generally the short one; the full symbol is adopted when the same short symbol could refer
to different settings). The third column gives the Schoenflies symbol. The fourth column the
number of the corresponding space group, according to the International Tables for Crystal-
lography. The number of the space group is written in parentheses when the orientation of the
symmetry operators does not correspond to the first setting in the I. T.
IGR Hermann Schoenflies N IGR Hermann Schoenflies N
Mauguin Mauguin
9
Oblique lattices (P) 41 P bam D2h 55
3
1 P1 C11 1 42 P maa D2h (49)
7
2 P 1̄ Ci1 2 43 P man D2h (53)
1
3 P 112 C21 (3) 44 P bma D2h (57)
8
4 P 11m Cs1 (6) 45 P baa D2h (54)
4
5 P 11a Cs2 (7) 46 P ban D2h 50
1 9
6 P 112/m C2h (10) 47 Cmmm D2h 65
4 2
7 P 112/a C2h (13) 48 Cmma D2h 67
Rectangular lattices (P or C) Square lattices (P)
8 P 211 C21 (3) 49 P 4 C41 75
9 P 21 11 C22 (4) 50 P 4̄ S41 81
1
10 C211 C23 (5) 51 P 4/m C4h 83
3
11 P m11 Cs1 (6) 52 P 4/n C4h 85
12 P b11 Cs2 (7) 53 P 422 D41 89
13 Cm11 Cs3 (8) 54 P 421 2 D42 90
1 1
14 P 2/m11 C2h (10) 55 P 4mm C4v 99
2 2
15 P 21 /m11 C2h (11) 56 P 4bm C4v 100
3 1
16 C2/m11 C2h (12) 57 P 4̄2m D2d 111
4 3
17 P 2/b11 C2h (13) 58 P 4̄21 m D2d 113
5 5
18 P 2/b11 C2h (14) 59 P 4̄m2 D2d 115
7
19 P 222 D21 16 60 P 4̄b2 D2d 117
1
20 P 221 2 D22 (17) 61 P 4/mmm D4h 123
3
21 P 21 21 2 D23 18 62 P 4/nbm D4h 125
5
22 C222 D26 21 63 P 4/mbm D4h 127
1 7
23 P mm2 C2v 25 64 P 4/nmm D4h 129
4
24 P ma2 C2v 28 Hexagonal lattices (P)
8
25 P ba2 C2v 32 65 P 3 C31 143
1 1
26 Cmm2 C2v 35 66 P 3̄ C3i 147
1
27 P 2mm C2v (25) 67 P 312 D31 149
2
28 P 21 am C2v (26) 68 P 321 D32 150
2 1
29 P 21 ma C2v (26) 69 P 3m1 C3v 156
4 2
30 P 2mb C2v (28) 70 P 31m C3v 157
7 1
31 P 21 mn C2v (31) 71 P 3̄1m D3d 162
3 3
32 P 2aa C2v (27) 72 P 3̄m1 D3d 164
5 1
33 P 21 ab C2v (29) 73 P 6 C6 168
6 1
34 P 2an C2v (30) 74 P 6̄ C3h 174
1 1
35 C2mm C2v (38) 75 P 6/m C6h 175
5
36 C2mb C2v (39) 76 P 622 D61 177
1 1
37 P mmm D2h 47 77 P 6mm C6v 183
5 1
38 P mam D2h (51) 78 P 6̄m2 D3h 187
5 3
39 P mma D2h 51 79 P 6̄2m D3h 189
3 1
40 P mmn D2h 59 80 P 6/mmm D6h 191

239
A.3 Labels of the rod groups (polymers)
The available rod groups belong to a subset of the 230 space groups; the symmetry operators
are generated for the space groups (principal axis z) and then rotated by 90◦ through y, to
have the polymer axis along x (CRYSTAL convention).
In the table, the first column gives the label to be used in the input card for identifying the
rod group (IGR variable).
The second column gives the ”polymer” symbol, according to the the following convention: x
is the first symmetry direction, y the second.
The third column gives the Schoenflies symbol.
The fourth column gives the Hermann-Mauguin symbol (generally the short one; the full symbol
is adopted when the same short symbol could refer to different settings) of the corresponding
space group (principal axis z).
The fifth column gives the number of the corresponding space group, according to the Interna-
tional Tables for Crystallography; this number is written in parentheses when the orientation
of the symmetry operators does not correspond to the first setting in the I. T.

”Polymer” Hermann Number of

IGR symbol Schoenflies Mauguin space group
(x direction) (z direction)
1 P1 C11 P1 1
2 P 1̄ Ci1 P 1̄ 2
3 P 211 C21 P 112 (3)
4 P 21 11 C22 P 1121 (4)
5 P 121 C21 P 121 (3)
6 P 112 C21 P 211 (3)
7 P m11 Cs1 P 11m (6)
8 P 1m1 Cs1 P 1m1 (6)
9 P 1a1 Cs2 P 1c1 (7)
10 P 11m Cs1 P m11 (6)
11 P 11a Cs2 P c11 (7)
1
12 P 2/m11 C2h P 112/m (10)
2
13 P 21 /m11 C2h P 1121 /m (11)
1
14 P 12/m1 C2h P 12/m1 (10)
4
15 P 12/a1 C2h P 12/c1 (13)
1
16 P 112/m C2h P 2/m11 (10)
4
17 P 112/a C2h P 2/c11 (13)
18 P 222 D21 P 222 16
19 P 21 22 D22 P 2221 17
1
20 P 2mm C2v P mm2 25
2
21 P 21 am C2v P mc21 26
2
22 P 21 ma C2v P cm21 (26)
3
23 P 2aa C2v P cc2 27
1
24 P m2m C2v P m2m (25)
4
25 P m2a C2v P c2m (28)
1
26 P mm2 C2v P 2mm (25)
4
27 P ma2 C2v P 2cm (28)
1
28 P mmm D2h P mmm 47
3
29 P 2/m2/a2/a D2h P ccm 49
5
30 P 21 /m2/m2/a D2h P cmm (51)
5
31 P 21 /m2/a2/m D2h P mcm (51)

240
 ”Polymer” Hermann Number of
IGR symbol Schoenflies Mauguin space group
(x direction) (z direction)
32 P4 C41 P4 75
33 P 41 C42 P 41 76
34 P 42 C43 P 42 77
35 P 43 C44 P 43 78
36 P 4̄ S41 P 4̄ 81
1
37 P 4/m C4h P 4/m 83
2
38 P 42 /m C4h P 42 /m 84
39 P 422 D41 P 422 89
40 P 41 22 D43 P 41 22 91
41 P 42 22 D45 P 42 22 93
42 P 43 22 D47 P 43 22 95
1
43 P 4mm C4v P 4mm 99
3
44 P 42 am C4v P 42 cm 101
5
45 P 4aa C4v P 4cc 103
7
46 P 42 ma C4v P 42 mc 105
1
47 P 4̄2m D2d P 4̄2m 111
2
48 P 4̄2a D2d P 4̄2c 112
5
49 P 4̄m2 D2d P 4̄m2 115
6
50 P 4̄a2 D2d P 4̄c2 116
1
51 P 4/mmm D4h P 4/mmm 123
2
52 P 4/m2/a2/a D4h P 4/mcc 124
9
53 P 42 /m2/m2/a D4h P 42 /mmc 131
10
54 P 42 /m2/a2/m D4h P 42 /mcm 132
55 P3 C31 P3 143
56 P 31 C32 P 31 144
57 P 32 C33 P 32 145
1
58 P 3̄ C3i P 3̄ 147
59 P 312 D31 P 312 149
60 P 31 12 D33 P 31 12 151
61 P 32 12 D35 P 32 12 153
62 P 321 D32 P 321 150
63 P 31 21 D34 P 31 21 152
64 P 32 21 D36 P 32 21 154
1
65 P 3m1 C3v P 3m1 156
3
66 P 3a1 C3v P 3c1 158
2
67 P 31m C3v P 31m 157
4
68 P 31a C3v P 31c 159
1
69 P 3̄1m D3d P 3̄1m 162
2
70 P 3̄1a D3d P 3̄1c 163
3
71 P 3̄m1 D3d P 3̄m1 164
4
72 P 3̄a1 D3d P 3̄c1 165

241
 ”Polymer” Hermann Number of
IGR symbol Schoenflies Mauguin space group
(x direction) (z direction)
73 P6 C61 P6 168
74 P 61 C62 P 61 169
75 P 65 C63 P 65 170
76 P 62 C64 P 62 171
77 P 64 C65 P 64 172
78 P 63 C66 P 66 173
1
79 P 6̄ C3h P 6̄ 174
1
80 P 6/m C6h P 6/m 175
2
81 P 63 /m C6h P 63 /m 176
82 P 622 D61 P 622 177
83 P 61 22 D62 P 61 22 178
84 P 65 22 D63 P 65 22 179
85 P 62 22 D64 P 62 22 180
86 P 64 22 D65 P 64 22 181
87 P 63 22 D66 P 63 22 182
1
88 P 6mm C6v P 6mm 183
2
89 P 6aa C6v P 6cc 184
3
90 P 63 am C6v P 63 cm 185
4
91 P 63 ma C6v P 63 mc 186
1
92 P 6̄m2 D3h P 6̄m2 187
2
93 P 6̄a2 D3h P 6̄c2 188
3
94 P 6̄2m D3h P 6̄2m 189
4
95 P 6̄2a D3h P 6̄2c 190
1
96 P 6/mmm D6h P 6/mmm 191
2
97 P 6/m2/a2/a D6h P 6/mcc 192
3
98 P 63 /m2/a2/m D6h P 63 /mcm 193
4
99 P 63 /m2/m2/a D6h P 63 /mmc 194

242
A.4 Labels of the point groups (molecules)
The centre of symmetry is supposed to be at the origin; for the rotation groups the principal
axis is z.
The first column gives the label to be used in the input card for identifying the point group
(IGR variable). The second column gives the short Hermann-Mauguin symbol. The third
column gives the Schoenflies symbol; for the C2 , C2h and Cs groups the C2 direction or the
direction orthogonal to the plane is indicated. The fourth column gives the number of pure
rotations for molecules (σ).

IGR Hermann Schoenflies σ

Mauguin
1 1 C1 1
2 1̄ Ci 1
3 2 (x) C2 (x) 2
4 2 (y) C2 (y) 2
5 2 (z) C2 (z) 2
6 m (x) Cs (x) 1
7 m (y) Cs (y) 1
8 m (z) Cs (z) 1
9 2/m (x) C2h (x) 2
10 2/m (y) C2h (y) 2
11 2/m (z) C2h (z) 2
12 222 D2 4
13 2mm C2v (x) 2
14 m2m C2v (y) 2
15 mm2 C2v (z) 2
16 mmm D2h 4
17 4 C4 4
18 4̄ S4 2
19 4/m C4h 4
20 422 D4 8
21 4mm C4v 4
22 4̄2m D2d (σv planes along x+y and x-y) 4
23 4̄m2 D2d (σv planes along x and y) 4
24 4/mmm D4h 8
25 3 C3 3
26 3̄ C3i 3
27 321 D3 (one C2 axis along y) 6
28 312 D3 (one C2 axis along x) 6
29 3m1 C3v (one σv plane along x) 3
30 31m C3v (one σv plane along y) 3
31 3̄m1 D3d (one σd plane along x) 6
32 3̄1m D3d (one σd plane along y) 6
33 6 C6 6
34 6̄ C3h 3
35 6/m C6h 6
36 622 D6 12
37 6mm C6v 6
38 6̄m2 D3h (one C2 axis along x) 6
39 6̄2m D3h (one C2 axis along y) 6
40 6/mmm D6h 12
41 23 T 12
42 m3̄ Th 12
43 432 O 24
44 4̄3m Td 12
45 m3̄m Oh 24

243
A.5 From conventional to primitive cells:
transforming matrices
The matrices describing the transformations from conventional (given as input) to primitive
(internally used by CRYSTAL) cells of Bravais lattices are coded in CRYSTAL. A point called x
in the direct lattice has xP coordinates in a primitive cell and xC coordinates in a conventional
cell. The relation between xP and xC is the following:

W xP = xC (A.1)

Likewise, for a point in the reciprocal space the following equation holds:

W̃ −1 x∗P = x∗C (A.2)

The W transforming matrices adopted in CRYSTAL, and reported below, satisfy the following
relation between the two metric tensors GP and GC :

GP = W GC W̃ (A.3)

The values of the elements of the metric tensors GP and GC agree with those displayed in
Table 5.1 of the International Tables of Crystallography (1992 edition).
0 1 0 1 1 1
1 0 0 2
0 2
B C B C
B 1 1̄
C B C
P →A B 0 C P →B B 0 1 0 C
B 2 2 C B C 0 1 0 1
@ A @ A 1 0 0 1 0 −1
1 1 1̄ 1
0 2 2 2
0 2 A→P @ 0 1 1 A B→P @ 0 1 0 A
0 −1 1 1 0 1
1 1̄ 1 1
0 1 0 1
2 2
0 0 2 2 0 1 0 1
B
B
C
C
B
B
C
C 1 1 0 −1 1 1
1 1 1 1
P →C B
B 2 2
0 C
C P →F B
B 2
0 2
C
C C→P @ −1 1 0 A F →P @ 1 −1 1 A
@ A @ A 0 0 1 1 1 −1
1 1
0 0 1 2 2
0
0 1 0 1
0 1 1 1 0 1
1̄ 1 1 2 1̄ 1̄ @ −1
I→P @ 1 0 1 A H→R 1 1 A
0 1 0 1
2 2 2 3 3 3
B C B C 1 1 0 0 −1 1
B 1 1̄ 1
C B C
P →I B C R→H B 1 1 2̄ C
B 2 2 2 C B 3 3 3 C
@ A @ A
1 1 1̄ 1 1 1
2 2 2 3 3 3

Table A.1: W matrices for the transformation from conventional to primitive and from prim-
itive to conventional cells. P stands for primitive, A, B and C for A-, B- and C-face centred,
I for body centred, F for all-face centred, R for primitive rhombohedral (‘rhombohedral axes’)
and H for rhombohedrally centred (‘hexagonal axes’) cell (Table 5.1, ref. [64]).

244
Appendix B

Summary of input keywords

All the keywords are entered with an A format; the keywords must be typed left-justified, with
no leading blanks. The input is not case sensitive.

Geometry (Input block 1)

Symmetry information

ATOMSYMM printing of point symmetry at the atomic positions 34 –

MAKESAED printing of symmetry allowed elastic distortions (SAED) 45 –
PRSYMDIR printing of displacement directions allowed by symmetry. 53 –
SYMMDIR printing of symmetry allowed geom opt directions 60 –
SYMMOPS printing of point symmetry operators 60 –
TENSOR print tensor of physical properties up to order 4 60 I

Symmetry information and control

BREAKELASsymmetry breaking according to a general distortion 35 I

BREAKSYM allow symmetry reduction following geometry modifications 34 –
KEEPSYMM maintain symmetry following geometry modifications 44 –
MODISYMM removal of selected symmetry operators 45 I
PURIFY cleans atomic positions so that they are fully consistent with the 53 –
group
SYMMREMO removal of all symmetry operators 60 –
TRASREMO removal of symmetry operators with translational components 60 –

Modifications without reduction of symmetry

ATOMORDE reordering of atoms in molecular crystals 32 –

NOSHIFT no shift of the origin to minimize the number of symmops with 52 –
translational components before generating supercell
ORIGIN shift of the origin to minimize the number of symmetry operators 52 –
with translational components
PRIMITIV crystallographic cell forced to be the primitive cell 52 –
ROTCRY rotation of the crystal with respect to the reference system cell 54 I

Atoms and cell manipulation (possible symmetry reduction (BREAKSYMM)

245
ATOMDISP displacement of atoms 32 I
ATOMINSE addition of atoms 32 I
ATOMREMO removal of atoms 32 I
ATOMROT rotation of groups of atoms 33 I
ATOMSUBS substitution of atoms 34 I
ELASTIC distortion of the lattice 38 I
POINTCHG point charges input 52 I
SCELPHONO generation of supercell for phonon dispersion 58 I
SUPERCEL generation of supercell - input refers to primitive cell 56 I
SUPERCON generation of supercell - input refers to conventional cell 56 I
USESAED given symmetry allowed elastic distortions, reads δ 61 I

From crystals to slabs (3D→2D)

SLABINFO definition of a new cell, with xy k to a given plane 56 I

SLABCUT generation of a slab parallel to a given plane (3D→2D) 55 I

From slabs to nanotubes (2D→1D)

NANOTUBE building a nanotube from a slab 47 I

SWCNT building a nanotube from an hexagonal slab 50 I

From periodic structure to clusters

CLUSTER cutting of a cluster from a periodic structure (3D→0D) 36 I

HYDROSUB border atoms substituted with hydrogens (0D→0D) 44 I

Molecular crystals

MOLECULE extraction of a set of molecules from a molecular crystal 46 I

(3D→0D)
MOLEXP variation of lattice parameters at constant symmetry and molec- 46 I
ular geometry (3D→3D)
MOLSPLIT periodic structure of non interacting molecules (3D→3D) 47 –
RAYCOV modification of atomic covalent radii 53 I

BSSE correction
MOLEBSSE counterpoise method for molecules (molecular crystals only) 45 I
(3D→0D)
ATOMBSSE counterpoise method for atoms (3D→0D) 31 I

Auxiliary and control keywords

ANGSTROM sets input units to Ångstrom 31 –

BOHR sets input units to bohr 34 –
BOHRANGS input bohr to Å conversion factor (0.5291772083 default value) 34 I
BOHRCR98 bohr to Å conversion factor is set to 0.529177 (CRY98 value) –
END/ENDG terminate processing of geometry input –
FRACTION sets input units to fractional 44 –
NEIGHBOR number of neighbours in geometry analysis 51 I
PARAMPRT printing of parameters (dimensions of static allocation arrays) 52 –
PRINTCHG printing of point charges coordinates in geometry output 52
PRINTOUT setting of printing options by keywords 53 –
SETINF setting of inf array options 54 I
SETPRINT setting of printing options 54 I
STOP execution stops immediately 56 –
TESTGEOM stop after checking the geometry input 60 –

Output of data on external units

246
COORPRT coordinates of all the atoms in the cell 37 –
EXTPRT write file in CRYSTAL geometry input format 41 –
FINDSYM write file in FINDSYM input format 44 –
STRUCPRT cell parameters and coordinates of all the atoms in the cell 56 –

External electric field - modified Hamiltonian

FIELD electric field applied along a periodic direction 41 I
FIELDCON electric field applied along a non periodic direction 43 I

Geometry optimization - see index for keywords full list

OPTGEOM Geometry optimization input block - closed by END 105 I

Type of optimization (default: atom coordinates)
FULLOPTG full geometry optimization –
CELLONLY cell parameters optimization –
INTREDUN optimization in redundant internal coordinates 115 –
ITATOCEL iterative optimization (atom/cell) –
CVOLOPT full geometry optimization at constant volume 117 –
Initial Hessian
HESSIDEN initial guess for the Hessian - identity matrix –
HESSMOD1 initial guess for the Hessian - model 1 (default) –
HESSMOD2 initial guess for the Hessian - model 2 –
HESSNUM initial guess for the Hessian - numerical estimate –
Convergence criteria modification
TOLDEG RMS of the gradient [0.0003] I
TOLDEX RMS of the displacement [0.0012] I
TOLDEE energy difference between two steps [10−7 ] I
MAXCYCLE max number of optimization steps I
Optimization control
FRAGMENT partial geometry optimization 119 I
RESTART data from previous run –
FINALRUN Wf single point with optimized geometry I
Gradient calculation control
NUMGRATO numerical atoms first derivatives 114 –
NUMGRCEL numerical cell first derivatives 114 –
NUMGRALL numerical atoms and cell first derivatives 114 –
External stress
EXTSTRESS apply external stress 121 I
EXTPRESS apply external hydrostatic pressure 121 I
Printing options
PRINTFORCES atomic gradients –
PRINTHESS Hessian –
PRINTOPT optimization procedure –
PRINT verbose printing –

Frequencies at Γ - Harmonic calculation

247
FREQCALC Frequencies at Γ (harmonic calculation) input block - closed by END 127 I
ANALYSIS –
[NOANALYSIS] –
DIELISO I
DIELTENS I
END[FREQ] –
FRAGMENT I
INTENS –
INTLOC - IR intensities through Wannier functions 134 –
[INTPOL] - IR intensities through Berry phase [default] 134 –
ISOTOPES 129 I
[MODES] –
[NOINTENS] –
NOMODES –
NORMBORN –
NOUSESYMM –
NUMDERIV I
PRESSURE I
PRINT –
RESTART –
SCANMODE I
STEPSIZE I
TEMPERAT I
TESTFREQ –
[USESYMM] –

ANHARM Frequency at Γ (anharmonic calculationinput block - closed by END142 I

END[ANHA] –
ISOTOPES modification of atomic mass 143 I
KEEPSYMM 44 –
NOGUESS 143 –
POINTS26 143 –
PRINT –
TEST[ANHA] –

CONFCNT - Configuration counting and cluster expansion 147

CPHF - Coupled Perturbed Hartree-Fock 145

ELASTCON - Second order elastic constants 149

EOS - Equation of state 145

Basis set input (Input block 2)

Symmetry control

ATOMSYMM printing of point symmetry at the atomic positions 34 –

Basis set modification

248
CHEMOD modification of the electronic configuration 62 I
GHOSTS eliminates nuclei and electrons, leaving BS 64 I

Auxiliary and control keywords

CHARGED allows non-neutral cell 62 –

NOPRINT printing of basis set removed 64 –
PARAMPRT printing of parameters controlling code dimensions 52 –
PRINTOUT setting of printing options 53 I
SETINF setting of inf array options 54 I
SETPRINT setting of printing options 54 I
STOP execution stops immediately 56 –
SYMMOPS printing of point symmetry operators 60 –
END/ENDB terminate processing of basis set definition keywords –

Output of data on external units

GAUSS98 printing of an input file for the GAUSS94/98 package 64 –

General information, hamiltonian, SCF (Input block 3)

All DFT related keyword are collected under the heading ”DFT”, closed b END[DFT]

Single particle Hamiltonian

RHF Restricted Closed Shell 96 –

UHF Unrestricted Open Shell 102 –

249
DFT DFT Hamiltonian 102 –
SPIN spin-polarized solution 75 –
Choice of the exchange-correlation functionals
CORRELAT correlation functional 74 I
LYP Lee-Yang-Parr [78] (GGA)
P86 Perdew 86 [88] (LDA)
PBE Perdew-Becke-Ernzerhof [90] (GGA)
PBESOL GGA. PBE functional revised for solids [96]
PWGGA Perdew-Wang 91 (GGA)
PWLSD Perdew-Wang 92 [92, 93, 91] (GGA)
PZ Perdew-Zunger [94] (LDA)
VBH von Barth-Hedin [127] (LDA)
WL GGA - Wilson-Levy [133]
VWN Vosko,-Wilk-Nusair [128] (LDA)
EXCHANGE exchange functional 75 I
BECKE Becke [13] (GGA)
LDA Dirac-Slater [34] (LDA)
PBE Perdew-Becke-Ernzerhof [90] (GGA)
PBESOL GGA. PBE functional revised for solids [96]
PWGGA Perdew-Wang 91 (GGA)
SOGGA second order GGA. [136]
VBH von Barth-Hedin [127] (LDA)
WCGGA GGA - Wu-Cohen [135]
Global keywords: exchange+correlation
B3PW B3PW parameterization 76 –
B3LYP B3LYP parameterization 76 –
PBE0 Adamo and Barone [2] 76
SOGGAXC see [136] 76
Custom made dft exchange-local exchange
HYBRID hybrid mixing 76 I
NONLOCAL local term parameterization 76 I
Numerical accuracy control
ANGULAR definition of angular grid I
BATCHPNT grid point grouping for integration I
[BECKE] selection of Becke weights (default) –
[CHUNKS] max n. points in a batch for numerical int. I
LGRID ”large” predefined grid I
RADSAFE safety radius for grid point screening I
RADIAL definition of radial grid I
SAVIN selection of Savin weights –
TOLLDENS density contribution screening 6 I
TOLLGRID grid points screening 14 I
XLGRID ”extra large” predefined grid I
XXLGRID ”extra extra large” predefined grid I
Atomic parameters control
RADIUS customized atomic radius 81 I
FCHARGE customized formal atomic charge 81 I
Auxiliary
PRINT extended printing
END close DFT input block

Numerical accuracy and computational parameters control

250
BIPOLAR Bipolar expansion of bielectronic integrals 73 I
BIPOSIZE size of coulomb bipolar expansion buffer 73 I
EXCHSIZE size of exchange bipolar expansion buffer 73 I
INTGPACK Choice of integrals package 0 92 I
NOBIPCOU Coulomb bielectronic integrals computed exactly 95 –
NOBIPEXCH Exchange bielectronic integrals computed exactly 95 –
NOBIPOLA All bielectronic integrals computed exactly 95 –
POLEORDR Maximum order of multipolar expansion 4 96 I
TOLINTEG Truncation criteria for bielectronic integrals 6 6 6 6 12 102 I
TOLPSEUD Pseudopotential tolerance 6 102 I

Type of execution
MPP MPP execution 93

Type of run

ATOMHF Atomic wave functions 72 I

SCFDIR SCF direct (mono+biel int computed) 96 –
EIGS S(k) eigenvalues - basis set linear dependence check 81 –
FIXINDEX Reference geometry to classify integrals 86 –

Basis set - AO occupancy

FDAOSYM f and d degeneracies analysis 83 I

FDAOCCUP f and d orbital occupation guess 84 I
GUESDUAL Density matrix guess - different Basis set 89 I

Integral file distribution

BIESPLIT writing of bielectronic integrals in n files n = 1 ,max=10 73 I

MONSPLIT writing of mono-electronic integrals in n file n = 1 , max=10 93 I

Numerical accuracy control and convergence tools

ANDERSON Fock matrix mixing 71 I

BROYDEN Fock matrix mixing 73 I
FMIXING Fock/KS matrix (cycle i and i-1) mixing 0 87 I
LEVSHIFT level shifter no 92 I
MAXCYCLE maximum number of cycles 50 93 I
SMEAR Finite temperature smearing of the Fermi surface no 99 I
TOLDEE convergence on total energy 5 102 I
TOLDEP convergence on density matrix 16 102 I

Initial guess

EIGSHIFT alteration of orbital occupation before SCF no 82 I

EIGSHROT rotation of the reference frame no 82 I
GUESSP density matrix from a previous run 91 –
GUESSPAT superposition of atomic densities 91 –

Spin-polarized system

ATOMSPIN setting of atomic spin to compute atomic densities 72 I

BETALOCK beta electrons locking 73 I
SPINLOCK spin difference locking 101 I
SPINEDIT editing of the spin density matrix used as SCF guess 100 I

Auxiliary and control keywords

251
END terminate processing of block3 input –
KSYMMPRT printing of Bloch functions symmetry analysis 92 –
NEIGHBOR number of neighbours to analyse in PPAN 51 I
PARAMPRT output of parameters controlling code dimensions 52 –
PRINTOUT setting of printing options 53 I
NOSYMADA No Symmetry Adapted Bloch Functions 95 –
SYMADAPT Symmetry Adapted Bloch Functions (default) 101 –
SETINF setting of inf array options 54 I
SETPRINT setting of printing options 54 I
STOP execution stops immediately 56 –
TESTPDIM stop after symmetry analysis 101 –
TEST[RUN] stop after integrals classification and disk storage estimate 101 –

Output of data on external units

NOFMWF wave function formatted output not written in file fort.98. 95 –

SAVEWF wave function data written every two SCF cycles 97 –

Post SCF calculations

POSTSCF post-scf calculations when convergence criteria not satisfied 96 –
EXCHGENE exchange energy evaluation (spin polarized only) 83 –
GRADCAL analytical gradient of the energy 87 –
PPAN population analysis at the end of the SCF no 96

Properties
RDFMWF wave function data conversion formatted-binary (fort.98 → fort.9)
Preliminary calculations

NEWK Eigenvectors calculation 179 I

NOSYMADA No symmetry Adapted Bloch Functions 95 –
PATO Density matrix as superposition of atomic (ionic) densities 180 I
PBAN Band(s) projected density matrix (preliminary NEWK) 180 I
PGEOMW Density matrix from geometrical weights (preliminary NEWK) 181 I
PDIDE Energy range projected density matrix (preliminary NEWK) 181 I
PSCF Restore SCF density matrix 187 –

Properties computed from the density matrix

ADFT Atomic density functional correlation energy 155 I

BAND Band structure 157 I
BIDIERD Reciprocal form factors 158 I
CLAS Electrostatic potential maps (point multipoles approximation) 159 I
ECHG Charge density and charge density gradient - 2D grid 164 I
ECH3 Charge density - 3D grid 163 I
EDFT Density functional correlation energy (HF wave function only) 164 I
EMDLPG Electron momentum distribution from density matrix 166 I
PMP2 MP2 correction to the Valence Density Matrix 181
POLI Atom and shell multipoles evaluation 182 I
POTM Electrostatic potential - 2D grid 185 I
POT3 Electrostatic potential - 3D grid 183 I
POTC Electrostatic properties 184 I
PPAN Mulliken population analysis 96
XFAC X-ray structure factors 188 I

Properties computed from the density matrix (spin-polarized systems)

252
ANISOTRO Hyperfine electron-nuclear spin tensor 156 I
ISOTROPIC Hyperfine electron-nuclear spin interaction - Fermi contact 168 I
POLSPIN Atomic spin density multipoles 182 I

Properties computed from eigenvectors (after keyword NEWK)

ANBD Printing of principal AO component of selected CO 155 I

BWIDTH Printing of bandwidth 159 I
DOSS Density of states 162 I
EMDL Electron momentum distribution - line 165 I
EMDP Electron momentum distribution - plane maps 166 I
PROF Compton profiles and related quantities 186 I

New properties

POLARI Berry phase calculations 190 I

SPOLBP Spontaneous polarization (Berry phase approach) 192 –
SPOLWF Spontaneous polarization (localized CO approach) 193 –
PIEZOBP Piezoelectricity (Berry phase approach) preliminary 189 –
PIEZOWF Piezoelectricity (localized CO approach) - preliminary 189 –
LOCALWF Localization of Wannier functions 169 I
DIEL Optical dielectric constant 160 I

Auxiliary and control keywords

ANGSTROM Set input unit of measure to Ångstrom 31 –

BASISSET Printing of basis set, Fock/KS, overlap and density matrices 157 –
BOHR Set input unit of measure to bohr 34 –
CHARGED Non-neutral cell allowed (PATO) 62 –
END Terminate processing of properties input keywords –
FRACTION Set input unit of measure to fractional 44 –
MAPNET Generation of coordinates of grid points on a plane 177 I
NEIGHBOR Number of neighbours to analyse in PPAN 51 I
PRINTOUT Setting of printing options 53 I
RAYCOV Modification of atomic covalent radii 53 I
SETINF Setting of inf array options 54 I
SETPRINT Setting of printing options 54 I
STOP Execution stops immediately 56 –
SYMMOPS Printing of point symmetry operators 60 –

Info - Output of data on external units

ATOMIRR Coordinates of the irreducible atoms in the cell 156 –

ATOMSYMM Printing of point symmetry at the atomic positions 34 –
COORPRT Coordinates of all the atoms in the cell 37 –
CRYAPI OUT geometry, BS, direct lattice information 160 –
EXTPRT Explicit structural/symmetry information 41 –
FMWF Wave function formatted outputi in file fort.98. Section 8.2 167 –
INFOGUI Generation of file with wf information for visualization 168 –
CRYAPI OUT Reciprocal lattice information + eigenvalues 160 –

253
Appendix C

Printing options

Extended printing can be obtained by entering the keywords PRINTOUT (page 53) or SET-
PRINT (page 54).
In the scf (or scfdir) program the printing of quantities computed is done at each cycle if the
corresponding LPRINT value is positive, only at the last cycle if the LPRINT value is negative.
The LPRINT options to obtain intermediate information can be grouped as follows. The follow-
ing table gives the correspondence between position number, quantity printed, and keyword.

crystal Keyword inp

• direct lattice - geometry information: 1 GLATTICE –

• symmetry operators : 4, 2 SYMMOPS –

• atomic functions basis set : 72 BASISSET –

• DF auxiliary basis set for the fitting: 79 DFTBASIS –

• scale factors and atomic configuration: 75 SCALEFAC –

• k-points geometrical wheight: 53 KWEIGHTS –

• shell symmetry analysis : 5, 6, 7, 8, 9

• Madelung parameters: 28

• multipole integrals: 20

• Fock/KS matrix building - direct lattice: 63, 64, 74 FGRED FGIRR N

• Total energy contributions: 69 ENECYCLE –

crystal - properties

• shell and atom multipoles: 68 MULTIPOLE N

• reciprocal space integration to compute Fermi energy: 51, 52, 53, 54, 55, 78

• density matrix - direct lattice: irreducible (58); reducible (59) PGRED PGIRR N
(reducible P matrix in crystal if PPAN requested only)

• Fock/KS eigenvalues : 66 EIGENVAL N

EIGENALL –

• Fock/KS eigenvectors : 67 EIGENVEC N

• symmetry adapted functions : 47 KSYMMPRT –

254
 • Population analysis: 70, 73, 77 MULLIKEN N

• Atomic wave-function: 71

properties

• overlap matrix S(g) - direct lattice: 60 (keyword PSIINF) OVERLAP N

• Densities of states: 105, 107 DOSS –

• Projected DOSS for embedding: 36, 37, 38

• DF correlation correction to total energy: 106

• Compton profile and related quantities: 116, 117, 118

• Fermi contact tensor : 18 FTENSOR –

• rotated eigenvectors (keyword ROTREF): 67 EIGENVEC –

• Charge density and electrostatic potential maps: 119 MAPVALUES –

Example
To print the eigenvalues at each scf cycle enter:
PRINTOUT
EIGENALL
END

To print the eigenvalues at the first 5 k points at the end of scf only, enter in any input block:
SETPRINT
1
66 -5

Eigenvectors printed by default are from the first valence eigenvector up to the first 6 virtual
ones. Core and virtual eigenvectors are printed by ”adding” 500 to the selected value of
LPRINT(67). To obtain print all the eigenvectors at the end of scf insert in any input block:
SETPRINT
1
66 -505

255
Printing options LPRINT array values
subroutine value printed information keyword input
1 GCALCO N up N=6 stars of direct lattice vectors GLATTICE
2 CRYSTA 6= 0 crystal symmetry operators SYMMOPS
3 EQUPOS 6= 0 equivalent positions in the reference cell EQUIVAT
4 CRYSTA 6= 0 crystal symmops after geometry editing
5 GILDA1 N>0 g vector irr- first n set type of couples
N<0 g vector irr- n-th set type of couples
6 GROTA1 6= 0 information on shells symmetry related
7 GV N>0 stars of g associated to the first n couples
7 N<0 stars of g associated to the n-th couple
8 GORDSH 6= 0 information on couples of shells symmetry related
9 GSYM11 6= 0 intermediates for symmetrized quantities
10 GMFCAL 6= 0 nstatg, idime, idimf, idimcou
11 MAIN2U 6= 0 exchange energy EXCHGENE
MAIND EXCHGENE
12 IRRPR 6= 0 symmops (reciprocal lattice) SYMMOPSR
13 MATVIC N n stars of neighbours in cluster definition
14 GSLAB 6= 0 coordinates of the atoms in the slab
15 symdir 6= 0 print symmetry allowed directions PRSYMDIR
18 TENSOR 6= 0 extended printing for hyperfine coupling cost FTENSOR
19
20 MONIRR N multipole integrals up to pole l=n
21
24 POINTCH printing of point charges coordinates
28 MADEL2 6= 0 Madelung parameters
29
30 CRYSTA 6= 0 write file FINDSYM.DAT
31 6 = 0 values of the dimension parameters PARAMETERS
32 N > 0 printing of ccartesian coordinates of the atoms
33 COOPRT N > 0 cartesian coordinates of atoms in file fort.33 ATCOORDS
34 FINE2 N >0 KNETOUT
READ2 output of reciprocal space information KNETOUT
35 N > 0 printing of symmops in short fomr
36 XCBD 6= properties - exchange correlation printing
37
38
39
40
41 SHELL* 6=0 printing of bipolar expansion parameters
47 KSYMBA n Symmetry Adapted Bloch Functions printing level
48 KSYMBA 6=0 Symmetry Adapted Bloch Functions printing active KSYMMPRT
51 AB 6= 0 B functions orthonormality check
52 DIF >0 Fermi energy - Warning !!!! Huge printout !!!
53 SCFPRT 6= 0 k points geometrical weights KWEIGHTS
54 CALPES >0 k points weights- Fermi energy
55 OMEGA >0 f0 coefficients for each band
56
57 PDIG N p(g) matrices-first n g vectors PGIRR N
58 PROT1 6 0
= mvlu, ksh, idp4
59 RROTA N > 0 P(g) matrices - first N vectors at the end of SCF, if PGRED N
PPAN present
NEWK N P(g) matrices - first N vectors PGRED N
PSIINF >0 P(g) matrices - first N vectors PGRED N
N <0 P(g) matrix for g=N PGRED N
60 PSIINF >0 overlap matrix S(g) - first N vectors OVERLAP N
N <0 overlap matrix S(g) for g = N N
61
63 TOTENY 6= 0 bielectronic contribution to irred. F(g) matrix
64 FROTA N F(g) matrix - first N g vectors FGRED N
PSIINF N> 0 FGRED N
N <0 f(g) matrix - for g = N (N-th g vector only) FGRED N

256
 subroutine value printed information keyword input
65
66 AOFK N e(k)- fock eigenvalues- first N k vectors EIGENVAL N
ADIK N
BANDE N
DIAG N EIGENALL
FDIK N
FINE2 N
NEWK N
67 AOFK N a(k) - fock eigenvectors - first N k vectors EIGENVEC N
ADIK N
DIAG N
FINE2 N
NEWK N
68 POLGEN N <0 shell and atom multipoles up to pole l=N MULTIPOL N
POLGEN N >0 atom multipoles up to pole l=N MULTIPOL N
QGAMMA N shell multipoles up to pole l=N MULTIPOL N
69 TOTENY 6= 0 contributions to total energy at each cycle ENECYCLE
FINE2 6 = 0 Mulliken population analysis
70
NEIGHB at the end of scf cycles
POPAN calls PPBOND, to perform Mulliken analysis
PDIBAN
71 PATIRR 6= 0 atomic wave function ATOMICWF
PATIR1 6 0
= ”” ATOMICWF
72 INPBAS 6= 0 basis set BASISSET
INPUT2 6 0
= basis set BASISSET
READFG SET
=1
73 POPAN 6= 0 Mulliken matrix up to N direct lattice vector MULLIKEN N
PPBOND
PDIBAN N
74 TOTENY N f(g) irreducible up to g=N FGIRR N
DFTTT2 N FGIRR N
75 INPBAS 6= 0 printing of scale factor and SCALEFAC
atomic configuration CONFIGAT
76
77 PPBOND 0 printing of neighbouring relationship
6= 0 no printing of neighbours relationship
78 FERMI 6= 0 informations on Fermi energy calculation
EMIMAN 6= 0
79
! 79 DFGPRT 6= 0 dft auxiliary basis set - default no printing DFTBASIS
! 80 ROTOP >0 printing of atoms coord. in rotated ref. frame ROTREF

92 INPBAS G94 deck on ft92 GAUSS94

MOLDRW input deck to MOLDRAW
93

105 DENSIM <0 DOSS along energy points DOSS

DFFIT3 >0 DFT intermediate printout
106
(keyword PRINT in dft input)
107 STARIN 6 = 0 DOSS information
112 PROFCA 6= 0 projected DOSS coefficients
116 PROFI 6 = 0 Compton profile information
117 PROFI 6 = 0
118 PROFI N
119 INTEG 6 = 0 charge density at grid points MAPVALUES
JJTEG 6 = 0 charge density at grid points MAPVALUES
MAPNET 6 = 0 electrostatic potential at grid points MAPVALUES
NAPNET 6 = 0 charge density gradient components MAPVALUES
120 LIBPHD 6 = 0 extended printing in berny optimizer
121 reserved for geometry optimizer
122 reserved for geometry optimizer
123 reserved for geometry optimizer
124 reserved for geometry optimizer
125 reserved for geometry optimizer

257
Appendix D

External format

Formatted data are written in files according to the following table:

program keyword ftn filename pg
OPTGEOM 34 optaxxx Geometry input - opt atoms coord. only - 41
See EXTPRT
OPTGEOM 34 optcxxx Geometry input - opt cell [atoms] - See 41
EXTPRT
66 HESSOPT.DAT Hessian matrix
68 OPTINFO.DAT Information to restart optimization
crystal GAUSS98 92 GAUSSIAN.DAT Input for GAUSS98 64
FINDSYM 26 FINDSYM.DAT data in crystallographic format - read by 37
program findsym(IUCR)
STRUCPRT 33 STRUC.INCOOR Cell parameters, coordinates of atoms 56
COORPRT 33 fort.33 Coordinates of the atoms in the cell 37
crystal EXTPRT 34 fort.34 Geometry input 41
properties PPAN 24 PPAN.DAT Mulliken population analysis 96
properties BAND 25 fort.25 Bands (Crgra2006) 157
24 BAND.DAT Bands data 157
CLAS 25 fort.25 Classical potential 159
DIEL 24 DIEL.DAT Dielectric constant 160
DOSS 25 fort.25 Density of states (IPLOT=1) 162
24 DOSS.DAT Density of states (IPLOT=2) 162
ECHG 25 fort.25 Electronic charge density - 2D grid 164
25 RHOLINE.DAT Electronic charge density - 1 grid 164
ECH3 31 — Electronic charge density - 3D grid 163
DENS CUBE.DAT Electronic charge density - 3D grid CUBE 163
format
SPIN CUBE.DAT Spin density - 3D grid CUBE format 163
POT3 POT CUBE.DAT Electostatic potential - 3D grid CUBE for- 183
mat
EMDL 25 fort.25 EMD line (IPLOT=1) 165
24 EMDL.DAT EMD line(IPLOT=2) 165
EMDP 25 EMD - 2D grid 261
INFOGUI 32 Data for the graphical user interface 168
POTC 24 POTC.DAT Electrostatic potential V, Electric field, 184
Electric field gradient
POTM 25 fort.25 Electrostatic potential - 2D grid 185
PROF 25 fort.25 Compton profile and related quantities 186
(IPLOT=1)
24 PROF.DAT Compton profile and related quantities 186
(IPLOT=2)

Please refer to the standard script for running CRYSTAL09 as to handle input/output file
names. See:

258
http://www.crystal.unito.it/tutorials =¿ How to run
Data in file fort.25 are read by the programs maps06, doss06, band06 of the package
Crgra2006. In the same run bands, density of states, value of a function in a 2D grid of points
can be computed. The appropriate command (maps06, doss06, band06) selects and plots
the selected data .
The package can be downloaded from:
http://www.crystal.unito.it/Crgra2006/Crgra2006.html

CLAS - ECHG - POTM - Isovalue maps

The value of the function chosen (classic electrostatic potential (CLAS), charge(+spin) density
(ECHG), electrostatic potential (POTM)) is computed in a given net of points. The data are
written in file fort.25.
If the system is spin polarized, total density data are followed by spin density data.

Structure of the file fort.25

1ST RECORD : -%-,IHFERM,TYPE,NROW,NCOL,DX,DY,COSXY format: A3,I1,A4,2I5,3E12.5
2ND RECORD : XA,YA,ZA,XB,YB,ZB format: 1P,6E12.5
3RD RECORD : XC,YC,ZC,NAF,LDIM format: 1P,3E12.5,4X,2I4
4TH RECORD
AND FOLLOWING : ((RDAT(I,J),I=1,NROW),J=1,NCOL) format: 1P,6E12.5
Meaning of the variables:
1 ’-%-’ 3 character string marks the beginning of a block of data;
1 IHFERM: 0 : closed shell, insulating system
1 : open shell, insulating system
2 : closed shell, conducting system - Fermi level can be drawn
3 : open shell, conducting system - Fermi level can be drawn
1 TYPE 4 characters string corresponding to the type of data "MAPN"
1 NROW number of rows of the data matrix RDAT
1 NCOL number of columns of the data matrix RDAT
1 DX increment of x (\AA ngstrom) in the plane of the window
1 DY increment of y (\AA ngstrom) in the plane of the window
1 COSXY cosine of the angle between x and y axis;
2 XA,YA,ZA coordinates of the points A,B (see keyword MAPNET) (\AA ngstrom)
2 XB,YB,ZB defining the window where the functions is computed (\AA ngstrom)
3 XC,YC,ZC coordinates of point C (\AA ngstrom)
3 NAF number of atoms in the cell
3 LDIM dimensionality (0 molecule; 1 polymer, 2 slab, 3 bulk)
4-> ncol*nrow values of the function (a.u.) at the nodes of the grid
naf records follow, with atomic number, symbol, coordinates (Ångstrom) of the atoms in the cell:
NAT,SYMBAT,XA,YA,ZA format: I4,1X,A,1P,3E20.12
NAT atomic number
SYMBAT Mendeleev symbol
XA,YA,ZA cartesian coordinates of the atoms in the cell (\AA ngstrom)

Cartesian components of cell parameters follow (Ångstrom)

AX, AY, AZ cartesian component of vector a format: 3E20.12
BX, BY, BZ cartesian component of vector b format: 3E20.12
CX, CY, CZ cartesian component of vector c format: 3E20.12
The program maps06 looks for the atoms lying in the windows used to compute the function, and it
can draw the symbol of the atoms, the van der Waals sphere, or the bonds between atoms closer than
the sum of their vdW radii.

ECHG Charge (spin) density - 1D profile

When points B and C coincides in ECHG 8.2 input, coordinates relative to the origin of the segment
and charge density value [coordinate along the line, charge density: charge density derivative x,y,z
components] are written with format (2E20.12:3E20.12) in file RHOLINE.DAT. A second set of data,
spin density, is written for spin polarized systems, after a blank line.

259
BAND - Band structure
Hamiltonian eigenvalues are computed at k points corresponding to a given path in the Brillouin zone.
Data are written in file BAND.DAT and processed by DLV; see http://www.cse.clrc.ac.uk/cmg/DLV)
and in file fort.25 (processed by Crgra2006/band06)

Structure of the file fort.25

One block is written for each segment of the path in k reciprocal space: the segment is defined by two
k points, whose crystallographic coordinates (I1,I2,I3) and (J1,J2,J3) are given as integers in ISS units
(see keyword BAND).
If the system is spin polarized, α electrons bands are followed by β electrons bands.
For each segment:

1ST RECORD : -%-,IHFERM,TYPE,NBAND,NKP,DUM,,DK,EF format: A3,I1,A4,2I5,3E12.5

2ND RECORD : EMIN,EMAX format: 1P,6E12.5
3RD RECORD : I1,I2,I3,J1,J2,J3 format: 6I3
4TH RECORD
AND FOLLOWING : ((RDAT(I,J),I=1,NROW),J=1,NCOL) format: 1P,6E12.5

Meaning of the variables:

1 ’-%-’ 3 character string marks the beginning of a block of data;
1 IHFERM: 0 : closed shell, insulating system
1 : open shell, insulating system
2 : closed shell, conducting system
3 : open shell, conducting system
1 TYPE 4 characters string corresponding to the type of data "BAND"
1 NBAND number of bands
NKP number of k points along the segment
DUM not used
DK distance in k space between two adjacent sampling points
along the segment
EF Fermi energy (hartree)
2 EMIN minimum energy of the bands in the explored path (hartree)
EMAX maximum energy (hartree)
3 I1,I2,I3,J1,J2,J3 : coordinates of the segment extremes in iunit of ISS
4 EPS(I,J) eigenvalues (hartree): eps(i,j) corresponds to the i-th
.. band, and the j-th k point of the segment.

DIEL
The data computed are written in file DIEL.DAT according to the following format:
#
@ XAXIS LABEL "DISTANCE(BOHR)"
@ YAXIS LABEL "MACRORHO MACROE MACROV RHOPLANE"
5 columns - format(08E15.7)
last record is blank

DOSS Density of states

Total and projected density of states are written in file DOSS.DAT (processed by DLV; see
http://www.cse.clrc.ac.uk/cmg/DLV) and in file fort.25 (processed by Crgra2006).
One block is written for each projected density of states, including the total one: so NPRO
(number pf projections) +1 blocks are written per each run.
If the system is spin polarized, α electrons bands are followed by β electrons bands.

260
Structure of the file written in file fort.25
1ST RECORD : -%-,IHFERM,TYPE,NROW,NCOL,DX,DY,COSXY
format : A3,I1,A4,2I5,1P,(3E12.5)
2ND RECORD : X0,Y0 format : 1P,6E12.5
3RD RECORD : I1,I2,I3,I4,I5,I6 format : 6I3
4TH RECORD
AND FOLLOWING : ((RDAT(I,J),I=1,NROW),J=1,NCOL) format : 1P,6E12.5
Meaning of the variables:
1 NROW 1 (DOSS are written one projection at a time)
NCOL number of energy points in which the DOS is calculated
DX energy increment (hartree)
DY not used
COSXY Fermi energy (hartree)
2 X0 energy corresponding to the first point
Y0 not used
3 I1 number of the projection;
I2 number of atomic orbitals of the projection;
I3,I4,I5,I6 not used
4 RO(J),J=1,NCOL DOS: density of states ro(eps(j)) (atomic units).

Structure of the file written in file DOSS.DAT

Data written in file DOSS.DAT:

1ST RECORD : NPUNTI, NPRO1, IUHF

format : ’# NEPTS’,1X,I5,1X,’NPROJ’,1X,I5,1X,’NSPIN’,1X,I5
2ND RECORD : ’#’
3RD RECORD : ’@ YAXIS LABEL "DENSITY OF STATES (STATES/HARTREE/CELL)"’
4TH RECORD : (ENE(I),DOSS(IPR,I),IPR=1,NPRO1)
AND FOLLOWING :
format : 1P,15E12.4
PROF
The computed quantities are written following the same sequence of the printout. Each record
contains:
4F coordinate, all electron, core, valence contribution

EMDL
The computed quantities are written following the same sequence of the printout.
NPUNTI records are written. Each records contains (FORMAT: 10E12.4)
p (emdl(p,ipro), ipro=1,nprojections))

EMDP
1ST RECORD : -%-,IHFERM,TYPE,NMAX1,NMAX2,PMAX1,PMAX2,COS12
format : A3,I1,A4,2I5,1P,(3E12.5)
2ND RECORD : XDUM,YDUM format : 1P,6E12.5
3RD RECORD : I11,I12,I13,I21,I22,I23 format : 6I3
4TH RECORD
AND FOLLOWING : ((RDAT(I,J),I=1,NMAX1),J=1,NMAX2) format : 1P,6E12.5

Meaning of the variables:

1 ’-%-’ 3 character string marks the beginning of a block of data;
1 IHFERM: 0 : closed shell, insulating system
1 : open shell, insulating system
2 : closed shell, conducting system
3 : open shell, conducting system

261
1 TYPE 4 characters string corresponding to the type of data "EMDP"
1 NMAX1 number of points in the first direction
NMAX2 number of points in the second direction
PMAX1 maximum p value along the first direction
PMAX2 maximum p value along the first direction
COSXY angle between the two vectors defining the plane
2 X0 not used
Y0 not used
3 I11,I12,I13 fractional coordinates of the first reciprocal lattice
vector defining the plane
I21,I22,I23 fractional coordinates of the second reciprocal lattice
vector defining the plane
4 RO(J),J=1,NMAX1*NMAX2 electron momentum density at the grid points
(atomic units).

POTC
When ICA 6= 0; NPU 6= 0 (2D or 3D systems) the data computed are written in file POTC.DAT
according to the following format:

#
@ XAXIS LABEL "Z (AU)"
@ YAXIS LABEL "ELECTROSTATIC PROPERTIES (AU)"
@ TITLE "String in the first record in crystal input "
@ SUBTITLE "ELECTRIC FIELD INTENSITY: 0.100 AU" ! if external field applied
@ LEGEND ON
@ LEGEND LENGTH 3
@ LEGEND X1 0.87
@ LEGEND Y1 0.8
@ LEGEND STRING 0 "V"
@ LEGEND STRING 1 "E"
@ LEGEND STRING 2 "DE/DZ"
@ LEGEND STRING 3 "RHO"
@ LEGEND STRING 4 "V+VEXT" ! if external field applied
@ LEGEND STRING 5 "VEXT" ! if external field applied
NPU records of 5 (7 when external field applied) columns - format 08E15.7

COORPRT
The keyword COORPRT, entered in geometry input or in properties writes in file fort.33
(append mode) the following data:
record data content
# type
1 I number of atoms (NAF)
2 A Title - If written after an SCF calculation, on the same line; totalenergy,
convergence on energy, number of cycles
3 A,3F Mendeleev symbol of the atom; x, y, z cartesian coordinates (Å)
······
NAF+2 A,3F Mendeleev symbol of the atom; x, y, z cartesian coordinates (Å)

The coordinates of the atoms are written at each geometry optimization cycle (keyword OPT-
GEOM
The file ”fort.33” is read by the program MOLDEN [116] which can be downloaded from:
www.cmbi.kun.nl/ schaft/molden/molden.html

262
STRUCPRT
The file STRUC.INCOOR is written according to the format given in the example (output for
bulk MgO, 2 atoms per cell).

$cell vectors cartesian components of cell parameters (bohr)

0.00000000000000 3.97787351190423 3.97787351190423
3.97787351190423 0.00000000000000 3.97787351190423
3.97787351190423 3.97787351190423 0.00000000000000
$coordinates cartesian coordinates of atoms (bohr)
MG 0.00000000000000 0.00000000000000 0.00000000000000 12
O 0.00000000000000 0.00000000000000 -3.97787351190423 8
$END

PPAN
# Mulliken Populations:
# NSPIN,NATOM n. determinants, number of atoms
---- for each atom
# IAT,NSHELL atomic number, number o shells
# Xiat,Yiat,Ziat (AU) cartesian coordinates (bohr)
# QTOT, QSHELL,I=1,NSHELL atom total electronic charge, (shell charges)
# NORB, QORB, I=1,NORB number of orbitals, (orbital electronic charges)

Example:

graphite STO-3G basis set, RHF (1 eterminant)

2 atoms, 2 shells per atom, 5 AO per atom

1 2 | 1 determinant, 2 atoms
6 2 | 1st atom: atomic number 6, 2 shells
-1.320 -2.287 0.000 | cartesian coordinates 1st atom
6.000 1.993 4.007 | 6, electronic charge of 1st atom
| 1.993 electronic charge of 1st shell (1s)
| 4.007 electronic charge of 2nd shell (2sp)
5 | 5 atomic orbitals
1.993 1.096 0.956 0.956 1.000| 1.993 electronic charge of 1st AO (1s)
| 1.096 electronic charge of 2nd AO (2s)
| 0.956 electronic charge of 3rd AO (px)
| 0.956 electronic charge of 4th AO (py)
| 1.000 electronic charge of 5th AO (pz)
6 2 | 2nd atom: atomic number 6, 2 shells
-2.640 0.000 0.000 | cartesian coordinates 2nd atom
6.000 1.993 4.007 | 6, electronic charge of 1st atom
| 1.993 electronic charge of 1st shell (1s)
| 4.007 electronic charge of 2nd shell (2sp)
5 | 5 atomic orbitals
1.993 1.096 0.956 0.956 1.000| 1.993 electronic charge of 1st AO (1s)
| 1.096 electronic charge of 2nd AO (2s)
| 0.956 electronic charge of 3rd AO (px)
| 0.956 electronic charge of 4th AO (py)
| 1.000 electronic charge of 5th AO (pz)
6 2 | second atom: atomic number 6, 2 shells
-2.640 0.000 0.000 | cartesian coordinates 2nd atom

263
EXTPRT / EXTERNAL - file fort.34
Geometry information can be read from an external file, fort.34, by entering the keyword
EXTERNAL. The system can be a molecule, a polymer, a slab or a crystal. The file is
written by entering the keyword EXTPRT in the input block 1. The file is written at the end
of successful geometry optimization. The ”history” of the optimization process is written in
files optaxxx (xxx number of optimization cycle) or optcxxx. //[0.2cm] The structure of the
file is as follow:
¯
rec # data type contents
¯
2 3I 9 dimensionality, centring and crystal type
3 3F =
4 3F cartesian components of the direct lattice vectors
5 3F
;
¯
6 1I number of symmetry operators

9For each symmetry operator 4 records:

7 3F =
8 3F symmetry operators matrices in cartesian coordinates
9 3F
;
¯
10 3F cartesian components of the translation
¯
n 1I number of atoms in the primitive cell
For each atom, 1 record:
¯
n+1 I,3F conventional atomic number, cartesian coordinates of the atoms
The keyword EXTERNAL and END must be inserted at the top and bottom of the deck
to use it as CRYSTAL geometry input.
Example - Test05 - Graphite 2D - standard geometry input
SLAB dimensionality
77 layer group number
2.42 lattice parameter
1 number of irreducible atoms in the cell
6 -0.33333333333 0.33333333333 0. coordinates of the atoms
EXTPRT
TESTGEOM
END

Data written in file fort.34 (Ångstrom):

2 1 5 ! dimensionality, centring and crystal type
0.2095781E+01 -0.1210000E+01 0.0000000E+00 ! cartesian components of direct lattice vectors
0.0000000E+00 0.2420000E+01 0.0000000E+00 !
0.0000000E+00 0.0000000E+00 0.5000000E+03 ! 2D system - formal value 500. \AA
12 ! number of symmetry operators
0.1000000E+01 0.0000000E+00 0.0000000E+00 ! 1st symmetry operator - 3x3 transformation matrix
0.0000000E+00 0.1000000E+01 0.0000000E+00 !
0.0000000E+00 0.0000000E+00 0.1000000E+01 !
0.0000000E+00 0.0000000E+00 0.0000000E+00 ! 1st symmetry operator - 3x1 translation component
-0.1000000E+01 0.0000000E+00 0.0000000E+00 ! 2nd symmetry operator
0.0000000E+00 -0.1000000E+01 0.0000000E+00 !
0.0000000E+00 0.0000000E+00 0.1000000E+01 !
0.0000000E+00 0.0000000E+00 0.0000000E+00 !
-0.5000000E+00 -0.8660254E+00 0.0000000E+00 ! 3rd symmetry operator
0.8660254E+00 -0.5000000E+00 0.0000000E+00 !
0.0000000E+00 0.0000000E+00 0.1000000E+01 !
0.0000000E+00 0.0000000E+00 0.0000000E+00 !
-0.5000000E+00 0.8660254E+00 0.0000000E+00 ! 4th symmetry operator
-0.8660254E+00 -0.5000000E+00 0.0000000E+00 !
0.0000000E+00 0.0000000E+00 0.1000000E+01 !
0.0000000E+00 0.0000000E+00 0.0000000E+00 !
0.5000000E+00 -0.8660254E+00 0.0000000E+00 ! 5th symmetry operator
0.8660254E+00 0.5000000E+00 0.0000000E+00 !
0.0000000E+00 0.0000000E+00 0.1000000E+01 !
0.0000000E+00 0.0000000E+00 0.0000000E+00 !
0.5000000E+00 0.8660254E+00 0.0000000E+00 ! 5th symmetry operator
-0.8660254E+00 0.5000000E+00 0.0000000E+00 !
0.0000000E+00 0.0000000E+00 0.1000000E+01 !
0.0000000E+00 0.0000000E+00 0.0000000E+00 !
-0.5000000E+00 0.8660254E+00 0.0000000E+00 ! 7th symmetry operator
0.8660254E+00 0.5000000E+00 0.0000000E+00 !
0.0000000E+00 0.0000000E+00 0.1000000E+01 !

264
 0.0000000E+00 0.0000000E+00 0.0000000E+00 !
0.1000000E+01 0.0000000E+00 0.0000000E+00 ! 8th symmetry operator
0.0000000E+00 -0.1000000E+01 0.0000000E+00 !
0.0000000E+00 0.0000000E+00 0.1000000E+01 !
0.0000000E+00 0.0000000E+00 0.0000000E+00 !
-0.5000000E+00 -0.8660254E+00 0.0000000E+00 ! 9th symmetry operator
-0.8660254E+00 0.5000000E+00 0.0000000E+00 !
0.0000000E+00 0.0000000E+00 0.1000000E+01 !
0.0000000E+00 0.0000000E+00 0.0000000E+00 !
0.5000000E+00 0.8660254E+00 0.0000000E+00 ! 10th symmetry operator
0.8660254E+00 -0.5000000E+00 0.0000000E+00 !
0.0000000E+00 0.0000000E+00 0.1000000E+01 !
0.0000000E+00 0.0000000E+00 0.0000000E+00 !
-0.1000000E+01 0.0000000E+00 0.0000000E+00 ! 11th symmetry operator
0.0000000E+00 0.1000000E+01 0.0000000E+00 !
0.0000000E+00 0.0000000E+00 0.1000000E+01 !
0.0000000E+00 0.0000000E+00 0.0000000E+00 !
0.5000000E+00 -0.8660254E+00 0.0000000E+00 ! 12th symmetry operator
-0.8660254E+00 -0.5000000E+00 0.0000000E+00 !
0.0000000E+00 0.0000000E+00 0.1000000E+01 !
0.0000000E+00 0.0000000E+00 0.0000000E+00 !
1 ! number of irreducible atoms in the primitive cell
6 -0.6985938 -1.2100000 0.0000000 ! conventional atomic number, cartesian coordinate

ECH3/POT3/GRID3D
Functions values computed at 3D grid of points by the keywords ECH3 (page 163), POT3
(page183), GRID3D (page167) are written according to two formats: . All data in atomic
units.
1. Fortran unit 31 is written According to the following format. All data in atomic units.
¯
rec # data type contents
¯
1 A title: charge density /spin density
¯
2 3I npa,npb,npc, number of points along the 3 directions
¯
3 3E x,y,z cartesian coordinates of the point (1,1,1)
¯
4 3E dxa, dya, dza cartesian components of the step along a
¯
5 3E dxb, dyb, dzb cartesian components of the step along b
¯
6 3E dxc, dyc, dzc cartesian components of the step along c
¯
7 ... 5E npa*npb*npc floating point data, 5/record

2. Function data computed at 3D grid points are written according to GAUSSIAN CUBE
format in files:

DENS CUBEḊAT charge density

SPIN CUBEḊAT spin density
POT CUBEḊAT electrostatic potential

INFOGUI
Fortran unit 32 is written through the keyword INFOGUI (page 168). The format is almost
self-explaining. The following data are written for MgO bulk (test11).

2 atom(s) per cell

6 shells
18 atomic orbitals
20 electrons per cell
12 core electrons per cell
No eigenvalue level shifting
No Alpha-Beta Spin locking
No N. Beta Spin locking
Type of Calculation: RESTRICTED CLOSED SHELL
Total Energy = -0.27466415E+03H

265
Fermi Energy = -0.31018989E+00H
1 -0.31018989E+00
6 18 20 12 | # shells, # AO, # electrons, # core electrons
2 | # atoms
1 12 1 0.000000 0.000000 0.000000 | sequence number, atomic number,?,cartesian coor(bohr)
3 # shells attributed to the first atom
0 shell type (s) of the 1st shell
1 shell type (sp) of the 2nd shell
1 shell type (sp) of the 3rd shell
2 8 2 3.977874 3.977874 3.977874 | sequence number, atomic number,?,cartesian coor(bohr)
3 # shells attributed to the second atom
0 shell type (s) of the 1st shell
1 shell type (sp) of the 2nd shell
1 shell type (sp) of the 3rd shell

REFLECTANCE.DAT
Written by REFLECTANCE in FREQCALC input block.
REFLECTANCE.DAT contains: 1 column reporting frequencies and 3 columns for each ex-
plored direction, containing R, Im(1/) and Im(), respectively.
Suppose we have the following reflectance input block

. . . .
FREQCALC
INTENS
[options for INTENS]
DIELTENS or DIELISO
. . . .
[optional FREQCALC keywords]
. . . .
REFLECTANCE
REFLDIR
2
1 2
END

The first column in the generated REFLECTANCE.DAT contains frequencies, from column 2
to 4 R, Im(1/) and Im() corresponding to reflection direction 1 are listed, and from column
5 to 7 the same quantities corresponding to reflection direction 2 are reported.
Suppose we want to plot R in the two directions. Once gnuplot is opened on the terminal (type
gnuplot), it is sufficient to type

plot ’REFLECTANCE.DAT’ using 1:2

where 1 stands for ”first column assigned to x axis and second column to y axis”. The plot of R
corresponding to the first direction appears on the screen and can be saved with the command

save ’name_plot’

If we want to save the plot as a Post-Script

set size 1.0, 0.6

set terminal postscript portrait enhanced mono dashed lw 1 "Helvetica" 14
set output "my-plot.ps"
replot

266
Further details about these commands and manipulation of files at http://www.duke.edu/ hp-
gavin/gnuplot.html and http://www.gnuplot.info/documentation.html.
In a similar way, we can type

plot ’REFLECTANCE.DAT’ using 1:5

in order to obtain R along the second reflection direction.

Interface to external programs

The keyword CRYAPI OUT, present into properties input stream writes formatted wave
function information, both in direct and reciprocal space, in file GRED.DAT and KRED.DAT
The scripts runcry06 and runprop06 save them in the current directory as inpfilename.GRED
and inpfilename.KRED .
The program cryapi inp, written in fortran 90, is distributed as source code
(http://www.crystal.unito.it => documentation => utilities). It reads and prints the data,
showing the meaning of the variables and the organization of data
cryapi inp should be compiled by any fortran 90 compiler: comments and request for more
information are welcome (mail to [email protected]).

GRED.DAT
The file GRED.DAT contains:

• Geometry, symmetry operators;

• Local functions basis set (including ECP)

• Overlap matrix in direct lattice

• Hamiltonian matrix in direct lattice

• Density matrix in direct lattice

• Wannier functions (if file fort.80, written by LOCALWF when localization is successful,
is present)

Overlap, hamiltonian, density matrices are written as arrays of non-zero elements. GRED.DAT
contains the information to build full matrices.
All data are printed executing cryapi inp

KRED.DAT
The file KRED.DAT is written if eigenvectors have been computed (keyword NEWK 8.2) by
properties.
CRYSTAL works in the irreducible Brillouin (IBZ) zone only: eigenvectors in the full Brillouin
zone (BZ) are computed by rotation, and by time reversal symmetry, when necessary. The file
KRED.DAT contains:

• Coordinates of k points in irreducible Brillouin zone, according to Pack-Monkhorst net

• Symmetry operators in reciprocal lattice

267
 • Geometrical weight of k points

• Hamiltonian eigenvalues

• Weight of k points for each band (computed by Fermi energy calculation)

• Eigenvectors in full Brillouin zone

Structure of matrices in direct lattice

Overlap, hamiltonian, and density matrices in direct lattice are arrays of non-zero elements:
cryapi inp prints the matrices as triangular (hamiltonian) or square matrices of size (local BS
x local BS), for a limited number of direct lattice vectors, to show the structure of the arrays.

From IBZ to BZ
CRYSTAL works on irreducible Brillouin zone (IBZ), full information is generated by applying
rotation operators.
Time reversal symmetry is exploited in reciprocal lattice: the inversion symmetry is always
present, even if the inversion operator is not present in direct lattice.
Given a shrinking factor according to Pack-Monkhorst sampling, to total number of k points
is for instance:
System n. symmops shrink factors IBZ NOSYMM BZ

graphite (2D) 12 3 3 5 9
SiC (3D) 24 4 8 36 64
MgO (3D) 48 4 8 36 64

IBZ number of points in IBZ

NOSYMM number of points removing direct lattice symmetry
BZ number of points in Brillouin zone

268
Appendix E

Normalization coefficients

A. Bert - Thesis 2002

The aim of this appendix is to show how normalization coefficients of the basis functions are
defined in CRYSTAL and to describe how they are stored in the program.

Basic Definitions
Let us consider a function, f (r); we have in general:
Z
2
dr |f (r)| 6= 1; (E.1)

however, we can always define a related f 0 (r), multiplying f (r) by a constant N :

f 0 (r) = N f (r), (E.2)

such that: Z
2
dr |f 0 (r)| = 1. (E.3)

f 0 (r) is said to be a normalized function and N is its Normalization Coefficient (NC). Substi-
tuting eq. E.2 in E.3, we have:
Z −1/2
2
N= dr |f (r)| . (E.4)

Gaussians: Product Theorem and Normalization

Let us define Gaussian functions as:

G(αi ; r − A) = exp(−αi (r − A)2 ), (E.5)

where A is the centroid of the function.

The Gaussian product theorem states that the product of two Gaussians, is still a Gaussian
function:1  
αβ
G(α; r − A)G(β; r − B) = exp − |R|2 G(ξ; r − P); (E.8)
ξ
1 Let us prove the Gaussian product theorem:
G(α; r − A)G(β; r − B) = exp(−αi (r − A)2 ) exp(−αj (r − B)2
= exp −α(r2 + A2 + 2rA) − β(r2 + B2 + 2rB)
` ´

αA2 + βB2
» „ «–
= exp −ξ (r − P)2 + P2 − . (E.6)
ξ

269
with:

ξ = α + β, (E.9)
αA + βB
P= , (E.10)
ξ
R = A − B. (E.11)

From eq. E.4, the NC of Gaussian functions, gi , can be written as:

Z −1/2
2
gi = dr (G(αi ; r))
Z −1/2
= dr G(2αi ; r)
 −3/4
π
= , (E.12)
2αi

where the Gaussian product theorem and the Gaussian integral [130] have been used. G0 (αi ; r),
defined as:
G0 (αi ; r) = gi G(αi ; r), (E.13)

is a normalized function.

Harmonic Gaussians
The Definition
The Solid Harmonic Functions, Y`m , [110] are defined as:
|m|
Y`m (r) = r` P` (cosϑ)eimφ , (E.14)

where P`m is the Legendre Polynomial Function characterized by the integers ` and m, such
that: ` ≥ 0 and −` ≤ m ≤ `. [1]
Starting from Y`m , the Real Solid Harmonic, X`m , can be defined:
|m| −|m|
|m| |m| Y` (r) + Y` (r)
X` (r) = <(Y` )= , (E.15)
2
|m| −|m|
−|m| |m| Y` (r) − Y` (r)
X` (r) = =(Y` )= . (E.16)
2i
We report some examples of X functions.
` = 0:
X00 (r) = 1; (E.17)

` = 1:
X10 (r) = z, X11 (r) = x, X1−1 (r) = y; (E.18)

` = 2:
X20 (r) = z 2 − 0.5(x2 − y 2 ), X21 (r) = 3zx, X2−1 (r) = 3zy, (E.19)

X22 (r) = 3(x2 + y 2 ), X2−2 (r) = 3xy. (E.20)

Using eqs. E.9, E.10 and E.11, eq. E.6 can be rewritten as:
„ «
αβ
G(α; r − A)G(β; r − B) = exp − |R|2 G(ξ; r − P). (E.7)
ξ

270
We have now the tools required to define the Solid Harmonic Gaussian, [110] ξ:

ξ n`m (αi ; r) = |r|2n Y`m (r)Gi (αi ; r), (E.21)

where n is a non-negative integer number (n ≥ 0). We are interested here only in n = 0

harmonic Gaussians (that is, ξ 0`m ), so we shall simply write (omitting the n = 0 index):

ξ `m (αi ; r) = Y`m (r)G(αi ; r). (E.22)

Substituting Y with X (eqs. E.15 and E.16) in eq. E.22, Real Harmonic Gaussians, γ, can be
defined:
γ `m (αi ; r) = X`m (r)G(αi ; r). (E.23)

γ are used as basis functions in the CRYSTAL program and are related to the ξ ones by
followings relations:

ξ `|m| + ξ ` −|m|
γ `|m| = , (E.24)
2
ξ `|m| − ξ ` −|m|
γ `−|m| = , (E.25)
2i
where eqs. E.15 and E.16 have been used.
Note that, when ` is equal to 0, ξ and γ functions degenerate to simple Gaussians:

ξ 00 = γ 00 = G, (E.26)

where eq. E.17 has been used and ξ degenerates to γ when m = 0:

ξ `0 = γ `0 , (E.27)

where eqs. E.24 and E.25 have been used.

The Normalization Coefficient

Let us consider now ξ and γ’s normalization coefficients (b and c, respectively), from eq. E.4,
follows:
−1/2
b`m
i = (Ξ) (E.28)
−1/2
c`m
i = (Υ) , (E.29)

where
Z
2
dr ξ `m (αi ; r)

Ξ= (E.30)
Z

2
Υ= dr γ `m (αi ; r) . (E.31)

Using eqs. E.5, E.8, E.14, E.22 and a spherical polar coordinate system,2 the Ξ integral can be
factorized as:
Z
∗
Ξ = dr [Y`m (r)G(αi ; r)] Y`m (r)G(αi ; r)
Z
= dr Y`−m (r) Y`m (r) G(2αi ; r)

= Ξr Ξϑ Ξφ , (E.32)
2 dr = r2 sinϑ dr dϑ dφ

271
with:
Z ∞
Ξr = dr exp(−2αi r2 )r2`+2
0
Γ(` + 3/2)
=
2(2αi )`+3/2
π 1/2 (2` + 1)!!
= , (E.33)
2`+2 (2αi )`+3/2

where we have used the Γ function’s properties; [1]

Z π  2
|m|
Ξϑ = dϑ P` (cos ϑ) sin ϑ
0
2(` + |m|)!
= , (E.34)
(2` + 1)(` − |m|)!

where the Legendre polynomials’ properties have been used, [1] and
Z 2π
Ξφ = dφ = 2π. (E.35)
0

Substituting eqs. E.32, E.33, E.34 and E.35 in the b definition (eq. E.28) we obtain:

π 1/2 (2` + 1)!! 2(` + |m|)!

b`m
i = 2π
2`+2 (2αi )`+3/2 (2` + 1)(` − |m|)!
!−1/2
π 3/2 (2` − 1)!! (` + |m|)!
= `+3/2
. (E.36)
22`+3/2 αi (` − |m|)!

Note that b is independent from the sign of m (as Ξ is), that is:
`|m| ` −|m|
bi = bi . (E.37)

In order to deduce the explicit expression for c, we are interested now in solving the integral
of eq. E.31: Z
2
Υ = dr (X`m (r)) G(2αi ; r), (E.38)

where eqs. E.8 and E.23 have been used. Substituting eq. E.24 (γ functions with m ≥ 0) in
previous equation, we have:
Z 2 Z 2
1  
Υm≥0 = dr ξ `|m| (αi ; r) + dr ξ ` −|m| (αi ; r) +
   
4
Z !
`|m| ` −|m|
+2 dr ξ (αi ; r)ξ (αi ; r) . (E.39)

The first two integrals in eq. E.39 can be recognized as Ξ (eq. E.32, reminding that Ξ is
independent from the m sign); the last one, if m 6= 0, is null for the orthogonality properties
of the Harmonic functions, [1] therefore:

Ξ
Υm>0 = . (E.40)
2
The same result is found for negative m, substituting eq. E.25 (instead of eq. E.24, as done)
in eq. E.31:
Ξ
Υm<0 = , (E.41)
2

272
so Υ (as Ξ is) is independent from the m sign. If m = 0, the last integral in eq. E.39 is equal
to Ξ, as the first two ones:
Υm=0 = Ξ; (E.42)

the previous equation can be deduced also from eq. E.27.

Summarizing, from eqs. E.40, E.41 and E.42, we get:
Ξ
Υ= (E.43)
2 − δm0
and, finally, substituting eqs. E.32 and E.43 in eq. E.29, we obtain:
!−1/2
π 3/2 (2` − 1)!! (` + |m|)!
c`m
i = `+3/2
. (E.44)
22`+3/2 (2 − δm0 ) αi (` − |m|)!

The c expression (eq. E.44) can be reorganized in a two factors formula:

c`m
i = a`i f `m , (E.45)

with:
−1/2
π 3/2

a`i = , (E.46)
(2αi )`+3/2
the α-dependent term, and
 −1/2
(2` − 1)!! (` + |m|)!
f `m = , (E.47)
2` (2 − δm,0 )(` − |m|)!

the m dependent term.

Note that,

• If ` = 0, γ degenerates in a simple Gaussian (eq. E.26),

f 00 = 1 and c00 0
i = ai = gi , (E.48)

where gi is the G’s NC (eq. E.12).

• If ` = 1, f 1m = 1/2 for the three m-values:

5/4
a1i α 27/4
f 1m = 1/2 and c1m
i = = i 3/4 , ∀ m = −1, 0, 1. (E.49)
2 π

• If ` = 2, we have:
7/4 7/4 7/4
αi 211/4 2−1 αi 211/4 2−2 αi 27/4
c20
i = √ ; c21
i = ci = ; c22
i = ci = . (E.50)
π 3/4 3 π 3/4 3 π 3/4 3

Let us verify, for two examples, that

γ0 = c γ (E.51)

is a normalized function, proving that the following integral, I, is equal to 1,

Z

2
Ii`m = dr c`m i γ
`m
(αi ; r) . (E.52)

273
The s Case (` = 0, m = 0)
Z

2
2
Ii00 = c00
i dr γ 00 (αi ; r)
Z
2
= (gi )2 dr (G(αi ; r))
Z −1 Z
2 2
= (G(αi ; r)) dr (G(αi ; r)) dr = 1, (E.53)

where eqs. E.48, E.27 and E.12 have been used.

A d Case (` = 2, m = 1)
Z

2
2
Ii21 = dr c21 21
i γ (αi ; r) = c21
i J, (E.54)

with: Z
2
J= dr (3zxG(2αi ; r)) , (E.55)

where eqs. E.23 and E.19 have been used.

Gaussians are separable functions, that is:

G(αi ; r) = Gx (αi ; x) Gy (αi ; y) Gz (αi ; z), (E.56)

with:
Gx (αi ; x) = exp(−αi x2 ) (E.57)

and similarly for y and z. Substituting eq. E.56 in eq. E.55, we have:

J = 9Jx Jy Jz , (E.58)

with:
Z √
π
Jx = 2
(2αi )−3/2 ,
x Gx (2αi ; x)dx = (E.59)
2
Z  1/2
π
Jy = Gy (2αi ; y)dy = , (E.60)
2αi
Z √
π
Jz = z 2 Gz (2αi ; z)dz = (2αi )−3/2 , (E.61)
2

where ref. [130] has been used in solving the integrals. Substituting now eqs. E.49 and E.58 in
eq. E.54, we obtain:
7/2 √ 2  1/2
α 211/2 π π
Ii21 = i 3/2 9 (2αi )−3/2 = 1. (E.62)
π 9 2 2αi

Atomic Orbitals Normalization

The variational basis functions of the CRYSTAL program (AOs), ϕµ , are normalized contrac-
tions (fixed linear combinations) of normalized real solid harmonic Gaussian type functions
(primitive functions), γ 0 (eq. E.51). The AOs are organized in shells, ϕµ belonging to the
same shell, λ, have same radial part, that is, same contraction coefficients, dλi , same Gaussian
exponents, αiλ and different angular part, X`m :
X X
ϕ`m
λ = Nλ dλi c`m
i γ
`m
(αiλ ; r) = Nλ dλi c`m m λ
i X` (r) G(αi ; r). (E.63)
i i

274
The index i runs over the primitive functions of the contraction, dλi is the contraction coefficient
of the i-th primitive in shell λ and, as we have seen, it is the same for all the AOs of λ, that
is, it does not depend on ` or m. γ and c are the primitive function and its NC (eq. E.29),
respectively. Nλ is the NC of AOs belonging to λ and is defined as:

Z !2 −1/2
X
Nλ =  dr dλi c`m
i γ
`m
(αiλ ; r)  , (E.64)
i

in the following will be shown that N depends only on the shell, λ.

We report, as an example, the three AOs of a p-type shell (` = 1), supposing that λ is classified
as the fourth shell of the unitary cell and each AO is a contraction of two primitives.

pz = ϕ10 4
d41 c10 10 4 4 10 10 4


4 =N 1 γ (α1 ; r) + d2 c2 γ (α2 ; r) , (E.65)
px = ϕ11 4
d41 c11 11 4 4 11 11 4


4 =N 1 γ (α1 ; r) + d2 c2 γ (α2 ; r) , (E.66)
ϕ1−1 4
d41 c1−1 1−1
(α14 ; r) d42 c1−1 1−1
(α24 ; r)


py = 4 =N 1 γ + 2 γ . (E.67)

Let us put our attention on Nλ . Eq. E.64 can be rewritten as:

 −1/2
X
`m 0 
Nλ =  dλi dλj c`m
i cj Υ , (E.68)
i,j

with: Z
0
Υ = dr γ `m (αi ; r) γ `m (αj ; r), (E.69)

where the shell index on α has been omitted for simplicity. Substituting eq. E.23 in eq. E.69,
we have:
Z Z
2
Υ0 = X`m (r) G(αi ; r) X`m (r) G(αj ; r)dr = (X`m (r)) G[(αi + αj ); r]dr, (E.70)

where the Gaussian product theorem (eq. E.8) has been used.
From eq. E.31, it can be seen that Υ0 differs from Υ only in the Gaussian exponent (αi + αj
instead of 2αi ), using then eqs. E.43, E.32, E.34 and E.35, Υ0 is rewritten as:

Υ0r Ξϑ Ξϕ
Υ0 = , (E.71)
2 − δm0
with:
Z ∞
Υ0r = dr exp[−(αi + αj )r2 ]r2`+2
0
Γ(` + 3/2)
=
2(αi + αj )`+3/2
π 1/2 (2` + 1)!!
= . (E.72)
2`+2 (αi + αj )`+3/2

Substituting eqs. E.44, E.71 and E.72 in eq. E.68, we obtain:

  q `+3/2 −1/2
2 α λ αλ
X i j
Nλ =  dλi dλj  λ , (E.73)
  
α + α λ 
i,j i j

where it is clear that N depends only on λ.

275
The Code
In order to explain easily the organization of NCs in CRYSTAL, eq. E.63 is reorganized as
follows: X
ϕ`m
λ = n`m
λ,i γ
`m
(αiλ ; r), (E.74)
i

with:
n`m λ `m
λ,i = Nλ di ci . (E.75)

Note that, while the AO is normalized, the function γ 00 = n γ is not; in fact n is not a
normalization factor, and it will be referred as the pre-Gaussian factor.
At the moment the CRYSTAL code is able to treat four type of shells: s, sp, p and d.3 An
s shell has only an AO, that is a contraction of simple Gaussians (` = 0); in a p one there
are three AOs (different for the m value, px , py and pz ) with ` = 1 primitives; d shells are
obviously formed by five ` = 2 AOs. The three basis functions of a sp shell are contractions of
one s primitive function and several ps’.
In the calculation of the integrals required in the SCF process, n must be very often multiplied
by the constant factor π 5/8 21/4 ; [103] therefore, in the code, pre-Gaussian factors are not
stored, but the following quantities, that we shall call code pre-Gaussian constants:

Sλi = π 5/8 21/4 n00

λ,i (E.76)
Pλi = π 5/8 21/4 n1m
λ,i ∀ m = 0, 1, −1 (E.77)
s
(2 + |m|)!
Dλi = π 5/8 21/4 n2m ∀ m = 0, 1, −1, 2, −2. (E.78)
(2 − δm0 )(2 − |m|)! λ,i

Note that the square root in eq. E.78 (the inverse of the m-dependent part of c, eq. E.44)
makes D independent from the m value, whereas n2m λ,i depends from it. In such a way, S, P
and D are m-independent
In the inpbas routine, contraction coefficients (as defined in input), dλi , related to s, p and d
AOs, are loaded in the two dimension packed arrays c1, c2 and c3, respectively (they are stored
in the module basato module). Their length corresponds to the total number of primitives
in the unit cell and is the same for the three arrays. The first elements are the contraction
coefficients for the first shell (d1i ), then the d2i s (second shell) follows, and so on; the contraction
index, i, is the internal one. For an s shell, for example, the elements of c2 and c3 are null, of
course.
In the gaunov routine, c1, c2 and c3 are redefined and loaded with the code pre-Gaussian
constants S, P and D, respectively; naturally they maintain the described organization and
module basato module is overwritten.
In gaunov two further arrays, c2w and c3w (that follow the convention used in the ATMOL
program) are also defined and loaded in basato module. They have the same organization as
c1, c2 and c3 and contain P0λ 0λ
i and Di coefficients, respectively:

π 5/8 21/4 1m
P0λ
i = nλ,i ∀ m = 0, 1, −1 (E.79)
2αi
s
π 5/8 21/4 (2 + |m|)!
D0λ
i = n2m ∀ m = 0, 1, −1, 2, −2. (E.80)
(2αi ) 2 (2 − δm0 )(2 − |m|)! λ,i
3 The implementation of higher polynomial functions is now in progress.

276
We give an example of evaluation of an overlap integral Sµν over an s and a m = 0 d AO
(ϕµ ≡ ϕλ00 , ϕν ≡ ϕσ20 ) sitting in the reference cell:
Z
Sµν = dr ϕλ00 (r)ϕσ20 (r). (E.81)

Substituting eq. E.74 in the previous equation, we have:

X Z
Sµν = nλ,i nσ,j dr γ 00 (αiλ ; r)γ 20 (αjσ ; r).
00 20
(E.82)
ij

Since in the code, S and D are available (but not the n coefficients), we express n as a function
of code pre-Gaussian constants, using eqs. E.76 and E.78, and we rewrite the overlap integral
as:
s
 −2 (2 − δ )(2 − |m|)! X Z
5/8 1/4 m0
Sµν = π 2 Sλi Dσi dr γ 00 (αiλ ; r)γ 20 (αjσ ; r). (E.83)
(2 + |m|)! ij

Note that the m-dependent term contained in n, for d shells, must be multiplied a posteriori,
because is not included in D. This operation is performed in the dfac3 routine, that provides
McMurchie-Davidson coefficients multiplied by code pre-Gaussian constants and, when λ is a
d shell, by the m-dependent part of n2m
λ,i .

277
Appendix F

CRYSTAL09 versus CRYSTAL06

Geometry
• Roto-translational symmetry
In the case of polymers it can treat helical structures (translation followed by a rota-
tion around the periodic axis). See keyword HELIX (page 15) and examples therein.
CRYSTAL06 allowed commensurate rotations only, by adopting a suitably large unit cell.

• Nanotubes
A special input option allows generation of 1D structures (nanotubes) from 2D one. See
keyword NANOTUBE (page 47).

Geometry optimization
• Default choice modified: TRUSTRADIUS scheme active.
Use keyword NOTRUSTR in OPTGEOM input block (page 111) to run geometry
optimization with CRYSTAL06 default.

Frequencies calculation
• default value for SCF convergence on total energy is 10−9

• default choice for numerical integration grid (DFT Hamiltonian): XLGRID. To run
DFT Hamiltonian cases with CRYSTAL06 numerical integration accuracy insert LGRID
in DFT input block (page 78)

• default choice to compute IR intensities: Berry phase approach. To compute IR intensi-

ties with CRYSTAL09 as with CRYSTAL06 insert keyword INTLOC in FREQCALC
input block.

Basis set
• f orbitals
- f orbitals (local functions basis set) with non-zero occupancy allowed. This new feature
allows study of systems with Lanthanides.
d and f orbitals occupation guess
- FDOCCUP (input block3, page 84) defines the occupation of specific f or d orbitals

278
 in a given shell in the SCF initial guess, according to the local atomic symmetry in the
crystal lattice.

• - Effective core pseudo potentials Projector operators up to angular quantum number

` = 4 are allowed. Input deck has been changed

279
Test cases
crystal09 versus crystal06 total energies (hartree)
CRYSTAL06 CRYSTAL09 diff
TEST 0 -110.7649354541 -110.7649354541 0.00E+00
TEST 1 -39.7267242374 -39.7267242374 0.00E+00
TEST 2 -223.7874756819 -223.7874756819 0.00E+00
TEST 3 -893.8746580004 -893.8746580004 0.00E+00
TEST 4 -1400.1776585535 -1400.1776585535 0.00E+00
TEST 5 -74.8333583570 -74.8333583570 1.01E-12
TEST 6 -58.4208255980 -58.4208255980 1.00E-12
TEST 7 -2800.7355953744 -2800.7355953744 0.00E+00
TEST 8 -571.3207540595 -571.3207540595 3.00E-11
TEST 9 -29.2566111179 -29.2566111179 0.00E+00
TEST10 -577.8265583253 -577.8265583271 -1.86E-09
TEST11 -274.6641530559 -274.6641530559 -3.00E-11
TEST12 -447.6810664796 -447.6810664796 -6.00E-11
TEST13 -23.9856901143 -23.9856901143 0.00E+00
TEST14 -159.6970601598 -159.6970601598 0.00E+00
TEST15 -5229.8366027793 -5229.8366027783 1.00E-09
TEST16 -2995.2869386583 -2995.2869386582 1.00E-10
TEST17 -2674.3752958019 -2674.3752958111 -9.20E-09
TEST18 -679.2766564082 -679.2766564082 0.00E+00
TEST19 -223.8070777853 -223.8070777853 0.00E+00
TEST20 -89.9552981103 -89.9552981101 1.79E-10
TEST21 -447.5749511978 -447.5749511978 0.00E+00
TEST22 -460.7186326563 -460.7186326563 -3.00E-11
TEST23 -8.0613160317 -8.0613160317 0.00E+00
TEST24 -1400.1776188146 -1400.1776188146 0.00E+00
TEST25 -74.8442039913 -74.8442039913 -9.95E-13
TEST26 -58.4208255860 -58.4208255860 0.00E+00
TEST27 -2800.7355409839 -2800.7355409839 0.00E+00
TEST28 -8.0115274157 -8.0115274157 -9.95E-14
TEST29 -2047.6430862971 -2047.6430862970 9.98E-11
TEST30 -109.0441458665 -109.0441458665 0.00E+00
TEST31 -4095.2867581787 -4095.2867581681 1.06E-08
TEST32 -92.1408103960 -92.1408103960 0.00E+00
TEST33 -92.1416129818 -92.1416129818 0.00E+00
TEST34 -1117.5230436113 -1117.5230436113 0.00E+00
TEST35 -936.5017511475 -936.5017511475 3.00E-11
TEST36 -112.5648952230 -112.5648952230 1.00E-11
TEST37 -3028.3682392877 -3028.3682392877 0.00E+00
TEST38 -2279.1395902376 -2279.1395902366 1.00E-09

280
Appendix G

Acronyms

AFM – Anti ferromagnetic

AO – Atomic Orbital
APW – Augmented Plane Wave
a.u. – atomic units
BF – Bloch Function
BS – Basis set
BSSE – Basis Set Superposition Error
BZ – Brillouin Zone (first)
B3PW – Becke Perdew Wang
B3LYP – Becke - Lee - Yang - Parr
CO – Crystalline Orbital
CPU – Central Processing Unit
DF(T) – Density Functional (Theory)
DM – Dipole Moment (see Wannier Functions)
DOS – Density of States
ECP – Effective Core Potentials
EFG – Electric Field Gradient
EMD – Electron Momentum Density
FM – Ferromagnetic
GC – Gradient-Corrected
GGA – Generalised Gradient Approximation
GS(ES) – Ground State (Electronic Structure)
GT(O) – Gaussian Type (Orbital)
GT(F) – Gaussian Type (Function)
GUI – Graphical User Interface
KS – Kohn and Sham
HF – Hartree-Fock
IBZ – Irreducible Brillouin zone
IR – Irreducible Representation
LAPW – Linearized Augmented Plane Wave
LCAO – Linear Combination of Atomic Orbitals
LDA – Local Density Approximation
LP – Local Potential
LSDA – Local Spin Density Approximation
LYP – GGA Lee-Yang-Parr
MO – Molecular Orbital

281
MPP – Massive Parallel Processor
MSI – Molecular Simulation Inc.
NLP – Non-local potential (correlation)
PBE – GGA Perdew-Burke-Ernzerhof
PDOS – Projected Density of States
PP – Pseudopotential
PVM – Parallel Virtual Machine
PW – Plane Wave
PWGGA – GGA. Perdew-Wang
PWLSD – LSD Perdew-Wang
PZ – Perdew-Zunger
P86 – GGA Perdew 86
P91 – Perdew 91
QM – Quantum Mechanics
RCEP – Relativistic Compact Effective Potential
RHF – Restricted Hartree-Fock
ROHF – Restricted Open-shell Hartree-Fock
SAED – Symmetry Allowed Elastic Distortions
SABF – Symmetry Adapted Bloch Functions SC – Supercell
SCF – Self-Consistent-Field
STO – Slater Type Orbital
UHF – Unrestricted Hartree-Fock
VBH – von Barth-Hedin
VWN – Vosko-Wilk-Nusair
WnF – Wannier Functions 0D – no translational symmetry
1D – translational symmetry in 1 direction (x, CRYSTAL convention)
2D – translational symmetry in 2 directions (x,y, CRYSTAL convention)
3D – translational symmetry in 3 directions (x,y,z CRYSTAL convention)

282
Bibliography

[1] M. Abramovitz and I. Segun, editors. Handbook of Mathematical Functions.

Dover, New York, 1965.
[2] C. Adamo and V. Barone. Toward reliable density functional methods
without adjustable parameters: the pbe0 model. J. Chem. Phys., 110:6158–
6170, 1999.
[3] D. G. Anderson. J. Assoc. Comput. Mach., 12:547, 1964.
[4] E. Aprà. Metodi quanto-meccanici per lo studio di solidi cristallini. PhD
thesis, University of Torino, Torino, 1993.
[5] T. Asthalter, W. Weyrich, A.H. Harker, A.B. Kunz, R. Orlando, and
C. Pisani. Comparison of quasi-hartree-fock wave-functions for lithium
hydride. Solid State Comm., 83:725, 1992.
[6] A. Veillard B. Roos, C. Salez and E. Clementi. A general program for
calculation of atomic SCF orbitals by the expansion method. Tech. Rep.
RJ518 IBM Res., 1968.
[7] E. B. Barrosa, A. Joriob, G. G. Samsonidzef, R. B. Capazc, A. G. Souza
Filhoa, J. Mendes Filhoa, G. Dresselhause, and M. S. Dresselhaus. Re-
view on the symmetry-related properties of carbon nanotubes. Phys. Rep.,
431:261–302 (and references therein), 2006.
[8] J.C. Barthelat. Private communication.
[9] J.C. Barthelat and P. Durand. Gazz. Chim. Ital., 108:225, 1978.
[10] J.C. Barthelat, P. Durand, and A. Serafini. Molec. Phys., 33:159, 1977.
[11] A. D. Becke. J. Chem. Phys., 88:2547, 1988.
[12] A. D. Becke. J. Chem. Phys., 88:1053, 1988.
[13] A. D. Becke. Density-functional exchange-energy approximation with cor-
rect asymptotic behavior. Phys. Rev. A, 38:3098–3100, 1988.
[14] A. D. Becke. Density-functional thermochemistry. iii the role of exact ex-
change. J. Chem. Phys., 98:5648, 1993.
[15] J.S. Binkley, J.A. Pople, and W.J. Hehre. J. Am. Chem. Soc., 102:939,
1980.
[16] Y. Bouteiller, C. Mijoule, M. Nizam, J.C. Barthelat, J.P. Daudey, M. Pel-
lissier, and B. Silvi. Mol. Phys., 65:295, 1988.

283
[17] S.F. Boys and F. Bernardi. The calculation of small molecular interaction
by the difference of separate total energies. some procedures with reduced
errors. Mol. Phys., 19:553, 1970.
[18] C. G. Broyden. Math. Comput., 19:577, 1965.
[19] C.G. Broyden. The convergence of a class of double-rank minimization
algorithms 1. general considerations. IMA J. Appl. Math., 6:76–90, 1970.
[20] C.G. Broyden. The convergence of a class of double-rank minimization
algorithms 1. the new algorithm. IMA J. Appl. Math., 6:222–231, 1970.
[21] H. Burzlaff and A. Hountas. J. Appl. Cryst., 15:464, 1982.
[22] S. Casassa, C. M. Zicovich-Wilson, and C. Pisani. Symmetry-adapted lo-
calized wannier functions suitable for periodic calculations. Theor. Chem.
Acc., 116:726–733, 2006.
[23] M. Catti, R. Dovesi, A. Pavese, and V. R. Saunders. Elastic constants and
electronic structure of fluorite (caf2 ): an ab initio hartree-fock study. J.
Phys.: Condens. Matter, 3:4151, 1991.
[24] M. Causà, R. Dovesi, R. Orlando, C. Pisani, and V.R. Saunders. Treatment
of the exchange interactions in hartree-fock lcao calculations of periodic
systems. J. Phys. Chem., 92:909, 1988.
[25] M. Causà, R. Dovesi, C. Pisani, and C. Roetti. Electronic structure and
stability of different crystal phases of magnesium oxide. Phys. Rev. B,
33:1308, 1986.
[26] M. Causà, R. Dovesi, and C. Roetti. Pseudopotential hartree-fock study
of seventeen iii-v and iv-iv semiconductors. Phys. Rev. B, 43:11937, 1991.
[27] B. Civalleri, C. M. Zicovich-Wilson, L. Valenzano, and P. Ugliengo. B3lyp
augmented with an empirical dispersion term (b3lyp-d*) as applied to
molecular crystal. Crys. Eng. Comm., 10:405–410, 2008.
[28] B. Civalleri, C. M. Zicovich-Wilson, L. Valenzano, and P. Ugliengo. Errata
corrige to: b3lyp augmented with an empirical dispersion term (b3lyp-d*)
as applied to molecular crystals. cryst. eng. comm., 10, 405-410 (2008).
Cryst. Eng. Comm., 10:1693–1694, 2008.
[29] E. Clementi. MOTECC 91. ESCON, Leiden, 1991.
[30] S. Dall’Olio, R. Dovesi, and R. Resta. Spontaneous polarization as a berry
phase of the hartree-fock wave function: the case of knbo3 . Phys. Rev.,
B56:10105–10114, 1997.
[31] P. D’Arco, Y. Noel, R. Demichelis, and R. Dovesi. Single-layered chrysotile
nanotubes: A quantum mechanical ab initio simulation. J. Chem. Phys.,
131:Art. N. 204701, 2009.
[32] C. Darrigan, M. Rerat, G. Mallia, and R. Dovesi. Implementation of the
finite field perturbation method in the crystal program for calculating the
dielectric constant of periodic systems. J. Comp. Chem., 24:1305–1312,
2003.

284
[33] R. Demichelis, Y. Noel, P. D’Arco, and R. Dovesi. Structure and energetics
of imogolite: a quantum mechanical ab initio study with the CRYSTAL
code. 2010.
[34] P. A. M. Dirac. Proc. Cambridge Phil. Soc., 26:376, 1930.
[35] R. Dovesi. On the role of symmetry in the ab initio hartree-fock linear com-
bination of atomic orbitals treatment of periodic systems. Int. J. Quantum
Chem., 29:1755, 1986.
[36] R. Dovesi, M. Causà, R. Orlando, C. Roetti, and V.R. Saunders. Ab initio
approach to molecular crystals: a periodic hartree-fock study of crystalline
urea. J. Chem. Phys., 92:7402, 1990.
[37] R. Dovesi, E. Ferrero, C. Pisani, and C. Roetti. Ab initio study of the
electron momentum distribution of metallic lithium. Z. Phys. B, 51:195,
1983.
[38] R. Dovesi and R. Orlando. Convergence properties of the supercell ap-
proach in the study of local defects in solids. Phase Trans., 52:151, 1994.
[39] R. Dovesi, C. Pisani, C. Roetti, M. Causà, and V.R. Saunders. CRYS-
TAL88, An ab initio all-electron LCAO-Hartree-Fock program for periodic
systems. QCPE Pgm N.577. Quantum Chemistry Program Exchange, In-
diana University, Bloomington, Indiana, 1989.
[40] R. Dovesi, C. Pisani, C. Roetti, and V.R. Saunders. Treatment of coulomb
interactions in hartree-fock calculations of periodic systems. Phys. Rev. B,
28:5781, 1983.
[41] R. Dovesi, C. Roetti, C. Freyria Fava, M. Prencipe, and V.R. Saunders. On
the elastic properties of lithium, sodium an potassium oxide. an ab initio
study. Chem. Phys., 156:11, 1991.
[42] R. Dovesi, C. Roetti, and V.R. Saunders. CRYSTAL92 User’s Manual.
Università di Torino and SERC Daresbury Laboratory, 1992.
[43] R. Dovesi, V.R. Saunders, C. Roetti, M. Causà, N. M. Harrison, R. Or-
lando, and E. Aprà. CRYSTAL95 User’s Manual. Università di Torino,
1996.
[44] R. Dovesi, V.R. Saunders, C. Roetti, R. Orlando, C.M. Zicovich-Wilson,
F. Pascale, B. Civalleri, K. Doll, N.M. Harrison, I.J. Bush, Ph. DArco, and
M. Llunell. CRYSTAL06 User’s Manual. Università di Torino, Torino,2006.
[45] P. Durand and J.C. Barthelat. Theor. Chim. Acta, 38:283, 1975.
[46] V.R. Saunders et al. Electronic structure theory: from molecules to crystals.
private communication, 1999.
[47] M. Ferrero, M. Rérat, R. Orlando, and R. Dovesi. The calculation of
static polarizabilities in 1-3d periodic compounds. the implementation in
the crystal code. J. Comput. Chem., 29:1450–1459, 2008.

285
[48] M. Ferrero, M. Rérat, R. Orlando, and R. Dovesi. Coupled perturbed
hartree-fock for periodic systems: the role of symmetry and related com-
putational aspects. J. Chem. Phys., 128:Art. N. 014100, 2008.
[49] M. Ferrero, M. Rérat, R. Orlando, R. Dovesi, and I. J. Bush. coupled
perturbed kohn-sham calculation of static polarizabilities of periodic com-
pounds. J. Phys.: Conf. Ser., 117:Art. N. 012016, 2008.
[50] R. Fletcher. Comput. J, 13:317, 1970.
[51] M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb,
J. R. Cheeseman, J. A. Montgomery, Jr., T. Vreven, K. N. Kudin, J. C.
Burant, J. M. Millam, S. S. Iyengar, J. Tomasi, V. Barone, B. Mennucci,
M. Cossi, G. Scalmani, N. Rega, G. A. Petersson, H. Nakatsuji, M. Hada,
M. Ehara, K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima,
Y. Honda, O. Kitao, H. Nakai, M. Klene, X. Li, J. E. Knox, H. P. Hratchian,
J. B. Cross, V. Bakken, C. Adamo, J. Jaramillo, R. Gomperts, R. E. Strat-
mann, O. Yazyev, A. J. Austin, R. Cammi, C. Pomelli, J. W. Ochterski,
P. Y. Ayala, K. Morokuma, G. A. Voth, P. Salvador, J. J. Dannenberg,
V. G. Zakrzewski, S. Dapprich, A. D. Daniels, M. C. Strain, O. Farkas,
D. K. Malick, A. D. Rabuck, K. Raghavachari, J. B. Foresman, J. V. Or-
tiz, Q. Cui, A. G. Baboul, S. Clifford, J. Cioslowski, B. B. Stefanov, G. Liu,
A. Liashenko, P. Piskorz, I. Komaromi, R. L. Martin, D. J. Fox, T. Keith,
M. A. Al-Laham, C. Y. Peng, A. Nanayakkara, M. Challacombe, P. M. W.
Gill, B. Johnson, W. Chen, M. W. Wong, C. Gonzalez, , and J. A. Pople.
Gaussian 03, Revision 02. Gaussian, Inc., Wallingford CT, 2004, 2004.
[52] M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb,
J. R. Cheeseman, V. G. Zakrzewski, J. A. Montgomery, Jr., R. E. Strat-
mann, J. C. Burant, S. Dapprich, J. M. Millam, A. D. Daniels, K. N.
Kudin, M. C. Strain, O. Farkas, J. Tomasi, V. Barone, M. Cossi, R. Cammi,
B. Mennucci, C. Pomelli, C. Adamo, S. Clifford, J. Ochterski, G. A. Pe-
tersson, P. Y. Ayala, Q. Cui, K. Morokuma, D. K. Malick, A. D. Rabuck,
K. Raghavachari, J. B. Foresman, J. Cioslowski, J. V. Ortiz, B. B. Ste-
fanov, G. Liu, A. Liashenko, P. Piskorz, I. Komaromi, R. Gomperts, R. L.
Martin, D. J. Fox, T. Keith, M. A. Al-Laham, C. Y. Peng, A. Nanayakkara,
C. Gonzalez, M. Challacombe, P. M. W. Gill, B. Johnson, W. Chen, M. W.
Wong, J. L. Andres, C. Gonzalez, M. Head-Gordon, E. S. Replogle, and
J. A. Pople. Gaussian 98, Revision A.3. Pittsburgh PA, 1998.
[53] M.J. Frisch, G.W. Trucks, H.B. Schlegel, P.M.W. Gill, B.G. Johnson,
M.A. Robb, J.R. Cheeseman, T.A. Keith, G.A. Pettersson, J.A. Mont-
gomery, K. Raghavachari, M.A. Al-Laham, V.G. Zakrzewski, J.V. Ortiz,
J.B. Foresman, J. Cioslowski, B.B. Stefanov, A. Nanayakkara, M. Challa-
combe, C.Y. Peng, P.Y. Ayala, W. Chen, M.W. Wong, J.L. Andres, E.S.
Replogle, R. Gomperts, R.L. Martin, D.J. Fox, J.S. Binkley, D.J. Defrees,
J. Baker, J.J.P. Stewart, M. Head-Gordon, C. Gonzalez, , and J.A. Pople.
Gaussian 94 (Revision A.I). Gaussian Inc., Pittsburgh PA, 1995.
[54] L. Fu, E. Yaschenko, L. Resca, and R. Resta. Hartree–fock approach
to macroscopic polarisation: Dielectric constant and dynamical charge of
knbo3 . Phys. Rev. B, 57:6967–6971, 1988.

286
[55] C. Giacovazzo. Fundamentals of Crystallography. IUC-Oxford University
Press, Oxford, 1992.
[56] G. Gilat. J. Comp. Phys., 10:432, 1972.
[57] G. Gilat and L.J. Raubenheimer. Accurate numerical method for calculat-
ing frequency-distribution in solids. Phys. Rev., 144:390, 1966.
[58] P.M.W. Gill, B.G. Johnson, and J.A. Pople. A standard grid for density
function calculations. Chem. Phys. Lett., 209:506–512, 1993.
[59] D. Goldfarb. Math. Comput., 24:23, 1970.
[60] M.S. Gordon, J.S. Binkley, J.A. Pople, W.J. Pietro, and W.J. Hehre.
J. Am. Chem. Soc., 104:2797, 1982.
[61] S. Grimme. Semiempirical gga-type density functional constructed with a
long range dispersion correction. J. Comput. Chem, 27:1787–, 2006.
[62] M.F. Guest and V.R. Saunders. On methods for converging open-shell
hartree-fock wave-functions. Mol. Phys., 28:819–828, 1974.
[63] G. Igel-Mann H. Stoll and H. Preuss. Arbeitsbericht 27 des Instituts für
Theoretische Chemie, Universität Stuttgart, 1990.
[64] T. Hahn. International Tables for Crystallography. Reidel Publishing Com-
pany, 1987.
[65] D. R. Hamann. Semiconductor charge densities with hard-core and soft-
core pseudopotentials. Phys. Rev. Lett., 42:662–665, 1979.
[66] P.C. Hariharan and J.A. Pople. Theor. Chim. Acta, 28:213, 1973.
[67] P.J. Hay and W.R. Wadt. Ab initio effective core potentials for molec-
ular calculations. potentials for k to au including the outermost orbitals.
J. Chem. Phys., 82:299, 1985.
[68] P.J. Hay and W.R. Wadt. Ab initio effective core potentials for molecular
calculations. potentials for main group elements na to bi. J. Chem. Phys.,
82:284, 1985.
[69] P.J. Hay and W.R. Wadt. Ab initio effective core potentials for molecular
calculations. potentials for transition metal atoms sc to hg. J. Chem. Phys.,
82:270, 1985.
[70] D.J. Hehre, W.A. Lathan, M.D. Newton, R. Ditchfield, and J.A.
Pople. GAUSSIAN70 Program number 236, QCPE. Indiana Univer-
sity,Bloomington, Indiana, 1972.
[71] D.J. Hehre, L. Radom, P.v.R. Schleyer, and J.A. Pople. Ab initio molecular
orbital theory. Wiley, New York, 1986.
[72] W.J. Hehre, R. Ditchfield, R.F. Stewart, and J.A. Pople. J. Chem. Phys.,
52:2769, 1970.
[73] W.J. Hehre, R.F. Stewart, and J.A. Pople. J. Chem. Phys., 51:2657, 1969.

287
[74] P Hohenberg and W. Kohn. Inhomogeneous electron gas. Phys. Rev.,
136:B864, 1964.
[75] D. D. Johnson. Modified broyden’s method for accelerating convergence in
self-consistent calculations. Phys. Rev B, 38:12807–12813, 1988.
[76] W. Kohn and L.J. Sham. Self-consistent equations including exchange and
correlation effects. Phys. Rev., 140:A1133, 1965.
[77] L. Kronsjo. Algorithms, their complexity and efficiency, volume 2nd ed.
Wiley, New York, 1986.
[78] C. Lee, W. Yang, , and R. G. Parr. Development of the colle-salvetti
correlation-energy formula into a functional of the electron density. Phys.
Rev. B, 37:785–789, 1988.
[79] K.H. Lee, M. Causà, and S.S. Park. Ab initio periodic hartree-fock calcula-
tions for interpretation of the scanning tunneling microscope (stm) images
of graphite. J. Phys. Chem. B, 102:6020, 1998.
[80] B. Lindberg. J. Chem. Phys., 88:3805–xxxx, 1988.
[81] R. Lindh, A. Bernhardsson, G. Katlström, and P.A. Malmqvist. On the use
of a hessian model function in molecular geometry optimizations. Chem.
Phys. Letters, 241:423–428, 1995.
[82] V. Magnasco and A. Perico. J. Chem. Phys., 47:971, 1967.
[83] N.D. Mermin. Thermal properties of the inhomogeneous electron gas. Phys.
Rev. A, 137:1441, 1965.
[84] D. Moncrieff and V.R. Saunders. Atmol. University of Manchester Regional
Computing Centre, Doc NAT648, 1986.
[85] H.J. Monkhorst and J.D. Pack. Phys. Rev., 13:5188, 1976.
[86] R. Orlando, R. Dovesi, C. Roetti, and V.R. Saunders. Ab initio hartree-
fock calculations of periodic compounds: application to semiconductors. J.
Phys. Condens. Matter, 2:7769, 1990.
[87] D. Viterbo P. Ugliengo and G. Chiari. Moldraw: Molecular graphics on a
personal computer. Zeitschrift fur Kristallographie, 207:9, 1993.
[88] J. P. Perdew. Density-functional approximation for the correlation energy
of the inhomogeneous electron gas. Phys. Rev. B, 33:8822, 1986.
[89] J. P. Perdew. Electronic Structure of Solids 1991. Akademie Verlag, Berlin,
1991.
[90] J. P. Perdew, K. Burke, and M. Ernzerhof. Generalized gradient approxi-
mation made simple. Phys. Rev. Lett., 77:3865–3868, 1996.
[91] J. P. Perdew and Y. Wang. Accurate and simple analytic representation
of the electron gas correlation energy. Phys. Rev. B, 45:13244, 1992.
[92] J. P. Perdew and Wang Yue. Accurate and simple density functional
for the electronic exchange energy: Generalized gradient approximation.
Phys. Rev. B, 33:8800–8802, 1986.

288
 [93] J. P. Perdew and Wang Yue. Erratum: Accurate and simple density func-
tional for the electronic exchange energy: Generalized gradient approxima-
tion. Phys. Rev. B, 40:3399, 1989.
[94] J. P. Perdew and A. Zunger. Self-interaction correction to density-
functional approximations for many-electron systems. Phys. Rev. B,
23:5048–5079, 1981.
[95] J.P. Perdew, J.A. Chevary, S.H. Vosko, K.A. Jackson, M.R. Pederson, D.J.
Singh, and C. Fiolhais. Atoms, molecules, solids and surfaces: applications
of the generalized gradient approximation for exchange and correlation.
Phys. Rev. B, 46:6671, 1992.
[96] J.P. Perdew, A. Ruzsinszky, G.I. Csonka, O.A. Vydrov, G.E. Scuseria, L.A.
Constantin, X. Zhou, and K. Burke. Phys. Rev. Letters, 100:Art. N. 184109,
2008.
[97] W.J. Pietro, E.S. Blurock, R.F. Hout, W.J. Hehre, W.J. DeFrees, and R.F.
Stewart. Inorg. Chem., 20:3650, 1981.
[98] W.J. Pietro, B.A. Levi, W.J. Hehre, and R.F. Stewart. Inorg. Chem.,
19:2225, 1980.
[99] J. Pipek and P. G. Mezey. J. Chem. Phys., page 4916, 1989.
[100] C. Pisani. Quantum-mechanical ab-initio calculation of the properties of
crystalline materials. volume 67 of Lecture Notes in Chemistry, Berlin,
1996. Springer Verlag.
[101] C. Pisani, E. Aprà, , and M. Causà. Density matrix of crystalline systems.
i. long range behavior and related computational problems. Int. J. Quan-
tum Chem., 38:395, 1990.
[102] C. Pisani, E. Aprà, M. Causà, and R. Orlando. Density matrix of crystalline
systems. ii. the influence of structural and computational parameters. Int.
J. Quantum Chem., 38:419, 1990.
[103] C. Pisani, R. Dovesi, and C. Roetti. Hartree-Fock ab initio Treatment of
Crystalline Systems, volume 48 of Lecture Notes in Chemistry. Springer
Verlag, Heidelberg, 1988.
[104] J.A Pople and R.K. Nesbet. Self-consistent orbitals for radicals.
J. Chem. Phys., 22:571, 1954.
[105] M.J.D Powell, editor. Academic Press, New York, 1982.
[106] M. Rérat, M. Ferrero, E. Amzallag, I. Baraille, and R. Doveesi. Comparison
of the polarizability of periodic systems computed by using the length and
velocity operators. J. Phys.: Conf. Ser., 117:Art. N. 012023, 2008.
[107] R. Resta. The geometric phase approach to macroscopic polarization in
crystalline dielectrics. Rev. Mod. Phys., 66:809, 1994.
[108] C.C.J. Roothaan. Self-consistent field theory for open shells of electronic
systems. Rev. Mod. Phys., 32:179–185, 1960.

289
[109] L. Salasco, R. Dovesi, R. Orlando, C. Pisani, M. Causà, and V.R. Saunders.
A periodic ab initio extended basis set study of α-al2 o3 . Mol. Phys., 72:267,
1992.
[110] V. R. Saunders. From molecules to solids. Book in preparation.
[111] V.R. Saunders. Ab initio hartree-fock calculations for periodic systems.
Faraday Symp. Chem. Soc., 19:79–84, 1984.
[112] V.R. Saunders, R. Dovesi, C. Roetti, M. Causà, N. M. Harrison, R. Or-
lando, and C.M. Zicovich-Wilson. CRYSTAL98 User’s Manual. Università
di Torino, Torino,1998.
[113] V.R. Saunders, R. Dovesi, C. Roetti, R. Orlando, C.M. Zicovich-Wilson,
N. M. Harrison, K. Doll, B. Civalleri, I. Bush, Ph. D’Arco, and M. Llunell.
CRYSTAL03 User’s Manual. Università di Torino, Torino,2003.
[114] V.R. Saunders, C. Freyria-Fava, R. Dovesi, L. Salasco, and C. Roetti. On
the electrostatic potential in crystalline systems where the charge density
is expanded in gaussian functions. Mol. Phys., 77:629, 1992.
[115] A. Savin. Int. J. Quantum Chem., S22:457, 1988.
[116] G. Schaftenaar. MOLDEN a pre- and post processing program of molecular
and electronic structure. CMBI, the Netherlands.
[117] H. B. Schlegel. Optimization of equilibrium and transition structures. J.
Comp. Chem., 3:214–218, 1982.
[118] H.B. Schlegel. Estimating the hessian for gradient-type geometry optimiza-
tions. Theoret. Chim. Acta, 66:333–340, 1984.
[119] B.G. Searle. Dl visualize. Comp. Phys. Communications, 137:25–32, 2001.
[120] D.F. Shanno. Math. Comput., 24:647, 1970.
[121] W.J. Stevens, H. Bash, and M. Krauss. Compact effective potentials and
efficient shared-exponent basis sets for the first- and second-row atoms. J.
Chem. Phys, 81:6026–6033, 1984.
[122] W.J. Stevens, M. Krauss, H. Basch, and P.G. Jasien. Relativistic compact
effective potentials and efficient, shared-exponent basis sets for the third-,
fourth-, and fifth-row atoms. Can. J. Chem, 70:612–630, 1992.
[123] J. Tersoff and D.R. Hamann. Theory of scanning tunneling microscope.
Phys. Rev. B, 31:805, 1985.
[124] M.D. Towler, A. Zupan, and M. Causà. Density functional theory in pe-
riodic systems using local gaussian basis sets. Comp. Phys. Commun.,
98:181–205, 1996.
[125] O. Treutler and R. Ahlrichs. Efficient molecular numerical integration
schemes. J. Chem. Phys., 102:346–354, 1995.

290
[126] P. Ugliengo, C.M. Zicovich-Wilson, S. Tosoni S, and B. Civalleri. Role of
dispersive interactions in layered materials: a periodic b3lyp and b3lyp-d*
study of mg(oh)2 , ca(oh)2 and kaolinite. J. Mater. Chem., 19:2564–2572,
2009.
[127] U. von Barth and L. Hedin. J. Phys. C: Solid State Phys., 5:1629, 1972.
[128] S. H. Vosko, L. Wilk, and M. Nusair. Accurate spin-dependent electron
liquid correlation energies for local spin density calculations: a critical anal-
ysis. Can. J. Phys., 58:1200, 1980.
[129] R. Dovesi V.R. Saunders, C. Freyria-Fava and C. Roetti. On the electro-
static potential in linear periodic polymers. Comp. Phys. Comm., 84:156,
1993.
[130] R. C. Weast, M. J. Astle, and W. H. Beyer, editors. CRC Handbook of
Chemistry and Physics, pages A–61 and A–62. CRC Press, Inc., Boca
Raton, Florida, 1986-87.
[131] J.A. Weil, J.R. Bolton, and E. Wertz. Electron ParamagneticResonance
- Elementary Theory and Practical Applications. John Wiley, New York,
1994.
[132] C. T. White, D. H. Robertson, and J. W. Mintmire. Helical and rotational
symmetries of nanoscale graphitic tubules. Phys. Rev. B, 47:5485–5488,
1993.
[133] L.C. Wilson and M. Levy. Nonlocal wigner-like correlation-energy density
functional through coordinate scaling. Phys. Rev. B, 41:12930–12932, 1990.
[134] J.M. Wittbrodt and H.B. Schlegel. Estimating stretching force constants
for geometry optimization. J. Mol. Structure - Theochem, 398-399:55–61,
1997.
[135] Zhigang Wu and R.E. Cohen. More accurate generalized gradient approx-
imation for solids. Phys. Rev. B, 73:Art. n. 235116, 2006.
[136] Y. Zhang and D.G. Truhlar. J. Chem. Phys., 128:Art. N. 184109, 2008.
[137] C. Zicovich-Wilson and R. Dovesi. On the use of symmetry adapted crys-
talline orbitals in scf-lcao p eriodic calculations.
ii. implementation of the self-consistent-field scheme and examples. Int. J.
Quantum Chem., 67:311–320, 1998.
[138] C. Zicovich-Wilson and R. Dovesi. On the use of symmetry adapted crys-
talline orbitals in scf-lcao periodic calculations.
i. the construction of the symmetrized orbitals. Int. J. Quantum Chem,
67:299–310, 1998.
[139] C. M. Zicovich-Wilson and R. Dovesi. Localized functions in crystalline
systems and their variational manifolds. Beyond standard Quantum Chem-
istry: From molecules to extended systems, page Chapter 8, India, 2007.
Transworld Research Network.

291
[140] C.M. Zicovich-Wilson, R. Dovesi, and V.R. Saunders. A general method
to obtain well localized wannier functions for composite energy bands in
lcao periodic calculations. J. Chem. Phys., 115:9708–9718, 2001.

292
Index

ADFT see EDFT AUTO(ROTCRY), 54

NEWBASIS, 154 B3LYP(DFT), 76
ALLOWTRUSTR(OPTGEOM), 111 B3PW(DFT), 76
ALL(ANISOTRO), 155 BANDLIST(LOCALWF), 168
ANALYSIS(FREQCALC), 128 BAND, 156
ANBD, 154 BARTHE, 65
ANDERSON, 71 BASE(FIXINDEX), 86
ANGLES, 30 BASISSET, 156
ANGROT(ROTCRY), 54 BATCHPNT(DFT), 80
ANGSFROZEN(OPTGEOM), 120 BECKE(DFT), 75, 77
ANGSTROM, 31, 155 BECKE(EDFT), 154, 164
ANGTODOUBLE(OPTGEOM), 115 BERNY(OPTGEOM), 110
ANGULAR(DFT), 78 BETALOCK, 73
ANGULAR(EDFT), 154, 164 BFGS(OPTGEOM), 110
ANHARM, 141 BIDIERD, 157
END, 141 CONV, 157
ISOTOPES, 142 DIFF, 157
KEEPSYMM, 142 DIR, 157
NOGUESS, 142 END, 157
POINTS26, 142 PROF, 157
PRINTALL, 142 BIESPLIT, 73
PRINT, 142 BIPOLAR, 73
TEST[ANHA], 142 BIPOSIZE, 73
ANISOTRO BOHRANGS, 34
ALL, 155 BOHRCR98, 34
PRINT, 155 BOHR, 34, 157
SELECT, 155 BOYSCTRL(LOCALWF, 170
UNIQUE, 155 BREAKELAS, 35
ATOMBSSE, 31 BREAKSYM, 26, 34
ATOMDISP, 32 BROYDEN, 73
ATOMHF, 72 BR(PROF), 185
ATOMINSE, 32 BUNITSDECO(FREQCALC), 131
ATOMIRR, 155 BWIDTH, 158
ATOMORDE, 32 CAPTURE(LOCALWF), 171
ATOMREMO, 32 CELLONLY(OPTGEOM), 109
ATOMROT, 33 CHARGED, 62, 158
ATOMSPIN, 72 CHARGE (GRID3D), 166
ATOMSUBS, 34 CHEMOD, 62
ATOMSYMM, 34, 62, 155 CHNGTSFOL(OPTGEOM), 124
ATOMS(EIGSHROT), 82 CHUNKS(DFT), 81
ATOMS(MAPNET), 176 CLAS, 158

293
CLUSPLUS(LOCALWF), 173 PWLSD, 75
CLUSTER, 36 PZ, 75
CMPLXFAC, 104 RADIAL, 78
CONFCNT, 146 RADIUS, 81
CONV(BIDIERD), 157 REPLGRID, 80
CONV(PROF), 185 SAVIN, 77
COORDINA(MAPNET) , 176 SOGGAXC, 76
COORPRT, 37, 158 SOGGA, 75
CORRELAT(DFT), 74 SPIN, 75
CPHF, 144 TOLLDENS, 80
END, 144 TOLLGRID, 80
MAXCYCLE, 144 VBH, 75
RESTART, 144 VWN, 75
TOLALPHA, 144 WCGGA, 75
TOLUDIK, 144 WL, 75
CP(PROF), 185 XLGRID, 79
CRYAPI OUT, 159 XXLGRID, 79
CRYDEF(OPTGEOM), 112 DIELISO(FREQCALC), 128
CRYSTAL, 13 DIELTENS(FREQCALC), 128
CVOLOPT(OPTGEOM), 117 DIEL/DIELECT, 159
CYCTOL(LOCALWF), 169 DIFF(BIDIERD), 157
DBANGLIST(OPTGEOM), 116 DIFF(PROF), 185
DCDIAG, 104 DIPOMOME(FREQCALC), 134
DEFANGLS(OPTGEOM), 116 DIR(BIDIERD), 157
DEFLNGS(OPTGEOM), 116 DISPERSION(FREQCALC), 143
DENSMAT, 158 DISTGRID(DFT), 79
DFT DLVINPUT, 15
ANGULAR, 78 DOSS, 161
B3LYP, 76 DURAND, 65
B3PW, 76 ECH3, 162
BATCHPNT, 80 RANGE, 162
BECKE, 75, 77 SCALE, 162
CHUNKS, 81 ECHG, 163
CORRELAT, 74 EDFT, 163
DISTGRID, 79 ANGULAR, 154, 164
END[DFT], 74 BECKE, 154, 164
EXCHANGE, 75 PRINTOUT, 154, 164
FCHARGE, 81 PRINT, 154, 164
HYBRID, 76 RADIAL, 154, 164
LDA, 75 SAVIN, 154, 164
LGRID, 78 EIGSHIFT, 82
LYP, 75 EIGSHROT, 82
NEWTON, 81 ATOMS, 82
NONLOCAL, 76 MATRIX, 82
PB86, 75 EIGS, 81
PBE0, 76 ELASTCON, 148
PBESOL, 75 END, 148
PBE, 75 NUMDERIV, 148
PWGGA, 75 PREOPTGEOM, 148

294
 PRINT, 148 FIXDELTE(OPTGEOM), 112
RESTART, 148 FIXDELTX(OPTGEOM), 112
STEPSIZE, 148 FIXINDEX, 86
ELASTIC, 38 BASE, 86
EMDLPG, 165 GEBA, 86
EMDL, 164 GEOM, 86
EMDP, 165 FMIXING, 87
END FMWF, 166
DFT, 74 FRACTCOOR(OPTGEOM), 112
ENDB basis set input, 63 FRACTION, 166
ENDG geometry input, 39 geometry input, 44
general information input, 74 FRACTION(OPTGEOM), 112
properties input, 166 FRACTIOO(OPTGEOM), 112
END(ANHARM), 141 FRAGMENT(FREQCALC), 128
END(BIDIERD), 157 FRAGMENT(OPTGEOM), 119, 208
END(CPHF), 144 FREEZDIH(OPTGEOM), 121
END(ELASTCON), 148 FREEZINT(OPTGEOM), 120
END(EOS), 39 FREQCALC, 126
END(FREQCALC), 127 ANALYSIS, 128
END(OPTGEOM), 105 BUNITSDECO, 131
END(PROF), 185 DIELISO, 128
END(SYMMWF), 171 DIELTENS, 128
EOS, 39 DIPOMOME, 134
END, 39 DISPERSION, 143
PREOPTGEOM, 39 END, 127
PRINT, 39 FRAGMENT, 128
RANGE, 39 INTENS, 132
RESTART, 39 INTLOC, 133
EXCHANGE(DFT), 75 INTPOL, 133
EXCHGENE, 83 ISOTOPES, 128
EXCHSIZE, 83 MODES, 130
EXPDE(OPTGEOM), 112 NOANALYSIS, 130
EXTERNAL, 15, 263 NOINTENS, 130
EXTPRESS(OPTGEOM), 121 NOKSYMDISP, 143
EXTPRT, 41, 166 NOMODES, 130
EXTSTRESS(OPTGEOM), 121 NOOPTGEOM, 127
FCHARGE(DFT), 81 NORMBORN, 130
FDAOSYM, 83 NOUSESYMM, 130
FDOCCUP, 84 NUMDERIV, 130
FIELDCON, 43 PREOPTGEOM, 127
FIELD, 41 PRESSURE, 130
FINALRUN (OPTGEOM), 112, 209 PRINT, 130
FINDSYM, 44 REFLECTANCE, 139
FITDEGR(OPTGEOM), 113 RESTART, 131
FITTOPATH(OPTGEOM), 124 SCANMODE, 135
FIXCELL(OPTGEOM), 116 STEPSIZE, 131
FIXCOOR(OPTGEOM), 119 TEMPERAT, 131
FIXDEF(OPTGEOM), 118 TEST[FREQ], 131
FIXDEIND(OPTGEOM), 112 USESYMM, 131

295
FULLBOYS(LOCALWF), 174 LDA(DFT), 75
FULLOPTG(OPTGEOM), 109 LEVSHIFT, 92
FULLTIME, 103 LGRID(DFT), 78
GAUSS98, 63 LNGSFROZEN(OPTGEOM), 120
GEBA(FIXINDEX), 86 LOCALWF, 168
GEOM(FIXINDEX), 86 BANDLIST, 168
GHOSTS, 64 BOYSCTRL, 170
GRADCAL, 87 CAPTURE, 171
GRID3D, 166 CLUSPLUS, 173
CHARGE, 166 CYCTOL, 169
POTENTIAL, 166 FULLBOYS, 174
GRIMME, 88 IGSSBNDS, 171
GUESDUAL, 89 IGSSCTRL, 170
GUESSF, 91 IGSSVCTS, 170
GUESSPAT, 91 INIFIBND, 168
GUESSP, GUESSP0, 91 MAXCYCLE, 170
HAYWLC, 65 OCCUPIED, 168
HAYWSC, 65 ORTHNDIR, 174
HELIX, 13 PHASETOL, 169
HESEVLIM(OPTGEOM), 113 PRINTPLO, 173
HESSFREQ(OPTGEOM), 109 RESTART, 169
HESSIDEN(OPTGEOM), 109 SYMMWF, 171
HESSMOD1(OPTGEOM), 109 VALENCE, 168
HESSMOD2(OPTGEOM), 109 WANDM, 174
HESSNUM(OPTGEOM), 110 LYP(DFT), 75
HESSOPT(OPTGEOM), 109 MAKESAED, 45
HYBRID(DFT), 76 MAPNET, 176
HYDROSUB, 44 ATOMS, 176
IGSSBNDS(LOCALWF), 171 COORDINA, 176
IGSSCTRL(LOCALWF), 170 MARGINS, 176
IGSSVCTS(LOCALWF), 170 PRINT, 176
INFOGUI, 167 RECTANGU, 176
INFO see INFOGUI, 167 MARGINS(MAPNET), 176
INIFIBND(LOCALWF), 168 MATRIX(EIGSHROT), 82
INPUT(ECP input), 65 MATROT(ROTCRY), 54
INTENS(FREQCALC), 132 MAXCYCLE
INTGPACK, 92 scf, 93
INTLOC(FREQCALC), 133 MAXCYCLE(CPHF), 144
INTPOL(FREQCALC), 133 MAXCYCLE(LOCALWF), 170
INTREDUN(OPTGEOM), 115 MAXCYCLE(OPTGEOM), 113
ISOTOPES(ANHARM), 142 MAXCYCLE(scf), 93
ISOTOPES(FREQCALC), 128 MAXITACE(OPTGEOM), 113
ISOTROPIC, 167 MAXTRADIUS(OPTGEOM), 111
ITACCONV(OPTGEOM), 113 MODEFOLLOW(OPTGEOM), 124
ITATOCEL(OPTGEOM), 109 MODES(FREQCALC), 130
KEEPSYMM, 26, 44 MODISYMM, 45
KEEPSYMM(ANHARM), 142 MOLDRAW, 45
KNETOUT (obsolete), 168 MOLEBSSE, 45
KSYMMPRT, 92 MOLECULE, 13

296
 from 3D structure, 46 OPTGEOM, 105
MOLEXP, 46 ALLOWTRUSTR, 111
MOLSPLIT, 47 ANGSFROZEN, 120
MONSPLIT, 93 ANGTODOUBLE, 115
MPP, 93 BERNY, 110
MULPOPAN, 96, 184 BFGS, 110
MYBIPOLA, 94 CELLONLY, 109
NANOJMOL, 51 CHNGTSFOL, 124
NANORE, 51 CRYDEF, 112
NANOTUBE, 47 CVOLOPT, 117
NEGLEFRE DBANGLIST, 116
NEGLEFRE, 130 DEFANGLS, 116
NEIGHBOR, 51, 95, 177 DEFLNGS, 116
NEIGHPRT see NEIGHBOR, 51 END, 105
NEWBASIS(ADFT), 154 EXPDE, 112
NEWK, 178 EXTPRESS, 121
NEWTON(DFT), 81 EXTSTRESS, 121
NOANALYSIS(FREQCALC), 130 FINALRUN, 112, 209
NOBICOU, 95 FITDEGR, 113
NOBIPEXC, 95 FITTOPATH, 124
NOBIPOLA, 95 FIXCELL, 116
NOFMWF, 95 FIXCOOR, 119
NOGUESS(ANHARM), 142 FIXDEF, 118
NOGUESS(OPTGEOM), 113 FIXDEIND, 112
NOINTENS(FREQCALC), 130 FIXDELTE, 112
NOKSYMDISP(FREQCALC), 143 FIXDELTX, 112
NOMODES(FREQCALC), 130 FRACTCOOR, 112
NOMONDIR, 95 FRACTION, 112
NONLOCAL(DFT), 76 FRACTIOO, 112
NOOPTGEOM(FREQCALC), 127 FRAGMENT, 119, 208
NOPRINT, 64 FREEZDIH, 121
NORMBORN(FREQCALC), 130 FREEZINT, 120
NOSHIFT, 52 FULLOPTG, 109
NOSYMADA, 95, 179 HESEVLIM, 113
NOSYMAP(SYMMWF), 171 HESSFREQ, 109
NOSYMMOPS(OPTGEOM), 114 HESSIDEN, 109
NOTRUSTR(OPTGEOM), 111 HESSMOD1, 109
NOUSESYMM(FREQCALC), 130 HESSMOD2, 109
NOXYZ(OPTGEOM), 114 HESSNUM, 110
NRSTEPS(OPTGEOM), 113 HESSOPT, 109
NUMDERIV(ELASTCON), 148 INTREDUN, 115
NUMDERIV(FREQCALC), 130 ITACCONV, 113
NUMGRALL(OPTGEOM), 114 ITATOCEL, 109
NUMGRATO(OPTGEOM), 114 LNGSFROZEN, 120
NUMGRCEL(OPTGEOM), 114 MAXCYCLE, 113
Nanotube rebuild, 50 MAXITACE, 113
OCCUPIED(LOCALWF), 168 MAXTRADIUS, 111
OLDCG(OPTGEOM), 110 MODEFOLLOW, 124
ONELOG(OPTGEOM), 114 NOGUESS, 113

297
 NOSYMMOPS, 114 POLI, 181
NOTRUSTR, 111 POLSPIN, 181
NOXYZ, 114 POLYMER, 13
NRSTEPS, 113 POSTSCF, 96
NUMGRALL, 114 POT3, 182
NUMGRATO, 114 RANGE, 182
NUMGRCEL, 114 SCALE, 182
OLDCG, 110 POTC, 183
ONELOG, 114 POTENTIAL (GRID3D), 166
PATHFOLLOW, 124 POTM, 184
POWELL, 110 POWELL(OPTGEOM), 110
PRINTFORCES, 114 PPAN, 96, 184
PRINTHESS, 114 PREOPTGEOM(ELASTCON), 148
PRINTOPT, 114 PREOPTGEOM(EOS), 39
PRINT, 114 PREOPTGEOM(FREQCALC), 127
RENOSAED, 112 PRESSURE(FREQCALC), 130
RESTART, 113 PRIMITIV, 52
SORT, 113 PRINTALL(ANHARM), 142
STEPBMAT, 116 PRINTCHG, 53
STEPSIZE, 114 PRINTFORCES(OPTGEOM), 114
TESTREDU, 117 PRINTHESS(OPTGEOM), 114
TOLDEE, 110 PRINTOPT(OPTGEOM), 114
TOLDEG, 110 PRINTOUT, 53, 64, 96, 184
TOLDEX, 110 ATCOORDS, 255
TOLREDU, 117 ATOMICWF, 256
TRUSTRADIUS, 111 BASISSET, 256
TSOPT, 123 CONFIGAT, 256
ORIGIN, 52 DFTBASIS, 256
ORTHNDIR(LOCALWF), 174 DOSS, 256
P86(DFT), 75 EIGENALL, 256
PARAMPRT, 52, 64, 95, 179 EIGENVAL, 256
PATHFOLLOW(OPTGEOM), 124 EIGENVEC, 256
PATO, 179 ENECYCLE, 256
PBAN, 179 EQUIVAT, 255
PBE0(DFT), 76 EXCHGENE, 255
PBESOL(DFT), 75 FGIRR, 256
PBE(DFT), 75 FGRED, 255
PDIBAN see PBAN, 179 GAUSS94, 256
PDIDE, 180 GLATTICE, 255
PGEOMW, 180 KNETOUT, 255
PHASETOL(LOCALWF), 169 KSYMMPRT, 255
PHASE, 189 KWEIGHTS, 255
PIEZOBP, 188 MAPVALUES, 256
PIEZOWF, 188 MULLIKEN, 256
PMP2, 180 MULTIPOL, 256
POINTCHG, 52 OVERLAP, 255
POINTS26(ANHARM), 142 PARAMETERS, 255
POLARI, 189 PGIRR, 255
POLEORDR, 96 PGRED, 255

298
 ROTREF, 256 REFLSTEPS, 140
SCALEFAC, 256 REFLECTANCE(FREQCALC), 139
SYMMOPSR, 255 REFLECTANCEREFLDIR, 139
SYMMOPS, 255 REFLNSTEP(REFLECTANCE), 140
PRINTOUT(EDFT), 154, 164 REFLRANGE(REFLECTANCE), 140
PRINTPLO(LOCALWF), 173 REFLSTEPS(REFLECTANCE), 140
PRINT (DIEL), 159 RENOSAED(OPTGEOM), 112
PRINT(ANHARM), 142 REPLGRID(DFT), 80
PRINT(ANISOTRO), 155 RESTART(CPHF), 144
PRINT(EDFT), 154, 164 RESTART(ELASTCON), 148
PRINT(ELASTCON), 148 RESTART(EOS), 39
PRINT(EOS), 39 RESTART(FREQCALC), 131
PRINT(FREQCALC), 130 RESTART(LOCALWF), 169
PRINT(MAPNET), 176 RESTART(OPTGEOM), 113
PRINT(OPTGEOM), 114 RHF, 96
PRINT(SYMMWF), 171 ROTATE see SLABINFO, 56
PROF, 185 ROTCRY, 54
BR, 185 ANGROT, 54
CONV, 185 AUTO, 54
CP, 185 MATROT, 54
DIFF, 185 ROTREF, 186
END, 185 SAVEWF, 97
PROF(BIDIERD), 157 SAVIN(DFT), 77
PRSYMDIR, 53 SAVIN(EDFT), 154, 164
PSCF, 186 SCALE (ECH3), 162
PURIFY, 53 SCALE (POT3), 182
PWGGA(DFT), 75 SCANMODE(FREQCALC), 135
PWLSD(DFT), 75 SCELPHONO, 58
PZ(DFT), 75 SCFDIR, 96
QVRSGDIM, 104 SELECT(ANISOTRO), 155
RADIAL(DFT), 78 SETINF, 54, 64, 97, 187
RADIAL(EDFT), 154, 164 SETPRINT, 54, 64, 97, 187
RADIUS(DFT), 81 SHRINK, 97
RANGE (ECH3), 162 SLABCUT/SLAB, 55
RANGE (POT3), 182 SLABINFO, 56
RANGE(EOS), 39 SLAB, 13
RAYCOV/RAYC/RCOVFACT, 53, SMEAR, 99
186 SOGGAXC(DFT), 76
RDFMWF, 166 SOGGA(DFT), 75
RECTANGU(MAPNET), 176 SORT(OPTGEOM), 113
REDEFINE see SLABINFO, 56 SPINEDIT, 100
REFLANGLE(REFLECTANCE), 140 SPINLOCK, 101
REFLDAMP(REFLECTANCE), 140 SPIN(DFT), 75
REFLDIR(REFLECTANCE), 139 SPOLBP, 191
REFLECTANCE SPOLWF, 192
REFLANGLE, 140 STEPBMAT(OPTGEOM), 116
REFLDAMP, 140 STEPSIZE(ELASTCON), 148
REFLNSTEP, 140 STEPSIZE(FREQCALC), 131
REFLRANGE, 140 STEPSIZE(OPTGEOM), 114

299
STOP, 56, 65, 101, 187 VWN(DFT), 75
STRUCPRT, 56 WANDM(LOCALWF), 174
SUPERCEL, 56 WCGGA(DFT), 75
SUPERCON, 60 WL(DFT), 75
SWCNTRE, 51 XFAC, 187
SWCNT, 50 XLGRID(DFT), 79
SYMADAPT, 101, 187 XXLGRID(DFT), 79
SYMMDIR, 60 ZCOR see EDFT, 163
SYMMOPS, 60, 65 0D systems input, 14
SYMMREMO, 60 1D systems input, 14
SYMMWF 2D from 3D, 55
END, 171 2D systems input, 14
NOSYMAP, 171 3D systems input, 14
PRINT, 171
TOLBOND, 171 adjoined gaussian, 220
TOLSYM, 171 adsorbed molecule rotation, 33
SYMMWF(LOCALWF), 171 adsorption of molecules, 32
TEMPERAT(FREQCALC), 131 Anderson method for accelerating con-
TENSOR, 60 vergence, 71
TESTGEOM, 60 angles printing, 30
TESTPDIM, 101 angular integration (DFT), 78, 154, 164
TESTREDU(OPTGEOM), 117 anharmonic calculation, 141
TESTRUN, 101 anisotropic tensor, 155
TEST[ANHA](ANHARM), 142 anisotropy shrinking factor, 24, 99
TEST[FREQ](FREQCALC), 131 anti ferromagnetic systems, 103
TEST, 65 Aragonite, 194
TOLALPHA(CPHF), 144 asymmetric unit, 16
TOLBOND(SYMMWF), 171 ATMOL integral package, 92
TOLDEE, 102 atomic
TOLDEE(OPTGEOM), 110 density matrix, 72, 179
TOLDEG(OPTGEOM), 110 wave function, 72
TOLDEP, 102 atomic energy
TOLDEX(OPTGEOM), 110 (correlation) a posteriori, 154
TOLINTEG, 102 atomic number conventional, 18
TOLLDENS(DFT), 80 Atomic Orbital
TOLLGRID(DFT), 80 definition, 220
TOLPSEUD, 102 order, 21
TOLREDU(OPTGEOM), 117 atomic units
TOLSYM(SYMMWF), 171 bohr, 34
TOLUDIK(CPHF), 144 charge, 181
TRASREMO, 60 conversion factor, 34
TRUSTRADIUS(OPTGEOM), 111 atoms
TSOPT(OPTGEOM), 123 (group of) rotation, 33
UHF, 102 addition, 32
UNIQUE(ANISOTRO), 155 displacement, 32
USESAED, 61 removal, 32
USESYMM(FREQCALC), 131 reordering, 32
VALENCE(LOCALWF), 168 substitution, 34
VBH(DFT), 75 autocorrelation function theory, 226

300
band structure calculation, 156 Broyden method for accelerating con-
band width, 158 vergence, 73
BAND.DAT, 259 buffer
basis set, 214, 220 Coulomb bipolar expansion, 73
f and d AO occupation, 84 exchange bipolar expansion, 83
all electron, 19, 20 bulk modulus, 231
AO symmetry analysis, 83 BZ - Brillouin Zone, 221
criteria for selection, 214
crystal, 20 Calcite, 195
Effective Core Pseudopotential, 67 cell
input, 19 centred, 18
input examples, 200 charged, 22
libraries, 214 conventional, 17
linear dependence check, 81 conventional/primitive transforma-
metals, 217 tion, 243
orbital ordering, 21 crystallographic, 17
Pople, 19 minimum set parameters, 16
printing, 256 neutrality, 62
printing removed, 64 non neutral, 62
type, 19 primitive, 17, 52
valence only, 19, 20 redefinition, 56
basis set superposition error Cesium Chloride, 194
molecular, 45 Chabazite, 196
atomic, 31 check
periodic, 64 basis set input, 101
Beryllium slab, 197 complete input deck, 101
BF - Bloch Functions, 220 disk storage to allocate, 101
bi-electronic integrals geometry input, 60
indexing, 86 chemisorption, 32
bielectronic integrals Cholesky reduction, 81, 219
file split, 73 cluster expansion mapping of CRYS-
package, 92 TAL calculations, 146
bipolar expansion cluster from 3D, 36
bielectronic integrals, 73, 94, 224 CO - Carbon Monoxide
Coulomb buffer, 73 molecule, 200
coulomb integrals, 95 on MgO (001), 198
elimination, 95, 224 CO - Crystalline Orbital, 220
exchange buffer, 83 Compton profile
exchange integrals, 95 average, 227
Bloch Functions directional, 227
definition, 220 input, 185
Symmetry Adapted, 225 theory, 226
Symmetry Adapted - printing, 92 constraint sp, 220
Boys contour maps, 176
localization, 168 contraction
Bravais lattice, 17, 56, 243 coefficients, 19
Brillouin zone, 221 of gaussians, 19, 220
sampling, 23, 97, 226 conventional atomic number, 18–20
conventional cell, 17

301
convergence substitutional, 34
acceleration techniques, 24, 99 vacancies, 32
tools density functional
Anderson method, 71 see DFT, 74
Broyden method, 73 density matrix
dual basis, 89 atomic, 179
Fock/KS matrix mixing, 87 band projected, 179
level shifter, 92 behaviour, 223
convergence criteria core electrons, 152
cycles overflow, 93 direct space, 221
density matrix, 102 editing, 100
conversion factors, 34 energy projected , 180
length, 34 from geometrical weights, 180
conversion factors (CR98), 34 initial guess, 91
conversion wave function data, 166 output, 159
coordinates restore, 186
of equivalent atoms, 18 rotation, 186
output, 37, 158 valence electrons, 152
units, 31 density of states
angstrom, 155 calculation, 161
bohr, 34, 157 Fourier-Legendre expansion, 162,
fraction, 166 226
fractional, 44 integrated, 162
units of measure, 14 DFT
Corundum functionals, 74
(0001) surface, 197 Hamiltonian, 74
(1010) surface, 197 input, 74
bulk, 195 integration scheme, 77
Coulomb energy, 221 Diamond, 193
Coulomb series, 222 (100) Surface, 198
bielectronic contribution, 222 DIEL.DAT, 161, 259
Coulomb series threshold, 102 dielectric constant (optical), 159
Coupled Perturbed HF/KS dielectric tensor, 144
in a crystal, 144 dual basis, 89
through a slab, 144
covalent radii ECP - see Effective Core Pseudopoten-
customised, 53, 186 tial, 65
default value, 53 ECP input examples, 201
Crystalline Orbital (CO) Edingtonite, 196
definition, 220 Effective Core Pseudopotential
crystallographic cell, 17 input, 65
crystals BARTHE, 65
(3D) input, 14 DURAND, 65
Cuprite, 194 HAYWLC, 65
HAYWSC, 65
defects input examples, 201
displacement, 32 truncation criteria, 102
in supercell, 56 eigenvalues (Hamiltonian), 220
interstitial, 32 eigenvalues (Hamiltonian) printing, 253

302
eigenvectors equivalent atoms coordinates, 18
calculation, 178, 220 exchange energy
output, 159 calculation, 83
principal components printout, 154 theory, 221
printing, 253 exchange series threshold, 102, 223
printing (core), 254
rotation, 186 Faujasite, 196
elastic constant, 38, 228 Fermi contact, 167
elastic constants, 148 Fermi energy, 221, 226
elastic distortion, 38 smear, 99
elastic moduli theory, 228 findsym, 26
elastic strain, 228 FINDSYM.DAT, 26
elastic tensor, 229 Fluorite, 194
electric field, 184 Fock matrix
along non-periodic direction, 43 definition in direct space, 221
in a crystal, 41 Formamide polymer, 199
through a slab, 41 formatted data
electric susceptibility, 144 3D for visualization, 264
electron charge density for visualization, 263
3D maps, 162 general info, 264
calculation, 163 formatted files
gradient, 163 POINTCHG.INP, 52
electron momentum density formatted wave function, 95, 166
line, 164, 165 fort.33, 108
plane, 165 fort.34, 108
theory, 226 fragment
electron spin density, 163 frequency calculation, 128
electronic configuration geometry optimization, 119
ions, 22 frequencies, 58
open shell atoms, 21 frequency calculation at Γ, 126
electronic properties, 151 functionals
electrostatic potential DFT, 74
2D maps, 184 GAUSS70 integral package, 92
3D maps, 182 Gaussian 98 input deck, 63
first derivative, 183 gaussian primitives contraction, 220
maps, 158 gaussian type functions definition, 220
second derivative, 183 geometry
with an electric field, 183 exported, 56
EMD theory, 226 space group identification, 44
energy visualization, 41, 45, 166, 263
(correlation) a posteriori, 163 geometry optimization, 105
atomic, 22 cartesian coordinates, 106
Coulomb, 221 convergence criteria, 106
exchange (definition), 223 example, 205
exchange contribution, 83 Hessian update, 107
Fermi, 221 Initial hessian, 109
energy derivatives (elastic constants), partial, 119, 207
228 trust radius, 111
equation of state, 39

303
ghost atoms IS, 23, 97, 226
atoms converted to, 64 ISP, 24, 98, 226
input deck, 20 ITOL1, 102, 222
Gilat net, 97, 178 ITOL2, 102, 223
Graphite, 194, 197 ITOL3, 102, 223
Grimme dispersion correction, 88 ITOL4, 102, 223
ground state electronic properties, 151 ITOL5, 102, 223
groups - see symmetry groups, 235
GTF keywords list, 244
definition, 220 Kohn - Sham Hamiltonian, 74
primitives, 220 lattice
primitives-input, 19 centred, 18
Hamiltonian definition, 16
closed shell, 96 vectors, 18
DFT, 74 layer groups, 238
open shell, 102 LCAO, 220
Hamiltonian matrix Lebedev accuracy levels, 77
elements selective shift, 82 level shifter, 92
Hay and Wadt pseudo-potentials, 66 linear dependence catastrophe, 81, 218
hessian localization
default, 109 Boys, 168
from external file, 109 Wannier, 168
identity, 109 LPRINT, 253
model Lindh 1996, 109 mapping, 146
HESSOPT.DAT, 108 maps (contour), 176
hydrogen metals basis set, 217
(border atoms substitution with), Methane molecule, 200
44 MgO
anharmonic vibrations, 141 (001) surface, 198
hyperfine electron nucleus interaction (110) surface, 197
anisotropic, 155 molecular crystals
isotropic, 167 non interacting units, 47
hyperpolarizability tensors, 144 lattice parameters modification, 46
INF molecules
setting values, 54, 64, 97, 187 from 3D, 46
initial guess input, 14
atomic densities, 91 non interacting, 47
input density matrix, 91 Monkhorst net, 178, 226
input examples shrinking factor, 23, 97
0D geometry, 200 monoelectronic integral file split, 93
1D geometry , 199 MP2 Density Matrix, 180
2D geometry, 198 Mulliken population analysis, 96, 184
3D geometry, 196 multipolar expansion
basis set, 200 definition, 222
Effective Core Pseudopotential, 201 maximum order, 96
integral evaluation criteria, 221 multipole moments
integration in reciprocal space, 226 printing, 181
spin, 181

304
 calculation, 181 neighbour list, 51, 95, 177
ordering, 181 parametrized dimensions, 52
spherical harmonics, 181 setting environment, 53, 64, 96, 184
setting options, 54, 64, 97, 187
Nanotubes, 47, 50, 51 properties
neighbour printing, 51, 95, 177 ground state electronic, 151
NiO anti ferromagnetic - input, 101 pseudopotential
NOSYMAP, 172 Durand-Barthelat, 66
one electron integrals Hay and Wadt, 66
kinetic, 221 Stevens et al., 68
nuclear, 221 Stuttgart-Dresden, 68
OPTINFO.DAT, 108 Pyrite, 194
orientation convention radial integration (DFT), 78, 154, 164
polymer, 18 RCEP, 68
slab, 18 reciprocal form factor, 157, 226
origin reciprocal space integration, 226
moving, 52 reference frame rotation, 82
setting, 18 REFLECTANCE.DAT, 265
output files refractive index, 144
fort.33, 108 Restricted HF, 96
fort.34, 108 Rock Salt structure, 193
HESSOPT.DAT, 108 rod groups, 239
optaxxx, 108 rotation
OPTINFO.DAT, 108 of the crystal, 54
SCFOUT.LOG, 127 density matrix, 186
overlap matrix eigenvectors, 186
definition, 220 of adsorbed molecules, 33
printing, 254 reference frame, 82
phonon Rutile, 194
dispersion, 143 SAED – Symmetry Allowed Elastic
phonons, 58 Distortions, 45
physisorption, 32 scale factor, 19, 20
point charges SCF
input, 52 acceleration techniques, 24, 99
printing, 53 convergence
point groups, 242 density matrix, 102
polarizability tensor, 144 total energy, 102
polarization functions, 20 cycles control, 93
polymer direct bielectronic integrals, 96
input, 14, 15 dual basis, 89
orientation, 18 input, 23
population analysis (Mulliken), 96, 184 level shifter, 92
POTC.DAT, 183, 261 mixing Fock/KS matrices, 87
primitive cell, 17 SCF convergence acceleration
PRINT, 172 Anderson, 71
printing Broyden, 73
keywords, 255 dual basis, 89
multipole moments, 181

305
 level shifter, 92 surfaces
SCFOUT.LOG, 127 2D slab model, 14
geometry optimization, 108 slab from 3D structure, 55
shell symmetry
definition, 220 allowed directions, 60
formal charge, 19, 21 analysis in K space, 92
type, 19, 21 breaking, 26, 34, 35
shift of Hamiltonian matrix elements, electric field, 42
82 maintaining, 26, 44
shrinking factor, 23, 24, 97, 98, 226 modification, 45
Gilat, 97 point operators printing, 60
Pack Monkhorst, 97 point symmetry, 224
slab related atoms printing, 34
information, 56 removal, 60
input, 14 translational, 224
orientation, 18 translational components removal,
SN polymer, 199 60
Sodalite, 196 Symmetry Adapted Bloch Functions,
Sodium Chloride, 193 225
sp constraint, 220 symmetry groups
space group layer, 238
monoclinic input, 18 point groups, 242
orthorhombic input, 18 rod groups, 239
setting, 18 space, 235
symbol, 17 SYMMFLAG, 172
space groups tables, 235
spherical harmonic multipole moments, tensor of physical properties, 60
181 threshold
spin Coulomb series, 222
DFT, 75 exchange series, 223
multipole moments, 181 TOLBOND, 172
spin configuration tolerances
locking α − β electrons, 101 bipolar expansion, 73, 94
locking β electrons, 73 Coulomb series, 222
setting, 72 DFT, 80
spin density matrix editing, 100 Effective Core Pseudopotential, 102
spin polarized systems, 25 exchange series, 223
Spinel, 203 integrals, 102
spontaneous polarization, 189 ITOL1, 102, 222
Berry phase, 191 ITOL2, 102, 223
localized orbitals approach, 192 ITOL3, 102, 223
Stevens et al. pseudopotential, 68 ITOL4, 102, 223
STM topography, 180 ITOL5, 102, 223
STRUC.INCOOR, 262 TOLSYM, 172
structure factors, 187 total energy, 221
Stuttgart-Dresden pseudopotential, 68 transformation matrices in crystallog-
supercell raphy, 243
creation, 56, 58 two electron
input examples, 57 Coulomb contribution, 221

306
 exchange contribution, 221
units
Ångstrom, 31
bohr, 34
fractional, 44
Unrestricted HF, 102
Urea molecule, 200
vibrationa modes
scanning example, 210
vibrational modes, 126
visualization
geometry, 41, 166
MOLDRAW, 45
Voigt convention, 229
Wadt (see Hay), 66
Wannier functions, 168
Water chain, 199
Wurtzite, 193
X-ray structure factors, 187
Zeolites
Chabazite, 196
Edingtonite, 196
Faujasite, 196
Sodalite, 196
Zinc Blend, 193
Zirconia
cubic, 195
monoclinic, 195
tetragonal, 195

307